JPH0746137A - Highly efficient sound encoder - Google Patents

Abstract

PURPOSE:To provide the highly efficient sound encoder which better satisfies the auditory mentality and improves the tone quality. CONSTITUTION:An audio signal has several samples segmented by a blind segmenting part 1, and the audio signal of each sample is divided into plural subbands by an orthogonal transformation part 2. An auditory mentality analysis part 3 analyzes the audio signal with the analysis band width in each frequency area, which is narrower than the subband width and narrower than the critical band width on auditory mentality and is ideally 1/2 to 1/3 of the critical band width, to determine the number of quantization bits in each subband. A quantizing and encoding part 4 quantizes and encodes the audio signal of each divided subband with this number of quantization bits by an orthogonal transformation part 2.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a high-efficiency speech coding apparatus for dividing an audio signal into a plurality of frequency bands (subbands) and quantizing and coding the divided signals for each subband. The present invention relates to a speech efficient coding apparatus that determines the number of quantization bits for each subband based on psychoacoustic analysis.

[0002]

2. Description of the Related Art High-efficiency audio coding in a mini disc (MD), a digital compact cassette (DCC), a karaoke CD, etc. is called music compression because it compresses the data amount of an audio signal. In such an encoding method, an audio signal is divided into a plurality of subbands by a digital filter or orthogonal transformation, and the number of quantization bits for each subband is determined based on psychoacoustic analysis in the frequency domain. In the following description, the term “encode” may be used to mean compression in addition to encoding.

FIGS. 10 (a) to 10 (d) show an example in which the frequency band is divided by orthogonal transformation in such an encoding method. FIG. 10A shows the target 16 of encoding.
This shows that 512 samples of a bit PCM audio signal are cut out, and here it is assumed that the total amount of information enclosed by the rectangle in the figure is 16 bits * 512 = 8192 bits. Of course, the number of samples to be cut out and the number of PCM bits are not limited to this value.

FIG. 10B shows a signal obtained by frequency-converting the signal shown in FIG. 10A by orthogonal transform such as DCT (discrete cosine transform) or FFT (fast Fourier transform). The curve in the figure shows the frequency spectrum. The envelope is shown. Here, assuming that the amount of information is preserved by orthogonal transformation, this total amount of information can also be expressed by the rectangular area in the figure. On the other hand, according to the psychoacoustic model, FIG.
When the signal shown in (b) is present, the signal level masked by the signal and inaudible can be defined as a curve, which is generally called a masking effect (described in detail later).

A masking curve can be drawn from FIG. 10 (b) as shown in FIG. 10 (c).
Considering requantization of the signal shown in FIG. 10B, if the quantization noise level generated by the requantization is equal to or lower than the level defined by the masking curve, the noise can be heard by the human ear. It can be said that there is no. Therefore, as shown in FIG. 10 (d), the spectrum is divided into sub-bands for each plurality of data, the maximum signal level for each sub-band is set to S, and the allowable noise level from FIG. 10 (c) is N. If requantization is performed with a bit number that satisfies this S / N, the quantization noise at that time is masked and cannot be heard.

The rectangle in FIG. 10 (d) shows the amount of information required at the time of compression and decompression. In particular, the deformed rectangle in the center of the figure shows the main information, and the elongated rectangle at the bottom of the figure shows the auxiliary information.
The auxiliary information is information indicating the maximum value (scale value) and the number of quantization bits of each subband necessary for decoding. Therefore, the total information amount shown in FIG. 10D is the sum of the main information amount and the auxiliary information amount, and FIG.
It can be seen that the total amount of information in FIG. By repeating the above processing for every predetermined section (512 sample sections in this example), encoding can be performed with almost no deterioration in sound quality.

Next, a method of calculating a masking curve based on auditory psychology will be described. The masking effect means that when a certain frequency spectrum is present, sounds around a certain level below a certain level cannot be detected, and masking curves for various frequency spectra are shown in FIG. As shown in FIG. 11, the slope of the curve is steeper in the lower range and slower in the higher range. Further, it is known that when the horizontal axis (frequency) in FIG. 11 is converted into a scale proportional to the auditory critical bandwidth, these curves have almost the same shape and slope as shown in FIG. . Also,
This critical bandwidth is DC to 20 kHz, as shown in FIG.
z can be expressed by being divided into 25 bands, and auditory characteristics such as masking often behave in proportion to this critical bandwidth.

The masking curve when a general signal as shown in FIG. 10 (b) is present is represented by the sum (overlap) of the masking curves for each frequency spectrum as shown in FIG. 11 or 12. Fig. 10 (c)
Although it can be represented by a curve as shown in FIG. 10, it is difficult in actual calculation to calculate the masking curve as a smooth curve as shown in FIG. Therefore, as an approximation, the spectrum is replaced with the power for each analysis band, and the masking curve is evaluated as a polygonal line waveform for each analysis band.

Next, referring to FIG. 10 (d), a conventional example of deriving the noise level N from the calculation of the masking curve, where the minimum value in each subband section of the masking curve is the noise level N allowed in that subband, will be described. To do. (1) First, the frequency spectrum N obtained by orthogonal transformation
Band total power P for each of m analysis bands i from the book
[I] is calculated.

[0010]

[Equation 1]

Next, the masking level M in each analysis band i is calculated by performing the convolution operation of the masking reference curve B and the band total power P [i] according to the analysis band as in the following equation (Equation 2). [I] is calculated. Here, the masking reference curve B is shown in FIG. 14 when a curve having a different shape for each analysis band i is required.
It can be represented by B [i] [k] (k is an integer) as shown in (b), and in the case of a constant shape irrespective of the analysis band i, B as shown in FIG. [K] (k is an integer)
Can be expressed as

[0012]

[Equation 2]

(3) When the analysis band i is different from the subband s, the minimum masking level M [i] in the section of the subband s is set as the allowable noise level N [s] of the subband s (subband Number n).

[0014]

## EQU00003 ## N [s] = min [M [i]] where i indicates any section included in subband s s = 0 ... n-1

When the analysis band i and the subband s are the same,

[0016]

## EQU3 ## N [s] = M [i] where s = 0 ... n-1

[0017]

[Problems to be Solved by the Invention] The above processes (1) to
Due to (3), the allowable noise level N of each subband s
[S] can be obtained. Here, the relationship between the analysis band i and the subband s will be described. Generally, there are many cases of the following three cases. (A) When the analysis band i is equal to the subband s and the bandwidth is equal to (in proportion to) the critical band (b) When the analysis band i is equal to the subband s and the bandwidth is not proportional to the critical band ( c) When the analysis band i is different from the subband s and the bandwidth is equal to (proportional to) the critical band.

In the case of the above (a), the allowable noise level N [s] and the true allowable noise level Na [obtained from the masking curve, that is, the true masking curve when the bandwidth is calculated sufficiently small. The difference between
5, the true permissible noise level Na [s] of each band needs to be the minimum value of the true masking curve of that band. The curve in FIG. 15 shows a true masking curve, the solid polygonal line shows the allowable noise level N [s], and the dotted polygonal line shows the true allowable noise level Na [s].

Further, FIG. 15A shows the case where the signal spectrum is flat (noise-like), and FIG. 15B shows the signal spectrum being peaky (tone-like) and the center power of the signal is near the center of the band. The case of
FIG. 15C shows the case where the signal spectrum is peaky and the center power of the signal is close to the band boundary.
In the case shown in FIG. 15A, the allowable noise level N
The difference between [s] and the true allowable noise level Na [s] is small, but there is a large error in the cases shown in FIGS. 15 (b) and 15 (c). The reason for this error is that the masking reference curve is generally 10 to 20 per critical band as shown in FIG.
This is because it has a slope of dB. Therefore, encoding with such an allowable noise level N [s] may deteriorate the sound quality.

The above case (b) corresponds to the case where the subband width cannot be made proportional to the critical band due to the convenience of the system and the analysis band is made common with the subband. In this case, the bandwidth is wider than the critical band in a certain band,
In a certain band, it is almost narrower than the critical band. Further, in the region where the bandwidth is equal to or wider than the critical band, the problem described in (a) above is aggravated.

In the case of the above (c), the problem described in the above (a) applies as it is. Further, when a situation occurs in which the subband s is narrower than the analysis band i, the masking level M [i] = N [s] obtained from the analysis band i as shown in FIG. Resolution is poor, the true allowable noise level Na
The difference from [s] becomes large. Here, if the analysis bandwidth is set to be equal to or smaller than the sub-bandwidth, the obtained allowable noise level N [s] is a more true allowable noise level Na [s].
Therefore, the advantage cannot be utilized even though it has the ability to mitigate the problem (a).

Next, other problems will be described. In the series of processes for obtaining the allowable noise level N [s] by the processes (1) to (3) described above, another important role is the shape and inclination of the masking reference curve described in FIG. The relative index k of the analysis band is taken on the abscissa of FIG. 14, but as is clear from FIG. 12, the masking amount-critical band characteristic when the critical band is taken on the abscissa is that the center of masking is Since it is considered to be constant in any critical band, when the power P is represented in the relative critical band on the horizontal axis as shown in FIG. 17, the masking curves of all frequency bands are uniquely defined. Can be expressed. In this sense, the reason why the masking reference curve is represented in FIG. 14 without depending on B [i] and the analysis band i is that the analysis band i is proportional to the critical band, and thus B [i] It is necessary to represent [k] when the analysis band i is not proportional to the critical band.

However, as described above, the masking amount-
With the idea that the critical band characteristic is constant, deterioration may be detected particularly in the low frequency side in the sound quality when actually encoded and decoded. The cause is
It is conceivable that the curve shape of the masking amount-critical band characteristic in the low frequency range is inappropriate. 18 (a) (b)
Shows the true permissible noise level Na [s] of each subband when the masking curve is wide (when the slope is gentle) and when it is narrow (when the slope is steep). This figure assumes the case where the signal is peaky (tone-like), and there is a big difference from the obtained allowable noise level N [s]. Therefore, in the conventional method, the sound quality may be deteriorated because the encoding is performed by an inappropriate masking curve.

In view of the above-mentioned conventional problems, it is an object of the present invention to provide a high-efficiency speech coding apparatus capable of satisfying auditory psychology and improving sound quality.

[0025]

In order to achieve the above object, the present invention is set to a subband width or less in each frequency range and a critical psychoacoustic bandwidth, and at least in a low range. The audio signal is analyzed with a frequency-dependent analysis bandwidth that sets the bandwidth sufficiently narrower than those, and the number of quantization bits of each subband is determined. That is, according to the present invention,
Dividing means for dividing an audio signal into a plurality of frequency band subbands; and quantizing / encoding means for quantizing and encoding the audio signal of each subband divided by the dividing means with a variable number of quantization bits. And an audio signal is analyzed with an analysis bandwidth that is less than or equal to the subband width in each frequency domain and less than or equal to the critical psychoacoustic bandwidth, and the number of quantization bits of each subband of the quantization and encoding means is A high-efficiency speech coding apparatus having a psychoacoustic analysis means for determining is provided.

The present invention also analyzes the audio signal of each sub-band by selecting a plurality of masking amount-critical band characteristics representing a noise masking level in the frequency domain when analyzing psychoacoustics according to the applied critical band. , The number of quantization bits for each subband is determined. That is, according to the present invention, a dividing unit that divides an audio signal into subbands of a plurality of frequency bands, and an audio signal of each subband divided by the dividing unit is quantized and encoded with a variable number of quantization bits. The quantizing / encoding means for encoding and the relative critical band curve slopes are selected to be different in at least some of the respective critical bands for representing the noise masking level in the frequency domain when analyzing auditory psychology. High-efficiency speech coding with audio-acoustic analysis means for analyzing the audio signal of each sub-band by using the masking amount-critical band characteristic and determining the number of quantization bits of each sub-band of the quantization and coding means. A device is provided.

[0027]

In the present invention, the audio signal is analyzed with the analysis bandwidth which is less than the sub-bandwidth in each frequency domain and less than the critical psychoacoustic bandwidth to determine the number of quantization bits of each subband. . Therefore, the relationship between the subband, which is the unit of quantization and coding, and the analysis band, which is the unit of auditory psychological analysis (calculation of masking curve) in the frequency domain, is optimized, so that the auditory psychology is more satisfied and the sound quality is improved. Can be improved.

In the present invention, a plurality of masking amount-critical band characteristics representing a noise masking level in the frequency domain when analyzing psychoacoustic characteristics are selected according to the applied critical band, and the audio signal of each subband is analyzed. , The number of quantization bits for each subband is determined. Therefore, since the optimum masking curve is selected according to the applied critical band, the psychology of hearing can be more satisfied and the sound quality can be improved.

[0029]

Embodiments of the present invention will be described below with reference to the drawings. 1 is a block diagram showing an embodiment of a high-efficiency speech coding apparatus according to the present invention, FIG. 2 is a flowchart for explaining processing of the high-efficiency speech coding apparatus of FIG. 1, and FIG. 3 is a speech diagram of FIG. FIG. 4 is a block diagram showing a modified example of the high-efficiency coding apparatus, FIG. 4 is an explanatory diagram showing the relationship between the sub-bandwidth, the critical bandwidth and the analysis bandwidth, and FIG. 5 is the analysis bandwidth of 1/3 of the critical bandwidth. FIG. 6 is an explanatory diagram showing the allowable noise level N [s] in the case, and FIG. 6 is an explanatory diagram showing the allowable noise level N [s] in the case where the analysis bandwidth and the critical bandwidth are the same.

In the high-efficiency speech coding apparatus shown in FIG. 1, for example, a 16-bit PCM audio signal is cut out by 512 samples by the windowing cutout unit 1, and the audio signal of each sample is output by the orthogonal transformation unit 2 by DCT or FFT. It is orthogonally transformed and divided into a plurality of subbands s (step S1 in FIG. 2). Then, the psychoacoustic analysis unit 3 determines the maximum value (scale value) of each subband s (step S2), and the analysis bandwidth in each frequency region is equal to or less than the subband width, and the psychoacoustic criticality. Below bandwidth, ideally 1/2 to 1 of critical bandwidth
Each sub-band s is auditory-analyzed with a width of / 3 to determine the allowable noise level N [s] of each sub-band s (step S3), and then determine the S / N ratio required for each sub-band s. (Step S4), and then the number of quantization bits is determined from this S / N ratio (step S5).

The quantizing / encoding unit 4 quantizes and encodes the audio signal of each sub-band s divided by the orthogonal transforming unit 2 with this quantized bit number, and the quantizing / encoding unit 4 quantizes The data compressed and encoded and the number of quantization bits determined by the psychoacoustic analysis unit 3 are multiplexed by the multiplex unit 5 and output to MD, DCC or the like (step S6). During decompression, the compressed data is decoded and dequantized based on the number of quantization bits for each subband s.

The high-efficiency speech coding apparatus shown in FIG. 3 is configured so that the input audio signal is divided into subbands s by the digital filter 6. Here, the sub-band division method using the filter bank cannot obtain the low-frequency band resolution required by the present invention. Therefore, as with the case shown in FIG. The signal is divided into a plurality of subbands s by the orthogonal transformation unit 2 and the psychoacoustic analysis unit 3
Is used to determine the number of quantization bits of the quantization / encoding unit 4.

Here, in the above method of determining the number of quantization bits, the analysis band is actually determined in consideration of the constraints of the system and the amount of calculation that can be assigned when calculating the masking curve. FIG. 4 shows the sub-bandwidth divided by the orthogonal transformation unit 2, the critical auditory bandwidth, and the psychoacoustic analysis unit 3.
3 shows the relationship between the analysis bandwidths of. This figure shows an example where the critical bandwidth is narrower than the sub-bandwidth on the low frequency side and wider than the sub-bandwidth on the high frequency side. It is analyzed so that it is below the bandwidth. Further, the actual masking curve is calculated by the above-described methods (1) to (3), and the allowable noise level N [s] for each subband s obtained as a result is the same as the above-mentioned (a) to (c). Set to fix the problem.

FIGS. 5 (a) and 5 (b) show the case where the analysis bandwidth is set to 1/3 of the critical bandwidth, and each corresponds to FIGS. 15 (b) and 15 (c) described in the conventional example. There is. That is, FIG. 5A shows the case where the signal spectrum is peaky (tone-like) and the center power of the signal is near the center of the band, and FIG. 5C shows the case where the signal spectrum is peaky and the center of the signal. It shows the case where the power is close to the band boundary. The solid broken line in the figure indicates the masking level M [i], the alternate long and short dash line indicates the allowable noise level N [s] of each subband s, and the dotted line indicates the true allowable noise level Na [s] of each subband. ing. 5 and 15
As is clear from the comparison of the above, in the present embodiment, the difference between the allowable noise level N [s] and the true allowable noise level Na [s] becomes small, and therefore the problem (b) above can be solved. it can.

FIG. 6 shows the case where the analysis band width is made equal to the sub band width, and corresponds to FIG. 16 in which the analysis band width is wider than the sub band width. As is clear from comparison between FIG. 6 and FIG. 16, in FIG. 6, the allowable noise level N [s] of each subband s is the true allowable noise level Na [s].
Therefore, it is possible to solve the problem (c), that is, the problem of frequency resolution.

Therefore, according to the above embodiment, the relationship between the subband, which is a unit of quantization and coding, and the analysis band, which is a unit of psychoacoustic analysis (calculation of masking curve) in the frequency domain, is optimized. Therefore, the evaluation accuracy of the allowable noise level N [s] can be improved,
It is possible to improve the sound quality when encoded using the psychoacoustic characteristics. Specifically, it is assumed that the improvement range of accuracy is 1 to 3 dB on average, and 6 dB or more when it is large. Further, considering that the S / N ratio of the quantizer is generally 6 dB steps, this improvement is a large value.

A second embodiment of the present invention will be described with reference to FIGS. Figure 7 shows the critical bandwidth and the equivalent noise band Widus.
ctangular noise Bandwidth) (hereinafter, simply abbreviated as ERB), FIG. 8 is an explanatory view showing a masking amount-critical band characteristic, and FIG. 9 is an S / N for each masking curve.
It is explanatory drawing which shows a ratio characteristic. In the second embodiment, the masking amount-critical band characteristic of the masking curve used in the psychoacoustic analysis is not fixed, but is appropriately changed according to each critical band.

Here, according to the recent research on auditory psychology,
It is said that ERB is anatomically more suitable as an auditory analysis band than the critical band that has been conventionally said. FIG. 7 compares the critical bandwidth and the ERB by taking the frequency (log) on the horizontal axis and the bandwidth on the vertical axis, and in particular, the difference between the critical bandwidth and the ERB is large at a low frequency of 400 Hz or less, 100 Hz.
Below, the critical bandwidth is about 3 to 4 times wider. In other words, when the masking amount-critical band characteristic and the masking amount-ERB characteristic are almost the same, and when both are constant regardless of the frequency domain, the masking amount-E
It can be said that the slope of the RB characteristic is much larger than the slope of the masking amount-critical band characteristic.

Therefore, when considering the critical band on the horizontal axis, the reproduced sound quality is good by making the masking curve variable with respect to the critical band rather than being constant, or by making the masking curve constant with respect to values outside the critical band. It is expected that, for example, three (generally a plurality of) masking amount-critical band characteristics A to C having different slopes are prepared as shown in FIG. 8, and the sound quality is improved by changing them according to the frequency band. I was able to do it.

FIG. 9 shows the S / N ratio when the masking curve is switched to a steep slope at 400 Hz or less in FIG. 8 and when the masking curve is constant, and the S / N ratio greatly rises at 400 Hz or less. . 18 (a) (b)
As described above, when the masking curve becomes narrow, the allowable noise level N [s] decreases in that region, and therefore, when the signal level S is the same, the S / N ratio improves. Since the amount of information assigned to other areas decreases, the S / N ratio slightly deteriorates in those areas, but the sound quality improvement effect in the low frequency range is larger in terms of hearing.

[0041]

As described above, according to the present invention, the sub-bandwidth is set to be less than or equal to the sub-bandwidth in each frequency range, and is set to be less than or equal to the critical bandwidth in psychoacoustics, and the bandwidth is set at least in these low frequencies. Compare the audio signal with a frequency-dependent analysis bandwidth that is set sufficiently narrow, for example, an analysis bandwidth that is subdivided from 1/2 to 1/3 of the critical bandwidth, and determine the number of quantization bits for each subband. Since it is determined, the relationship between the subband, which is the unit of quantization and coding, and the analysis band, which is the unit of auditory psychological analysis (calculation of the masking curve) in the frequency domain, can be optimized. Can be more satisfied and the sound quality can be improved.

According to the present invention, the slope of the relative critical band curve is set to be large at least in the low range of 1 kHz or less for representing the noise masking level in the frequency domain when analyzing the psychology of hearing. Since the audio signal of each subband is analyzed using the masking amount-critical band characteristics selected so that the slopes of the relative critical band curves are different in different critical bands, the number of quantization bits of each subband is determined. , You can select the optimal masking curve according to the applied critical band,
Therefore, the psychology of hearing can be more satisfied and the sound quality can be improved.

[Brief description of drawings]

FIG. 1 is a block diagram showing a first embodiment of a speech efficient coding apparatus according to the present invention.

FIG. 2 is a flowchart for explaining the high-efficiency speech coding processing of FIG.

FIG. 3 is a block diagram showing a modified example of the high-efficiency speech coding apparatus of FIG.

FIG. 4 is an explanatory diagram showing a relationship between a sub-bandwidth, a critical bandwidth and an analysis bandwidth.

FIG. 5 is an explanatory diagram showing an allowable noise level N [s] when the analysis bandwidth is 1/3 of the critical bandwidth.

FIG. 6 is an explanatory diagram showing an allowable noise level N [s] when the analysis bandwidth and the critical bandwidth are the same.

FIG. 7 is an explanatory diagram comparing the critical bandwidth with the ERB in the second embodiment.

FIG. 8 is an explanatory diagram showing a masking amount-critical band characteristic of the second embodiment.

FIG. 9 is an explanatory diagram showing S / N ratio characteristics for each masking curve.

FIG. 10 is an explanatory diagram schematically showing a high-efficiency voice encoding method.

FIG. 11 is an explanatory diagram showing an example of a masking curve in various frequency spectra.

12 is an explanatory diagram showing a masking curve in which the frequency on the horizontal axis of FIG. 11 is replaced with a critical band.

FIG. 13 is an explanatory diagram showing a critical bandwidth of 25 bands.

FIG. 14 is an explanatory diagram showing a masking reference curve.

FIG. 15 is a conventional allowable noise level N [s] and a true allowable noise level N when the analysis band and the subband are equal.
It is explanatory drawing which shows a [s].

FIG. 16 is a conventional allowable noise level N [s] and a true allowable noise level N when the subband is wider than the analysis band.
It is explanatory drawing which shows a [s].

FIG. 17 is an explanatory diagram showing a conventional masking amount-critical band characteristic.

FIG. 18 is an explanatory diagram showing the relationship between the slope of the masking curve shown in FIG. 17 and the allowable noise level N [s].

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Window cut-out section 2 Orthogonal transformation section (dividing means) 3 Perceptual psychological analysis section (perceptual psychological analysis means) 4 Quantization and encoding section (quantization / encoding means) 5 Multiplex section 6 Subband filter section

Claims (2)

[Claims] 1. A dividing unit for dividing an audio signal into subbands of a plurality of frequency bands, and a quantum for quantizing and encoding the audio signal of each subband divided by the dividing unit with a variable number of quantization bits. Encoding / encoding means and a frequency that is less than or equal to the subband width in each frequency region and less than or equal to the critical psychoacoustic bandwidth and that sets the bandwidth sufficiently narrower than those at least in the low frequency range. A high-efficiency speech coding apparatus having a psychoacoustic analysis means for analyzing an audio signal with a dependent analysis bandwidth and determining the number of quantization bits of each subband of the quantization and coding means. 2. A dividing unit for dividing an audio signal into subbands of a plurality of frequency bands, and a quantum for quantizing and encoding the audio signal of each subband divided by the dividing unit with a variable number of quantization bits. Encoding / encoding means, and a masking amount selected so that the slope of the relative critical band curve is different in at least some of the respective critical bands for representing the noise masking level in the frequency domain when analyzing the psychology of hearing. A high-efficiency speech coding apparatus having a psychoacoustic analysis means for analyzing an audio signal of each subband by using a critical band characteristic and determining the number of quantization bits of each subband of the quantization and coding means.