JPH0472908A - Quantization error reduction device for audio signal - Google Patents

Abstract

PURPOSE:To reduce noise in an audible sense with simple constitution by setting a filter characteristic of a noise filter based on information obtained by varying a synthesis ratio of information of an equi-loudness curve and frequency analysis information of an input audio signal in response to a level of the input audio signal. CONSTITUTION:A filter characteristic of a noise filter 13 is set based on information obtained by varying a synthesis ratio of information of an equi-loudness curve and frequency analysis information of an input audio signal in response to a level of the input audio signal. That is, the filter characteristic of the noise filter 13 for noise shaping is adaptively changed in matching with the level of an input audio signal and its frequency characteristic. A quantization error is fed back to an input of a quantizer 11 via the noise filter 13. Thus, it is possible to reduce noise in an audible sense and a reproduction sound with a higher dynamic range in an audible sense is obtained.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to an audio signal quantization error reduction device that reduces quantization errors that occur during quantization of audio signals.

[Summary of the invention]

The present invention provides an audio signal quantization error reduction device in which quantization errors are fed back to the input side of a quantizer via a noise filter. By setting the synthesis ratio with the frequency analysis information of the signal based on the information obtained by varying the level of the input audio signal, it is possible to reduce auditory noise. The present invention provides a conversion error reduction device.

[Prior Art] Currently, digital audio devices that handle digital audio signals include, for example, so-called compact disc (CD) playback machines, so-called digital audio tape recorders (DAT), and the like. Various unified standards are defined for these digital audio devices, and for example, the bit length of digital audio signals handled by these devices is defined as 16 bits according to the unified standard. In addition, digital audio signals in these digital audio devices include:
A digital audio signal obtained by encoding an analog audio signal (voice waveform signal) using simple linear quantization such as so-called PCM (linear pulse encoding) is used.

[Problem to be solved by the invention C] In recent years, digital audio devices such as those described above have been able to produce reproduced sound that is audibly higher in quality than the reproduced sound that is actually obtained based on the above-mentioned unified standard. It is hoped that

In order to obtain better audible sound, for example, the digital audio signals handled by these digital audio devices should be treated as signals in which the noise components contained in the digital audio signals themselves have been reduced. It is possible to leave it there. The reproduced sound obtained from the digital audio signal whose noise components have been reduced in this way has less noise. As the noise component reduction process of the digital audio signal, for example, when the audio signal is quantized, there is a process in which quantization error is reduced by so-called error feedback. In other words, in the quantization error reduction process using error feedback, the quantization error (quantization noise or quantization distortion) generated by the quantizer is filtered through a noise filter when requantizing the digital audio signal. Noise shaving processing is known that reduces quantization errors by so-called error feedback such as feedback to the input side of the quantizer.

Here, the quantization distortion in linear quantization such as the above-mentioned PCM encoding has flat frequency characteristics over the entire frequency band of the audio signal. However, since the human ear has a difference in hearing sensitivity depending on the frequency of sound, it is difficult to say that the quantization error reduction process using the error feedback described above is necessarily effective in terms of human hearing.

Therefore, the present invention has been proposed in view of the actual situation as described above, and is an audio signal quantization error reduction device that can effectively reduce quantization errors (quantization noise) in terms of auditory sense. The purpose is to provide the following.

[Means for Solving the Problems] The audio signal quantization error reduction device of the present invention has been proposed in order to achieve the above-mentioned object, and it reduces the quantization errors generated in the quantizer by using a noise filter. The quantizer is fed back to the input side of the quantizer via the
Level detecting means for detecting the level of an input audio signal; Frequency analyzing means for frequency analyzing the input audio signal for each critical band; Equal loudness curve generating means for generating information based on an equal loudness curve according to auditory characteristics. , an allowable noise that varies the synthesis ratio of the output of the frequency analysis means and the output of the equal loudness curve generation means according to the output of the level detection means, and calculates an allowable noise spectrum based on the obtained synthesis information. The filter characteristic of the noise filter is set based on the output information of the allowable noise spectrum calculation means.

[Operation] According to the present invention, the equal loudness curve corresponds to the human auditory characteristics, and the synthesis ratio of the equal loudness curve information and the frequency characteristic information of the input audio signal is determined according to the level of the input audio signal. For example, when the level is high, the masking effect can be used by increasing the ratio of frequency characteristic information, and when the level is low, for example, by increasing the ratio of the equal loudness curve, this equal loudness can be used. You will be able to use curves effectively.

〔Example〕

Embodiments to which the present invention is applied will be described below with reference to the drawings.

FIG. 1 shows a block diagram of a schematic configuration of an audio signal quantization error reduction device according to this embodiment.

The device of this embodiment shown in FIG. a level detector 16 that detects the level of the input audio signal, a frequency analysis circuit 17 that analyzes the frequency of the input audio signal for each critical band, and a so-called equal loudness curve RC that is based on the so-called equal loudness curve RC as shown in FIG. The equal loudness curve generation circuit 15 generates information and the frequency analysis circuit 1 according to the output of the level detector 16.
7 and the output of the equal loudness curve generation circuit 15, and an allowable noise spectrum calculation circuit 18 that calculates an allowable noise spectrum based on the obtained synthesis information, the noise filter 13 The filter characteristics are set based on the output information of the allowable noise spectrum calculation circuit 18.

That is, the apparatus of this embodiment includes an adder 12 that obtains a quantization error generated during quantization in the quantizer 11 by subtracting the input to the quantizer 11 from the output of the quantizer 11. , the noise filter 13 that filters and outputs the output of the adder 12, and the adder 10 that adds the output of the noise filter 13 to the input side of the quantizer 11, forming a so-called error feedback circuit. are doing. Here, the filter characteristics of the noise filter 13 are determined by calculating filter coefficients based on the information of the allowable noise spectrum described later in the allowable noise spectrum calculating circuit 18 using the filter coefficient calculating circuit 14, Coefficient information is sent to the above noise filter 1.
It has been decided that the data should be sent to 3. Therefore, the error feedback circuit performs quantization error reduction processing (so-called noise shaving processing) based on the allowable noise spectrum, which will be described later, and then outputs the signal from the output terminal 2.

By the way, when performing quantization error reduction processing (noise shaving processing) on the audio signal using the error feedback circuit, the perceptual dynamic range can be increased by performing processing that takes into account so-called masking of the input signal spectrum. I can do it. Noise shaving that takes this masking into consideration is, for example, noise shaving according to the spectrum of an input audio signal whose signal spectrum pattern is fixed to some extent, that is, noise shaving that takes into account the so-called masking that will be described later of the input audio signal spectrum. One can mention noise shaving using an allowed noise spectrum.

Alternatively, when the spectrum of an input audio signal changes, there is noise shaving using an adaptive noise spectrum that is obtained by taking masking of the spectrum into consideration. The masking described above refers to a phenomenon in which one signal masks another signal and becomes inaudible due to the characteristics of human hearing.This masking effect includes a masking effect on signals on the time axis and a masking effect on signals on the frequency axis. There is a masking effect (or simultaneous masking, temporal masking) on the above signal.

Due to this masking effect, even if there is noise in the masked area, this noise becomes inaudible.

For example, as shown in Fig. 2, when the frequency response of a signal S at a certain frequency is OdB, the simultaneous masking effect described above is caused by the signal S causing a masking effect below the curve M (approximately -25 dB or below). It comes to work.

In addition, if the input signal is divided into critical bands using human auditory characteristics, and noise shaving is performed for each band using the allowable noise spectrum that takes into account the masking described above, it is possible to achieve even more audible effects. You can perform noise shaving. By doing this, it becomes possible to increase the perceptual dynamic range of reproduced sound.

For this reason, the frequency analysis circuit 17 divides the input audio signal into so-called critical bands using human auditory characteristics, and performs frequency analysis for each critical band. At this time, division using the above-mentioned critical bandwidth may be performed, for example, by converting the input audio signal into components (FFT coefficients) on the frequency axis using Fast Fourier Transform (FFT), and then converting the input audio signal into components (FFT coefficients) of the FFT coefficients (hereinafter referred to as .
For example, the hands can be grouped (divided) into groups Gn of 25 hands (hereinafter, n indicates the number of each hand, n = 0 to 24). In addition, frequency analysis for each of these critical bands can be obtained, for example, by calculating the sum of the amplitude terms Am for each band (peak or average or energy sum of the amplitude terms Am) using equation (1). The so-called Herxbek I Le (spectrum of summation) Bn
It is possible to conduct an analysis that requires the following.

B n = 1010g1o Cn (I'n)2[d
B] (],) However, n=o~24, and Cn
is the number of elements in the n-th hand, that is, the amplitude term (number of points), and Pn is the peak value of each band. - The Hark spectrum Bn of each of the above bands is as shown in FIG. 3, for example. However, in the example of FIG. 3, in order to simplify the illustration, the total number of bands in the critical band is set to 12, for example.
It is expressed as a hand (B, ~B1□). The frequency analysis circuit 17 performs division by the critical bandwidth and frequency analysis for each band as described above, and sends the output information to the allowable noise spectrum calculation circuit 18.

Further, the equal loudness curve generating circuit 15 generates and outputs information on the equal loudness curve RC. Here, the acid loudness curve RC is a curve that corresponds to the human auditory characteristics, and is a curve that connects the sound pressure of a sound at each frequency that can be heard as loud as a pure tone of 1 kHz. This is also called the iso-sensitivity curve. In this equal loudness curve RC, as shown in Figure 4, human hearing is sensitive around 4 kHz, so even if the sound pressure is 8 dB to 10 dB lower than 1 kHz, it can be heard as approximately the same loudness as 1 kHz, and vice versa. , for example 10
This shows that it is about 20 dB harder to hear at kHz than around 4 kHz. Therefore, by performing noise shaving on the audio signal by the error feedback circuit based on this equal loudness curve RC, it becomes possible to increase the perceptual dynamic range.

In other words, by performing noise shaving using the allowable noise spectrum obtained by taking this equal loudness curve RC into consideration, more audibly effective noise shaving can be performed, and the perceptually dynamic range of reproduced sound can be increased. It becomes possible. Information on this equal loudness curve RC (or its approximate curve) is transmitted to the equal loudness curve generation circuit 1.
5 and is sent to the allowable noise spectrum calculation circuit 18.

Therefore, the allowable noise spectrum calculation circuit 18 calculates the allowable noise spectrum based on the output information of the equal loudness curve generation circuit 15 and the output information of the frequency analysis circuit 17. At this time, for example, from the Hark spectrum Bn for each critical band in the frequency analysis circuit 17, convolution (convolving a predetermined weighting function) is performed using equation (2), taking into account the influence between bands. By doing so, the convolved park spectrum Sn for each band is calculated.

S n = H n * B n
(2) However, Hn is the convolution coefficient.

Through this convolution, the sum of the parts indicated by the dotted line in FIG. 3 is calculated. Furthermore, the convolved Hark spectrum Sn and the required S/N
Using the value 0n (n=o ~ 24), the (3)
, A convolved masking threshold Tn is calculated using equation (4).

On = N-K
(3) Tn=Sn-On
(4) For example, it is possible to set N=38 and K=1, and there is little deterioration in sound quality in this case. That is, as shown in FIG. 5, areas below the level indicated by each level of the convolved masking threshold Tn are masked. Then, the convolved masking threshold Tn is set to the (5th)
By performing deconvolution using the formula, an allowable noise level (allowable noise spectrum) TFn is calculated. Actually, for example, the DC (direct current) gain Dn of the convolution by the coefficient Hn is subtracted.

TFn=Tn-Dn (5) The allowable noise spectrum calculation circuit 18 synthesizes the output information of the frequency analysis circuit 17 obtained as described above and the output information from the equal loudness curve generation circuit 15 described above. The above-mentioned allowable noise spectrum is determined based on the synthesis information obtained.

Here, the allowable noise level in the allowable noise spectrum based on the above-mentioned equal loudness curve RC may be a level below the allowable noise level at which the masking effect acts depending on the level of the input audio signal. That is,
For example, if the level of the input audio signal is high,
Depending on the allowable noise level at which the masking effect of the input audio signal acts, the level of the allowable noise spectrum based on the equal loudness curve may also be masked at the same time.

For this reason, in the embodiment of the present invention, the level of the input audio signal is detected by the level detector 16,
Based on this level detection output, the synthesis ratio of the output information of the equal loudness curve generation circuit 15 and the output information of the frequency analysis circuit 17 is changed. here,
The synthesis of the output information of the equal loudness curve generation circuit 15 and the frequency analysis circuit 17 is performed for each frequency band, for example. In this case, the level detection by the level detector 16 will also be performed for each band, and therefore the synthesis ratio will be changed for each band based on the level detection output for each band. . That is, the synthesis information for calculating the allowable noise spectrum in the allowable noise spectrum calculation circuit 18 is such that, for example, when the level of the low range of the input audio signal is high and the masking effect in the low range is large, the low range is high. The synthesis information is obtained at a synthesis ratio that yields an allowable noise spectrum with low high frequencies. Also, for example, conversely, if the level of the high range is high (the masking effect in the high range is large),
On the other hand, synthesis information is created at a synthesis ratio that yields an acceptable noise spectrum in which the high range is high and the low range is low.

Information on the allowable noise spectrum obtained in this way is sent to the filter coefficient calculation circuit 14, and the filter coefficient calculation circuit 14 outputs a filter coefficient according to the allowable noise spectrum, and the noise filter 1
Sent to 3.

From the above, the filter characteristics of the noise filter 13 are determined according to the filter coefficients based on the allowable noise spectrum obtained by varying the synthesis ratio for each band according to the level of the input audio signal. Become. Here, it is assumed that, for example, when the level of the human-powered audio signal is flat, the filter characteristics of the noise filter 13 are as shown by the curve MR in FIGS. 6 to 9.

At this time, if the input audio signal becomes, for example, a signal SI with a slightly high level in the low range as shown in FIG. 6, the synthesis ratio is changed as described above, and the filter The characteristics can be changed to those shown by curve MR in FIG. 6, in which the low range of the curve MR is slightly raised and the high range is slightly lowered. Also, for example,
When the input audio signal is a signal S2 with a large level in the low range as shown in FIG. 7, the filter characteristics of the noise filter 13 are such that the low range of the curve MR is greatly increased and the high range is greatly decreased. The curve MR in Fig. 7,
It can be changed to the characteristics shown in .

Conversely, if the input audio signal is a signal S3 with a slightly high level in the high range as shown in FIG. 8, the filter characteristics will be such that the low range of the curve MR is slightly lowered and the high range If the characteristic is changed to a slightly raised curve MR in FIG. 8, and for example, if the signal S4 has a high level in the high range as shown in FIG. The characteristics are changed to those shown by curve MR4 in FIG. 9, in which the low range of MR is greatly lowered and the high range is greatly raised. As shown in FIGS. 6 to 9 above, by changing the filter characteristics, it becomes possible to perform noise shaving that is more responsive to the human auditory characteristics.

That is, in other words, in the present embodiment, when the level of the input audio signal is low, noise shaving is performed by changing the filter characteristic of the noise filter 13 to a characteristic similar to the equal loudness curve RC. . Also, as the signal level increases,
In order to make quantization noise less noticeable depending on the level of the input audio signal, the characteristics of the noise filter 13 are made flat in accordance with the signal level of the input audio signal. Furthermore, when the signal level is low, the noise filter 13 brings the characteristics like the equal loudness curve RC closer to flat according to the signal level, and changes it into noise shaving characteristics (masking effect, etc.) according to the signal characteristics. Make it. That is, the characteristics of the noise filter 13 described in "2" are such that when the signal level is low, the filter characteristics are like the equal loudness curve RC, and when the signal level is high, the filter characteristics are such that the masking effect is taken into consideration.

In addition, in the curve MR shown in FIGS. 6 to 9 above, which shows the filter characteristics when the level of the input audio signal is flat, considering the equal loudness curve RC shown in FIG. It is also possible to consider (
(In other words, it is possible to increase the allowable noise), but since the change in the equal loudness curve RC below 4 kHz is steep despite the fact that the bandwidth is not wide, the equal loudness curve RC below 4 kHz is If a combined noise filter 13 is created, the dimension of the filter becomes high. If the dimension of the filter is increased in this way, the configuration becomes complicated and large-scale.

However, at this time, an effect commensurate with the scale of the filter cannot be obtained, so in this embodiment, as described above, the 4 kHz
The following filter characteristics are assumed to be flat. In addition, in the noise filter 13 of this embodiment, the masking effect is effective in the middle and low ranges, which are usually used frequently in audio sounds, as described above. The curve MR in Fig. 9 is not lowered as much as the equal loudness curve RC in Fig. 4 (curve M
R is a curve that is gentler than the equal loudness curve RC). That is, in order to do this, the filter coefficients are set in consideration of the masking effect as described above.

FIG. 10 shows a specific example of a system configuration in which the apparatus of this embodiment is used in an encoder/decoder system for, for example, a so-called compact disc (CD). This first
In FIG. 0, an analog audio signal is supplied to the input terminal 31. This analog audio signal is A/D
After being converted into a 20-bit digital signal by a converter 32, it is sent to a 20-bit compatible encoder 33 incorporating the quantization error reduction device of this embodiment. This encoder 33
The data is subjected to quantization error reduction processing and encoded into 16-bit data, and then recorded on the CD. The data recorded on this CD is converted into an audio signal by a playback circuit 34 and a D/A converter 35 of an existing CD player, and the audio signal is output from an output terminal 36 and played back.

That is, since the data recorded on the CD has had the quantization error reduced by the quantization error reduction device of this embodiment, the sound obtained by playing this CD is as follows.
The audible dynamic range is high.

Further, FIG. 11 shows a specific example of a system configuration using a medium that records data in, for example, 10 bits, unlike the above-mentioned CD. In this case, the analog signal manually supplied to the input terminal 41 is converted into, for example, 16-bit digital data by the A/D converter 42, and is then converted into a 10-bit compatible encoder 43 incorporating the quantization error reduction device of this embodiment. sent to.

The encoder 43 performs quantization error reduction processing and encodes the data into 10-bit data.
recorded on the above media. The data recorded on this medium is converted into an analog signal by a reproducing circuit 44 and a D/A converter 45 of an existing reproducing device, and is outputted from an output terminal 46. In this case as well, the audible dynamic range of the obtained reproduced signal is increased.

FIG. 12 shows a specific example of using the device of this embodiment in a D/A conversion system that performs oversampling. In this case, the analog signal supplied manually to the input terminal 51 is
The data is converted into, for example, 20-bit digital data by an A/D converter 52 that performs oversampling, and sent to the quantization error reduction device 53 of this embodiment via a transmission line. The quantization error reduction device 53 performs quantization error reduction processing, and D
The signal is converted into an analog signal via the /A converter 54 and output from the output terminal 55. As a result, oversampling can be performed and the resolution of D/A conversion can be lowered, making it easier to create the D/A converter 54 in a manner that improves linearity accordingly.

As described above, according to this embodiment, in the audio signal quantization error reduction device in which the quantization error is fed back to the input side of the quantizer 11 via the noise filter 13, the filter of the noise filter 11 is By setting the characteristics based on the information obtained by varying the synthesis ratio of the information on the equal loudness curve in FIG. 4 and the frequency analysis information of the input audio signal according to the level of the input audio signal, that is, Noise filter 13 for noise shaving according to the input audio signal level and frequency characteristics
By adaptively changing the filter characteristics of, for example, the curves MR and M R+ =M R in FIGS.
A noise filter 13 having filter characteristics as shown in a is obtained, and the quantization error is fed back to the input side of the quantizer 11 through this noise filter 13.
It has become possible to reduce auditory noise. Therefore, the audio signal quantization error reduction device of this embodiment can be used, for example, in digital audio equipment with unified standards (
For example, the so-called compact disc.

If applied to a digital audio tape recorder, etc., it becomes possible to obtain reproduced sound with a higher dynamic range perceptually than the dynamic range actually obtained from the unified standard. For example, the unified standard (the above CD
, in the case of DAT, the 16-bit slot word length standard) (no changes are made to the playback side, maintaining compatibility), and the perceptual dynamic range of the reproduced sound of this audio signal is be able to raise it.

In this method, the filter characteristics of the noise filter are set based on the information obtained by varying the synthesis ratio of equal loudness curve information and frequency analysis information of the input audio signal according to the level of the input audio signal. This has made it possible to reduce auditory noise and increase the auditory dynamic range.

Therefore, if the audio signal quantization error reduction device of the present invention is applied to, for example, digital audio equipment that has a unified standard, it will be possible to reproduce a dynamic range that is perceptually higher than the dynamic range actually obtained from the unified standard. You will be able to get sound. For example, it becomes possible to increase the perceptual dynamic range of the reproduced sound of this audio signal while maintaining the unified standard (without making any changes to the reproduction side and maintaining compatibility).

〔Effect of the invention] Audio signal quantization error reduction device of the present invention

[Brief explanation of the drawing]

FIG. 1 is a block diagram of a schematic configuration of an audio signal quantization error reduction device according to an embodiment of the present invention, FIG. 2 is a diagram for explaining the masking effect, and FIG. 3 is a diagram showing a park spectrum. is a characteristic diagram showing the equal loudness curve, Figure 5 is a diagram showing the masking threshold, Figures 6 to 9
FIG. 10 is a characteristic diagram for explaining the filter characteristics, FIG.
FIG. 1 is a block diagram showing a specific example in which the device of this embodiment is applied to a 10-bit system, and FIG. 12 is a block diagram showing a specific example in which the device of this embodiment is applied to a D/A conversion system that performs oversampling. . 16... Level detector 17... Frequency analysis circuit

Claims (1)

[Claims] An audio signal quantization error reduction device in which a quantization error generated in a quantizer is fed back to the input side of the quantizer via a noise filter, comprising: A level detection means for detecting the frequency, a frequency analysis means for frequency-analyzing the input audio signal for each critical band, an equal-loudness curve generation means for generating information based on an equal-loudness curve according to auditory characteristics, and a level detection means for the above-mentioned level detection means. and an allowable noise spectrum calculation means for varying the synthesis ratio of the output of the frequency analysis means and the output of the equal loudness curve generation means according to the output, and calculating an allowable noise spectrum based on the obtained synthesis information. A quantization error reduction device for an audio signal, characterized in that filter characteristics of the noise filter are set based on output information of the allowable noise spectrum calculation means.