RU2777615C1 - Perceptual encoding of audio with adaptive non-uniform arrangement in time-frequency tiles using sub-band merging and spectral overlap reduction in the time domain - Google Patents

Abstract

FIELD: audio data processing.

SUBSTANCE: technical solution relates to the field of computer technology for processing audio data. Technical result is achieved by means of variants of an audio processor for processing audio signal, so as to obtain a sub-band representation of the audio signal with the possibility of performing a cascade overlapping critically sampled conversion, for partially overlapping audio signal sample blocks; wherein the cascade overlapping critically sampled conversion includes segmentation of a set of resolution elements; wherein the cascade overlapping critically sampled conversion with segmentation of a set of resolution elements where sets of resolution elements are processed with a second overlapping critically sampled conversion from a cascading overlapping critically sampled conversion.

EFFECT: reduction in the spectral overlap in the time domain of the audio signal due to the discarded redundant and perceptually irrelevant information.

26 cl, 27 dwg

Description

Embodiments relate to an audio processor/method for processing an audio signal to obtain a subband representation of an audio signal. Additional embodiments relate to an audio processor/method for processing a subband representation of an audio signal in order to obtain an audio signal. Some embodiments relate to perceptual audio coding with adaptive time-frequency tiling using sub-band combining and time-domain aliasing reduction. Some embodiments relate to a method for assisting in managing a plurality of non-uniform filter bank parameters and in extending the filter bank to a multi-channel mode of operation.

In perceptual coding, the entropy and therefore the bit rate are collectively reduced by discarding redundant and perceptually irrelevant information. This is achieved using a filter bank and quantization. This filterbank, quantizer and psychoacoustic model are used together to shape the quantization noise so that it is as close as possible to the masking threshold in order to maximize the coding efficiency and perceptual quality of the overall system [2].

During synthesis, quantization noise must be shaped in time and frequency by means of the spectral and temporal shape of the impulse and frequency response of the filterbank. It follows that, in order to allow highly detailed control over the shape of the quantization noise, it is desirable to use a filterbank with an impulse response that is compact in both time and frequency.

The most commonly used filterbank with these properties is the modified discrete cosine transform (MDCT), a filterbank that has uniform time-frequency resolution across all frequency bands.

However, the human auditory system exhibits non-uniform time-frequency resolution [3], resulting in different threshold masking patterns for different frequencies.

Therefore, it is an object of the present invention to provide a concept for driving a non-uniform filterbank that allows the quantization noise level to be increased while keeping audible artifacts low.

This problem is solved by independent claims.

Embodiments provide an audio processor for processing an audio signal so as to obtain a subband representation of the audio signal. The audio processor comprises a cascaded overlapped critically sampled transform stage configured to perform a cascaded overlapped critically sampled transform on at least two partially overlapping blocks of audio samples so as to obtain a set of subband samples based on the first block of audio samples and obtain a corresponding set of subband samples. based on the second block of audio samples. Additionally, the audio processor comprises a time domain aliasing reduction stage configured to weight-combine two respective sets of subband samples, one derived from the first block of audio samples and the other obtained from the second block of audio samples, so as to obtain a subband a reduced aliasing representation of the audio signal. Because of this, the cascaded overlapped critically sampled transform stage is configured to segment the bin set derived from the first block of samples using at least two window functions, and obtain at least two segmented bin sets based on segmented bin set corresponding to the first block of samples, wherein the cascaded overlapped critically sampled transform stage is configured to segment the bin set derived from the second block of samples using at least two window functions and obtain at least , two bin sets (e.g., at least two bands of combining coefficients) (e.g., 128 coefficients each) based on the segmented bin set corresponding to the second block of samples, and the bin sets o are processed (e.g., combined) using a second overlapped critically sampled transform stage from a cascaded overlapped critically sampled transform stage, wherein the second overlapped critically sampled transform stage is configured to perform overlapped critically sampled transforms having an identical frame length (e.g., pooling factor), at least one (eg, each) set of permission elements.

In embodiments, the second overlapped critical transform stage is configured to perform N _1.1 overlapped critically sampled transforms on N _1.1 subsets of the first set (e.g., 128 coefficients) from at least two bin sets derived from the segmented bin set corresponding to the first block of samples, where N _1.1 overlapping critically sampled transforms contain the same frame length (eg, pooling factor), where N _1.1 is a natural number greater than or equal to two.

In embodiments, the second overlapping critical transform stage is configured to perform N _1.2 overlapping critically sampled transforms for N _1.2 subsets (e.g., equal length) of the corresponding first set (e.g., 128 coefficients) of at least two sets bins derived from the segmented bin set corresponding to the second block of samples, where N _2.1 overlapping critically sampled transforms contain the same frame length (e.g., pooling factor), where N _2.1 is a natural number greater than or equal to two.

In embodiments, the audio processor is configured to individually select a frame length (eg, combining factor) for each bin set or for each of the respective bin sets.

For example, the processor may select for each set a combining factor, thereby achieving, for example, increased or even the highest possible coding efficiency.

In embodiments, the audio processor is configured to individually select a frame length (eg, pooling factor) for each block of samples.

For example, the processor may select for each block a combining factor, thereby achieving, for example, increased or even the highest possible coding efficiency.

In embodiments, the audio processor is configured to activate or keep the time domain aliasing reduction stage activated if identical frame lengths (e.g., combining coefficients) are used for two subsequent blocks, and/or the audio processor is configured to deactivate or keep the stage deactivated. reducing aliasing in the time domain if different frame lengths are used for the next two blocks.

For example, due to TDAR, there are dependencies between blocks. If the pooling factor is supported, then TDAR is possible. If the aggregation factor of two successive blocks is different, then TDAR is deactivated. Because of this (optionally), the processor may be configured to perform a general/combined optimization to adapt the combining coefficients, eg, a trellis mesh, per set of subband samples (bandwidth of the combining coefficients).

In embodiments, the processor is configured to perform a general/combined optimization for adapting the combining coefficients (eg, a trellis mesh per set of subband samples (bandwidth of the combining coefficients)).

In embodiments, the at least two window functions comprise the same window width (eg, to split the spectrum into at least two (eg, 8) equal-sized (eg, 128 coefficients) combining coefficient bands).

In embodiments, the at least two window functions comprise different window widths (eg, to split the spectrum into at least two bands of combining coefficients of different sizes).

In embodiments, the at least two window functions comprise substantially rectangular windows.

In embodiments, bin sets derived from at least two window functions are processed (e.g., combined) using a second overlapped critically sampled transform stage, wherein the second overlapped critically sampled transform stage is configured to perform at least at least two overlapping critically sampled transforms having the same frame length (eg, pooling factor) for at least one (eg, each) of the bin sets derived from the at least two window functions.

In embodiments, the audio processor is configured to perform joint channel coding.

In embodiments, the audio processor is configured to perform M/S or MCT as a combined channel processing.

In embodiments, the audio processor is configured to activate or deactivate the combined channel processing individually for each of the at least two segmented bin sets (e.g., individually activate or deactivate the combined channel processing for each band of combining coefficients; for example, between the first stage overlapped critically sampled transform and a second overlapped critically sampled transform stage from the cascaded overlapped critically sampled transform stage).

In embodiments, the audio processor is configured to generate a bitstream from an encoded subband dealiased representation of the audio signal, wherein the audio processor is configured to provide a bitstream containing at least one MF parameter signaling at least one length frame (eg, pooling factor) of the respective bin sets in the bitstream.

In embodiments, the audio processor is configured to entropy encode at least one MF parameter.

In embodiments, the audio processor is configured to provide a bitstream containing only a subset of MF parameters signaling the frame length (eg, pooling factor) of the respective bin sets in the bitstream.

In embodiments, the audio processor is configured to provide a bit stream containing a TDAR parameter for the respective bin sets.

In embodiments, the audio processor is configured to perform combined entropy encoding of the MF and TDAR parameters.

In embodiments, the audio processor is configured to provide a bitstream containing at least one MF parameter such that a TDAR parameter can be extracted from the at least one MF parameter (e.g., in a receiver or a web site decoder).

For example, instead of including the TDAR parameter in the bitstream provided by the audio processor, said TDAR parameter may be extracted from at least one MF parameter. For example, equal MF parameters in two subsequent frames may indicate that TDAR is active, while different MF parameters in two subsequent frames may indicate that TDAR is inactive.

In embodiments, the audio processor is configured to use a trellis grid per set of subband samples (bandwidth of the combining coefficients).

Additional embodiments provide an audio encoder, wherein the audio encoder comprises an audio processor according to one of the above-described embodiments, wherein the audio encoder comprises an encoder configured to encode a dealiased sub-band representation of an audio signal to obtain an encoded de-aliased sub-band representation of the audio signal, and wherein the audio encoder comprises a bit stream generator configured to generate a bit stream from an encoded subband dealiased representation of the audio signal.

Additional embodiments provide an audio processor for processing a subband representation of an audio signal so as to obtain an audio signal. The audio processor comprises a time-domain de-aliasing stage configured to weight-combine two respective de-aliased sub-band representations of an audio signal to obtain an aliased sub-band representation, wherein the aliased sub-band representation is a set of sub-band samples . Additionally, the audio processor includes a cascaded critically sampled inverse transform stage configured to perform a cascaded critically sampled inverse transform on a set of subband samples to obtain a set of samples associated with a block of audio signal samples. Because of this, the cascaded inverse lapped critically sampled transform stage comprises a first inverse lapped critically sampled transform stage, configured to perform an inverse lapped critically sampled transform on a set of subband samples to obtain a bin set associated with a given subband of the audio signal, when therein, the first inverse overlap critically sampled transform stage is configured to perform inverse overlap critically sampled transforms having the same frame length (eg, pooling factor) for the set of subband samples.

In embodiments, the first stage of the inverse overlapped critically sampled transform is configured to perform N _1.1 inverse overlapped critically sampled transforms on the N _1.1 subsets of the subband sample set, where the N _1.1 overlapped critically sampled transforms contain an identical frame length (e.g. , the pooling factor), while N _1,1 is a natural number greater than or equal to two.

Additional embodiments provide an audio decoder, wherein the audio decoder comprises a bitstream parser configured to parse the bitstream to obtain a de-aliased encoded sub-band representation, wherein the audio decoder comprises a decoder configured to decode the de-aliased encoded sub-band representation. aliasing so as to obtain a reduced aliasing sub-band representation of the audio signal, wherein the audio decoder comprises an audio processor according to one of the above-described embodiments.

Additional embodiments provide a method for processing an audio signal in order to obtain a subband representation of the audio signal. The method comprises the step of performing a cascaded overlapping critically sampled transform on at least two partially overlapping blocks of audio samples so as to obtain a set of subband samples based on the first block of audio samples and obtain a corresponding set of subband samples based on the second block of audio samples. Additionally, the method comprises the step of performing a weighted combination of two respective sets of subband samples, one of which is derived from the first block of audio samples and the other is derived from the second block of audio samples, so as to obtain a subband representation with reduced aliasing of the audio signal. Because of this, performing a cascaded overlapped critically sampled transform comprises segmenting the bin set derived from the first block of samples using at least two window functions so as to obtain at least two segmented bin sets (e.g. , at least two bands of combining coefficients) based on a segmented bin set corresponding to the first block of samples, wherein performing a cascaded overlapping critically sampled transform comprises segmenting the bin set derived from the second block of samples using at least , two window functions so as to obtain at least two bin sets (e.g., at least two bands of combining coefficients) based on the segmented bin set corresponding to the second block of samples, etc. and wherein the bin sets are processed (e.g., combined) using a second overlapped critically sampled transform from a cascaded overlapped critically sampled transform, wherein the second overlapped critically sampled transform comprises performing overlapped critically sampled transforms having the same frame length (e.g., pooling factor), at least one (eg, each) set of permission elements.

Additional embodiments provide a method for processing a subband representation of an audio signal so as to obtain an audio signal. The method comprises the step of performing a weighted combination of two respective reduced aliasing subband representations of an audio signal to obtain an aliased subband representation, wherein the aliased subband representation is a set of subband samples. Further, the method comprises the step of performing a cascaded inverse lap critically sampled transform on the set of subband samples to obtain a set of samples associated with a block of audio signal samples, wherein performing the cascaded inverse lap critically sampled transform comprises performing a first inverse lap critically sampled transform on the set of subband samples. samples to obtain a set of bins associated with a given audio subband, wherein performing the first inverse overlapped critically sampled transform comprises performing inverse overlapped critically sampled transforms having the same frame length (e.g., pooling factor) for the set of subband samples.

Embodiments provide a non-uniform filter bank with compact impulse responses that allows, by virtue of being able to more closely adhere to the masking threshold, at both high and low frequencies, to introduce more quantization noise without introducing audible artifacts, thereby providing a lower bit rate than uniform filterbank.

In embodiments, banding of the combining coefficients is performed. Because of this, many completely different combination schemes are possible. However, due to the high flexibility, it is very difficult to optimize the overall system. Therefore, in embodiments, the spectrum is divided into bands of combining factors (MFBs), in which the identical combining factor (MF) is always used (cf. section 2.1.1 below). Rectangular non-overlapping windows at the edges of the combining coefficient band can be used since all the combining coefficient bands can be independently controlled in this manner (cf. FIG. 17).

In embodiments, combined channel processing (M/S or multi-channel coding engine (MCT) [12]) is performed. In general, M/S or MCT can also be used with subband combining and TDAR. The middle channels and side channels can independently use different MF parameters and TDAR parameters per frame.

In embodiments, M/S or MCT are individually enabled/disabled per combining coefficient bandwidth. The combined stereo extension is to turn on/off the combined stereo for each frequency band between the first MDCT and the second MDCT to be combined. This provides frequency selective activation of MS/MCT, for example, per MFB. However, TDAR is only possible between two frames with the same combined stereo configuration (eg, no TDAR between L/R and M/S).

In embodiments, a window switch solver is used from existing methods to solve the pooling factor. In existing methods, long/short window solver controls are provided. In some circumstances, these solvers may also be used for MF.

In embodiments, the pooling factor is signaled in the bit stream (including entropy encoding and interprocessing of multiple MF parameters). Generally speaking, a small number of bits in the bitstream are required in order to be signaled in the MF signaling for the current frame. These bits can also be entropy encoded. In addition, the bits can also be encoded among themselves. Input: Various MFs such as 2, 8, 1, 2, 16, 32 are expected to be less likely than more uniform MFs such as 4, 4, 8, 8, 16, 16. This correlation can be used to to reduce the amount of data, for example, through differential coding.

In embodiments, fewer pooling coefficients are transmitted, whereby the missing pooling coefficients can be extracted/interpolated from neighboring nodes. If the MFs really need to be as uniform as in the previous sections, all MFs can also be interpolated from a small number of MFs.

In embodiments, TDAR parameters are derived from MF parameters. TDAR may be transmitted as 1 bit per MFB. Alternatively, the TDAR bit may also be extracted from the MF bits (identical MF parameter in two consecutive frames=TDAR enabled). In such a case, no ancillary information is required for TDAR.

In embodiments, cross-entropy encoding of MF parameters and TDAR parameters is performed. MF values and TDAR values may be entropy encoded in a reciprocal fashion.

In this case, with 8 MFB and 6 MF, it is not required:

but only:

bits.

In embodiments, the bands of the combining coefficients are divided in a non-uniform manner. For simplicity, the following description assumes uniform MFBs. However, they can also become uneven. A reasonable separation would presumably be identical to the scaling factor (SFB) frequency bands. In such a case, one scaling factor and one combining factor are transmitted per SFB.

In embodiments, a lattice optimization of the pooling coefficients is performed. The use of trellis grids for audio coding is already prior art [7]. However, traditional systems may only use one grid; on the other hand, according to embodiments, one grating can be used per MFB.

Additional embodiments provide an audio processor for processing the audio signal so as to obtain a subband representation of the audio signal. The audio processor includes a cascaded overlapped critically sampled transform and a time domain aliasing reduction stage. The cascaded overlapped critically sampled transform stage is configured to perform a cascaded overlapped critically sampled transform on at least two partially overlapping blocks of audio samples so as to obtain a set of subband samples based on the first block of audio samples and obtain a corresponding set of subband samples based on the second block of audio samples. block of samples of the audio signal. The time domain aliasing reduction stage is configured to weight-combine two respective sets of sub-band samples, one derived from the first block of audio samples and the other derived from the second block of audio samples, so as to obtain a reduced sub-band representation. overlaying the spectrum of the audio signal.

Additional embodiments provide an audio processor for processing a subband representation of an audio signal so as to obtain an audio signal. The audio processor comprises an inverse time domain aliasing reduction stage and a cascaded inverse overlapping critically sampled transform stage. The time-domain de-aliasing stage is configured to weight (and shift) two respective aliased sub-band representations (different blocks of partially overlapping samples) of an audio signal to obtain an aliased sub-band representation, wherein the sub-band representation aliased is a set of subband samples. The cascaded inverse overlapped critically sampled transform stage is configured to perform a cascaded inverse overlapped critically sampled transform on a set of subband samples to obtain a set of samples associated with a block of audio signal samples.

According to the principle of the present invention, an additional post-processing stage is added to an overlapped critically sampled transform (e.g., MDCT) pipeline, wherein the additional post-processing stage comprises another overlapped critically sampled transform (e.g., MDCT) along the frequency axis and de-aliasing in the time domain along each sub-band time axes. This allows the extraction of arbitrary frequency scales from the overlapped critically sampled transform (eg, MDCT) spectrogram with improved temporal compactness of the impulse response, without introducing additional redundancy and with the introduction of reduced frame delay of the overlapped critically sampled transform.

Additional embodiments provide a method for processing an audio signal so as to obtain a subband representation of the audio signal. The method contains:

performing a cascaded overlapping critically sampled transform on at least two partially overlapping blocks of audio samples so as to obtain a set of subband samples based on the first block of audio samples and obtain a corresponding set of subband samples based on the second block of audio samples; and

- performing a weighting combination of two respective sets of sub-band samples, one of which is derived from the first block of audio samples and the other is derived from the second block of audio samples, so as to obtain a sub-band representation with reduced aliasing of the audio signal.

Additional embodiments provide a method for processing a subband representation of an audio signal so as to obtain an audio signal. The method contains:

- performing a weighting (and shifting) combination of two respective reduced aliasing subband representations (different blocks of partially overlapping samples) of the audio signal to obtain a aliased subband representation, wherein the aliased subband representation is a set of subband samples; and

- performing a cascaded inverse overlap critically sampled transform on the set of subband samples to obtain a set of samples associated with a block of audio samples.

Preferred implementations are covered in the dependent claims.

The following describes preferred implementations of an audio processor for processing an audio signal so as to obtain a subband representation of the audio signal.

In embodiments, the cascaded overlapped critically sampled transform stage may be a cascaded MDCT (MDCT=modified discrete cosine transform), MDST (MDST=modified discrete sine transform) or MLT (MLT=modulated lapped transform) stage.

In embodiments, the cascaded overlapped critically sampled transform stage may comprise a first overlapped critically sampled transform stage configured to perform overlapped critically sampled transforms on the first block of samples and the second block of samples of at least two partially overlapping blocks of audio samples, so that to obtain a first set of bins for the first block of samples and a second set of bins (overlapping critically sampled coefficients) for the second block of samples.

The first stage of the overlapped critically sampled transform may be the first MDCT, MDST or MLT stage.

The cascaded overlapped critically sampled transform stage may further comprise a second overlapped critically sampled transform stage configured to perform overlapped critically sampled transform on a segment (self subset) of the first bin set and perform overlapped critically sampled transform on the segment (either subset) of the second bin set , each segment being associated with a subband of the audio signal so as to obtain a set of subband samples for the first set of bins and a set of subband samples for the second set of bins.

The second stage of the overlapped critically sampled transform may be the second MDCT, MDST or MLT stage.

Because of this, the first and second stages of the overlapped critically sampled transform may be of the same type, i.e. represent one of the MDCT-, MDST- or MLT-cascades.

In embodiments, the second overlapped critically sampled transform stage may be configured to perform overlapped critically sampled transforms on at least two partially overlapping segments (either subsets) of the first bin set and perform overlapped critically sampled transforms on at least two partially overlapping segments (own subsets) of a second bin set, each segment being associated with a subband of the audio signal so as to obtain at least two sets of subband samples for the first bin set, and at least two subband sets samples for the second set of resolution elements.

Therefore, the first set of subband samples may be the result of a first overlapped critically sampled transform based on the first segment of the first bin set, while the second set of subband samples may be the result of a second overlapped critically sampled transform based on the second segment of the first bin set, wherein the third set of subband samples may be the result of a third overlapped critically sampled transform based on the first segment of the second bin set, wherein the fourth set of subband samples may be the result of a fourth overlapped critically sampled transform based on the second segment of the second bin set. The time domain aliasing reduction stage may be configured to perform weighting of the first set of subband samples and the third set of subband samples to obtain a first subband dealiased representation of the audio signal, and perform weighting of the second set of subband samples and the fourth set of subband samples. a set of sub-band samples so as to obtain a second sub-band representation with reduced aliasing of the audio signal.

In embodiments, the cascaded overlapped critically sampled transform stage may be configured to segment the bin set derived from the first block of samples using at least two window functions and obtain at least two subband sample sets based on segmented bin set corresponding to the first block of samples, wherein the cascaded overlapping critically sampled transform stage may be configured to segment the bin set derived from the second block of samples using at least two window functions, and obtain, by at least two sets of subband samples based on the segmented bin set corresponding to the second block of samples, wherein the at least two window functions comprise different window widths.

In embodiments, the cascaded overlapped critically sampled transform stage may be configured to segment the bin set derived from the first block of samples using at least two window functions and obtain at least two subband sample sets based on segmented bin set corresponding to the first block of samples, wherein the cascaded overlapped critically sampled transform stage may be configured to segment the bin set derived from the second block of samples using at least two window functions and obtain at least at least two subband sample sets based on the segmented bin set corresponding to the second block of samples, wherein the window function filter slopes corresponding to adjacent subband sample sets are symmetrical.

In embodiments, the cascaded overlapped critically sampled transform stage may be configured to segment the audio signal samples into a first block of samples and a second block of samples using a first windowing function, wherein the overlapped critically sampled transform stage may be configured to segment the bin set obtained based on the first block of samples, and a bin set derived from the second block of samples using a second window function to obtain respective sub-band samples, the first window function and the second window function having different window widths.

In embodiments, the cascaded overlapped critically sampled transform stage may be configured to segment the audio signal samples into a first block of samples and a second block of samples using a first windowing function, wherein the overlapped critically sampled transform stage may be configured to segment the bin set obtained based on the first block of samples, and a bin set obtained based on the second block of samples using a second window function, so as to obtain the corresponding sub-band samples, wherein the window width of the first window function and the window width of the second window function are different from each other, wherein the window width of the first window function and the window width of the second window function differ from each other by a factor different from a power of two.

The following describes preferred implementations of an audio processor for processing a subband representation of an audio signal so as to obtain an audio signal.

In embodiments, the inverse cascaded overlay critically sampled transform stage may be an inverse cascaded MDCT (MDCT=modified discrete cosine transform), MDST (MDST=modified discrete sine transform), or MLT (MLT=modulated lapped transform).

In embodiments, the cascaded inverse lap critically sampled transform stage may comprise a first inverse lap critically sampled transform stage configured to perform an inverse lap critically sampled transform on a set of subband samples to obtain a bin set associated with a given audio subband.

The first stage of the inverse overlap critically sampled transform may be the first inverse MDCT, MDST or MLT stage.

In embodiments, the cascaded inverse lap critically sampled transform stage may comprise a first lap summation stage configured to concatenate a bin set associated with a plurality of audio subbands, which comprises weighting a bin set associated with a given audio subband , with a bin set associated with a different audio subband, to obtain a bin set associated with a block of audio samples.

In embodiments, the cascaded critically sampled inverse transform stage may comprise a second critically sampled inverse lapped transform stage configured to perform an inverse overlapped critically sampled transform on a set of bins associated with a block of audio samples to obtain a set of samples associated with block of samples of the audio signal.

The second stage of the inverse overlap critically sampled transform may be the second inverse MDCT, MDST or MLT stage.

Because of this, the first and second stages of the inverse overlapped critically sampled transform may be of the same type, i.e. represent one of the reverse MDCT, MDST, or MLT cascades.

In embodiments, the cascaded critically sampled inverse transform stage may comprise a second lap-add stage configured to overlap-add a set of samples associated with a block of audio samples and another set of samples associated with another block of audio samples, wherein the block of samples and the other block of audio samples are partially overlapped so as to obtain an audio signal.

Embodiments of the present invention are described herein with reference to the accompanying drawings.

Fig. 1 shows a schematic block diagram of an audio processor configured to process an audio signal so as to obtain a subband representation of the audio signal, according to an embodiment;

Fig. 2 shows a schematic block diagram of an audio processor configured to process an audio signal so as to obtain a subband representation of the audio signal, according to a further embodiment;

Fig. 3 shows a schematic block diagram of an audio processor configured to process an audio signal so as to obtain a subband representation of the audio signal, according to a further embodiment;

Fig. 4 shows a schematic block diagram of an audio processor for processing subband representation of an audio signal so as to obtain an audio signal, according to an embodiment;

Fig. 5 shows a schematic block diagram of an audio processor for processing subband representation of an audio signal so as to obtain an audio signal, according to a further embodiment;

Fig. 6 shows a schematic block diagram of an audio processor for processing subband representation of an audio signal so as to obtain an audio signal, according to a further embodiment;

Fig. 7 shows diagrammatically an example of sub-band samples (upper plot) and the spread of their samples in time and frequency (lower plot);

Fig. 8 shows in a diagram the spectral and temporal uncertainty obtained through several different transformations;

Fig. 9 shows a diagrammatic comparison of two exemplary impulse responses generated by sub-band combining with and without TDAR, simple short MDCT blocks, and Hadamard sub-band combining;

Fig. 10 shows a flowchart of a method for processing an audio signal so as to obtain a subband representation of an audio signal, according to an embodiment;

Fig. 11 shows a flowchart of a method for processing a subband representation of an audio signal so as to obtain an audio signal, according to an embodiment;

Fig. 12 shows a schematic block diagram of an audio encoder, according to an embodiment;

Fig. 13 shows a schematic block diagram of an audio decoder according to an embodiment;

Fig. 14 shows a schematic block diagram of an audio analyzer according to an embodiment;

Fig. 15 shows a schematic block diagram of the overlapped critically sampled transform substages for the second overlapped critically sampled transform stage of the audio processor shown in FIG. 2 and 3 according to an embodiment of the present invention;

Fig. 16 shows a schematic block diagram of the inverse overlapped critically sampled transform substages of the first inverse overlapped critically sampled transform stage of the audio processor shown in FIG. 5 and 6 according to an embodiment of the present invention;

Fig. 17 is a diagram showing window functions used to segment a bin set in order to obtain bin sets, according to an embodiment of the present invention;

Fig. 18 shows in distribution diagrams the combination factor (MF) and time domain aliasing reduction (TDAR) selections made by the encoder;

Fig. 19 shows in diagrams the average bit rates of the two systems for various distortion parameters q for 39 test elements;

Fig. 20 lists in the table the quality settings and their distortion parameter q and the resulting average bit rate;

Fig. 21 lists various test items in a table;

Fig. 22 lists in a table the results of the Shapiro-Wilk normality test for pairwise MUSHRA score differences between a window switching (WS) filter bank and a subband combining (SM) filter bank at slightly degraded (MQ) and moderately degraded (LQ) quality settings;

Fig. 23 shows distribution diagrams and kernel density estimates of MUSHRA score differences between a window switching (WS) filterbank and a subband combining (SM) filterbank at slightly degraded (MQ) and moderately degraded (LQ) quality settings;

Fig. 24 lists results by mean, standard deviation (SD), and Wilcoxon signed-rank test for quantitative MUSHRA scores comparing window switching (WS) filterbank and subband merging (SM) filterbank at slightly degraded (MQ) and moderately degraded (LQ) quality settings;

Fig. 25 diagrammatically shows the mean and 95% confidence intervals of individual element MUSHRA score differences, window switching filterbanks, and subband combining filterbanks at slightly degraded (MQ) and moderately degraded (LQ) quality settings;

Fig. 26 shows a flowchart of a method for processing an audio signal so as to obtain a subband representation of an audio signal, according to an embodiment; and

Fig. 27 shows a flowchart of a method for processing a subband representation of an audio signal so as to obtain an audio signal, according to an embodiment.

Identical or equivalent elements, or elements with identical or equivalent functionality, are referred to in the following description by identical or equivalent reference numerals.

In the description below, many details are set forth in order to provide a more complete explanation of the embodiments of the present invention. However, those skilled in the art will appreciate that embodiments of the present invention may be practiced without these specific details. In other instances, known structures and devices are shown in block diagram form rather than in detail so as not to obscure the embodiments of the present invention. In addition, the features of the various embodiments described below may be combined with each other, unless expressly stated otherwise.

First, in Section 1, a non-uniform orthogonal filterbank based on cascading two MDCTs and time domain aliasing reduction (TDAR) is described, which allows one to achieve impulse responses that are compact in time and frequency [1]. Thereafter, in Section 2, the perceptual quality of such a non-uniform filter bank in an audio encoder scenario is evaluated and compared to the performance of a uniform window-switching filter bank used in current encoders such as advanced audio coding (AAC) [2].

1. Non-uniform orthogonal filter bank based on cascading two MDCTs and time domain aliasing reduction (TDAR)

Fig. 1 shows a schematic block diagram of an audio processor 100 configured to process an audio signal 102 so as to obtain a subband representation of the audio signal, according to an embodiment. The audio processor 100 includes a cascaded overlapped critically sampled transform (LCST) stage 104 and a time domain alias reduction (TDAR) stage 106 .

The cascaded overlapped critically sampled transform stage 104 is configured to perform a cascaded overlapped critically sampled transform on at least two partially overlapping blocks 108_1 and 108_2 of samples of the audio signal 102 so as to obtain a subband sample set 110_1,1 based on the first block 108_1 samples ( from at least two overlapping blocks 108_1 and 108_2 samples) of the audio signal 102 and obtain a corresponding set 110_2,1 subband samples based on the second block 108_2 samples (from at least two overlapping blocks 108_1 and 108_2 samples) of the audio signal 102.

The time-domain aliasing reduction stage 104 is configured to weight-combine two respective sets 110_1,1 and 110_2,1 of subband samples (i.e., subband samples corresponding to the same subband), one of which is derived from the first block. 108_1 samples of the audio signal 102, and another is derived based on the second block 108_2 of audio samples so as to obtain a sub-band de-aliased representation 112_1 of the audio signal 102.

In embodiments, the cascaded overlapped critically sampled transform stage 104 may comprise at least two cascaded overlapped critically sampled transform stages, or in other words, the two overlapped critically sampled transform stages are connected in a cascaded fashion.

The cascaded overlapped critically sampled transform cascade may be a cascaded MDCT (MDCT=Modified Discrete Cosine Transform) cascade. A cascaded MDCT cascade may comprise at least two MDCT cascades.

Naturally, the cascaded overlapped critically sampled transform stage can also be a cascaded MDST (MDST=modified discrete sine transform) or MLT (MLT=modulated lapped transform) stage comprising at least two MDST or MLT stages, respectively.

The two respective sets 110_1,1 and 110_2,1 of subband samples may be subband samples corresponding to the same subband (ie frequency band).

Fig. 2 shows a schematic block diagram of an audio processor 100 configured to process an audio signal 102 to obtain a subband representation of the audio signal, according to a further embodiment.

As shown in FIG. 2, the cascaded overlapped critically sampled transform stage 104 may comprise a first overlapped critically sampled transform stage 120 configured to perform overlapped critically sampled transforms for the first block 108_1 (2M) of samples (x _i-1 (n), 0≤n≤2M- 1) and a second block 108_2 (2M) of samples (x _i (n), 0≤n≤2M-1) from at least two partially overlapping blocks 108_1 and 108_2 of samples of the audio signal 102 so as to obtain a first set 124_1 of (M) bins (LCST coefficients) (X _i-1 (k), 0≤k≤M-1) for the first block 108_1 samples and the second set 124_2 of (M) bins (LCST coefficients) (X _i (k), 0≤k≤M-1) for the second block 108_2 samples.

The cascaded overlapped critically sampled transform stage 104 may comprise a second overlapped critically sampled transform stage 126 configured to perform overlapped critically sampled transform for the (own subset) segment 128_1,1 (X _{v, i-1} (k)) of the first bin set 124_1 and perform an occluded critically sampled transform on the (own subset) segment 128_2,1 (X _{v, i} (k)) of the second bin set 124_2, with each segment associated with a subband of the audio signal 102, so as to obtain a subband sample set 110_1,1 (ŷ _{v, i-1} (m)) for the first bin set 124_1 and subband samples set 110_2.1 (ŷ _{v, i} (m)) for the second bin set 124_2.

Fig. 3 shows a schematic block diagram of an audio processor 100 configured to process an audio signal 102 to obtain a subband representation of the audio signal, according to a further embodiment. In other words, FIG. 3 shows a schematic of an analytical filter bank. Because of this, appropriate window functions are assumed. It should be noted that for simplicity, in FIG. 3 indicates (only) the processing of the first half of the subband frame (y[m], 0<=m<N/2) (ie, only the first line of equation (6)).

As shown in FIG. 3, the first overlapped critically sampled transform stage 120 may be configured to perform the first overlapped critically sampled transform 122_1 (eg, MDCT i-1) on the first block 108_1 (2M) of samples (x _i-1 (n), 0≤n≤ 2M-1) so as to obtain a first set 124_1 of (M) bins (LCST coefficients) (X _i-1 (k), 0≤k≤M-1) for the first sample block 108_1, and perform a second overlapping critically sampled transform 122_2 (eg, MDCT i) for the second block 108_2 (2M) samples (x _i (n), 0≤n≤2M-1) so as to obtain a second set 124_2 of (M) bins (LCST- coefficients) (X _i (k), 0≤k≤M-1) for the second block 108_2 samples.

In detail, the second overlapped critically sampled transform stage 126 may be configured to perform overlapped critically sampled transforms on at least two partially overlapping segments 128_1,1 and 128_1,2 (own subsets) (X _{v, i-1} (k) ) of the first bin set 124_1 and perform overlapping critically sampled transforms on at least two partially overlapping segments 128_2,1 and 128_2,2 (own subsets) (X _{v, i} (k)) of the second bin set, each segment associated with a subband of the audio signal so as to obtain at least two sets 110_1,1 and 110_1,2 subband samples (ŷ _{v, i-1} (m)) for the first set 124_1 bins and at least two set 110_2,1 and 110_2,2 subband samples (ŷ _{v, i} (m)) for the second bin set 124_2.

For example, the first subband sample set 110_1,1 may be the result of a first overlapped critically sampled transform 132_1,1 based on the first segment 132_1,1 of the first bin set 124_1, while the second subband sample set 110_1,2 may be the result of a second critically overlapped sampled transform 132_1,2 based on the second segment 128_1,2 of the first bin set 124_1, wherein the third subband sample set 110_2,1 may be the result of a third overlapping critically sampled transform 132_2,1 based on the first segment 128_2,1 of the second bin set 124_2 resolution, wherein the fourth subband sample set 110_2,2 may be the result of a fourth overlapped critically sampled transform 132_2,2 based on the second segment 128_2,2 of the second bin set 124_2.

Because of this, the time domain de-aliasing stage 106 can be configured to weight-combine the first set of subband samples 110_1,1 and the third set of subband samples 110_2,1 to obtain a first dealiased subband representation 112_1 ( y _1,i [m ₁ ]) of the audio signal, wherein the domain de-aliasing stage 106 can be configured to weight-combine the second subband sample set 110_1,2 and the fourth subband sample set 110_2,2 to obtain a second subband representation 112_2 with reduced aliasing (y _2,i [m ₂ ]) of the audio signal.

Fig. 4 shows a schematic block diagram of an audio processor 200 for processing a subband representation of an audio signal so as to obtain an audio signal 102, according to an embodiment. The audio processor 200 includes a time domain inverse aliasing reduction (TDAR) stage 202 and a cascaded inverse overlapping critically sampled transform (LCST) stage 204 .

The inverse time domain de-aliasing stage 202 is configured to weight (and shift) combine the two respective de-aliased subband representations 112_1 and 112_2 (y _{v, i} (m), y _{v, i-1} (m)) audio signal 102 to obtain a sub-band aliased representation 110_1 (ŷ _{v, i} (m)), wherein the sub-band aliased representation is a set 110_1 of sub-band samples.

The cascaded inverse overlapped critically sampled transform stage 204 is configured to perform a cascaded inverse overlapped critically sampled transform on subband sample set 110_1 to obtain a set of samples associated with sample block 108_1 of audio signal 102.

Fig. 5 shows a schematic block diagram of an audio processor 200 for processing a subband representation of an audio signal so as to obtain an audio signal 102, according to a further embodiment. The cascaded inverse overlapped critically sampled transform stage 204 may comprise a first inverse overlapped critically sampled transform (LCST) stage 208 and a first overlap summation stage 210 .

v,i(k))..The first stage 208 of the inverse overlap critically sampled transform may be configured to perform an inverse overlap critically sampled transform on the set 110_1,1 subband samples to obtain the set 128_1,1 bins associated with a given subband of the audio signal (

v,i(k))).

v,i(k))разрешения, ассоциированных с данной субполосой (v) частот аудиосигнала 102, с набором 128_1,2 элементов (

v,i(k))разрешения, ассоциированных с другой субполосой (v-1) частот аудиосигнала 102, с тем чтобы получать набор 124_1 элементов разрешения, ассоциированных с блоком 108_1 выборок аудиосигнала 102.The first overlap summation stage 210 may be configured to concatenate the bin sets associated with the multiple subbands of the audio signal, which comprises weighting the bin set 128_1.1 (

v,i(k)) of the resolution associated with a given subband (v) of the audio signal 102, with a set of 128_1,2 elements (

v,i(k)) of the resolution associated with another subband (v-1) of the audio signal 102 in order to obtain a set 124_1 of the bins associated with the block 108_1 of samples of the audio signal 102.

As shown in FIG. 5, the cascaded inverse overlap critically sampled transform (LCST) stage 204 may comprise a second inverse overlap critically sampled transform (LCST) stage 212 configured to perform an inverse overlap critically sampled transform on the bin set 124_1 associated with the sample block 108_1 of the audio signal 102 so as to receive a set 206_1,1 samples associated with block 108_1 samples of the audio signal 102.

Additionally, the cascaded inverse overlap critically sampled transform stage 204 may comprise a second lap-add stage 214 configured to overlap-add a set 206_1,1 of samples associated with block 108_1 of samples of audio signal 102 and another set 206_2,1 of samples associated with another an audio sample block 108_2, wherein the sample block 108_1 and the other sample block 108_2 of the audio signal 102 partially overlap so as to obtain the audio signal 102.

Fig. 6 shows a schematic block diagram of an audio processor 200 for processing a subband representation of an audio signal so as to obtain an audio signal 102, according to a further embodiment. In other words, FIG. 6 shows a schematic of a synthesis filter bank. Because of this, appropriate window functions are assumed. It should be noted that for simplicity, in FIG. 6 indicates (only) the processing of the first half of the subband frame (y[m], 0<=m<N/2) (ie, only the first line of equation (6)).

As described above, audio processor 200 comprises a time domain inverse aliasing reduction stage 202 and an inverse lapped critically sampled transform stage 204 comprising a first inverse lapped critically sampled transform stage 208 and a second inverse lapped critically sampled transform stage 212.

The time domain inverse reduction stage 104 is configured to perform a first weighted and shifted combination 220_1 of the first and second subband representations y _1,i-1 [ _m1 ] and y _1,i [ _m1 ] with reduced aliasing so as to obtain the first sub-band representation 110_1,1 ŷ _1,i [m ₁ ] with aliasing, while the sub-band representation with aliasing is a set of sub-band samples, and perform the second combination 220_2 with weighting and shifting the third and fourth sub-band representations y _2,i-1 [ _m1 ] and y _2,i [ _m1 ] with reduced aliasing so as to obtain a second aliased subband representation 110_2,1 ŷ _2,i [m ₁ ], wherein the aliased subband representation is a set of subband samples .

v,i(k)), и выполнять второе обратное перекрывающееся критически дискретизированное преобразование 222_2 для второго набора 110_2,1 ŷ_2,i[m₁] субполосных выборок, с тем чтобы получать набор 128_2,1 элементов разрешения, ассоциированных с данной субполосой частот аудиосигнала (

v,i(k)).The first inverse overlap critically sampled transform stage 208 is configured to perform the first inverse overlap critically sampled transform 222_1 on the first set 110_1.1 ŷ _1,i [m ₁ ] of subband samples to obtain a set 128_1.1 bins associated with a given subband of the audio signal (

v,i(k)) and perform a second inverse overlap critically sampled transform 222_2 on the second set 110_2,1 ŷ _2,i [m ₁ ] of subband samples to obtain a set 128_2,1 bins associated with that subband. audio signal (

v,i(k)).

The second inverse overlap critically sampled transform stage 212 is configured to perform an inverse overlap critically sampled transform on the overlapped and summed bin set obtained by overlapping and summing the bin sets 128_1,1 and 128_2,1 provided by the first inverse overlapping critically sampled stage 208 transformations so as to obtain a sample block 108_2.

The following describes embodiments of the audio processors shown in FIG. 1-6, in which it is roughly assumed that the cascaded overlapped critically sampled transform stage 104 is an MDCT stage, i. e. the first and second overlapped critically sampled transform stages 120 and 126 are MDCT stages, and the inverse staged overlapped critically sampled transform stage 204 is an inverse staged MDCT stage, i. e. the first and second stages 120 and 126 of the inverse overlapped critically sampled transform are inverse MDCT stages. Naturally, the description below is also applicable to other embodiments of the cascaded overlapped critically sampled transform stage 104 and the inverse overlapped critically sampled transform stage 204, such as a cascaded MDST or MLT stage, or an inverse cascaded MDST or MLT stage.

Because of this, the described embodiments can work for a sequence of MDCT spectra of limited length and use MDCT and time domain aliasing reduction (TDAR) as a subband combining operation. The resulting non-uniform filterbank is overlapped, orthogonal, and provides subband widths k=2 ⁿ with n∈N. Due to TDAR, a temporally and spectrally more compact sub-band impulse response can be achieved.

The following describes embodiments of the filter bank.

The implementation of the filterbank is directly based on the general schemes of the lapped MDCT transform: The original lapped and weighted transform remains unchanged.

Without loss of generality, the following notation allows orthogonal MDCT transformations, for example, in which the analytic and synthesis windows are identical.

where k (k, n, M) is the kernel of the MDCT transform and h(n) is the appropriate analytic window.

субполос частот с отдельными ширинами N_v и снова преобразуется с использованием MDCT Это приводит к гребенке фильтров с перекрытием во временном и спектральном направлении.The output value of this transformation X _i (k) is then segmented into

subbands with separate widths N _v and again converted using MDCT This results in a filter bank with overlap in the temporal and spectral direction.

To simplify the notation in this document, one common pooling factor N is used for all subbands; however, any valid switching/ordering of the MDCT window may be used to implement the desired time-frequency resolution. The following contains additional information regarding resolution design.

where w(k) is an appropriate analytic window and is generally different from h(n) in size and may be different in window type. Since the embodiments use a window in the frequency domain, it should be noted that the time and frequency selectivity of the window are swapped.

For proper edge handling, an additional offset from N/2 can be introduced in equation (4), combined with the rectangular halves of the start/end window at the edges. Also, to simplify the notation, this offset is not taken into account here.

представляет собой список v-векторов отдельных длин N_v коэффициентов с соответствующими полосами

пропускания и временным разрешением, пропорциональным этой полосе пропускания.Conclusion

is a list of v-vectors of distinct lengths of N _v coefficients with corresponding bands

bandwidth and a temporal resolution proportional to that bandwidth.

However, these vectors contain aliasing from the original MDCT transform and therefore exhibit poor temporal compactness. To compensate for this aliasing, TDAR can be simplified.

The samples used for TDAR are drawn from two adjacent blocks v of subband samples in the current and previous MDCT frame i and i-1. The result is a reduced aliasing in the second half of the previous frame and the first half of the second frame.

с:for

With:

,

,

и

могут быть спроектированы с возможностью минимизировать остаточное наложение спектров. Ниже вводится простой способ оценки на основе синтезирующего окна g(n) синтезирующего.TDAR ratios

,

,

and

can be designed to minimize residual aliasing. Below we introduce a simple estimation method based on the synthesis window g(n) of the synthesis.

и

, например, оба MDCT являются ортогональными, и матрица A является ортогональной, полный конвейер составляет ортогональное преобразование.It should also be noted that if A is non-singular, operations (6) and (8) correspond to a biorthogonal system. Additionally, if

and

eg both MDCTs are orthogonal and matrix A is orthogonal, the full pipeline constitutes an orthogonal transformation.

To calculate the inverse transform, the first inverse TDAR is performed,

after which an inverse MDCT is performed and time-domain aliasing suppression (TDAC, although aliasing suppression is performed along the frequency axis here) must be performed in order to suppress the aliasing generated in Equation 5:

Finally, the initial MDCT in Equation 2 is inverted and TDAC is performed again:

The following describes the design limitations of the time-frequency resolution. Although any desired time-frequency resolution is possible, some constraints on the design of the resulting window functions must be observed in order to ensure reversibility. In particular, the slopes of two adjacent subbands may be symmetrical such that Equation (6) satisfies the Prinzen-Bradley condition [J. Princen, A. Johnson and A. Bradley, "Subband/transform coding using filter bank designs based on time domain aliasing cancellation", in Acoustics, Speech and Signal Processing, IEEE International Conference on ICASSP '87, April 1987, Volume 12, pp. 2161-2164]. The window switching scheme introduced in [B. Edler, "Codierung von Audiosignalen mit überlappender Transformation und adaptiven Fensterfunktionen", Frequenz, Vol. 43, pp. 252-256, September 1989], originally designed to resist pre-echo effects, can be applied here. See [Olivier Derrien, Thibaud Necciari and Peter Balazs, "A quasi-orthogonal, invertible and perceptually relevant time-frequency transform for audio coding", in EUSIPCO, Nice, France, August 2015].

Second, the sum of all lengths of the second MDCT transforms must add up to the total length of the provided MDCT coefficients. The frequency bands may be chosen to not be transformed using a unit staggered window with zeros in the desired coefficients. However, the symmetry properties of adjacent windows must be tracked [B. Edler, "Codierung von Audiosignalen mit überlappender Transformation und adaptiven Fensterfunktionen", Frequenz, vol. 43, pp. 252-256, September 1989]. The resulting transform should result in zeros in these frequency bands so that the original coefficients can be directly used.

As a possible time-frequency resolution, scale factor bands from most modern audio encoders can be directly used.

Next, the calculation of time domain aliasing reduction (TDAR) coefficients is described.

According to the above temporal resolution, each sub-band sample corresponds to M/N _v original samples, or an interval of N _v times the size, as one original sample.

In addition, the amount of aliasing in each subband sample depends on the amount of aliasing in the interval it represents. Because the aliasing is weighted by the analytical window h(n), using an approximation of the fusynthesizing window at each subband sampling interval is assumed to be a good first estimate for the TDAR coefficient.

длины 2N_v.Experiments demonstrate that two very simple coefficient calculation schemes provide good initial values with improved temporal and spectral compactness. Both methods are based on a hypothetical synthesis window

length 2N _v .

1) For parametric windows such as sinusoidal or Kaiser-Bessel extracted windows, a simple shorter window of the same type can be specified.

2) For both parametric and tabular windows without a closed view, the window can be divided into 2N _v sections of equal size, allowing coefficients to be obtained using the average of each section:

Taking into account the MDCT boundary conditions and aliasing mirroring, in this case, the result is the TDAR coefficients:

or in case of orthogonal transformation:

; тем не менее, не в сигнале x(n), синтезированном посредством гребенки обратных фильтров.No matter which coefficient fitting solution is chosen, as long as A is non-singular, the perfect reconstruction of the entire filterbank is preserved. In other respects, the suboptimal choice of coefficients should only affect the amount of residual aliasing in the subband signal.

; however, not in the signal x(n) synthesized by the inverse filter bank.

Fig. 7 shows diagrammatically an example of sub-band samples (upper plot) and the spread of their samples in time and frequency (lower plot). The annotated sample has a wider bandwidth but less time spread than the lower samples. Analytical windows (lower graph) have a full resolution of one coefficient per original time sample. The TDAR coefficients therefore need to be approximated (provided with a dot) for each time domain of the subband samples (m=256:::384).

The results (simulations) are described next.

Fig. 8 shows the spectral and temporal uncertainty obtained through several different transformations, as shown in [Frederic Bimbot, Ewen Camberlein and Pierrick Philippe, "Adaptive filter banks using fixed size mdct and subband merging for audio coding-comparison with the mpeg aac filter banks", in Audio Engineering Society Convention 121, October 2006].

It can be seen that the Hadamard matrix transforms offer highly limited time-frequency trade-off characteristics. For growing pool sizes, the additional temporal resolution results in a disproportionately high cost in spectral uncertainty.

In other words, FIG. 8 shows a comparison of the spectral and temporal energy multiplexing of various transforms. Built-in labels indicate frame lengths for MDCT, split factors for Heisenberg split, and pool factors for everything else.

However, subband combining with TDAR has a linear trade-off between temporal and spectral uncertainty, parallel to simple uniform MDCT. Their product is constant, although slightly higher than the simple uniform MDCT. For this analysis, the sinusoidal analysis window and the Kaiser-Bessel-derived window at sub-band combining show the most compact results and are therefore chosen.

However, it seems that using TDAR for the pooling factor N _v =2 reduces temporal and spectral compactness. This is because the coefficient calculation scheme introduced in section II-B is too simplistic and does not adequately approximate values for steep window function slopes. The numerical optimization scheme should be presented in a subsequent publication.

квадратной формы импульсной характеристики

, заданной как [Athanasios Papoulis, "Signal analysis", Electrical and electronic engineering series, McGraw-Hill, Нью-Йорк, Сан-Франциско, Париж, 1977 год]These compactness values are calculated using the center of gravity and the effective length

square wave impulse response

given as [Athanasios Papoulis, "Signal analysis", Electrical and electronic engineering series, McGraw-Hill, New York, San Francisco, Paris, 1977]

The average values of all impulse responses of each individual filter bank are shown.

Fig. 9 shows a comparison of two exemplary impulse responses generated by sub-band combining with and without TDAR, simple short MDCT blocks, and sub-band combining based on the Hadamard matrix as proposed in [O.A. Niamut and R. Heusdens, "Flexible frequency decompositions for cosine-modulated filter banks", in Acoustics, Speech and Signal Processing, 2003. Proceedings (ICASSP '03), 2003 IEEE International Conference, April 2003, Volume 5, Page V -449-52, volume 5].

The poor temporal compactness of the Hadamard-based union transformation is clearly visible. It can also be clearly seen that most subband aliasing artifacts are greatly reduced by TDAR.

In other words, FIG. 9 shows exemplary impulse responses of a combined subband filter containing 8 out of 1024 original bins using the method proposed here without TDAR, with TDAR, the method proposed in [O.A. Niamut and R. Heusdens, "Subband merging in cosine-modulated filter banks", Signal Processing Letters, IEEE, Vol. 10, No. 4, pp. 111-114, April 2003], and using a shorter MDCT frame length of 256 samples.

Fig. 10 shows a flow diagram of a method 300 for processing an audio signal so as to obtain a subband representation of the audio signal. Method 300 comprises 302 performing a cascaded overlapping critically sampled transform on at least two partially overlapping blocks of audio samples to obtain a set of subband samples based on the first block of audio samples and obtain a corresponding set of subband samples based on the second block of audio samples. Additionally, the method 300 comprises 304 performing a weighted combination of two respective sets of sub-band samples, one of which is derived from the first block of audio samples and the other is derived from the second block of audio samples, so as to obtain a sub-band representation with reduced aliasing of the audio signal. .

Fig. 11 shows a flowchart of a method 400 for processing a subband representation of an audio signal so as to obtain an audio signal. The method 400 comprises a step 402 of performing a weighted (and shifted) combination of two respective reduced aliasing subband representations (different blocks of partially overlapping samples) of an audio signal to obtain a aliased subband representation, where the aliased subband representation is a set subband samples. Additionally, method 400 comprises 404 performing a cascaded inverse overlap critically sampled transform on a set of subband samples to obtain a set of samples associated with a block of audio samples.

Fig. 12 shows a schematic block diagram of an audio encoder 150, according to an embodiment. The audio encoder 150 comprises an audio processor (100) as described above, an encoder 152 configured to encode a sub-band de-aliased representation of an audio signal so as to obtain a de-aliased sub-band encoded representation of the audio signal, and a bit stream generator 154 configured to generate a stream of 156 bits from the encoded subband representation with reduced aliasing of the audio signal.

Fig. 13 shows a schematic block diagram of an audio decoder 250, according to an embodiment. The audio decoder 250 includes a bitstream parser 252 configured to parse the bit stream 154 to obtain a de-aliased sub-band encoded representation, a decoder 254 configured to decode the de-aliased sub-band encoded representation to obtain a sub-band de-aliased representation. a de-aliased representation of the audio signal, and an audio processor 200 as described above.

Fig. 14 shows a schematic block diagram of an audio analyzer 180, according to an embodiment. The audio analyzer 180 includes an audio processor 100 as described above, an information extractor 182 configured to analyze the dealiased subband representation so as to provide information describing the audio signal.

Embodiments provide time domain aliasing reduction (TDAR) in subbands of non-uniform orthogonal modified discrete cosine transform (MDCT) filterbanks.

Embodiments add an additional post-processing step to the widely used MDCT pipeline, the step itself containing only another overlapping MDCT along the frequency axis and time domain aliasing reduction (TDAR) along each sub-band time axis, providing the ability to extract arbitrary frequency scales from MDCT spectrograms with improved temporal compactness of the impulse response, without the introduction of additional redundancy and with the introduction of only one MDCT frame delay.

2. Perceptual audio coding with adaptive non-uniform time-frequency tiling using sub-band combining and time-domain aliasing

Fig. 15 shows a schematic block diagram of the overlapped critically sampled transform substages 132_1,1, 132_1,2, 132_2,1, and 132_2,2 for the second overlapped critically sampled transform stage 126 of the audio processor 100 shown in FIG. 2 and 3 according to an embodiment of the present invention.

Because of this, at least one of the overlapped critically sampled transform substages 132_1,1, 132_1,2, 132_2.1, and 132_2,2 can be configured to perform overlapped critically sampled transforms having the same frame length (e.g., pooling factor ) for the corresponding set of 128_1,1, 128_1,2, 128_2,1 and 128_2,2 bins.

In embodiments, the first overlapped critical transform substage 132_1,1 may be configured to perform N _1,1 overlapped critical transforms on N _1,1 equal-length subsets of the first bin set 128_1,1 derived from the segmented bin set 124_1 , corresponding to the first block 108_1 samples, while N _1.1 overlapping critically sampled transforms contain the same frame length (eg, pooling factor), while N _1.1 is a natural number greater than or equal to two.

For example, assuming that the first bin set 128_1,1 contains 128 bins (or coefficients), the first overlapped critical transform substage 132_1,1 may be configured to perform, for example, N _1,1 =2 overlapped critical transforms for N _1,1 =2 subsets of 64 bins (i.e., 128 bins divided by N _1,1 ), where N _1,1 =2 overlapping critically sampled transforms contain identical frame length (e.g., pooling factor) , for example, 64. Naturally, the first substage 132_1,1 overlapped critical transform can also be configured to perform, for example, N _1,1 =4 (or 8) overlapped critical transforms for N _1,1 =4 (or 8) subsets of 32 (or 16) bins (i.e. 128 bins divided by N _1,1 ), with N _1,1 = 4 (or 8) overlapping critically sampled the transforms contain the same frame length (eg, pooling factor), eg 32 (or 16).

In embodiments, the second overlapped critical transform substage 132_1,2 may be configured to perform N _1,2 overlapped critical transforms for N _1,2 equal-length subsets of the second bin set 128_1,2 derived from the segmented bin set 124_1 , corresponding to the first block 108_1 samples, while N _1.2 overlapping critically sampled transforms contain the same frame length (eg, pooling factor), while N _1.2 is a natural number greater than or equal to two.

For example, assuming that the second bin set 128_1,2 contains 128 bins (or coefficients), the second overlapped critical transform substage 132_1,2 may be configured to perform, for example, N _1,2 =2 overlapped critical transforms for N _1,2 =2 subsets of 64 bins (i.e., 128 bins divided by N _1,1 ), where N _1,2 =2 overlapping critically sampled transforms contain identical frame length (e.g., pooling factor) , for example, 64. Naturally, the second substage 132_1,2 overlapped critical transform can also be configured to perform, for example, N _1,2 =4 (or 8) overlapped critical transforms for N _1,2 =4 (or 8) subsets of 32 (or 16) bins (i.e. 128 bins divided by N _1,2 ), with N _1,2 = 4 (or 8) overlapping critically sampled the transforms contain the same frame length (eg, pooling factor), eg 32 (or 16).

In embodiments, the third overlapped critical transform substage 132_2.1 may be configured to perform N _2.1 overlapped critical transforms for N _2.1 equal-length subsets of the third bin set 128_2.1 derived from the segmented bin set 124_2 , corresponding to the second block 108_2 samples, while N _2.1 overlapping critically sampled transforms contain the same frame length (eg, pooling factor), while N _2.1 is a natural number greater than or equal to two.

For example, assuming that the third bin set 128_2,1 contains 128 bins (or coefficients), the third overlapped critical transform substage 132_2,1 may be configured to perform, for example, N _2,1 =2 overlapped critical transforms for N _2,1 =2 subsets of 64 bins (i.e., 128 bins divided by N _1,1 ), where N _1,1 =2 overlapping critically sampled transforms contain identical frame length (e.g., pooling factor) , for example, 64. Naturally, the third substage 132_2,1 of the overlapped critical transform can also be configured to perform, for example, N _2,1 =4 (or 8) overlapped critically sampled transforms for N _2,1 =4 (or 8) subsets of 32 (or 16) bins (i.e. 128 bins divided by N _1,1 ), with N _2,1 = 4 (or 8) overlapping critically sampled the transforms contain the same frame length (eg, pooling factor), eg 32 (or 16).

In embodiments, the fourth overlapped critical transform substage 132_2,2 may be configured to perform N _2,2 overlapped critical transforms for N _2,2 equal-length subsets of the fourth bin set 128_2,2 derived from the segmented bin set 124_2 corresponding to the second bin block 108_2, where N _2.2 overlapping critically sampled transforms contain the same frame length (eg, pooling factor), where N _2.2 is a natural number greater than or equal to two.

For example, assuming that the fourth subband bin set 128_2,2 contains 128 bins (or coefficients), the fourth overlapped critical transform substage 132_2,2 may be configured to perform, for example, N _2,2 =2 overlapped critical bins. for N _2,2 =2 subsets of 64 bins (i.e., 128 bins divided by N _2.2 ), where N _2,2 =2 overlapping critically sampled transforms contain identical frame length (e.g., pooling factor ), for example, 64. Naturally, the fourth substage 132_2,2 overlapped critical transform can also be configured to perform, for example, N _2,2 =4 (or 8) overlapped critically sampled transforms for N _2,2 =4 (or 8 ) subsets of 32 (or 16) bins (i.e. 128 bins divided by N _2.2 ), with N _2.2 = 4 (or 8) overlapping critical The sampled transforms contain the same frame length (eg, pooling factor), for example, 32 (or 16).

In embodiments, the first bin set 128_1,1 and the second bin set 128_1,2 may contain the same or different length (ie, number of bins).

In embodiments, N _1.1 and N _1.2 may be identical or different natural numbers.

In embodiments, the third bin set 128_2,1 and the fourth bin set 128_2,2 may contain the same or different length (ie, number of bins).

In embodiments, N _2.1 and N _2.2 may be identical or different natural numbers.

In embodiments, if TDAR is activated, the first bin set 128_1,1 and the third bin set 128_2,1 may contain the same length (ie, the same number of bins). In addition, N _1.1 and N _2.1 may represent an identical natural number. Similarly, also the second bin set 128_1,2 and the fourth bin set 128_2,2 may contain the same length (ie, the same number of bins). In addition, N _2.1 and N _2.2 may represent an identical natural number.

In embodiments, if TDAR is disabled, the first bin set 128_1.1 and the third bin set 128_2.1 may also contain different lengths (i.e., different numbers of bins), and also N _1.1 and N _2.1 can be different natural numbers. Similarly, the second bin set 128_1.2 and the fourth bin set 128_2.2 may also contain different lengths (i.e., different numbers of bins), and also N _2.1 and N _2.2 may be different natural numbers. .

Fig. 16 shows a schematic block diagram of the inverse overlapped critically sampled transform substages 222_1 and 222_2 for the first inverse overlapped critically sampled transform stage 208 of the audio processor 200 shown in FIG. 5 and 6 according to an embodiment of the present invention.

Because of this, at least one of the inverse overlapped critically sampled transform substages 222_1 and 222_2 can be configured to perform inverse overlapped critically sampled transforms having the same frame length (e.g., pooling factor) for the respective set 110_1,1 and 110_1 .2 subband samples so as to obtain a corresponding bin set 128_1,1 and 128_1,2.

In embodiments, the first inverse overlapped critical transform substage 222_1 may be configured to perform N _1.1 inverse overlap critical transforms for N _1.1 equal-length subsets of the first set 110_1 subband samples, with N _1.1 overlapping critically sampled transforms. contain the same frame length (eg, pooling factor), with N _1,1 being a natural number greater than or equal to two.

In embodiments, the second inverse overlap critical transform substage 222_2 may be configured to perform N _{1.2 inverse overlap critical transforms for N 1.2 equal} -length subsets of the second subband sample set 110_1, with N _1.2 overlapping critical transforms. contain the same frame length (eg, pooling factor), with N _1,2 being a natural number greater than or equal to two.

The following describes detailed embodiments of the non-uniform filter bank. Additionally, the perceptual quality of such a non-uniform filterbank in an audio encoder scenario is evaluated and compared to the performance of a uniform windowed filterbank used in current encoders, such as advanced audio coding (AAC) [2].

2.1. Coding system

The scoring system models a simple perceptual encoder, with an analytic filterbank, a psychoacoustic model [4], a quantizer, a perceptual entropy estimator [5], and a synthesis filterbank. In the two competing systems, filter bank or uniform MDCT with window switching [6] (WS) or non-uniform MDCT with subband combining and (SM) TDAR[1].

The relevant filterbank parameters (window switching bounds for uniform MDCT, or pooling factors and TDAR bounds for nonuniform MDCT) are adaptively and optimally chosen to minimize the total remaining entropy.

Additional post-processing steps or coding tools cannot be used.

2.1.1. Filter Bank Options

, называемой коэффициентом объединения. Согласно проектированию системы, во время анализа оптимальный вариант выбора в коэффициенте объединения по-прежнему неизвестен, и каждая полоса частот коэффициентов объединения не знает коэффициент объединения ни одного из своих соседних узлов. Таким образом, окна на границах полосы коэффициентов объединения могут выбираться таким образом, что они всегда являются асимметричными и достаточно крутыми для того, чтобы приспосабливать самый крутой соседний коэффициент объединения, см. фиг. 17.The window switching filter bank can use MDCT with conventional AAC frame lengths of 1024-sample long frames or 8 128-sample short frames and appropriate weighted-coded transition windows in between. A cosine window can be used. The subband combining filter bank 120 may use an initial frame length MDCT of, for example, 1024 and then partition the spectrum into 8 combining coefficient bands (eg, 128_1.1, 128_1.2, etc.) of 128 coefficients. Each band of combining coefficients can then be combined with the MDCT for the frame length

, called the pooling factor. According to the system design, at the time of analysis, the best choice in the pooling factor is still unknown, and each band of pooling factors does not know the pooling factor of any of its neighbors. Thus, the windows at band edges of the combining coefficients can be chosen such that they are always asymmetric and steep enough to accommodate the steepest adjacent combining coefficient, see FIG. 17.

In detail, fig. 17 shows in a diagram window functions used to segment a set (124_1) of bins in order to obtain sets (or segments) of (128_1,1, 128_1,2) bins. In other words, FIG. 16 shows exemplary window selections in four bands of combining coefficients. Steep windows at the edge of the band of pooling coefficients are highlighted in black.

This design decision limits the overall flexibility of the filter bank and introduces non-ideal timing ripple for these asymmetric windows [1], but offers a way to efficiently and independently optimize the combining factor for each band of combining coefficients.

The cosine window may be used as the transform window, and a Kaiser-Bessel extracted window with an arbitrary choice of β=5.1 may be selected as the pooling window.

Finally, quantization step sizes can be controlled using a real-valued distortion parameter q that multiplicatively lowers or raises the estimated masking threshold from the perceptual model by a constant factor q. After quantization, the perceptual entropy estimator calculates the theoretical bit rate r, which naturally depends on q. For q=1.0, the psychoacoustic model predicts transparent encoding with no audible artifacts at all, for large values of q>1.0, the quantization step size increases, the bit rate r drops, and the perceived quality of the process is expected to deteriorate.

2.1.2. Parametric optimization

To perform optimal parameter tuning, each signal is converted and quantized using all possible combinations of parameters, and the perceptual entropy of each frame for each parameter is estimated. From among all the output coefficients, the optimal combination of parameters is calculated that minimizes the total perceptual entropy, and the output signal is then synthesized using these parameters.

To find the optimal filter bank parameters, each band of pooling coefficients in each frame (a pooling tile of 128 coefficients) is quantized and its entropy is calculated. A graph of all parameters of all pooling tiles in one band of pooling coefficients then forms a lattice in which the weights of each transition probability are set equal to the entropy of the next pooling tile [7].

As noted above, not all parameter combinations and transitions need to allow for perfect recovery during synthesis, for example, when switching from long to short frames, an initial asymmetric window should be used in between. Similar rules apply for using TDAR in a non-uniform filter bank [1]. To prevent these unauthorized parameter transitions, the transition probabilities in the lattice are multiplied by a mask that encodes all authorized and unauthorized transitions, i.e. 1 for authorized and 1 for unauthorized transitions.

Thereafter, the path with the minimum weights through the grating is computed using dynamic programming, resulting in the overall optimal parameter path in each individual band of combining coefficients, which also guarantees perfect recovery.

This approach requires multiple encoding passes, a very large look-ahead, and is therefore unsuitable for an actual online encoder, however, it ensures that both methods are performed at the highest possible efficiency at any given time. For online coding, there are ways to decode such trellis schemes with time delay constraints [8].

. Для субполосного объединения, 29 битов использованы в расчете на кадр, с тем чтобы передавать в служебных сигналах коэффициент объединения и TDAR-флаг (8 полос частот коэффициентов объединения с 6 коэффициентами объединения и 2 TDAR-значениями каждая,

). Коэффициенты масштабирования или пороговые значения маскирования известны на стороне декодера.In addition, the system assumes a simple and uncompressed transmission of the necessary ancillary information: For window switching, 1 bit is used for each frame in order to signal long and short blocks

. For subband combining, 29 bits are used per frame to signal the combining factor and TDAR flag (8 combining factor bands with 6 combining factors and 2 TDAR values each,

). The scaling factors or masking thresholds are known at the decoder side.

2.2. General Observations

By simply performing the encoding/decoding process, the following properties can be observed:

In the two to three highest bands of combining factors, in the range of 15-24 kHz, the encoder almost always selects a combining factor of 1, turning off the combining. In the middle section, bands 2-5 of the combining coefficients or the frequency range between 3-15 kHz, the encoder mainly selects the combining coefficient at 1 or at 32. In the lower frequency band of the combining coefficients, in the range of 0-3 kHz, the encoder mainly selects pooling factors at 1 and 2. Pooling factors at 4, 8, and 16 are rarely chosen. See fig. eighteen.

In detail, fig. 18 shows distribution diagrams of the combination factor (MF) and time domain alias reduction (TDAR) selections made by the encoder.

This observation is consistent with basic assumptions about the auditory system: due to high frequencies having a very high threshold in silence, almost everything is in fact quantized to zero, making the choice of pooling factor irrelevant. At mid frequencies, the auditory system has a high temporal resolution, while at lower frequencies the human ear has a higher frequency resolution.

Second, it should be noted that for any chosen distortion parameter q, the corresponding subband combining filter bank bit rate is lower than the window switching filter bank bit rate. On average, a non-uniform system requires 5-13% fewer bits per sample in order to encode signals, see FIG. 19.

In detail, fig. 19 shows in diagrams the average bit rates of the two systems for different distortion parameters q for 39 test elements.

2.3. Test layout based on listening

Three different quality settings are considered at different quantizer step size factors and therefore average bit rates: transparent (HQ), slightly degraded (MQ), and moderately degraded (LQ), see Table 1 in FIG. twenty.

In detail, Table 1 in FIG. 20 lists the quality settings and their distortion parameter q and the resulting average bit rate.

According to perceptual model design, audible artifacts are not expected for HQ [4]. In addition, in fact, in small-scale listening-based ABC/HR tests (ITU-R BS.1116-3) [9], expert listeners cannot detect significant differences between either method and the reference signal. Since it is unlikely that such a listening-based test would reveal any meaningful results, it is omitted in favor of the two remaining quality settings MQ and LQ.

For MQ and LQ, the distortion parameter q of the window switching filterbank system is chosen such that its output bit rate is the same as the output bit rate of the distortion parameter q of the subband combining filterbank, which means that the distortion parameter q for the subband filterbank pooling is lower than for the window switching filter bank. It follows that with a non-uniform filter bank, higher perceived quality can be achieved by providing a bit rate identical to the bit rate of the window switching filter bank. To test this, a listening-based test was performed using the multi-stimulus test method with hidden reference and anchor signal (MUSHRA, ITU-R BS.1534-3) [10].

2.4. Body of test signals

The test signals for this evaluation are derived from a test suite widely used for developing and tuning audio encoders. It contains male and female speech and several musical recordings containing harmonic and percussive sounds. All conditions are loudness normalized using ITU-R BS.1770-4 [11]. See Table 2 in FIG. 21. In detail, Table 2 in FIG. 21 lists various test elements.

2.5. Listening Test Results

A total of N=16 expert listeners took part in the test.

First, the Shapiro-Wilk test is used to test pairwise differences in MUSHRA scores between the two methods for normality. For LQ and MQ, the differences are substantially non-normal, see Table 3 in FIG. 22 and FIG. 4 in FIG. 23.

In detail, Table 3 in FIG. 22 lists the results of the Shapiro-Wilk normality test for pairwise MUSHRA score differences between a window switching (WS) filter bank and a subband combining (SM) filter bank at slightly degraded (MQ) and moderately degraded (LQ) quality settings. W denotes the statistical W-value, p denotes the p-value.

Therefore, Wilcoxon's parametric signed-rank test is used instead of the parametric paired t-test under all conditions. The summarized results of all tests are shown in Table 4 in FIG. 24.

In detail, Table 4 in FIG. 24 lists results by mean, standard deviation (SD), and Wilcoxon signed-rank test for quantitative MUSHRA scores comparing window switching (WS) filterbank and subband merging (SM) filterbank at slightly degraded (MQ) and moderately degraded (LQ) quality settings. W denotes the statistical W-value, p denotes the p-value.

The Wilcoxon sign-rank test was performed to compare the perceptual quality of the two systems at MQ. There is a significant difference in the MUSHRA scores for the window switching filter bank and the subband combining filter bank, p=0.000.

Second, the Wilcoxon sign-rank test was performed to compare the perceptual quality of the two systems at the LQ quality setting. There is a significant difference in the MUSHRA scores for the window switching filter bank and the subband combining filter bank, p=0.000.

Fig. 25 diagrammatically shows the mean and 95% confidence intervals of individual element MUSHRA score differences, window switching filterbanks, and subband combining filterbanks at slightly degraded (MQ) and moderately degraded (LQ) quality settings. Positive values favor sub-band combining over window switching.

2.6. Additional Embodiments

Embodiments provide a way to use non-uniform orthogonal filterbank based on MDCT analysis/synthesis and TDAR in a simple audio encoder. Above, its coding efficiency is compared to a uniform bank of MDCT window switching filters. On average, ripple requires 5-13% fewer bits per sample in order to encode test signals. This additional coding gain can be used to improve the perceived quality of the encoder at the same output bit rate.

In the test described above, an increased perceived quality of 6-7 MUSHRA scores was identified using the MUSHRA listening-based test and subsequent statistical analysis. The difference in perceived quality was found to be statistically significant.

Fig. 26 shows a flowchart of a method 500 for processing an audio signal so as to obtain a subband representation of an audio signal, according to an embodiment of the present invention. Method 500 comprises 502 performing a cascaded overlapping critically sampled transform on at least two partially overlapping blocks of audio samples to obtain a set of subband samples based on the first block of audio samples and obtain a corresponding set of subband samples based on the second block of audio samples. The method 500 further comprises 504 performing a weight-combining of two respective sets of sub-band samples, one derived from the first block of audio samples and the other derived from the second block of audio samples, so as to obtain a sub-band representation with reduced aliasing of the audio signal; wherein performing the cascaded overlapping critically sampled transform comprises segmenting the bin set derived from the first block of samples using at least two window functions so as to obtain at least two segmented bin sets based on the segmented set elements of permission corresponding to the first block of samples; wherein performing the cascaded overlapping critically sampled transform comprises segmenting the bin set derived from the second block of samples using at least two window functions so as to obtain at least two bin sets based on the segmented bin set resolution corresponding to the second block of samples; and wherein the bin sets are processed using a second overlapped critically sampled transform from a cascaded overlapped critically sampled transform, wherein the second overlapped critically sampled transform comprises performing overlapped critically sampled transforms having the same frame length for at least one bin set.

Fig. 27 shows a flowchart 600 for a method for processing a subband representation of an audio signal so as to obtain an audio signal, according to an embodiment of the present invention. The method 600 comprises 602 performing a weighted combination of two respective reduced aliasing subband representations of the audio signal to obtain an aliased subband representation, where the aliased subband representation is a set of subband samples. Additionally, method 600 comprises 604 performing a cascaded critically sampled inverse transform on a set of subband samples to obtain a set of samples associated with a block of audio samples, wherein performing the cascaded critically sampled inverse transform comprises performing a first critically sampled inverse lapping transform on a set of subband samples to obtain a bin set associated with a given subband of the audio signal, wherein performing the first inverse overlap critically sampled transform comprises performing inverse overlap critically sampled transforms having the same frame length for the set of subband samples.

The following describes further embodiments. Because of this, the following embodiments can be combined with the above embodiments.

Embodiment 1: An audio processor (100) for processing an audio signal (102) so as to obtain a subband representation of an audio signal (102), wherein the audio processor (100) comprises: sampled transform for at least two partially overlapping blocks (108_1; 108_2) of audio samples (102) so as to obtain a set (110_1,1) of sub-band samples based on the first block (108_1) of audio samples (102) and obtain the corresponding a set (110_2,1) of sub-band samples based on a second block (108_2) of audio samples (102); and a time-domain aliasing reduction stage (106) configured to weight-combine two respective sets (110_1.1; 110_1.2) of subband samples, one of which is obtained based on the first block (108_1) of audio signal samples (102 ) and the other is obtained based on the second block (108_2) of audio samples so as to obtain a sub-band representation (112_1) with reduced aliasing of the audio signal (102).

Embodiment 2: The audio processor (100) according to Embodiment 1, wherein the cascaded overlapped critically sampled transform stage (104) comprises: a first overlapped critically sampled transform stage (120) configured to perform overlapped critically sampled transforms for the first block (108_1) samples and a second block (108_2) of samples from at least two partially overlapping blocks (108_1; 108_2) of samples of the audio signal (102) so as to obtain a first set (124_1) of bins for the first block (108_1) of samples and a second set (124_2) bins for the second block (108_2) of samples.

Embodiment 3: The audio processor (100) according to Embodiment 2, wherein the cascaded overlapped critically sampled transform stage (104) further comprises: a second overlapped critically sampled transform stage (126) configured to perform overlapped critically sampled transform for segment (128_1, 1) the first bin set (124_1) and perform an overlap critically sampled transform on a segment (128_2,1) of the second bin set (124_2), each segment being associated with an audio subband (102) to obtain a set (110_1, 1) subband samples for the first bin set and a set of (110_2,1) subband samples for the second bin set.

Embodiment 4: The audio processor (100) according to Embodiment 3, wherein the first set (110_1,1) of subband samples is the result of the first overlapped critically sampled transform (132_1,1) based on the first segment (128_1,1) of the first set (124_1) ) bins, wherein the second set (110_1,2) subband samples is the result of the second overlapping critically sampled transform (132_1,2) based on the second segment (128_1,2) of the first set (124_1) bins, while the third set ( 110_2,1) subband samples is the result of a third overlapping critically sampled transform (132_2,1) based on the first segment (128_2,1) of the second set (128_2,1) bins, with the fourth set (110_2,2) subband samples representing is the result of the fourth overlapping critically sampled transform (132_2,2) based on the second the th segment (128_2.2) of the second set (128_2.1) of bins; and wherein the time domain aliasing reduction stage (106) is configured to weight-combine the first set (110_1,1) subband samples and the third set (110_2,1) subband samples to obtain the first subband representation (112_1) with reduced aliasing of the audio signal, wherein the time domain aliasing reduction stage (106) is configured to weight-combine the second set (110_1,2) subband samples and the fourth set (110_2.2) subband samples to obtain a second subband representation (112_2) with reduced aliasing of the audio signal.

Embodiment 5: An audio processor (100) according to one of embodiments 1-4, wherein the cascaded overlapping critically sampled transform stage (104) is configured to segment the bin set (124_1) obtained from the first block (108_1) of samples, with using at least two window functions and obtain at least two segmented sets (128_1,1; 128_1,2) subband samples based on the segmented set of bins corresponding to the first block (108_1) samples; wherein the cascaded overlapped critically sampled transform stage (104) is configured to segment the set (124_2) of bins obtained from the second block (108_2) of samples using at least two window functions so as to obtain at least at least two segmented sets (128_2.1; 128_2.2) of sub-band samples based on the segmented set of bins corresponding to the second block (108_2) of samples; and wherein the at least two window functions comprise different window widths.

Embodiment 6: An audio processor (100) according to one of embodiments 1-5, wherein the cascaded overlapped critically sampled transform stage (104) is configured to segment the bin set (124_1) obtained from the first block (108_1) of samples, with using at least two window functions and obtain at least two segmented sets (128_1,1; 128_1,2) subband samples based on the segmented set of bins corresponding to the first block (108_1) samples; wherein the cascaded overlapped critically sampled transform stage (104) is configured to segment the set (124_2) of bins obtained from the second block (108_2) of samples using at least two window functions so as to obtain at least at least two sets (128_2.1; 128_2.2) of sub-band samples based on the segmented bin set corresponding to the second block (108_2) of samples; and the filter slopes of the window functions corresponding to adjacent sets of subband samples are symmetrical.

Embodiment 7: An audio processor (100) according to one of embodiments 1-6, wherein the cascaded overlapped critically sampled transform stage (104) is configured to segment audio samples into a first block (108_1) of samples and a second block (108_2) of samples using first window function; wherein the overlapping critically sampled transform stage (104) is configured to segment the bin set (124_1) obtained from the first block (108_1) of samples and the bin set (124_2) obtained from the second block (108_2) of samples, with using a second window function so as to obtain the corresponding subband samples; and wherein the first window function and the second window function comprise a different window width.

Embodiment 8: An audio processor (100) according to one of embodiments 1-6, wherein the cascaded overlapped critically sampled transform stage (104) is configured to segment audio samples into a first block (108_1) of samples and a second block (108_2) of samples using first window function; wherein the cascade (104) of the cascaded overlapping critically sampled transform is configured to segment the set (124_1) of bins obtained from the first block (108_1) of samples and the set (124_2) of bins obtained from the second block (108_2) of samples, using a second window function so as to obtain the corresponding subband samples; and wherein the window width of the first window function and the window width of the second window function differ from each other, wherein the window width of the first window function and the window width of the second window function differ from each other by a factor other than a power of two.

Embodiment 9: An audio processor (100) according to one of Embodiments 1-8, wherein the time-domain de-aliasing stage (106) is configured to weight-combine two respective sets of subband samples according to the following equation:

с:for

With:

so as to obtain a sub-band audio de-aliasing representation, wherein y _{v, i} (m) is the first sub-band audio de-aliasing representation, y _{v, i-1} (N-1-m) is the second sub-band representation with reduced audio aliasing, ŷ _{v, i} (m) is a set of sub-band samples based on the second block of audio samples, ŷ _{v, i-1} (N-1-m) is a set of sub-band samples based on the first block of audio samples, a _v (m) is ..., b _v (m) is ..., c _v (m) is ..., and d _v (m) is ....

Embodiment 10: An audio processor (200) for processing a subband representation of an audio signal so as to obtain an audio signal (102), wherein the audio processor (200) comprises: a time-domain de-aliasing stage (202) configured to perform two-weighted combining corresponding sub-band aliased representations of the audio signal (102) to obtain a sub-band aliasing representation, the sub-band aliasing being a set of (110_1,1) sub-band samples; and a cascaded critically sampled inverse transform stage (204) configured to perform a cascaded critically sampled inverse lapped transform on a set of (110_1,1) subband samples to obtain a set of (206_1,1) samples associated with a block of audio signal samples ( 102).

Embodiment 11: The audio processor (200) according to embodiment 10, wherein the cascaded inverse overlapped critically sampled transform stage (204) comprises a first inverse overlapped critically sampled transform stage (208) configured to perform an inverse overlapped critically sampled transform for a set (110_1 ,1) subband samples to obtain a set of (128_1,1) bins associated with a given subband of the audio signal; and a first overlap summation stage (210) configured to concatenate the bin sets associated with the plurality of audio subbands, which comprises weighting the bin set (128_1,1) associated with a given audio subband (102), with a set (128_1,2) of bins associated with another audio subband (102) to obtain a set (124_1) of bins associated with a block of audio samples (102).

Embodiment 12: The audio processor (200) according to Embodiment 11, wherein the cascaded inverse overlapped critically sampled transform stage (204) comprises a second inverse overlapped critically sampled transform stage (212) configured to perform an inverse overlapped critically sampled transform for the set (124_1 ) bins associated with the block of audio samples (102) so as to obtain a set of samples associated with the block of audio samples (102).

Embodiment 13: The audio processor (200) according to embodiment 12, wherein the cascaded inverse overlap critically sampled transform stage (204) comprises a second overlap summation stage (214) configured to overlap add a set (206_1,1) of samples associated with a block of audio samples (102), and another set (206_2,1) of samples associated with another block of audio samples (102), wherein the block of samples and the other block of audio samples (102) partially overlap so as to obtain an audio signal (102) .

Embodiment 14: An audio processor (200) according to one of embodiments 10-13, wherein the inverse time domain de-aliasing stage (202) is configured to weight-combine two respective subband de-aliased representations of the audio signal (102) on based on the following equation:

с:for

With:

so as to obtain a sub-band aliasing representation, wherein y _{v, i} (m) is the first sub-band representation with reduced aliasing of the audio signal, y _{v, i-1} (N-1-m) is the second sub-band representation with reduced audio signal aliasing, ŷ _{v, i} (m) is a set of sub-band samples based on the second block of audio samples, ŷ _{v, i-1} (N-1-m) is a set of sub-band samples based on the first block of audio samples, a _v (m) is ..., b _v (m) is ..., c _v (m) is ..., and d _v (m) is ....

Embodiment 15: An audio encoder comprising: an audio processor (100) according to one of embodiments 1-9; an encoder configured to encode a sub-band de-aliased representation of the audio signal so as to obtain a de-aliased sub-band encoded representation of the audio signal; and a bitstream generator configured to generate a bitstream from the encoded subband dealiased representation of the audio signal.

Embodiment 16: An audio decoder, comprising: a bitstream parser configured to parse the bitstream to obtain a reduced aliasing encoded subband representation; a decoder configured to decode the encoded de-aliased sub-band representation so as to obtain a de-aliased sub-band representation of the audio signal; and an audio processor (200) according to one of the embodiments 10-14.

Embodiment 17: An audio analyzer comprising: an audio processor (100) according to one of embodiments 1-9; and an information extractor configured to analyze the de-aliased subband representation so as to provide information describing the audio signal.

Embodiment 18: A method (300) for processing an audio signal so as to obtain a subband representation of the audio signal, the method comprising: performing (302) a cascaded overlapping critically sampled transform on at least two partially overlapping blocks of audio signal samples so that obtaining a set of subband samples based on the first block of audio samples and obtaining a corresponding set of subband samples based on the second block of audio samples; and performing (304) weighting the two respective sets of subband samples, one of which is derived from the first block of audio samples and the other is obtained from the second block of audio samples, to obtain a subband audio dealiased representation.

Embodiment 19: A method (400) for processing a subband representation of an audio signal so as to obtain an audio signal, the method comprising: performing (402) weighting two respective subband representations with reduced aliasing of the audio signal so as to obtain an aliased subband representation spectra, wherein the subband aliased representation is a set of subband samples; and performing (404) a cascaded inverse overlap critically sampled transform on the set of subband samples to obtain a set of samples associated with the block of audio samples.

Embodiment 20: A computer program for carrying out the method according to one of embodiments 18 and 19.

Although some aspects are described in the context of equipment, it is obvious that these aspects also represent a description of the corresponding method, with the block or device corresponding to a method step or a feature of a method step. Likewise, aspects described in the context of a method step also provide a description of the associated block or element, or feature of the associated equipment. Some or all of the steps of the method may be performed by (or using) hardware such as, for example, a microprocessor, a programmable computer, or an electronic circuit. In some embodiments, one or more of the most important steps of the method may be performed by this equipment.

Depending on certain implementation requirements, embodiments of the invention may be implemented in hardware or in software. The implementation may be performed using a digital storage medium, such as a floppy disk, DVD, Blu-ray, CD, ROM, PROM, EPROM, EEPROM, or flash memory, having electronically readable control signals stored that interact (or are capable of interacting) with programmable computer system in such a way that the corresponding method is carried out. Therefore, the digital storage medium can be machine readable.

Some embodiments according to the invention comprise a storage medium having electronically readable control signals that are capable of interacting with a programmable computer system such that one of the methods described herein is carried out.

In general, embodiments of the present invention may be implemented as a computer program product with program code, wherein the program code is configured to perform one of the methods when the computer program product is running on the computer. The program code may, for example, be stored on a computer-readable medium.

Other embodiments comprise a computer program for carrying out one of the methods described herein, stored on a computer-readable medium.

In other words, an embodiment of the proposed method is therefore a computer program having program code for carrying out one of the methods described herein when the computer program is running on a computer.

Therefore, a further embodiment of the inventive methods is a storage medium (digital storage medium or computer-readable medium) containing a recorded computer program for carrying out one of the methods described herein. The storage medium, digital storage medium or recorded data medium is typically tangible and/or non-volatile.

Therefore, an additional embodiment of the inventive method is a stream of data or a sequence of signals representing a computer program for implementing one of the methods described herein. The data stream or signal sequence, for example, may be configured to be transmitted over a data connection, such as the Internet.

An additional embodiment comprises processing means, such as a computer or programmable logic device, configured to perform one of the methods described herein.

An additional embodiment comprises a computer having a computer program installed to implement one of the methods described herein.

A further embodiment according to the invention comprises equipment or a system capable of transmitting (eg, electronically or optically) a computer program for performing one of the methods described herein to a receiving device. The receiving device may, for example, be a computer, mobile device, storage device, or the like. The equipment or system, for example, may include a file server for transmitting a computer program to a receiving device.

In some embodiments, a programmable logic device (eg, a field programmable gate array) may be used to perform some or all of the functionality of the methods described herein. In some embodiments, a user-programmable gate array may interface with a microprocessor to perform one of the methods described herein. In general, the methods are preferably carried out by any hardware.

The equipment described herein may be implemented using hardware, either using a computer, or using a combination of hardware and a computer.

The equipment described herein, or any components of the equipment described herein, may be implemented at least in part in hardware and/or software.

The methods described herein may be performed using hardware, or using a computer, or using a combination of hardware and a computer.

The methods described herein, or any components of the equipment described herein, may be performed at least in part by hardware and/or by software.

The above described embodiments are merely illustrative of the principles of the present invention. It should be understood that modifications and changes to the arrangements and details described herein should be apparent to those skilled in the art. Therefore, they are intended to be limited only by the scope of the following claims, and not by the specific details provided by way of describing and explaining the embodiments herein.

Bibliographic list

[1] Nils Werner and Bernd Edler, "Nonuniform orthogonal filterbanks based on MDCT analysis/synthesis and time-domain aliasing reduction", IEEE Signal Processing Letters, Volume 24, Number 5, pp. 589-593, May 2017.

[2] Fernando C. Pereira and Touradj Ebrahimi, The MPEG-4 Book, Prentice Hall PTR, Upper Saddle River, NJ, USA, 2002.

[3] B. C. Moore and B. R. Glasberg, "Suggested formulae for calculating auditory-filter bandwidths and excitation patterns", J. Acoust. soc. Am., Vol. 74, Number 3, pp. 750-753, September 1983.

[4] A. Taghipour, M. C. Jaikumar, and B. Edler, "A psychoacoustic model with partial spectral flatness measure for tonality estimation", in Proc. 22nd Eur. signal processing. Conf. (EUSIPCO), 2014, pp. 646-650.

[5] J. D. Johnston, "Estimation of perceptual entropy using noise masking criteria", in ICASSP-88, International Conference on Acoustics, Speech and Signal Processing, April 1988, pp. 2524-2527, volume 5.

[6] B. Edler, "Codierung von Audiosignalen mit überlappender Transformation und adaptiven Fensterfunktionen", Frequenz, vol. 43, pp. 252-256, September 1989.

[7] V. Melkote and K. Rose, "Trellis-based approaches to rate-distortion optimized audio encoding", IEEE Transactions on Audio, Speech and Language Processing, Volume 18, Number 2, pp. 330-341, February 2010 .

[8] Mukund Narasimhan, Paul Viola and Michael Shilman, "Online decoding of markov models under latency constraints", in Proceedings of the 23rd International Conference on Machine Learning, New York, NY, USA, 2006, ICML' 06, pp. 657-664, ACM.

[9] ITU Radiocommunication Bureau, "BS.1116-3: methods for the subjective assessment of small impairments in audio systems", Recommendation ITU-R BS. 1116, 2015.

[10] ITU Radiocommunication Bureau, "BS.1534-3: method for the subjective assessment of intermediate quality level of coding systems", Recommendation ITUR BS. 1534, 2015.

[11] ITU Radiocommunication Bureau, "BS.1770-3: algorithms to measure audio program loudness and truepeak audio level", Recommendation ITU-R BS. 1770, 2015

[12] F. Schuh, S. Dick, R. Füg, C. R. Helmrich, N. Rettelbach, and T. Schwegler, "Efficient Multichannel Audio Transform Coding with Low Delay and Complexity", Audio Engineering Society, September 2016 [online]. At: http://www.aes.org/e-lib/browse.cfm? elib=18464

[13] WO 2018 019 909 A1

[14] EP 3 276 620 A1

Claims (120)

1. An audio processor (100) for processing an audio signal (102) to obtain a subband representation of an audio signal (102), wherein the audio processor (100) comprises: - a cascaded overlapped critically sampled transform (104) configured to perform a cascaded overlapped critically sampled transform on at least two partially overlapping blocks (108_1; 108_2) of audio signal samples (102) to obtain a set (110_1,1) of subband samples based on the first block (108_1) of audio samples (102) and obtaining a corresponding set (110_2,1) of subband samples based on the second block (108_2) of audio samples (102); and - a time domain aliasing reduction stage (106) configured to perform weighting combining two respective sets (110_1.1; 110_1.2) of subband samples, one obtained based on the first block (108_1) of audio signal samples (102), and the other is obtained based on the second block (108_2) of samples of the audio signal, to obtain a sub-band representation (112_1) with reduced aliasing of the audio signal (102); - at the same time, the cascade (104) of the cascaded overlapping critically sampled transform is configured to segment the set (124_1) of bins obtained from the first block (108_1) of samples using at least two window functions, and obtain at least , two sets (128_1,1; 128_1,2) bins based on the segmented bin set corresponding to the first block (108_1) samples; wherein the cascade (104) of the cascaded overlapping critically sampled transform is configured to segment the set (124_2) of bins obtained from the second block (108_2) of samples using at least two window functions and obtain at least , two bin sets (128_2.1; 128_2.2) based on the segmented bin set corresponding to the second block (108_2) of samples; and - while the sets (128_1,1; 128_1,2; 128_2,1; 128_2,2) of the bins are processed using the second stage of overlapped critically sampled transform from the stage of the cascade overlapped critically sampled transform, while the second stage of overlapped critically sampled transform is performed with the ability to perform overlapping critically sampled transforms having the same frame length for at least one bin set; wherein the audio processor (100) is configured to activate or keep activated the de-aliasing stage (106) in the time domain if identical frame lengths are used for two subsequent blocks (108_1, 108_2), and/or wherein the audio processor (100) is configured to deactivate or keep the de-aliasing stage (106) deactivated in the time domain if different frame lengths are used for the next two blocks (108_1, 108_2). 2. Audio processor (100) according to claim 1, - in which the second cascade (126) of overlapping critical transformation is configured to perform N _1.1 overlapping critically sampled transformations for N _1.1 subsets of the first set (128_1.1) of the at least two sets (128_1.1; 128_1,2) bins derived from the segmented bin set corresponding to the first block (108_1) of samples, where N _1,1 overlapping critically sampled transforms contain the same frame length, where N _1,1 is a natural number greater than or equal to two - while the second stage (126) of overlapping critical transformation is configured to perform N _1.2 overlapping critically sampled transformations for N _1.2 subsets of the corresponding first set (128_1.2) of the said at least two sets (128_2.1 ; 128_2,2) bins derived from the segmented bin set corresponding to the second block (108_2) of samples, wherein N _2.1 overlapping critically sampled transforms contain an identical frame length, in which N _2.1 is a natural number greater than or equal to two. 3. Audio processor (100) according to claim 1, - while the audio processor (100) is configured to individually select the frame length for each set (128_1.1, 128_1.2, 128_2.1, 128_2.2) of resolution elements or for each of the corresponding sets (128_1.1 and 128_2.1 ; 128_1,2 and 128_2,2) permission elements. 4. Audio processor (100) according to claim 1, - while the audio processor is configured to individually select the frame length for each block (108_1, 108_2) samples. 5. Audio processor (100) according to claim 1, wherein the processor (100) is configured to perform a general/combined optimization to adapt frame lengths. 6. Audio processor (100) according to claim 1, - wherein said at least two window functions contain the same window width. 7. Audio processor (100) according to claim 1, - wherein said at least two window functions comprise different window widths. 8. Audio processor (100) according to claim 1, - wherein said at least two window functions comprise substantially rectangular windows. 9. Audio processor (100) according to claim 6, - in which bin sets derived from at least two window functions are processed using a second stage of overlapped critically sampled transform, wherein the second overlapped critically sampled transform stage is configured to perform at least two overlapped critically sampled transforms having the same frame length for at least one of the bin sets derived from at least two windowed functions. 10. Audio processor (100) according to claim 1, wherein the audio processor (100) is configured to perform joint channel coding. 11. Audio processor according to claim 10, wherein the audio processor (100) is configured to perform M/S or multi-channel coding (MCT) as a combined channel processing. 12. Audio processor (100) according to claim 10, wherein the audio processor (100) is configured to activate or deactivate the combined channel processing individually for each of said at least two segmented sets (128_1.1; 128_1.2) of bins. 13. Audio processor (100) according to claim 1, - wherein the audio processor (100) is configured to generate a bit stream from the encoded subband representation (112_1, 112_2) with reduced aliasing of the audio signal, - while the audio processor (100) is configured to provide a bit stream containing at least one combination factor (MF) parameter signaling at least one frame length of the respective sets (128_1,1 and 128_2,1; 128_1, 2 and 128_2,2) bins in the bitstream. 14. Audio processor (100) according to claim 13, wherein the audio processor (100) is configured to entropy encode said at least one fusion factor (MF) parameter. 15. An audio processor (100) for processing an audio signal (102) to obtain a subband representation of an audio signal (102), wherein the audio processor (100) comprises: - a cascaded overlapped critically sampled transform (104) configured to perform a cascaded overlapped critically sampled transform on at least two partially overlapping blocks (108_1; 108_2) of audio samples (102) to obtain a set (110_1,1 ) subband samples based on the first block (108_1) of audio samples (102) and obtain a corresponding set (110_2,1) of subband samples based on the second block (108_2) of audio samples (102); and - a time domain aliasing reduction stage (106) configured to perform weighting combining two respective sets (110_1.1; 110_1.2) of subband samples, one obtained based on the first block (108_1) of audio signal samples (102), and the other is obtained based on the second block (108_2) of samples of the audio signal so as to obtain a sub-band representation (112_1) with reduced aliasing of the audio signal (102); - at the same time, the cascade (104) of the cascaded overlapping critically sampled transform is configured to segment the set (124_1) of bins obtained from the first block (108_1) of samples using at least two window functions, and obtain at least , two sets (128_1,1; 128_1,2) bins based on the segmented bin set corresponding to the first block (108_1) samples; wherein the cascade (104) of the cascaded overlapping critically sampled transform is configured to segment the set (124_2) of bins obtained from the second block (108_2) of samples using at least two window functions and obtain at least , two bin sets (128_2.1; 128_2.2) based on the segmented bin set corresponding to the second block (108_2) of samples; and - while the sets (128_1,1; 128_1,2; 128_2,1; 128_2,2) of the bins are processed using the second stage of overlapped critically sampled transform from the stage of the cascade overlapped critically sampled transform, while the second stage of overlapped critically sampled transform is performed with the ability to perform overlapping critically sampled transforms having the same frame length for at least one bin set; - wherein the audio processor (100) is configured to generate a bit stream from the encoded subband representation (112_1, 112_2) with reduced aliasing of the audio signal, - while the audio processor (100) is configured to provide a bit stream containing at least one combination factor (MF) parameter signaling at least one frame length of the respective sets (128_1,1 and 128_2,1; 128_1, 2 and 128_2,2) bins in the bitstream; - while the audio processor (100) is configured to provide a bit stream containing only a subset of the parameters of the combination factor (MF), signaling the frame length of the corresponding sets (128_1,1 and 128_2,1; 128_1,2 and 128_2,2) of the bins in the stream bits. 16. An audio processor (100) for processing an audio signal (102) to obtain a subband representation of an audio signal (102), wherein the audio processor (100) comprises: - a cascaded overlapped critically sampled transform (104) configured to perform a cascaded overlapped critically sampled transform on at least two partially overlapping blocks (108_1; 108_2) of audio samples (102) to obtain a set (110_1,1 ) subband samples based on the first block (108_1) of audio samples (102) and obtain a corresponding set (110_2,1) of subband samples based on the second block (108_2) of audio samples (102); and - a time domain aliasing reduction stage (106) configured to perform weighting combining two respective sets (110_1.1; 110_1.2) of subband samples, one obtained based on the first block (108_1) of audio signal samples (102), and the other is obtained based on the second block (108_2) of samples of the audio signal so as to obtain a sub-band representation (112_1) with reduced aliasing of the audio signal (102); - at the same time, the cascade (104) of the cascaded overlapping critically sampled transform is configured to segment the set (124_1) of bins obtained from the first block (108_1) of samples using at least two window functions, and obtain at least , two sets (128_1,1; 128_1,2) bins based on the segmented bin set corresponding to the first block (108_1) samples; wherein the cascade (104) of the cascaded overlapping critically sampled transform is configured to segment the set (124_2) of bins obtained from the second block (108_2) of samples using at least two window functions and obtain at least , two bin sets (128_2.1; 128_2.2) based on the segmented bin set corresponding to the second block (108_2) of samples; and - while the sets (128_1,1; 128_1,2; 128_2,1; 128_2,2) of the bins are processed using the second stage of overlapped critically sampled transform from the stage of the cascade overlapped critically sampled transform, while the second stage of overlapped critically sampled transform is performed with the ability to perform overlapping critically sampled transforms having the same frame length for at least one bin set; - wherein the audio processor (100) is configured to generate a bit stream from the encoded subband representation (112_1, 112_2) with reduced aliasing of the audio signal, - while the audio processor (100) is configured to provide a bit stream containing at least one combination factor (MF) parameter signaling at least one frame length of the respective sets (128_1,1 and 128_2,1; 128_1, 2 and 128_2,2) bins in the bitstream; wherein the audio processor is configured to provide a bitstream containing a time domain alias reduction (TDAR) parameter for the respective bin sets (128_1.1 and 128_2.1; 128_1.2 and 128_2.2) of bins. 17. Audio processor (100) according to claim 16, wherein the audio processor (100) is configured to perform joint entropy encoding of the combining factor (MF) and time domain aliasing reduction (TDAR) parameters. 18. An audio processor (100) for processing an audio signal (102) to obtain a subband representation of an audio signal (102), wherein the audio processor (100) comprises: - a cascaded overlapped critically sampled transform (104) configured to perform a cascaded overlapped critically sampled transform on at least two partially overlapping blocks (108_1; 108_2) of audio samples (102) to obtain a set (110_1,1 ) subband samples based on the first block (108_1) of audio samples (102) and obtain a corresponding set (110_2,1) of subband samples based on the second block (108_2) of audio samples (102); and - a time-domain aliasing reduction stage (106) configured to perform a weighting combination of two respective sets (110_1,1; 110_1,2) of subband samples, one of which is obtained based on the first block (108_1) of audio signal samples (102 ) and the other is obtained based on the second block (108_2) of audio samples so as to obtain a sub-band representation (112_1) with reduced aliasing of the audio signal (102); - at the same time, the cascade (104) of the cascaded overlapping critically sampled transform is configured to segment the set (124_1) of bins obtained from the first block (108_1) of samples using at least two window functions, and obtain at least , two sets (128_1,1; 128_1,2) bins based on the segmented bin set corresponding to the first block (108_1) samples; wherein the cascade (104) of the cascaded overlapping critically sampled transform is configured to segment the set (124_2) of bins obtained from the second block (108_2) of samples using at least two window functions and obtain at least , two bin sets (128_2.1; 128_2.2) based on the segmented bin set corresponding to the second block (108_2) of samples; and - while the sets (128_1,1; 128_1,2; 128_2,1; 128_2,2) of the bins are processed using the second stage of overlapped critically sampled transform from the stage of the cascade overlapped critically sampled transform, while the second stage of overlapped critically sampled transform is performed with the ability to perform overlapping critically sampled transforms having the same frame length for at least one bin set; - wherein the audio processor (100) is configured to generate a bit stream from the encoded subband representation (112_1, 112_2) with reduced aliasing of the audio signal, - while the audio processor (100) is configured to provide a bit stream containing at least one combination factor (MF) parameter signaling at least one frame length of the respective sets (128_1,1 and 128_2,1; 128_1, 2 and 128_2,2) bins in the bitstream; wherein the audio processor (100) is configured to provide a bitstream containing at least one combining factor (MF) parameter such that the time domain aliasing reduction (TDAR) parameter can be extracted from at least one parameter pooling factor (MF). 19. An audio processor (100) for processing an audio signal (102) to obtain a subband representation of an audio signal (102), wherein the audio processor (100) comprises: - a cascaded overlapped critically sampled transform (104) configured to perform a cascaded overlapped critically sampled transform on at least two partially overlapping blocks (108_1; 108_2) of audio samples (102) to obtain a set (110_1,1 ) subband samples based on the first block (108_1) of audio samples (102) and obtain a corresponding set (110_2,1) of subband samples based on the second block (108_2) of audio samples (102); and - a time domain aliasing reduction stage (106) configured to perform weighting combining two respective sets (110_1.1; 110_1.2) of subband samples, one obtained based on the first block (108_1) of audio signal samples (102), and the other is obtained based on the second block (108_2) of samples of the audio signal so as to obtain a sub-band representation (112_1) with reduced aliasing of the audio signal (102); - at the same time, the cascade (104) of the cascaded overlapping critically sampled transform is configured to segment the set (124_1) of bins obtained from the first block (108_1) of samples using at least two window functions, and obtain at least , two sets (128_1,1; 128_1,2) bins based on the segmented bin set corresponding to the first block (108_1) samples; wherein the cascade (104) of the cascaded overlapping critically sampled transform is configured to segment the set (124_2) of bins obtained from the second block (108_2) of samples using at least two window functions and obtain at least , two bin sets (128_2.1; 128_2.2) based on the segmented bin set corresponding to the second block (108_2) of samples; and - while the sets (128_1,1; 128_1,2; 128_2,1; 128_2,2) of the bins are processed using the second stage of overlapped critically sampled transform from the stage of the cascade overlapped critically sampled transform, while the second stage of overlapped critically sampled transform is performed with the ability to perform overlapping critically sampled transforms having the same frame length for at least one bin set; wherein the audio processor (100) is configured to perform a common/combined optimization for adapting frame lengths using a trellis grid per set of subband samples. 20. An audio encoder comprising: - audio processor (100) according to any one of paragraphs. 1, 15, 16, 18 or 19; an encoder configured to encode a sub-band de-aliased representation of the audio signal so as to obtain a de-aliased sub-band encoded representation of the audio signal; and - a bit stream generator configured to generate a bit stream from an encoded sub-band representation with reduced aliasing of the audio signal. 21. A method for processing an audio signal (102) to obtain a subband representation of an audio signal (102), the method comprising the steps of: - perform a cascaded overlapping critically sampled transform on at least two partially overlapping blocks (108_1; 108_2) of audio samples (102) in order to obtain a set (110_1,1) of sub-band samples based on the first block (108_1) of audio samples ( 102) and obtain a corresponding set (110_2,1) of subband samples based on the second block (108_2) of audio samples (102); and - performing a weighting combination of two respective sets (110_1,1; 110_1,2) of sub-band samples, one obtained based on the first block (108_1) of audio samples (102), and the other obtained based on the second block (108_2) of audio samples, with so as to obtain a sub-band representation (112_1) with reduced aliasing of the audio signal (102); wherein the execution of the cascaded overlapping critically sampled transform comprises the step of segmenting the set (124_1) of bins obtained from the first block (108_1) of samples using at least two window functions, and obtaining at least two a bin set (128_1,1; 128_1,2) based on the segmented bin set corresponding to the first block (108_1) of samples; wherein the execution of the cascaded overlapping critically sampled transform comprises the step of segmenting the set (124_2) of bins obtained from the second block (108_2) of samples using at least two window functions, and obtaining at least two sets (128_2.1; 128_2.2) of bins based on the segmented bin set corresponding to the second block (108_2) of samples; and wherein the sets (128_1.1; 128_1.2; 128_2.1; 128_2.2) of bins are processed using a second overlapped critically sampled transform from a cascaded overlapped critically sampled transform, the second overlapped critically sampled transform comprising: performing overlapping critically sampled transforms having the same frame length for at least one bin set; - wherein the weighted combination of the two respective sets (110_1,1; 110_1,2) of sub-band samples is performed if identical frame lengths are used for the next two blocks (108_1, 108_2), - and/or weighted combining of the two respective sets (110_1,1; 110_1,2) of subband samples is not performed if different frame lengths are used for the two subsequent blocks (108_1, 108_2). 22. A method for processing an audio signal (102) to obtain a subband representation of an audio signal (102), the method comprising the steps of: - perform a cascaded overlapping critically sampled transform on at least two partially overlapping blocks (108_1; 108_2) of audio samples (102) in order to obtain a set (110_1,1) of sub-band samples based on the first block (108_1) of audio samples ( 102) and obtain a corresponding set (110_2,1) of subband samples based on the second block (108_2) of audio samples (102); and - performing a weighting combination of two respective sets (110_1,1; 110_1,2) of sub-band samples, one obtained based on the first block (108_1) of audio samples (102), and the other obtained based on the second block (108_2) of audio samples, with so as to obtain a sub-band representation (112_1) with reduced aliasing of the audio signal (102); wherein the execution of the cascaded overlapping critically sampled transform comprises the step of segmenting the set (124_1) of bins obtained from the first block (108_1) of samples using at least two window functions, and obtaining at least two a bin set (128_1,1; 128_1,2) based on the segmented bin set corresponding to the first block (108_1) of samples; wherein the execution of the cascaded overlapping critically sampled transform comprises the step of segmenting the set (124_2) of bins obtained from the second block (108_2) of samples using at least two window functions, and obtaining at least two sets (128_2.1; 128_2.2) of bins based on the segmented bin set corresponding to the second block (108_2) of samples; and wherein the sets (128_1.1; 128_1.2; 128_2.1; 128_2.2) of bins are processed using a second overlapped critically sampled transform from a cascaded overlapped critically sampled transform, the second overlapped critically sampled transform comprising: performing overlapping critically sampled transforms having the same frame length for at least one bin set; - form a bitstream from the encoded subband representation (112_1, 112_2) with reduced aliasing of the audio signal, - provide a bitstream containing at least one fusion factor (MF) parameter signaling at least one frame length of the respective sets (128_1,1 and 128_2,1; 128_1,2 and 128_2,2) of bins in bitstream; wherein providing the bitstream comprises a step of providing a bitstream containing only a subset of the combining factor (MF) parameters signaling the frame length of the respective sets of (128_1,1 and 128_2,1; 128_1,2 and 128_2,2) bins in bit stream. 23. A method for processing an audio signal (102) to obtain a subband representation of an audio signal (102), the method comprising the steps of: - perform a cascaded overlapping critically sampled transform on at least two partially overlapping blocks (108_1; 108_2) of audio samples (102) in order to obtain a set (110_1,1) of sub-band samples based on the first block (108_1) of audio samples ( 102) and obtain a corresponding set (110_2,1) of subband samples based on the second block (108_2) of audio samples (102); and - performing a weighting combination of two respective sets (110_1,1; 110_1,2) of subband samples, one obtained based on the first block (108_1) of audio samples (102), and one obtained based on the second block (108_2) of audio samples, with so as to obtain a sub-band representation (112_1) with reduced aliasing of the audio signal (102); wherein the execution of the cascaded overlapping critically sampled transform comprises the step of segmenting the set (124_1) of bins obtained from the first block (108_1) of samples using at least two window functions, and obtaining at least two sets (128_1.1; 128_1.2) of bins based on the segmented bin set corresponding to the first block (108_1) of samples; wherein the execution of the cascaded overlapping critically sampled transform comprises the step of segmenting the set (124_2) of bins obtained from the second block (108_2) of samples using at least two window functions, and obtaining at least two sets (128_2.1; 128_2.2) of bins based on the segmented bin set corresponding to the second block (108_2) of samples; and wherein the sets (128_1.1; 128_1.2; 128_2.1; 128_2.2) of bins are processed using a second overlapped critically sampled transform from a cascaded overlapped critically sampled transform, the second overlapped critically sampled transform comprising: performing overlapping critically sampled transforms having the same frame length for at least one bin set; - form a bitstream from the encoded subband representation (112_1, 112_2) with reduced aliasing of the audio signal, - provide a bitstream containing at least one fusion factor (MF) parameter signaling at least one frame length of the respective sets (128_1,1 and 128_2,1; 128_1,2 and 128_2,2) of bins in bitstream; wherein providing the bitstream comprises a step of providing a bitstream containing a time domain alias reduction (TDAR) parameter for the respective bin sets (128_1,1 and 128_2,1; 128_1,2 and 128_2,2) bins. 24. A method for processing an audio signal (102) to obtain a subband representation of an audio signal (102), the method comprising the steps of: - perform a cascaded overlapping critically sampled transform on at least two partially overlapping blocks (108_1; 108_2) of audio samples (102) in order to obtain a set (110_1,1) of sub-band samples based on the first block (108_1) of audio samples ( 102) and obtain a corresponding set (110_2,1) of subband samples based on the second block (108_2) of audio samples (102); and - performing a weighting combination of two respective sets (110_1,1; 110_1,2) of sub-band samples, one obtained based on the first block (108_1) of audio samples (102), and the other obtained based on the second block (108_2) of audio samples, with so as to obtain a sub-band representation (112_1) with reduced aliasing of the audio signal (102); wherein the execution of the cascaded overlapping critically sampled transform comprises the step of segmenting the set (124_1) of bins obtained from the first block (108_1) of samples using at least two window functions, and obtaining at least two sets (128_1.1; 128_1.2) of bins based on the segmented bin set corresponding to the first block (108_1) of samples; wherein the execution of the cascaded overlapping critically sampled transform comprises the step of segmenting the set (124_2) of bins obtained from the second block (108_2) of samples using at least two window functions, and obtaining at least two sets (128_2.1; 128_2.2) of bins based on the segmented bin set corresponding to the second block (108_2) of samples; and wherein the sets (128_1.1; 128_1.2; 128_2.1; 128_2.2) of bins are processed using a second overlapped critically sampled transform from a cascaded overlapped critically sampled transform, the second overlapped critically sampled transform comprising: performing overlapping critically sampled transforms having the same frame length for at least one bin set; - form a bitstream from the encoded subband representation (112_1, 112_2) with reduced aliasing of the audio signal, - provide a bitstream containing at least one fusion factor (MF) parameter signaling at least one frame length of the respective sets (128_1,1 and 128_2,1; 128_1,2 and 128_2,2) of bins in bitstream; wherein providing the bitstream comprises providing a bitstream containing at least one combining factor (MF) parameter such that the time domain aliasing reduction (TDAR) parameter can be extracted from at least one pooling factor parameter (MF). 25. A method for processing an audio signal (102) to obtain a subband representation of an audio signal (102), the method comprising the steps of: - perform a cascaded overlapping critically sampled transform on at least two partially overlapping blocks (108_1; 108_2) of audio samples (102) in order to obtain a set (110_1,1) of sub-band samples based on the first block (108_1) of audio samples ( 102) and obtain a corresponding set (110_2,1) of subband samples based on the second block (108_2) of audio samples (102); and - performing a weighting combination of two respective sets (110_1,1; 110_1,2) of sub-band samples, one obtained based on the first block (108_1) of audio samples (102), and the other obtained based on the second block (108_2) of audio samples, with so as to obtain a sub-band representation (112_1) with reduced aliasing of the audio signal (102); wherein the execution of the cascaded overlapping critically sampled transform comprises the step of segmenting the set (124_1) of bins obtained from the first block (108_1) of samples using at least two window functions, and obtaining at least two sets (128_1.1; 128_1.2) of bins based on the segmented bin set corresponding to the first block (108_1) of samples; wherein the execution of the cascaded overlapping critically sampled transform comprises the step of segmenting the set (124_2) of bins obtained from the second block (108_2) of samples using at least two window functions, and obtaining at least two sets (128_2.1; 128_2.2) of bins based on the segmented bin set corresponding to the second block (108_2) of samples; and wherein the sets (128_1.1; 128_1.2; 128_2.1; 128_2.2) of bins are processed using a second overlapped critically sampled transform from a cascaded overlapped critically sampled transform, the second overlapped critically sampled transform comprising: performing overlapping critically sampled transforms having the same frame length for at least one bin set; - perform a common/combined optimization for adapting frame lengths using a trellis grid per set of subband samples. 26. A storage medium containing a computer program stored thereon for implementing the method according to any one of paragraphs. 21-25.

Images