JP2005010337A - Audio signal compression method and apparatus - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To speed up psychological acoustic sense analysis processing while conforming the processing to MODEL1 of a psychological acoustic sense model by ISO/IEC 11172-3. <P>SOLUTION: An effective band estimation part 11 estimates a band (a valid band variable) effectively used for psychological acoustic sense analysis processing on the basis of the number of subbands in use which is determined according to the bit rate and sampling frequency of an audio signal, a critical subband number determination part 12 determines the critical number of usable subband ( a critical band variable) as to nonlinear subbands used to make a noise choice according to the estimated valid band variable, and after an effective band correction part 13 corrects the valid band variable, the psychological acoustic sense analysis processing is carried out by using the critical band variable and valid band variable. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to an audio signal compression method and an audio signal compression device, and more particularly, to an audio signal compression method and an audio signal compression device for compressing an audio signal by subband coding and psychoacoustic analysis processing using human auditory characteristics.
[0002]
[Prior art]
MPEG (Moving Picture Coding Experts Group) 1 Audio, which is one of the compression coding systems for digital audio signals (hereinafter referred to as audio signals), is defined by ISO / IEC 11172-3 of the international standard system. In particular, MPEG1 Audio layer 2 (hereinafter referred to as MP2) is used for audio recording of a digital video camera or the like.
[0003]
FIG. 7 is a block diagram showing a configuration of an audio signal compression circuit of MPEG 1 Audio (layer 1 and layer 2) defined by ISO / IEC 11172-3.
[0004]
As illustrated, the audio signal compression circuit 100 includes an audio signal input unit 101 that inputs a digital audio signal, a filter bank application unit 102 that divides the audio signal into subbands, and a scale factor calculation unit that calculates a scale factor. 103, a quantizer 104 that performs quantization, a psychoacoustic analysis unit 105b that performs masking (psychological auditory analysis processing) based on human auditory characteristics to reduce the amount of data, and bit allocation that calculates bit allocation The calculation unit 106 includes a bit stream generation unit 107 that generates an MPEG1 Audio bit stream.
[0005]
When an audio signal digitized by an external A / D converter or the like is input to the audio signal input unit 101 in units of frames, the input audio signal is passed to the filter bank application unit 102 and the psychoacoustic analysis unit 105b. The filter bank application unit 102 linearly divides the input audio signal into 32 subbands. The divided subbands are input to the scale factor calculation unit 103. The scale factor calculation unit 103 detects a sample having the maximum absolute value in each subband, calculates a scale factor obtained by converting the value into a logarithm and quantizing. Using this scale factor, normalization is performed so that the maximum amplitude becomes 1.0, and the dynamic range of each subband is made uniform.
[0006]
On the other hand, the psychoacoustic analysis unit 105b calculates a perception threshold value (hereinafter referred to as a masking threshold value) for human hearing for each subband from the input audio signal. The bit allocation calculation unit 106 calculates bit allocation using the calculated masking threshold.
[0007]
In the quantizer 104, the subband is quantized using the scale factor calculated by the scale factor calculation unit 103 and the output of the bit allocation calculation unit 106. Finally, the bitstream generation unit 107 generates an MPEG1 Audio bitstream.
[0008]
FIG. 8 is a flowchart illustrating an example of processing in the psychoacoustic analysis unit.
This procedure is based on the psychoacoustic model MODEL1 defined in ISO / IEC 11172-3, and consists of eight procedures.
[0009]
Step S100: Fourier transform
When an audio signal is input, the psychoacoustic analysis unit 105b first performs a fast Fourier transform (FFT) or the like to convert the audio signal into a frequency component. Note that MP2 handles 1152 audio samples per frame, but frequency conversion is performed on 2 factorial 1024 audio samples due to the nature of FFT.
[0010]
Here, the power spectrum is obtained by taking the sum of squares of the adjacent two frequency components. By combining adjacent two frequency components, the bandwidth changes from 1024 to 512.
[0011]
Step S101: Sound pressure level calculation
Here, the 512-band power spectrum calculated in step S100 is divided into 32 linear subbands. Each divided subband includes 16 power spectra. The sound pressure level of the subband is calculated by taking the sum of the power spectrum. Therefore, here, 16 sums are repeated 32 times.
[0012]
Step S102: Sound selection
Here, a component to be marked as a sound is searched from the power spectrum of 512 bands. In the search method, first, all power spectra are searched from the lower band to the upper band, and a place that becomes a variation point, that is, a place where the spectrum is convex upward is marked as a potential sound component. A place where the marked component is 7 dB larger than the surrounding components is finally determined as a sound component.
[0013]
Step S103: Noise selection
Here, the 512-band power spectrum is divided into non-linear subbands to detect noise components. The method of dividing the nonlinear subband depends on the sampling frequency, and is divided into 25 when the sampling frequency is 32 kHz. The others are divided into 27 pieces. Noise is detected for each non-linear subband, and it is determined whether or not it is noise by adding all power spectra belonging to each subband while weighting them. This process is a product-sum process.
[0014]
Step S104: Masker selection
The sound and noise components (these are called maskers) extracted in steps S102 and S103 are thinned out at predetermined intervals to select a masker. The thinning method is defined in ISO / IEC 11172-3.
[0015]
Step S105: Global mask threshold calculation
The 512-band power spectrum is divided into non-linear 133 subbands, and a mask threshold is calculated for each subband region using the masker selected in the process of step S104. The 133 non-linear subbands are hereinafter referred to as frequency bands. The frequency band division method is defined in ISO / IEC 11172-3.
[0016]
Step S106: Minimum mask threshold calculation
The minimum mask threshold value for the linear 32-divided subband calculated in step S101 is obtained. In this process, the mask threshold value for each nonlinear frequency band obtained in step S105 is made to correspond to a linear sub-band of 32 divisions.
[0017]
Step S107: SMR calculation
Here, for each of the 32 subbands, a difference (SMR: Signal Mask Ratio) between the minimum mask threshold obtained in step S106 and the maximum value of each subband signal is calculated.
[0018]
The processing in the psychoacoustic analysis unit 105b as described above has a drawback that the calculation amount is large and the calculation takes a very long time.
In MP2 encoding, an audio signal is divided into 32 subbands and encoded by subband encoding (also referred to as band division encoding). However, all 32 divided subbands are not used. The number of subbands used depends on the bit rate and sampling frequency, and only 30 at most are used. According to ISO / IEC 11172-3, the maximum number of subbands used is very small, 12 under the conditions of a bit rate of 32 kbps and a sampling frequency of 32 kHz. Conventionally, the processing other than the psychoacoustic analysis unit 105b has been speeded up by using / not using subbands in ISO / IEC 11172-3.
[0019]
Some methods for reducing the amount of computation of the psychoacoustic analysis unit 105b have been proposed (see, for example, Patent Document 1).
[0020]
[Patent Document 1]
Japanese Patent Laid-Open No. 2002-189499 (paragraph numbers [0010] to [0014], FIG. 1)
[0021]
[Problems to be solved by the invention]
However, the above-described conventional technique uses a method different from the psycho-acoustic model MODEL 1 according to ISO / IEC 11172-3, and has a problem that it is a method that does not comply with international standards.
[0022]
The present invention has been made in view of the above points, and is capable of speeding up the processing of the psychoacoustic analysis unit while conforming to MODEL1 of the psychoacoustic model according to ISO / IEC 11172-3. An object is to provide a signal compression method and an audio signal compression apparatus.
[0023]
[Means for Solving the Problems]
In the present invention, in order to solve the above-described problem, in the audio signal compression method for compressing the audio signal by the sub-band coding and the psychoacoustic analysis process using the characteristics of human auditory sense, Based on the number of subband usages determined accordingly, a band effective for use in the psychoacoustic analysis process is estimated by linear approximation, and noise is determined in the psychoacoustic analysis process according to the estimated effective band. In the non-linear sub-band used for selection, the critical sub-band number sufficient for use is determined, and the effective band is corrected and determined according to the linear or non-linear sub-band used in the psychoacoustic analysis processing. The psychoacoustic analysis processing is performed using the critical subband number and the corrected effective band. A method is provided.
[0024]
According to the above method, based on the number of subbands determined according to the bit rate and sampling frequency of the audio signal, the effective band to be used in the psychoacoustic analysis processing is estimated, and the estimated effective band The number of critical subbands that are sufficient for use is determined in the nonlinear subbands used for noise selection according to the above, and after correcting the effective band, the number of critical subbands and the effective band are used. Since psychoacoustic analysis processing is performed, the bandwidth used for the calculation is limited, the amount of calculation is reduced, and the processing speed is increased.
[0025]
DETAILED DESCRIPTION OF THE INVENTION
Embodiments of the present invention will be described below with reference to the drawings.
FIG. 1 is a functional block diagram showing the main part of an audio signal compression apparatus according to an embodiment of the present invention.
[0026]
The audio signal compression apparatus 10 according to the embodiment of the present invention speeds up the process performed in the psychoacoustic analysis model 1051 of the psychoacoustic model defined in ISO / IEC 11172-3 shown in FIG. And has an effective band estimation unit 11, a critical subband number determination unit 12, and an effective band correction unit 13.
[0027]
In FIG. 1, the conventional psychoacoustic analysis unit 105b in the configuration of the audio signal compression circuit of MPEG1 Audio (layer 1 and layer 2) defined by ISO / IEC 11172-3 shown in FIG. It corresponds to the psychoacoustic analysis unit 105a. The other components are the same as those shown in FIG. 7 and are not shown.
[0028]
The effective band estimation unit 11 is a band effective for use in the psychoacoustic analysis unit 105a based on the number of subbands used determined in accordance with the bit rate and sampling frequency of the audio signal (hereinafter, a valid band ) Called variables) by linear approximation.
[0029]
The critical subband number determination unit 12 uses a critical subband that is sufficient for use in the non-linear subband used in the noise selection in step S103 of FIG. 8 in the psychoacoustic analysis unit 105a according to the estimated valid band variable. The number of bands (hereinafter referred to as a critical band variable) is determined.
[0030]
The effective band correction unit 13 corrects the valid band variable according to the linear or non-linear subband used in the psychoacoustic analysis unit 105a.
As the linear subband used in the psychoacoustic analysis unit 105a, the sound pressure level calculation in step S101 is included in the processing performed by MODEL1 of the psychoacoustic model defined in ISO / IEC 11172-3 shown in FIG. There are subbands obtained by linearly dividing the 512 band to be used into 32 bands.
[0031]
As the non-linear sub-band, the sub-band obtained by nonlinearly dividing the 512 band used in the above-described noise selection into 25 or 27 or used in the global mask threshold calculation in the process of step S105 in FIG. There are 133 subbands which are nonlinearly divided (details will be described later).
[0032]
Hereinafter, the operation of the audio signal compression apparatus 10 shown in FIG. 1 will be described.
First, the valid band estimation unit 11 estimates a valid band variable.
In the audio signal compression apparatus 10, 1152 audio samples are used in the configuration shown in FIG. 7 except for the psychoacoustic analysis unit 105a. Therefore, the subbands other than the psychoacoustic analysis unit 105a are obtained by dividing the audio signal of 1152 audio samples into 32 frequency bands. When the bit rate and sampling frequency of the audio signal are determined, the number of subbands used is determined according to the specification of ISO / IEC 11172-3. For example, when the bit rate is 32 kbps and the sampling frequency is 32 kHz, the number of subbands used is twelve.
[0033]
In the processing by the psychoacoustic analysis unit 105a, the audio signal is represented by 512 bands. Therefore, the effective band estimation unit 11 estimates a valid band, that is, a valid band variable among the 512 bands used in the psychoacoustic analysis unit 105a by linear approximation based on the number of subbands used.
[0034]
FIG. 2 is a diagram illustrating how to estimate a valid band variable.
The figure shows the correspondence between the audio data of 1152 audio samples divided into 32 frequency bands and the 512 bands used in the psychoacoustic analysis unit 105a divided into 32 linear frequency bands. Yes.
[0035]
If the number of subbands used in the 1152 audio sample other than the psychoacoustic analysis unit 105a is “sblimit” and the valid band variable is expressed as “valid_band”, it becomes a simple proportional calculation. Therefore, the valid band variable is estimated as follows: it can.
[0036]
[Expression 1]
valid_band = sblimit × 512/32 (1)
The subband usage number “sblimit” is determined according to the bit rate and sampling frequency of the audio signal. For example, when the bit rate is 32 kbps and the sampling frequency is 32 kHz, the number of subbands used is “sblimit = 12”. As shown in FIG. 2, when this is made to correspond to the 512 band used in the psychoacoustic analysis unit 105a, “valid_band = 192” is obtained from the equation (1), and a valid band variable can be estimated.
[0037]
Next, the critical subband number determination unit 12 determines a critical band variable to be used in the noise selection in step S103 of FIG. 8 by the psychoacoustic analysis unit 105a according to the estimated valid band variable.
[0038]
FIG. 3 is a diagram of the critical band variable when the sampling frequency is 32 kHz and the bandwidth corresponding to the value of the critical band variable.
The method of dividing the non-linear subband is defined by ISO / IEC 11172-3 and depends on the sampling frequency. When the sampling frequency is 32 kHz, the 512 band is divided into 25, and the maximum critical band variable is “24”. Otherwise, the 512 band is divided into 27, and the maximum critical band variable is “26”. In the figure, in the case of 32 kHz, the maximum critical band variable is “24”.
[0039]
Conventionally, the critical band variable is calculated up to “24” which is the maximum, regardless of how much band is used (regardless of the number of subbands used), and up to 480 in terms of bandwidth. In the embodiment of the present invention, a critical band variable sufficient for use is determined in the noise selection in accordance with the valid band variable estimated by the effective band estimation unit 11. For example, as described above, when the bit rate is 32 kbps and the sampling frequency is 32 kHz, the valid band variable is “valid_band = 192”. Therefore, referring to FIG. 3, the critical band variable corresponding to the valid band variable is “19”. It becomes. The relationship between the critical band variable and the valid band variable is expressed by the following equation.
[0040]
[Expression 2]
band_width [crit_band-1] <valid_band ≦ band_width [crit_band] (2)
In the above equation, the critical band variable is expressed as “crit_band”, the valid band variable is expressed as “valid_band”, “band_width” indicates the bandwidth, for example, “band_width [crit_band]” is the critical band variable “crit_band”. Indicates bandwidth.
[0041]
As in Expression (2), a critical band variable in which the value of “valid_band” falls between the bandwidth of “crit_band-1” and the bandwidth of “crit_band” is searched.
[0042]
In this way, since the critical band variables need only be calculated according to the estimated valid band variables, the amount of calculation can be reduced.
[0043]
Next, the valid band correction unit 13 corrects the valid band. First, the valid band variable is corrected according to the critical band variable determined by the critical subband number determination unit 12. Specifically, correction is performed according to the following equation.
[0044]
[Equation 3]
valid_band = band_width [crit_band] (3)
As shown in Expression (3), the valid band variable is matched with the bandwidth of the critical band variable. For example, as shown in FIG. 3, the valid band variable is raised by “Δwd” to match the bandwidth of the critical band variable “19”.
[0045]
In the process of the psychoacoustic analysis unit 105a, in the global mask threshold calculation in step S105 of FIG. 8 showing the process of MODEL1 of the psychoacoustic model by ISO / IEC 11172-3, a non-linear subband obtained by dividing the 512 band into 133 pieces. (Frequency band) is used. In the following processing, the valid band variable is corrected in order to correspond to this frequency band.
[0046]
FIG. 4 is a diagram showing the relationship between the frequency band and the bandwidth when the sampling frequency is 32 kHz.
The frequency band division method is defined in ISO / IEC 11172-3.
[0047]
According to the frequency band shown here, it is possible to correspond to 133 non-linear frequency bands by correcting the valid band variable in the same manner as the correction according to the critical band variable described above. Specifically, correction is performed according to the following equation.
[0048]
[Expression 4]
band_width [frequency_band-1] <valid_band ≦ band_width [frequency_band] (4)
Here, the frequency band is described as “frequency_band”. “Band_width [frequency_band]” is the bandwidth of the frequency band.
[0049]
In Equation (4), the bandwidth of the frequency band is determined, and the determined bandwidth of the frequency band is a corrected valid band variable. That is, the following equation is obtained.
[0050]
[Equation 5]
valid_band = band_width [frequency_band] (5)
As described above, first, the valid band variable is corrected from the subband obtained by linearly dividing the 512 band into 32 bands, and then the accuracy is obtained by correcting with the subband obtained by nonlinearly dividing the 512 band into 133 bands. Good correction can be made.
[0051]
In the processing of the psychoacoustic analysis unit 105a, in the sound pressure level calculation of step S101 in FIG. 8 showing the processing of MODEL1 of the psychoacoustic model by ISO / IEC 11172-3, 512 bands are linearly associated with 32 subbands. Therefore, it is necessary to calculate every 16 samples. In order to cope with this, the valid band variable is further corrected to be a multiple of 16. For example, correction is performed according to the following equation.
[0052]
[Formula 6]
valid_band = valid_band_old− (valid_band_old% 16) +16 (6)
Here, “valid_band_old” represents a valid band variable before correction to a multiple of 16, and “valid_band_old% 16” represents a remainder when “valid_band_old” is divided by 16.
[0053]
As described above, MODE1 of the psychoacoustic model according to ISO / IEC 11172-3 as shown in FIG. 8 is performed using the corrected valid band variable and the critical band variable.
[0054]
Hereinafter, the flow of the above processing is summarized in a flowchart.
FIG. 5 is a flowchart for explaining the processing flow of the audio signal compression method according to the embodiment of the present invention.
[0055]
S1: Valid band variable estimation
Based on the number of subbands used determined in accordance with the bit rate and sampling frequency of the audio signal, a valid band effective for use in psychoacoustic analysis processing is estimated by linear approximation.
[0056]
S2: Critical band variable determination
In accordance with the estimated valid band variable, the psychoacoustic analysis unit 105a determines a critical band variable to be used at the time of noise selection in step S103 of FIG.
[0057]
S3: Valid band variable correction
The valid band variable is corrected according to the linear or nonlinear subband used in the psychoacoustic analysis processing. Specifically, the following three steps are performed. That is: Correction is made to match the determined critical band variable band. Correction is made in accordance with a frequency band which is a sub-band obtained by dividing the 2.512 band into 133 non-linearly. 3. Correct to be a multiple of 16.
[0058]
S4: Psychological auditory analysis processing
Using the critical band variable determined in the process of step S2 and the valid band variable corrected in the process of step S3, the psychoacoustic model MODEL1 process according to ISO / IEC 11172-3 shown in FIG. 8 is performed.
[0059]
The following effects can be expected by performing the psychoacoustic analysis process using the valid band variable calculated as described above and the critical band variable.
First, when obtaining the power spectrum after the Fourier transform process in step S100 of FIG. 8, the sum of squares is conventionally calculated over all 512 bands, but the valid band variable determined by the process of the embodiment of the present invention. By using, it is not necessary to perform the calculation for all 512 bands, and it is sufficient to calculate up to the valid band variable.
[0060]
Also in the sound pressure level calculation in step S101, it is not necessary to calculate the sound pressure for all 512 bands. Since the valid band variable is corrected to be a multiple of 16, the calculation frequency only needs to be repeated 16 times the sum of the valid band variables / 16 times.
[0061]
In the sound selection in step S102, a valid band variable is introduced to search for a variation point from the 512 band, that is, 512 comparisons between three adjacent components are performed in the region up to the valid band variable. In other words, the process of searching for the adjacent three components of the valid band variable times is sufficient.
[0062]
In addition, in the noise selection in step S103, by introducing the critical band variable determined in the process of step S2, for example, all the 25 non-linear subbands are not subjected to product-sum processing, but to the critical band variable. It is only necessary to perform product-sum processing for each non-linear subband.
[0063]
As described above, the amount of calculation in the psychoacoustic analysis unit 105a can be greatly reduced, and the processing can be speeded up.
Next, a specific hardware configuration example to which the audio signal compression apparatus according to the embodiment of the present invention is applied will be described.
[0064]
FIG. 6 is a schematic configuration diagram of an audio signal recording apparatus for recording an audio signal.
As shown in the figure, the audio signal recording apparatus 20 includes an A / D converter 21 that converts an input analog audio signal into a digital signal, and an MP2 encoder 22 that compresses and encodes the audio signal in the MP2 format. The audio signal compression apparatus 10 according to the embodiment of the present invention can be realized by the MP2 encoder 22 shown here.
[0065]
The audio signal recording device 20 as shown in FIG. 6 is mounted on, for example, a digital video camera.
The operation of the audio signal recording device 20 will be briefly described.
[0066]
When an analog audio signal is input, the A / D converter 21 converts the audio signal into a digital audio signal. After the conversion, the audio signal is input to the MP2 encoder 22. In the MP2 encoder 22, the audio signal is compressed into the MP2 format by the processing in each unit as shown in FIG. 7 and the processing in the psychoacoustic analysis unit 105a in which the calculation amount is reduced by each function shown in FIG. And recorded on the recording medium 30.
[0067]
Examples of the recording medium 30 include a magnetic recording device, an optical disc, a magneto-optical recording medium, and a semiconductor memory. Examples of the magnetic recording device include a hard disk device (HDD), a flexible disk (FD), and a magnetic tape. Examples of the optical disk include a DVD (Digital Versatile Disc), a CD-R (Recordable) / RW (ReWriteable), and the like. Magneto-optical recording media include MO (Magneto-Optical disc).
[0068]
As described above, according to the present invention, it is possible to reduce the amount of computation of psychoacoustic analysis processing, and it is possible to greatly reduce the processing time. Further, the processing time is shortened in conformity with ISO / IEC 11172-3. By shortening the processing time, the audio signal recording apparatus 20 as shown in FIG. 6 can be driven at a lower low frequency, and the power consumption and power supply apparatus of the audio signal recording apparatus 20 can be reduced. The audio signal recording apparatus 20 itself can be reduced in size.
[0069]
Although the present invention is characterized by conforming to ISO / IEC 11172-3, it is not limited to the above standards.
[0070]
【The invention's effect】
As described above, in the present invention, a band (valid band variable) that is effective for use in psychoacoustic analysis processing based on the number of subband usages determined according to the bit rate and sampling frequency of the audio signal. In the non-linear subband used for noise selection according to the estimated and estimated valid band, the critical number of subbands (critical band variable) that is sufficient for use is determined, and after validating the valid band variable, the critical band is corrected. Since the psychoacoustic analysis process is performed using the band variable and the valid band variable, the band used for the calculation is limited, and the calculation amount can be reduced. Thereby, it is possible to speed up the psychoacoustic analysis processing, which conventionally requires a large amount of calculation and takes time to process.
[Brief description of the drawings]
FIG. 1 is a functional block diagram showing a main part of an audio signal compression apparatus according to an embodiment of the present invention.
FIG. 2 is a diagram illustrating a method of estimating a valid band variable.
FIG. 3 is a diagram of a critical band variable in the case of a sampling frequency of 32 kHz and a bandwidth corresponding to the value of the critical band variable.
FIG. 4 is a diagram showing a relationship between a frequency band and a bandwidth when a sampling frequency is 32 kHz.
FIG. 5 is a flowchart illustrating a processing flow of the audio signal compression method according to the embodiment of the present invention.
FIG. 6 is a schematic configuration diagram of an audio signal recording apparatus for recording an audio signal.
FIG. 7 is a block diagram showing a configuration of an audio signal compression circuit of MPEG 1 Audio (layer 1 and layer 2) defined by ISO / IEC 11172-3.
FIG. 8 is a flowchart illustrating an example of processing in a psychoacoustic analysis unit.
[Explanation of symbols]
DESCRIPTION OF SYMBOLS 10 ... Voice signal compression apparatus, 11 ... Effective band estimation part, 12 ... Critical subband determination part, 13 ... Effective band correction part, 105a ... Psychological auditory analysis part

Claims (6)

In an audio signal compression method for compressing an audio signal by psycho-acoustic analysis processing using subband coding and human auditory characteristics,
Based on the number of subbands used determined according to the bit rate and sampling frequency of the audio signal, the effective band to be used in the psychoacoustic analysis processing is estimated by linear approximation,
In accordance with the estimated effective band, in the non-linear subband used for noise selection in the psychoacoustic analysis processing, the number of critical subbands sufficient for use is determined,
Correct the effective band according to the linear or non-linear subband used in the psychoacoustic analysis process,
The psychoacoustic analysis process is performed using the determined critical subband number and the corrected effective band.
An audio signal compression method. 2. The audio signal compression method according to claim 1, wherein the correction is performed so that the effective band is matched with a band in the determined critical number of subbands. 3. The audio signal compression method according to claim 2, wherein the correction is further performed according to a subband obtained by nonlinearly dividing the 512 band into 133 pieces. 4. The audio signal compression method according to claim 3, wherein the correction further corrects the effective band to a multiple of 16. The audio signal compression method according to claim 1, wherein the psychoacoustic analysis processing is based on ISO / IEC 11172-3. In an audio signal compression apparatus that compresses an audio signal by psycho-acoustic analysis processing using subband coding and human auditory characteristics,
Based on the number of subbands used determined according to the bit rate and sampling frequency of the audio signal, an effective band estimation unit that estimates a band effective for use in the psychoacoustic analysis process by linear approximation;
In accordance with the estimated effective band, a critical subband number determination unit that determines the number of critical subbands that are sufficient for use in a non-linear subband used in noise selection in the psychoacoustic analysis process;
An effective band correction unit that corrects the effective band according to a linear or non-linear subband used in the psychoacoustic analysis process;
An audio signal compression apparatus comprising:

Images