KR960003453B1 - Stereo digital audio coder with bit assortment - Google Patents

Abstract

The stereo digital audio coding device for coding a digital audio signal by adaptively allotting bit to each channel and frame comprises a perceptual entropy calculating unit for obtaining a first power density spectrum by inputting a group of frame composed of a plurality of frames of right, left, center, right and left surround channels and obtaining the perceptual for the above channels and the frames of each channel by using an auditory characteristic of human; an adaptive channel and frame bit allocation unit for adaptively allotting a bit to an audio signal for the channels and each frame of each channel in response to the perceptual entropy; and a coder for coding the audio signal by the bit supplied from the adaptive channel and frame bit allocation unit.

Description

Stereo digital audio encoding device that adaptively allocates and encodes channels and frames

1 is a block diagram showing a stereo digital audio encoding apparatus according to the present invention.

2 is a block diagram showing a frame group (GOF) portion of the stereo digital audio encoding apparatus of the present invention having five channels (L, R, center, L surround and R surround) shown in FIG.

3 is a graph showing cognitive information amount PE ₁ (X axis) versus frame bit allocation state (Y axis) of the present invention.

* Explanation of symbols for main parts of the drawings

100: L channel frame group 120: R channel frame group

130: center channel frame group 140: L surround channel frame group

150: R surround channel frame group 160: cognitive information amount calculation unit

170: adaptive channel and frame bit allocation unit 180-220: encoder

230: multiplexing unit

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a digital audio coder, and more particularly, to an amount of cognitive information measured using human auditory characteristics with respect to a digital audio signal input through a channel of L, R, center, L, and R surround. Entropy (PE) relates to a stereo digital audio encoding apparatus for adaptively bit allocation and encoding on a channel and each frame.

It is relatively narrow in high-definition television (HDTV) information transmission system, which is being developed for the purpose of reproducing sound quality signals such as compact discs (CDs) and digital audio tape recorders (DATs) that are currently in use. Since video and audio signals must be transmitted through a 6MHz transmission line, an efficient signal compressor method has been required for audio signals as in video signal processing.

For this purpose, the high-quality digital audio signal processing technology for HDTV has recently been applied to the digital audio device-level sound quality even with a relatively simple receiver at a low data rate by using an adaptive transform coding method that reflects human hearing characteristics. Actively researched for reproducible algorithms and hardware implementation.

As the above-described adaptive transform encoding technique, in a stereo digital audio encoding apparatus having a plurality of channels, for example, a conventional method of allocating bits to five channels includes a method of encoding each channel independently, and a first half. There are a method of allocating different bit amounts to three channels (that is, L, R and center channels) and the remaining two second channels (that is, L and R surround channels). As described above, the reason for allocating different bit amounts by dividing into two sections is that the important information amount of the audio signal is included in the first three channels, and the latter two channels contain relatively less important information amounts. Based on the above-described signal characteristics, the latter two-channel audio signal is composed of an anti-aliasing filter using a band pass filter having a bandwidth narrower than the first three channels, and the sampling frequency set in the first three channels. It is usually set smaller than (Sampling Frequency). However, these coding schemes are advantageous when the first channel contains more information than the latter channel as described above, but in the surround audio system, the coding efficiency when more important information is included in the latter, i.e., L and R surround channels. There was a problem of falling sound quality at the same time falling.

Accordingly, the main object of the present invention is to adaptively adapt bits in response to L, R, center, L surround and R surround channels and the amount of human cognitive information for each frame in order to increase coding efficiency and further improve sound quality. Disclosed is a channel to be allocated and encoded, and a stereo digital audio encoding apparatus for adaptively bit allocation and encoding for each frame.

In order to achieve the above object, the present invention adapts a digital audio signal consisting of a group of frames (GOF) having a plurality of frames for left (L), right (R), center, L and R surround channels. By assigning and encoding bits, a frame group having a plurality of frames of the L, R, center, L, and R surround channels is input to calculate a first power density spectrum, and the first power density spectrum is calculated. A cognitive information amount calculating unit configured to calculate a perceptual entropy for the five channels and the frames of each channel by using the power density spectrum of the human auditory characteristics; Adaptive to allocate a bit adaptively to the channel output from the frame group of each channel and the audio signal for each frame of each channel in response to the amount of cognition information for each frame of the channel and channel obtained by the cognitive information amount calculating unit Adaptive Channel And Frame Bit Allocation section; A coder for encoding an audio signal for a channel output from the frame group and each frame of each channel by the bits provided by the adaptive channel and frame bit allocation unit is provided.

In the present invention, the amount of cognitive information for audio signals input to L, R, center, L surround and R surround channels is typically different in size for each channel and also in sizes between frames in the same channel. For example, if the amount of cognitive information is large, the human ear has a high probability of feeling an error about the audible level. Therefore, more bits should be allocated and encoded. Based on the concept

As will be described later, in order to implement the present invention, the amount of cognitive information for each channel is first obtained, and then, the average and variance values of the cognitive information amount are obtained, and the average and variance values are used. The weight is assigned according to the amount of cognitive information of the channel, and the weight is assigned according to the amount of cognitive information for each frame in the same channel to allocate bits.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the drawings.

FIG. 1 is a block diagram showing a stereo digital audio encoding apparatus for adaptively bit allocation and encoding L, R, center, L surround and R surround channels and a plurality of frames within each channel according to the present invention. , Center, L, and R surround channel frame groups 110, 120, 130, 140, and 150, cognitive information amount calculating unit 160, adaptive channel and frame bit allocation unit 170, L, R, center, L, and R surround channels Encoders 180 to 220 and a multiplexer 230 that encode by a known encoding scheme are included. A digital audio signal of 5M frames having a plurality of frames of L, R, center, L, and R surround channels input by a stereo digital audio encoding apparatus including such devices is added to the amount of cognitive information measured for the channel and each frame. By adaptively allocating and encoding bits, encoding efficiency can be increased and sound quality can be improved.

The L, R, center, L surround, and R surround channel frame groups 110, 120, 130, 140, and 150 respectively receive digital audio signals input through the L, R, center, L, and R surround channels, respectively. To calculate, classify the data into frame group units having a plurality of frames (where M denotes the number of frames in the frame group of each channel) and provide the data to the encoders 100, 190, 200, 210, and 220 and the cognitive information amount calculating unit 160, respectively. do.

Referring to FIG. 2, FIG. 2 shows the GOF configuration of the aforementioned L, R, center, L surround and R surround channels. As shown, one frame consists of N samples, where N is a positive integer, and is typically in the range of 10 msec to about 40 msec. In addition, as shown, the GOF of each channel is composed of M frames, the GOF of one channel is composed of N × M samples, the GOF for five channels of the present invention is N × 5M samples Is done. Here, N value is 10msec which is a stationary process section of an audio signal using a delay time inevitably generated during encoding and decoding of an image signal input from a video camera (not shown). It can be configured in the unit of 40msec, M value is determined by the delay time for performing the coding and decoding (Codec).

Referring back to FIG. 1, the cognitive information amount calculator 160 analyzes an audio signal during a delay time of encoding and decoding an image signal to calculate an amount of cognitive information limited to human auditory characteristics. The frame bit allocation unit 170 may increase coding efficiency by varying the bit allocation amount for each of the L, R, center, L, and R surround channels and frames of each channel. The cognitive information amount calculation unit 150 is a finite digital audio signal source of one frame, i.e., x (n) consisting of N samples provided from the L, R, center, L, and R surround GOF units 110, 120, 130, 140, and 150 described above. Masking Threshold (M (w)) can be obtained by using the human auditory characteristics according to the power density spectrum of S _xx (w), and then the amount of cognitive information can be obtained by the following equation. have. For example, first, an approximate power density spectrum S _xx (w) for a finite digital audio signal source x (n) of one frame is calculated by the following equation (1).

In other words,

In addition, since M (w) described above corresponds to a power value of a region having a power density spectrum of S _xx (w) that cannot be detected by a human ear in a predetermined frequency component, If the signal is reproduced with an error value less than or equal to M (w), the area cannot be detected by the ear. Therefore, in order to obtain the power density spectrum (S _ee (w)) of the area, that is, the error signal, the input signal of the encoder Is x (n) and the output signal is y (n), and the error signal e (n) is calculated by the following equation (2).

E (n) = x (n) -y (n) (2)

Therefore, by replacing S _ee (w) for N finite signal sources with M (w), the amount of cognitive information (R _PE ) for transmitting the signal x (n) such that an error cannot be detected as the human ear is given by It can calculate by (3).

On the other hand, when the masking threshold M (w) uses the same band split coder having the same frequency every frame and infinite frequency band, the theoretically feasible minimum bit rate is the value obtained by the above equation (3).

However, in the case of configuring the encoder for the actual audio data, each audio parameter is analyzed for each frame time-domain signal having N samples, and then the quantization level is set differently for each frequency band divided accordingly. Since the auditory parameters vary every N samples, the substantially necessary bit rate will be greater than the amount of cognitive information calculated by equation (3) above.

For example, when the input signal x (n) is divided into L frequency bands, where L is a positive integer, and encoded, the power density spectrum S _xx (i) and masking of the i th frequency band are masked. The threshold M (i) can be calculated approximately by the following equations (4 and 5). In other words,

R _i is a frequency domain corresponding to the i-th division band, and S _xx (wi) is a value corresponding to a power density spectrum of the j-th frequency component in an N point discrete Fourier transform (DFT). Also,

Here, M (W _j ) represents the minimum value of the masking threshold value M (w) belonging to the j-th division band.

For example, in the case of obtaining a power density spectrum using a 1024-point DFT (that is, L = 1024) and dividing it into 32 frequency bands (that is, L = 32), the cognitive information amount R _PE is expressed as follows. Can be obtained by

Next, a concept for calculating the amount of cognitive information of 1 GOF for each channel will be described. Recognition amount of information PE ₁ by using the power density spectrum and masking threshold for the audio data of the i-th frame consisting of N samples, where i is a positive integer greater than 0 and smaller than the number of frames. And then the standard deviation representing the change in the average perceptual information amount PE ₁₀ and PE ₁ for each PE ₁₀ described above for the group of 1 frame having 5M total frames for L, R, center, L and R surround channels. PE _std is obtained by the following equations (7 and 8).

In other words,

(7), (8)

Next, the adaptive channel and frame bit allocator 170 receives L, R, center, L, and R surround channels obtained from the cognitive information amount calculating unit 160 and the amount of cognitive information for each frame of each channel. By adaptively allocating the bit amount to provide to the encoder (180, 190, 200, 210, 220) for the L, R, center, L and R surround channels, respectively.

Hereinafter, a method of allocating bits variably according to a channel obtained by the cognitive information amount calculating unit 160 and the cognitive information amount of a frame of each channel will be described in detail.

Referring to FIG. 3, FIG. 3 is referred to as PE _i for the amount of cognition information for the i-th frame in 1 GOF, which is composed of a total of 5M frames for L, R, center, L and R surround channels, and 1 When the average recognition information amount of GOF is PE ₁₀ , it is a graph showing a frame bit allocation state (Index) according to PE _i . In the same figure, the index of the vertical (Y) axis represents a bit allocation state having an integer value between -q and + q, and X _i of the horizontal (X) axis may have one frame determined by the following equation (9). Indicates a predetermined amount of cognitive information.

For example, the weight δ applied in the case of M = 8 can be obtained as shown in Table 1 below with reference to the experimental results for PE ₁₀ and PE _std obtained through the above equations (7 and 8).

TABLE 1

That is, when PE ₁₀ is 0 to 0.315 and PE _std is 0 to 0.625, δ is 1000. Referring to Table 1 for other values, δ may be obtained in the same manner as described above.

Then, using the weight δ obtained based on Table 1, the cognitive information amount X _i can be obtained by the following equation (9).

(a)

Here, there is a relationship of -q≤i≤q between i, -q and q, and the following conditions are assumed.

sign (i) = 1 if (i> 0)

sign (i) =-1 if (i <0)

sign (i) = 0 if (i = 0)

In addition, the δ value is a weight determined based on the PE ₁₀ and PE _std values calculated by the above equations (7 and 8) for the 5M PE ₁ values for the 1 GOF described above.

As an example of the present invention, when the q value is 8, that is, the number of frame bits corresponding to each index corresponding to the predetermined amount of recognition information X ₁ extracted through the above-described equation (9) is based on the above-described equations. It can be obtained as shown in Table 2.

TABLE 2

(Unit, Frame Bit: Bit / Frame, Bit Rate: K Bit / sec)

That is, referring to Table 2, for example, when Index is 0, the number of frame bits is 3072, that is, the number of bits allocated to one frame when encoded at an information rate of 128 Kbps per second in a frame unit composed of 1152 samples. Is 3072 bits, and the number of bits suggested by the audio section of the Motion Picture Expert Grpup (MPEG) is allocated, and as the Index increases, the number of frame bits far exceeds the average, whereas as the Index decreases, the number of frame bits exceeds the average. You will notice that it is less allocated.

Referring back to FIG. 1, the encoders 180, 190, 200, 210, and 220 provide an adaptive channel and frame bit allocation unit 170 for the audio signals for each channel provided by the L, R, center, L, and R surround channel GOF units 110, 120, 130, 140, and 150, respectively. Is encoded in a known encoding scheme by the channels provided in &lt; RTI ID = 0.0 &gt; and &lt; / RTI &gt; The MUX 230 multiplexes the encoded data provided by the above-described encoders 180, 190, 200, 210, and 220, the channel provided by the adaptive channel and frame bit allocator 170, and the bit allocation information allocated to the frames of each channel. Channel is converted into a bit stream and output as appropriate.

According to the stereo digital audio encoding apparatus for adaptively bit-allocating and encoding a channel and each frame according to the present invention described above, it is possible to respond to an average and a variance (or standard deviation) of the amount of cognitive information for a channel and a frame of each channel. By adaptively assigning and encoding bits to channels and frames within each channel, there is a great advantage of increasing coding efficiency and improving sound quality.

Claims (3)

By adaptively allocating and encoding a digital audio signal consisting of a group of frames (GOF) having a plurality of frames for the left (L), right (R), center, L, and R surround channels, A first power density spectrum is calculated by inputting a frame group having a plurality of frames of the L, R, center, L, and R surround channels, and using the human auditory characteristics of the first power density spectrum. Perceptual Entropy is calculated for the five channels and the frames of each channel, and the first power density spectrum is used for the five channels and the frames of each channel using the human auditory characteristics. A cognitive information amount calculating unit for calculating Perceptual Entropy; Adaptive to allocate a bit adaptively to the channel output from the frame group of each channel and the audio signal for each frame of each channel in response to the amount of cognition information for each frame of the channel and channel obtained by the cognitive information amount calculating unit Adaptive Chaunel And Frame Bit Allocatiom; Adaptively bit allocation to a channel and each frame including an encoder for encoding a channel output from the frame group and an audio signal for each frame of each channel by bits provided by the adaptive channel and frame bit allocation unit Stereo digital audio encoding device for encoding. The second power of claim 1, wherein the cognitive information amount calculation unit obtains a masking threshold based on the first power density spectrum, and is a region below the masking threshold and is not detected by the human ear. Obtaining a density spectrum, and then calculating L, R, center, L, and R surround channels and the amount of cognitive information for each channel and each frame using the first power density spectrum and the second power density spectrum. Stereo digital audio coding device. The method of claim 1, wherein the adaptive channel and frame bit allocation unit comprises one frame having 5M frames of L, R, center, L, and R surround channels, where M is the number of frames in GOF for one channel. The amount of cognitive information for the i th frame of the group (where i represents any frame of 1 GOF as a positive integer) is PE ₁ , and the average cognitive information amount PE ₁₀ in the 1 frame group and for the 1 frame group Using the standard deviation PE _std to obtain a weight (δ), and then using the average recognition information amount PE ₁₀ and the weight can be obtained by the following equation, the recognition information amount (X1) having a predetermined value, here, if (i> 0) sign (i) = 1 if (i <0) sign (i) =-1 if (i = 0) sign (i) = 0 i is within the Index range, and is then adapted to the final L, R, center, L and R surround channels and each frame of each channel by an Index representing a frame bit allocation state corresponding to the predetermined amount of cognitive information. Stereo digital audio decoding apparatus characterized in that the allocation of bits.