JPH08237130A - Method and device for signal coding and recording medium - Google Patents

Abstract

PURPOSE: To reduce sound quality deterioration in an audible sense even when quantization is applied to a signal where isolated spectral components are in existence apart in the entire frequency band with a few allocated bit numbers. CONSTITUTION: The device is provided with a mapping circuit 102 separating an audio signal into frequency components, a control circuit 113 obtaining a function relating to the magnitude of each frequency component, a high frequency gain attenuation circuit 103 attenuating a gain of frequency components within a frequency band symmetrical to the Nyquist frequency and in existence close to the Nyquist frequency depending on the function, and a quantization circuit 105 quantizing each frequency component. The band width to attenuate the gain is more extended and the gain reduction is increased more as the value of the function is larger. The function indicates the spread of the weight of the magnitude of each frequency component toward higher frequency bands, tonality based on a change in the magnitude of each frequency component and a required bit number when each frequency component is quantized.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a signal coding method and apparatus for coding a digital signal such as an audio signal, and a recording medium for recording the coded signal.

[0002]

2. Description of the Related Art There are various techniques and devices for high-efficiency coding of audio or voice signals. For example, a time domain audio signal is divided into blocks for each unit time, and a signal on the time axis of each block is used. Is orthogonally transformed into a signal on the frequency axis, and the obtained signal in the frequency domain (spectral component)
A so-called transform coding method, which is a blocking frequency band division method for coding, and a time domain audio signal, etc., is divided into a plurality of frequency bands and encoded without being blocked for each unit time as described above. A band division coding (sub-band coding: SBC) method, which is a non-blocking frequency band division method, can be used. Further, a high-efficiency coding method and device combining the above-described band-division coding and transform coding is also considered, and in this case, for example, the time-domain signal is frequency-converted by the band-division coding method. After the band division, the signal for each frequency band is orthogonally transformed into a signal (spectral component) in the frequency domain by the transform coding method, and the spectral component is coded.

Here, as a band division filter used in the above-mentioned band division coding method, for example, QMF is used.
There is a filter such as (Quadrature Mirror filter), which is, for example, the document "Digital coding of speech in subvans"("Digital coding of speech").
in subbands "RECrochiere, Bell Syst.Tech. J.,
Vol.55, No.8 1976). This QMF filter divides a band into two equal bandwidths, and is characterized in that so-called aliasing does not occur when the divided bands are combined later. In addition, the document "Polyphase Quadratic Filters-New Band Division Coding Technique"("Po
lyphase Quadrature filters -A new subband coding t
echnique ", Joseph H. Rothweiler ICASSP 83, BOSTON)
Describes an equal bandwidth filter partitioning technique. This polyphase quadrature filter is characterized in that when a signal is divided into a plurality of bands of equal width, it can be divided at once.

Further, as the orthogonal transform for transforming the above-mentioned time domain signal into a signal on the frequency axis, for example, discrete Fourier transform (DFT) or discrete cosine transform (DC)
T), modified discrete cosine transform (MDCT), and the like. For MDCT, refer to "Subband / Transform Coding Using Filter Bank Design Based on Time Domain Aliasing Cancellation"("Subband / Transfor
m Coding Using FilterBank Designs Based on Time Do
main Aliasing Cancellation, "JPPrincen ABBradl
ey, Univ. of Surrey Royal Melbourne Inst. of Tech.
ICASSP 1987).

If an audio signal or the like is divided into signals for each frequency band by the band dividing filter or the orthogonal transformation as described above, the signals are divided for each frequency band (that is, the signals in the time domain divided into a plurality of bands or In the case of quantizing (orthogonal-transformed spectrum component signal), it is possible to control the band in which the quantization noise is generated, and by utilizing properties such as the masking effect, aurally more preferable and highly efficient quantization can be achieved. It becomes possible to do. Further, for example, before performing the quantization, for each frequency band, if the signal component in the band is normalized by the maximum value of the absolute value of the signal component in that band, for example, the quantization with higher efficiency can be achieved. It is also possible to do.

Further, when a signal divided into frequency bands is quantized, the frequency division width is, for example, a band division width considering human auditory characteristics. That is, an audio signal may be divided into a plurality of bands with a band width that becomes wider in a higher band generally called a critical band. When quantizing the signal for each frequency band at this time, quantization is performed by predetermined bit allocation for each frequency band or adaptive bit allocation (bit allocation) for each frequency band. . For example, MDCT as an example of orthogonal transform
When the MDCT coefficient data obtained by the above is quantized by adaptive bit allocation, the number of bits adaptively allocated for the MDCT coefficient data for each frequency band obtained by the MDCT for each block. Will be used for quantization.

[0007]

By the way, when these techniques are applied to compress an audio signal, for example,
In order to realize efficient compression, it is effective to use a so-called masking effect or minimum audible limit characteristic as the auditory characteristic as described above, but in order to use this auditory characteristic, It is necessary to separate the audio signal for each frequency for analysis, and for that purpose, the above-mentioned orthogonal transform, band division filter, etc. are used. In other words, the spectral components on the frequency axis obtained by these orthogonal transforms and the signal components obtained by dividing into a plurality of bands by the band division filter can be treated as data on the frequency axis with certain accuracy. Therefore, by changing the characteristic on the frequency axis, it is possible to easily realize a configuration capable of performing efficient compression using the auditory characteristic.

However, in order to realize a higher bit compression, if the degree of reduction of the number of allocated bits when quantizing the spectrum component or the signal component is increased, for example, the auditory characteristic such as a masking effect is used. In the case of an audio signal in which isolated spectrum components stand sparsely in the difficult entire band, the quantization noise generated by the quantization with the small number of allocated bits becomes audible,
The deterioration of sound quality will be noticeable.

Therefore, the present invention has been made in view of such a situation, and even if quantization is performed with a smaller number of allocated bits for a signal in which isolated spectrum components are sparsely distributed over the entire band. It is an object of the present invention to provide a signal encoding method and apparatus, and a recording medium that can reduce auditory sound quality deterioration.

[0010]

The signal coding method of the present invention has been made in view of such circumstances, and in the signal coding method for coding an input digital signal,
The input digital signal is decomposed into a plurality of frequency components subdivided on the frequency axis, a function related to the magnitude of each frequency component is obtained, and a band that is symmetric with respect to the folding frequency and close to the folding frequency according to the above function. The gain of frequency components within the width is attenuated, and each frequency component is quantized.

Further, the signal coding apparatus of the present invention is, in the signal coding apparatus for coding an input digital signal, decomposing means for decomposing the input digital signal into a plurality of frequency components subdivided on the frequency axis, Function calculating means for obtaining a function related to the magnitude of each frequency component, and gain attenuating means symmetric with respect to the folding frequency and attenuating the gain of the frequency component within the bandwidth close to the folding frequency according to the function, And a quantizing means for quantizing each frequency component.

Furthermore, the recording medium of the present invention decomposes a digital signal into a plurality of frequency components subdivided on the frequency axis, and is symmetric with respect to the folding frequency according to a function related to the magnitude of each frequency component. The gain of frequency components within the bandwidth close to the folding frequency is attenuated, and the coded information obtained by quantizing each frequency component is recorded.

These signal encoding method and apparatus of the present invention,
In addition, in the recording medium, the larger the function is, the wider the bandwidth for attenuating the gain of the frequency component or the larger the amount of reducing the gain of the frequency component. In addition, the function is the spread of the weight of the size of each frequency component on the frequency axis to the high band, the tonality based on the change in the size of each frequency component, and the required bit amount at the time of quantization of each frequency component. Is.

[0014]

According to the present invention, before the quantization of each frequency component,
By attenuating the gain of the frequency component within the bandwidth that is symmetric with respect to the folding frequency and close to the folding frequency, it is possible to reduce the quantization noise while minimizing the signal deterioration. As a function related to the size of the subdivided multiple frequency components, the spread of the weight of the size of each frequency component on the frequency axis to the high band, the tonality based on the change in the size of each frequency component, The required bit amount at the time of quantization of each frequency component is obtained, and the gain of the frequency component is reduced according to this function, so it is possible to obtain an input digital signal in which isolated spectrum components stand sparsely in all bands. It is also possible to perform adaptive gain attenuation.

[0015]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENT A preferred embodiment of the present invention will be described below with reference to the drawings.

FIG. 1 shows an overall configuration in which an encoded signal obtained by encoding an audio signal or the like is recorded or transmitted and the recorded or transmitted encoded signal is decoded.

The high-efficiency coding apparatus 1 for realizing the signal coding method of the present invention shown in FIG. 1 is a decomposing means for decomposing an input digital signal such as an audio signal into a plurality of frequency components subdivided on the frequency axis. , A control circuit 113 also having a function as a function calculating means for obtaining a function related to the magnitude of each frequency component, and a folding frequency (in the present embodiment, in accordance with the function). A high-band gain attenuating circuit 103 as a gain attenuating means that is symmetric with respect to the Nyquist frequency) and attenuates the gain of a frequency component within a bandwidth close to the folding frequency, and a quantizing circuit 10 that quantizes each frequency component.
And at least 5.

Here, the larger the function is, the wider the bandwidth for attenuating the gain of the frequency component or the larger the amount of reducing the gain of the frequency component. Widening the bandwidth and increasing the amount of reducing the gain can be combined. In addition, the function is the spread of the weight of the size of each frequency component on the frequency axis to a high band (or the spread of the frequency component to a wide band),
Tonality based on changes in the magnitude of each frequency component,
This is the required bit amount when quantizing each frequency component.

Further, in FIG. 1, the high-efficiency coding apparatus 1 quantizes the normalization circuit 104 for normalizing the output signal from the high-frequency gain attenuation circuit 103 and the normalization circuit 104. It also has a quantization circuit 105.

The signal obtained by the coding process in the high efficiency coding apparatus 1 is sent to the decoding apparatus 2 via the transmission or recording / reproducing unit 107.

The decoding device 2 includes an inverse quantization circuit 109,
The inverse normalization circuit 110 and the inverse mapping circuit 111 are included.

The operation will be described below. High efficiency coding device 1
A digital audio signal on the time axis composed of a plurality of sampling data is supplied to the input terminal 101 of. The digital audio signal is sent to the mapping circuit 102 and the control circuit 113.

The mapping circuit 102 decomposes the digital audio signal on the time axis into a plurality of frequency components. As a method of decomposing a signal on the time axis in the mapping circuit 102 into a plurality of frequency components, a digital audio signal on the time axis is divided into blocks for each short unit time, and the signal of each block is orthogonally transformed to be on the frequency axis. , Or a method of decomposing the digital audio signal on the time axis into signal components on the time axis of a plurality of bands by a band division filter such as QMF. In addition, this block is
It corresponds to a block in which so-called block floating is performed. As a method of decomposing the signal on the time axis in the mapping circuit 102 into a plurality of frequency components, a digital audio signal on the time axis is decomposed into signal components on a time axis of a plurality of bands by a band division filter, and further, It is also possible to use a method in which the signal component is divided into blocks for each band and orthogonally transformed to decompose into spectral components on the frequency axis. Therefore, the frequency component obtained by decomposing the input digital signal on the time axis referred to in the present invention means the spectrum component on the frequency axis by orthogonal transformation, the signal component on the time axis by division by the band division filter, or This includes any spectrum component on the frequency axis obtained by further orthogonally transforming the signal component on the time axis by the band division filter. In the following description, a case will be described in which a digital audio signal on the time axis is decomposed into signal components of a plurality of bands by a band division filter, and these signal components are divided into blocks and orthogonally transformed into spectral components. Explaining. As the orthogonal transform, for example, discrete Fourier transform (DFT), discrete cosine transform (DCT), modified discrete cosine transform (MDCT), or the like can be used. The spectral component signal from the mapping circuit 102 is sent to the high band gain attenuation circuit 103.

In the high frequency gain attenuating circuit 103, according to the gain control signal from the control circuit 113,
The gain of the high-frequency part (spectral component signal on the high-frequency side) of the spectral component signal for each block obtained from the mapping circuit 102 is attenuated. Here, when similarly attenuating the gain to the signal component obtained by decomposing the digital audio signal into a plurality of bands by the band division filter,
The gain of the signal component in the high band will be attenuated. The specific configuration and operation of the control circuit 113 and the details of the gain control signal will be described later.
The spectrum component signal of the entire band including the spectrum component in which the gain of the high band portion is attenuated by the high band gain attenuation circuit 103 is sent to the normalization circuit 104.

The normalization circuit 104 normalizes the spectrum component signal for each block for so-called block floating. That is, the normalization circuit 1
In 04, by normalizing the spectrum component signal for each block, even if the size of the spectrum component in the block changes, if the quantization word length by the quantization circuit 105 in the subsequent stage is the same, the signal-to-noise ratio is constant. Try to be. In addition,
A more detailed operation of the normalization circuit 104 will be described later. The normalized spectrum component signal from the normalization circuit 104 and the normalization coefficient information (scale factor) at the time of the normalization are sent to the quantization circuit 105.

In the quantization circuit 105, in order to reduce the bit amount of the normalized spectrum component signal (data compression is performed) in which each spectrum component is represented by, for example, 24 bits, this normalized spectrum is used. Quantization is performed to shorten the word length of the component signal. The more detailed operation of the quantization circuit 105 will be described later.
From the quantization circuit 105, a quantized spectrum component signal (spectrum component signal with a shortened word length),
Quantization accuracy information indicating the degree of reduction of the bit amount (that is, quantized word length information indicating the word length of quantization), and the normalization circuit 1
The normalized coefficient information supplied from 04 is output.

The signal obtained by the orthogonal transform, the high-frequency gain attenuation, and the subsequent normalization and quantization as described above is an output terminal of the high-efficiency encoder 1 as an encoded signal of a predetermined bit stream. It is output from 106.

The encoded signal is transmitted or recorded / reproduced by the recording / reproducing unit 1.
It is transmitted or recorded / reproduced at 07. When recording / reproducing is performed by the transmission or recording / reproducing unit 107, specifically, for example, a disc-shaped recording medium such as an optical disc, a magneto-optical disc, a magnetic disc, or a tape-shaped recording medium such as a magnetic tape is used as a recording medium. Alternatively, the encoded signal is recorded in a semiconductor memory, an IC card or the like and then reproduced. Further, when transmitting or transmitting by the recording / reproducing unit 107, the encoded signal is transmitted via a transmission medium such as an electric wire or an optical cable, or is transmitted by radio waves or the like. The recording medium in the present invention also includes a transmission medium.

Here, when the coded signal is recorded on, for example, an optical disk as an example of a recording medium, the coded signal output from the output terminal 106 is sent to the transmission or recording / reproducing unit 107 and is sent to, for example, an ECC encoder. Then, the error correction code is added, and the so-called 8-14 modulation is performed by the EFM circuit, for example, and then recorded on the optical disk by the recording head. The signal reproduced from the optical disk by the reproducing head is demodulated by 8-14 modulation by the EFM demodulation circuit and further error-corrected by the ECC decoder, and then output from the transmission or recording / reproducing unit 107. Like

The coded signal transmitted through the transmission or recording / reproducing unit 107 is supplied to the input terminal 108 of the decoding device 2. The encoded signal supplied to the input terminal 108 is sent to the inverse quantization circuit 109. The dequantization circuit 109 dequantizes the quantized spectrum component signal in the encoded signal using the quantization accuracy information in the encoded signal. More specifically, the inverse quantization circuit 109 restores the word length of the spectrum component signal by filling the bits shortened by the quantization in the quantization circuit 105 with "0" bits. A more detailed operation of the inverse quantization circuit 109 will be described later. The spectral component signal whose word length has been restored by the dequantization circuit 109 and the normalized coefficient information that has passed through the dequantization circuit 109 are sent to the denormalization circuit 110.

The denormalization circuit 110 denormalizes the spectral component signal whose word length has been restored by the dequantization circuit 109 (that is, this signal is a normalized spectral component signal). Perform denormalization. Specifically, the inverse normalization circuit 110 uses the normalization coefficient information to restore the size of the spectral component normalized for each block to the original size. The spectral component signal restored to the original size by the denormalization circuit 110 is sent to the demapping circuit 111.

In the inverse mapping circuit 111, the spectral component signals for each block on the frequency axis are transformed (inverse orthogonal transformation) to obtain signal components of a plurality of bands on the time axis, and the signal components of these plurality of bands are further converted. A digital audio signal on the time axis is reproduced by synthesizing with a band synthesizing filter. The signal of the inverse mapping circuit 111 is output from the output terminal 112 as a decoded digital audio signal on the time axis.

Next, each component of FIG. 1 will be described in order.

First, FIG. 2 shows the configuration of the mapping circuit 102 of the high-efficiency coding apparatus 1 of FIG. The mapping circuit 102 shown in FIG. 2 includes a band division filter 202 and an orthogonal transformation circuit 204.

In FIG. 2, the input terminal 201 has:
A digital audio signal on the time axis is supplied via the input terminal 101 of FIG. 1, and the digital audio signal on the time axis is sent to the band division filter 202. The band division filter 202 decomposes the digital audio signal on the time axis into signal components of a plurality of bands. The signal component from the band division filter 202 is sent to the orthogonal transform circuit 204. The orthogonal transform circuit 204 divides the supplied signal component into blocks for each unit time, and performs orthogonal transform for each block. The spectral component signal for each block is output from the output terminal 205 to the subsequent stage configuration (the high-frequency gain attenuation circuit 103 in FIG. 1).

More specifically, the mapping circuit 102 is constructed as shown in FIG.

The mapping circuit 102 shown in FIG.
QMF circuits 207 and 208 as the band division filter 202 and MDCT circuit 2 as the orthogonal transform circuit 204
09 to 211.

In FIG. 3, the digital audio signal on the time axis supplied to the input terminal 201 is sent to the QMF circuit 207 including the above-mentioned QMF filter. The QMF circuit 207 divides the entire band of the digital audio signal into two bands. QMF circuit 2
Among the signal components of the two bands on the time axis obtained by 07, the signal component on the high frequency side is sent to the MDCT circuit 209, and the signal component on the low frequency side is sent to the QMF circuit 208.
In the QMF circuit 208, the signal component on the low frequency side is further reduced to 2
The signal component on the high frequency side (that is, the signal component in the middle frequency band from the original entire band) is divided into the signal components on the time axis of one band and the obtained signal components of each band are divided into MDCT circuit 210.
The signal component on the low frequency side is sent to the MDCT circuit 211. In this way, the high-frequency signal component of the entire band of the digital audio signal is sent to the MDCT circuit 209,
The mid-range signal component is sent to the MDCT circuit 210, and the low-range signal component is sent to the MDCT circuit 211. MDC
In the T circuits 209 to 211, the supplied signal components of each band are divided into blocks, and the modified discrete cosine transform is performed for each block to obtain the spectrum component on the frequency axis. The spectrum component signals from the MDCT circuits 209 to 211 are output from the corresponding output terminals 2 respectively.
It is output via 12 to 214. That is, the high frequency spectrum component signal is output from the output terminal 212 to the output terminal 2
The spectrum component signal in the middle range is output from the output terminal 21.
4 outputs a low-frequency spectrum component signal. The output terminal 205 in FIG. 2 is the output terminal 212 in FIG.
~ 214 are collectively shown.

Next, FIG. 4 shows the configuration of the high-frequency gain attenuation circuit 103 shown in FIG. In FIG. 4, the spectrum component signals of each band from the mapping circuit 102 are input to the input terminal 301, and the spectrum component signals of each band are sent to the high band gain attenuation circuit 103. Further, a gain control signal from the control circuit 113 of FIG. 1 is supplied to the high band gain attenuating circuit 103 via a terminal 303. The gain control signal from the control circuit 113 has a coefficient that reduces the gain of the high frequency spectrum component signal. In the high frequency gain attenuator circuit 103, the spectrum component signal input via the input terminal 301 is added to By multiplying by the gain control signal, the gain of the high frequency spectrum component signal of the inputted spectrum component signals is attenuated. The amount by which the gain of the high-frequency spectrum component signal is attenuated according to the gain control signal (attenuation level) is determined in consideration of the human sense of hearing, and a typical amount (attenuation level) is approximately 6 dB. / Octave to 12 dB /
Octave. Also, depending on the nature of the digital audio signal, it may be 6 dB / octave or less. In this way, the spectrum component signal that has been subjected to the high-frequency gain attenuation processing by the high-frequency gain attenuation circuit 103 according to the gain control signal is output from the output terminal 302 provided for each band.

Here, the gain control signal generated by the control circuit 113 and the operation of the high frequency gain attenuation circuit 103 according to the gain control signal will be described more specifically.

Input terminal 301 of high frequency gain attenuation circuit 103
It is assumed that the spectrum component signal supplied to the above is a spectrum component SP distributed in a band up to the Nyquist frequency as a folding frequency as shown in FIG. 5, for example. The curve GC _N in FIG. 6 is a control circuit 11
3 shows an example of the gain characteristic of the gain control signal set in No. 3. That is, the curve G of FIG.
As indicated by C _N , the gain control signal set in the control circuit 113 has a characteristic of attenuating the spectrum component signal in the high frequency region near the Nyquist frequency.

Here, the control circuit 113, as a function related to the magnitude of each spectral component, spreads the envelope of the spectral component of the digital audio signal on the frequency axis (ie, the spectral component of each spectral component on the frequency axis). Of the weight to a high band) and the tonality based on the magnitude of the change in the level of the spectrum component on the frequency axis (hereinafter, a large change in the level of the spectrum component is referred to as high tonality). The gain characteristic of the gain control signal is adaptively changed based on these functions.

Specifically, the control circuit 113
The wider the envelope of the spectrum component on the frequency axis is, or the higher the tonality is (the greater the tonality is), for example, the curve G in FIG.
The gain characteristic of the gain control signal, which is indicated by C _NL , is changed in the direction in which the frequency bandwidth on the high frequency side for reducing the gain of the spectrum component signal is widened, as indicated by the curve GC _W in FIG. 7. . Here, by changing the gain characteristic of the gain control signal such that the wider the envelope of the spectrum component on the frequency axis is, the wider the frequency bandwidth on the high frequency side that reduces the gain of the spectrum component signal becomes. By increasing the spread of the envelope on the frequency axis of the signal, it is possible to reduce the increase in the quantization noise level that occurs during quantization in the quantization circuit 105. That is, at the time of quantization in the quantization circuit 105, a predetermined number of bits for each block is distributed to each spectrum component signal for quantization, and the quantization is performed on the frequency axis of the spectrum component signal. When the spread becomes large, the quantization circuit 105
Then, the predetermined number of bits for each block must be distributed to the spectrum component signals spread on this frequency axis, and as a result, each spectrum component signal must be quantized with a small number of bits. However, when the spread of the envelope on the frequency axis of the spectrum component signal increases as described above, as shown by a curve GC _W in FIG. By using a gain control signal having a gain characteristic in which the frequency bandwidth on the side is widened, the spread of the envelope of the spectrum component signal on the frequency axis can be suppressed, and therefore, the quantization in the quantization circuit 105 can be suppressed. It becomes possible to reduce the increase in the quantization noise level that occurs at this time. Furthermore, it can be expected that the quantization noise level will fall below the masking threshold, which is a threshold value at which the masking effect acts, due to the masking effect of the spectrum component signal on the low frequency side. In the control circuit 113, the gain characteristic of the gain control signal is changed so that the higher the tonality, the wider the frequency bandwidth on the high frequency side that reduces the gain of the spectrum component signal.
Bit distribution that gives a sufficient signal-to-noise ratio in the low frequency range can be performed for spectral component signals with high masking effects and high tonality, thus enabling good quantization in terms of hearing. Becomes

The gain characteristic of the control circuit 113 before the change of the gain control signal is shown by the curve GC _{NH in} FIG.
When the control circuit 113 originally has a certain high gain characteristic over a wide range on the high frequency side as shown in (4), the wider the spread of the envelope of the spectrum component on the frequency axis and the higher the tonality, The gain characteristics of the gain control signal as shown by the curve GC _NH in FIG. 8 are changed so that the amount of gain reduction with respect to the spectrum component signal on the high frequency side becomes large as shown by the curve GC _L in FIG. It is also possible to allow it. Also at this time, as the spread of the envelope of the spectrum component on the frequency axis becomes wider, by changing the gain characteristic of the gain control signal so as to increase the amount of gain reduction for the spectrum component signal on the high frequency side,
By increasing the spread of the envelope of the spectrum component signal on the frequency axis, it is possible to reduce the increase in the quantization noise level generated during the quantization in the quantization circuit 105. That is, at the time of quantization in the quantization circuit 105, each spectrum component signal is quantized using a predetermined number of bits, but when the spread of the spectrum component signal on the frequency axis becomes large, Circuit 1
In 05, a predetermined number of bits must be allocated to each spread spectrum component signal, and as a result, each spectrum component signal must be quantized with a small number of bits, and the quantization noise level increases. However, when the spread of the envelope on the frequency axis of the spectrum component signal increases as described above, the curve GC _{L of} FIG.
As shown in Fig. 4, by suppressing the spread of the envelope of the spectrum component signal on the frequency axis, by using a gain control signal with a gain characteristic that increases the gain reduction amount on the high frequency side where human hearing ability is low. Therefore, it is possible to reduce the increase in the quantization noise level generated during the quantization in the quantization circuit 105. In this case as well, it can be expected that the quantization noise level falls below the masking threshold, which is a threshold at which the masking effect acts, due to the masking effect of the spectrum component signal on the low frequency side. In the control circuit 113, the gain characteristic of the gain control signal is changed so that the gain reduction amount with respect to the spectrum component signal on the high frequency side is increased as the tonality increases, so that the spectrum component signal with a high masking effect and high tonality is obtained. Against
It becomes possible to perform bit allocation such that a sufficient signal-to-noise ratio can be given in the low frequency band, and therefore, it is possible to perform good quantization in terms of hearing sense.

As described above, the processing speed for increasing the gain attenuation amount is preferably faster than the processing speed for reducing the gain attenuation amount, and the frequency bandwidth for attenuating the gain is set. It is preferable that the processing speed at the time of widening is also faster than the processing speed at the time of narrowing the frequency bandwidth for attenuating the gain. This is because increasing the attenuation amount of the gain and widening the frequency bandwidth for attenuating the gain have a great influence on the reduction of the quantization noise in hearing.

As described above with reference to FIGS. 6 to 8, the gain control circuit 113 capable of adaptively changing the gain characteristic of the gain control signal should be realized by the configuration shown in FIG. 9, for example. You can

The control circuit 113 shown in FIG.
Is an orthogonal transform circuit 401 that divides a digital audio signal into blocks and performs orthogonal transform for each block to generate a spectrum component signal, and a plurality of spectrum components obtained by the orthogonal transform in the orthogonal transform circuit 401. Energy calculation as a configuration for generating a band control circuit 402 for generating a band, and a gain control signal having an equivalent bandwidth considering the tonality using the spectrum component signal for each band from the band control circuit 402 as an element A circuit 403, an equivalent bandwidth arithmetic circuit 404,
The LPF cutoff frequency calculation circuit 408, the tonality calculation circuit 407, and the gain control signal in consideration of human auditory characteristics (that is, in consideration of psychoacoustic model) using the spectral component signal for each band from the banding circuit 402 are also used. Psychoacoustic model (psychoacoustic model) circuit 40 as a configuration for generating
5. A selector 411 which has a required bit amount calculation circuit 406 and an LPF cutoff frequency calculation circuit 409, and which switches and selects between a gain control signal having an equivalent bandwidth as an element and a gain control signal considering human hearing characteristics. It also has

In FIG. 9, the input terminal 400 has
The digital audio signal supplied to the input terminal 101 of FIG. 1 is supplied. This digital audio signal is sent to the orthogonal transformation circuit 401. The orthogonal transform circuit 401 divides a digital audio signal composed of sampling data into blocks for each unit time (that is, for each of a plurality of adjacent sampling data), and performs orthogonal transform on the digital audio signal for each block, so that it is on the frequency axis. A fine spectrum component is obtained, and a spectrum component signal composed of these spectrum components is output. Fourier transform is used for the orthogonal transform in the orthogonal transform circuit 401. The spectral component signal obtained by the Fourier transform in the orthogonal transform circuit 401 is converted into the banding circuit 4
Sent to 02. The banding circuit 402 groups a plurality of spectral components for each block by the orthogonal transform circuit 401 into a band. That is, in the banding circuit 402, the spectrum component signals on the frequency axis sent from the orthogonal transform circuit 401 are divided into frequency widths at equal intervals, and a plurality of adjacent spectrum components are collected into a band on the frequency axis. To convert. At this time, the number of bands in the block is, for example, 32, and the number of spectral components in each band is the same. The spectrum component signal for each band from the banding circuit 402 is sent to the energy calculation circuit 403, the tonality calculation circuit 407, and the psychoacoustic model circuit 405.

The tonality calculating circuit 407 calculates the height (magnitude) of the tonality from the spectrum component signal for each band sent from the banding circuit 402. A more detailed operation of the banding by the banding circuit 402 and the tonality calculation in the tonality calculation circuit 407 will be described later.

The energy calculation circuit 403 calculates the energy value for each band from the spectrum component signal for each band sent from the banding circuit 402. Specifically, the energy value of each band is calculated by taking the sum of the squares of the spectral components of each band. The energy calculation signal for each band calculated by the energy calculation circuit 403 is sent to the equivalent bandwidth calculation circuit 404.
In this equivalent bandwidth calculation circuit 404, the maximum energy value is obtained from the energy calculation signal of each band, and the highest frequency at which the ratio to the maximum energy value from the high frequency side to the low frequency side is within 40 dB is defined as the equivalent band. To do. Further, the equivalent bandwidth arithmetic circuit 404 includes a tonality arithmetic circuit 407.
A signal (hereinafter referred to as a tonality signal) indicating the degree of tonality calculated in step 1 is also supplied. The equivalent bandwidth arithmetic circuit 404 modifies the defined equivalent bandwidth by the tonality signal. The banding circuit 402 to the energy calculating circuit 403 and the equivalent bandwidth calculating circuit 40
More detailed operations up to 4 will be described later. The signal indicating the equivalent bandwidth obtained by the equivalent bandwidth calculation circuit 404 in this way (hereinafter referred to as the equivalent bandwidth signal) is LP.
F (low-pass filter) cutoff frequency calculation circuit 40
Sent to 8.

The LPF cutoff frequency calculation circuit 40.
Reference numeral 8 finds a parameter for attenuating the gain of the spectrum component signal on the high band side in the high band gain attenuator circuit 103 of FIG. 1 from the equivalent bandwidth signal modified in consideration of tonality. Here, the high frequency gain attenuation circuit 103
Is L that attenuates the gain of the spectrum component signal on the high frequency side.
Since it operates as a PF (low-pass filter), the parameter corresponds to the cutoff frequency of this LPF. Therefore, the LPF cutoff frequency calculation circuit 408 specifically has a ROM that stores information corresponding to the LPF cutoff frequency, and reads LPF cutoff frequency information according to the parameters from this ROM. To Since the LPF is a digital filter, the LPF cutoff frequency information stored in this ROM can be used as a multiplication coefficient for high frequency attenuation.

On the other hand, the psychoacoustic model circuit 405 obtains a masking threshold considering the so-called masking effect for each band from the spectrum component signal for each band sent from the banding circuit 402, and further, this masking threshold is obtained. The allowable noise level is calculated by adding the minimum audible limit characteristic to the hold. The signal indicating the allowable noise level from the psychoacoustic model circuit 405 is sent to the required bit amount calculation circuit 406. The required bit amount calculation circuit 406 obtains the bit amount corresponding to the allowable noise level, and obtains the actually required bit amount (requested bit amount) in each band. That is, since the masking effect and the minimum audible limit function act below the permissible noise level and the bit allocation for quantizing the spectrum component signal below the permissible noise level is not necessary, the bit amount corresponding to this permissible noise level From, the required bit amount actually required for each band is obtained. Further, the required bit amount calculation circuit 406 obtains the total required bit amount of the block from the required bit amount of each band. The banding circuit 402 to the psychoacoustic model circuit 405 and the required bit amount calculation circuit 40
More detailed operations up to 6 will be described later.

Information indicating the required bit amount and the total required bit amount from the required bit amount calculation circuit 406 is sent to the LPF cutoff frequency calculation circuit 409. The LPF
The cutoff frequency calculation circuit 409 obtains a parameter for attenuating the gain of the spectrum component signal on the high frequency side in the high frequency gain attenuation circuit 103 of FIG. 1 based on these pieces of information. This LPF cutoff frequency calculation circuit 4
Also in 09, the LPF cutoff frequency calculation circuit 4
Similar to 08, a parameter corresponding to the LPF cutoff frequency in the high-frequency gain attenuation circuit 103 that operates as an LPF (low-pass filter) is obtained. Specifically, the LPF cutoff frequency calculation circuit 409 has a ROM that stores information corresponding to the cutoff frequency of the LPF, and reads the LPF cutoff frequency information according to the parameters from this ROM. To do. The LPF cutoff frequency information stored in this ROM can also be used as the multiplication coefficient for high frequency attenuation, as described above.

LPF cutoff frequency calculation circuit 4 described above
The LPF cutoff frequency information (or the multiplication coefficient for high frequency attenuation) from 08 and 409 is sent to the selector 411. The selector 411 selects the LPF cutoff frequency information (or the multiplication coefficient for high frequency attenuation) of either the LPF cutoff frequency calculation circuit 408 or 409 as the gain control signal. That is, the selector 411 outputs either a gain control signal having an equivalent bandwidth considering the tonality as an element or a gain control signal considering the required bit amount and the total required bit amount. Via the output terminal 412, the high frequency gain attenuating circuit 1 of FIG.
It will be sent to 03. Note that this selector 411
The selection in is performed manually.

Next, more detailed operations of the banding circuit 402 and the tonality arithmetic circuit 407 of FIG. 9 will be described with reference to FIG.

In FIG. 10, first, step S70.
Then, the banding circuit 402 creates a band by combining a plurality of adjacent spectrum components on the frequency axis obtained from the orthogonal transform circuit 41.

Next, in the tonality arithmetic circuit 407,
The following processing is performed using the spectrum component signal for each band by the banding circuit 402. In step S71, the effective value of the spectral component of the band is obtained, and the process proceeds to step S72. In step S72, the peak value of the spectral component of each band is obtained. In the next step S73, the ratio between the effective value of the spectrum component of each band obtained in step S71 and the peak value of the spectrum component of each band obtained in step S72 is obtained, and the process proceeds to step S74. In step S74, the ratio between the effective value and the peak value obtained in step S73 is added over all bands. The result of this addition is sent to the equivalent bandwidth arithmetic circuit 404 of FIG. 9 as a tonality signal.

Next, more detailed operations of the banding circuit 402, the energy calculating circuit 403 and the equivalent bandwidth calculating circuit 404 of FIG. 9 will be described with reference to FIG.

In FIG. 11, first, step S40.
Then, the banding circuit 402 collects a plurality of adjacent spectral components on the frequency axis obtained from the orthogonal transformation circuit 41 to create a band, and then proceeds to step S41.

In this step S41, the energy calculation circuit 403 obtains the energy value for each band from the spectrum component signal for each band obtained by the banding circuit 402 as described above, and then the process proceeds to step S42.

In step S42, the equivalent bandwidth calculation circuit 404 calculates the maximum energy value using the energy calculation signal calculated by the energy calculation circuit 403,
After that, it advances to step S43. In step S43, the highest band is set first. In the next step S44,
By using the energy value of the band set in step S43 and the maximum energy value obtained in step S42, the energy value of the set band is −4 from the maximum energy value.
It is compared whether or not it has a value within 0 dB. When it is determined that the answer is NO in step S44, step S45
In step S45, the band on the lower frequency side than the band set in step S43 is set, and then the process returns to step S44. At this time, in step S44,
The energy value of the band set in step S45 and the maximum energy value are compared again. Therefore, the processing of steps S44 and S45 is
Until the energy value of the band set in 45 is determined to be a value within -40 dB from the maximum energy value, that is, until a band having an energy value within -40 dB from the maximum energy value is found, the low frequency side is sequentially added. Repeated towards the band. When it is determined in step S44 that a band having an energy value within -40 dB from the maximum energy value is found, the process proceeds to step S46. In this step S46, the band obtained in step S44, that is, the highest frequency band in which the ratio with the maximum energy value is within −40 dB from the high frequency side to the low frequency side is defined as the equivalent band, and the next step Proceed to S47. In this step S47, the equivalent bandwidth defined in step S46 is modified based on the tonality signal from the tonality calculation circuit 47. Specifically, a process of narrowing the equivalent bandwidth as the tonality is higher (that is, a process of widening the high frequency side from the Nyquist frequency) is performed. The equivalent bandwidth signal obtained in this way is sent to the LPF cutoff frequency calculation circuit 408 in FIG.

Next, the detailed operations of the banding circuit 402, the psychoacoustic model circuit 405 and the required bit amount calculation circuit 406 of FIG. 9 will be described with reference to FIG.

In FIG. 12, first, step S50.
Then, the banding circuit 402 creates a band by combining a plurality of adjacent spectrum components on the frequency axis obtained from the orthogonal transform circuit 41, and then proceeds to step S51.

In step S51, the psychoacoustic model circuit 405 calculates the energy value for each band. The energy value for each band in the psychoacoustic model circuit 405 can be obtained, for example, by taking the sum of squares of the spectral components of each band. Next, in step S52, the masking threshold is obtained in consideration of the so-called masking effect, and the process proceeds to step S53. In step S53, a so-called minimum audible limit curve is further added to the masking threshold obtained in step S52 to obtain an allowable noise level, and then the process proceeds to step S54.

In step S54, the required bit amount calculation circuit 406 calculates the ratio between the signal level of the spectrum component signal of each band and the allowable noise level. In the next step S55, the actually required bit amount (requested bit amount) is obtained from the value of the signal level to the allowable noise level ratio. Specifically, the requested bit amount is calculated on the assumption that a signal level-tolerable noise level ratio of 6 dB is obtained with 1 bit. In the next step S56, the required bit amount of each band is added to obtain the total required bit amount of each block. Information indicating the required bit amount of each band and the total required bit amount of the block obtained in this way is shown in FIG.
Will be sent to the LPF cutoff frequency calculation circuit 409.

Next, referring to FIG. 13, the selector 41 of FIG. 9 for manually selecting the gain control signal.
A more detailed operation of No. 1 will be described.

In FIG. 13, in step S101, it is determined whether or not to select a gain control signal having an equivalent bandwidth as an element. When manually selecting a gain control signal having an equivalent bandwidth as an element, the process proceeds to step S102. This step S10
In 2, the gain control signal having the equivalent bandwidth as an element is selected. On the other hand, when the gain control signal is selected in consideration of the required bit amount (and the total required bit amount) in step S101, step S804
Proceed to. In step S804, the gain control signal is selected in consideration of the required bit amount (and the total required bit amount).

Next, with reference to FIG. 14, a more detailed operation of the normalization circuit 104 of the high efficiency encoding apparatus 1 of FIG. 1 will be described.

In FIG. 14, the spectral component signal in the block that has passed through the high frequency band gain attenuating circuit 103 is input in step S10, and the peak value relating to the spectral component in the block is detected in step S11. In the next step S12, the normalization scale factor corresponding to the peak value, that is, the normalization coefficient information is obtained.
The process of obtaining the scale factor can be performed, for example, by selecting the scale factor corresponding to the peak value from a plurality of scale factors prepared in advance. In step S13 after step S12, it is determined whether or not the normalization processing is completed for all blocks. If it is determined that the normalization processing is not completed, the procedure returns to step S10, and if it is determined that the normalization processing is completed, the normalization processing is performed. The conversion process ends.

Next, with reference to FIG. 15, a more detailed operation of the quantization circuit 105 of the high efficiency encoding apparatus 1 of FIG. 1 will be described.

In FIG. 15, in step S21, the spectral component signal for each block supplied via the high frequency band gain attenuating circuit 103 and the normalizing circuit 104 and the quantization precision information indicating the degree of reduction of the bit amount, that is, Quantized word length information for shortening the word length of the spectral component is input. The quantized word length information can be generated based on the allowable noise level obtained in the same manner as the psychoacoustic model circuit 405 in FIG. 9, for example. In the next step S22, after the spectrum component signal whose spectrum component is represented by, for example, a 24-bit length is quantized by rounding using the quantized word length information to shorten the bit length. , And proceeds to step S23. In step S23, it is determined whether or not the quantization process has been completed for all blocks, and if it is determined that the quantization process has not been completed, step S2
Returning to 1, when it is judged that the processing is completed, the quantization processing is ended.

Next, referring to FIG. 16, the decoding device 2 of FIG.
A more detailed operation of the inverse quantization circuit 109 will be described.

In FIG. 16, in step S31, the coded signal transmitted or recorded / reproduced is input, and the quantized spectral component signal and the quantized word length information (quantization accuracy information) in the coded signal are input. ) And are taken out. In step S32 subsequent to step S31, the quantized spectral component signal is inversely quantized using the quantized word length information. Specifically, for example, the value of 0 is packed in the LSB (least significant bit) side of each spectral component in the block so that the length becomes 24 bits, based on the quantized word length information. Therefore, a spectral component having a long word length and being quantized has a length of zero padding.
That is, in the inverse quantization process of step S32, the bit portion of the quantized spectrum component and the zero-padded portion are combined to be 24 bits. In step S33 following step S32, it is determined whether or not the inverse quantization processing has been completed for all blocks. If it is determined that the inverse quantization processing has not been completed, step S31 is performed.
If it is determined that the process is finished, the inverse quantization process is finished.

Next, with reference to FIG. 17, the decoding device 2 of FIG.
A more detailed operation of the inverse normalization circuit 110 will be described.

In FIG. 17, in step S61, the dequantized spectrum component signal and the scale factor (normalization coefficient information) are input. In step S62 following step S61, inverse normalization (cancellation of normalization) using a scale factor is performed on the dequantized spectral component signal. Specifically, in the inverse normalization processing in step S62, the scale factor is used to restore the magnitude of the spectral component normalized for each block to the original magnitude. In the next step S63, it is determined whether or not the denormalization processing has been completed for all blocks. If it is determined that the denormalization processing has not been completed, the procedure returns to step S61, and if it is determined that the denormalization processing has been completed, the denormalization processing is performed. Ends the process.

Next, FIG. 18 shows a schematic configuration of the inverse mapping circuit 111 of the decoding device 2 of FIG. 1, and FIG. 19 shows a more specific configuration. Note that FIG. 18 corresponds to FIG. 2 and FIG. 19 corresponds to FIG.

First, as shown in FIG. 18, the inverse mapping circuit 111 has an inverse orthogonal transform circuit 501 as a basic configuration.
And band synthesis filter 502.

In FIG. 18, the denormalized spectral component signal from the denormalization circuit 110 of FIG. 1 is supplied to the input terminal 500. This spectrum component signal is sent to the inverse orthogonal transform circuit 501. The inverse orthogonal transform circuit 501 performs inverse orthogonal transform on the spectral component signal of each block to generate signal components in a plurality of bands on the time axis. The signal components on the time axis of the plurality of bands are sent to the band synthesis filter 502. The band synthesizing filter 502 reproduces a digital audio signal consisting of sampling data on the time axis by performing band synthesizing of signal components of a plurality of bands. The reproduced digital audio signal on the time axis is output terminal 503.
Is sent to the output terminal 112 of the decoding device 2 of FIG.

More specifically, the inverse mapping circuit 111 is, as shown in FIG. 19, IMDCT (inverse MDCT) circuits 604 to 606 as an inverse orthogonal transform circuit 501 and IQMF (inverse QMF) as a band synthesis filter 502. It is composed of circuits 607 and 608. In this FIG.
The input terminal 500 of FIG. 18 is actually composed of three input terminals 601-603 as shown in FIG.
Reference numeral 1 is for supplying a high-frequency spectrum component signal, input terminal 602 is for supplying a middle-range spectrum component signal, and input terminal 603 is for supplying a low-frequency spectrum component signal.
The spectral component signals supplied to the respective input terminals 601 to 603 are IMDCTs provided correspondingly.
(Inverse MDCT) circuit 604 to 606. Each IM
The DCT circuits 604 to 606 perform inverse MDCT processing on the supplied spectral component signals to generate time-axis signal components. Each of these IMDCT circuits 60
Among the signal components from 4 to 606, the IMDCT circuit 60
The signal components from 5 and 606 are sent to the IQMF (inverse QMF) circuit 607 and band-combined by the IQMF circuit 607. The band-combined signal component from the IQMF circuit 607 and the signal component from the IMDCT circuit 604 are sent to the IQMF circuit 608, and the IQMF circuit 6
At 08, band synthesis is further performed. This IQMF circuit 6
The output of 08 is sent to the output terminal 503.

In the example of FIG. 1, the digital audio signal from the input terminal 101 is the control circuit 113.
Are input to the mapping circuit 102.
It is also possible to input the spectrum component signal from the. In this case, the orthogonal transformation circuit 401 of FIG. 9 is unnecessary.

As described above, in the embodiment of the present invention, as a function related to the magnitude of the spectrum component obtained by orthogonally transforming the digital audio signal, the weight of the spectrum component on the frequency axis becomes higher in the weight band. , The tonality based on the change in the size of each spectral component, and the required bit amount at the time of quantization of each spectral component, and symmetric about the Nyquist frequency according to this function before the quantization of each spectral component. In addition, since the gain of the spectrum component signal within the bandwidth close to the Nyquist frequency is reduced, there is an adaptive gain attenuation that can be applied to a digital audio signal in which isolated spectrum components stand sparsely in the entire band. Therefore, it is possible to reduce the quantization noise while minimizing the signal deterioration due to the high frequency gain limitation. Further, in the present embodiment, it is possible to achieve reduction of quantization noise by using block floating together, and masking by a low frequency side signal is also added, which can be used for improving sound quality.

[0082]

As is apparent from the above description, according to the present invention, each function on the frequency axis is expressed as a function related to the magnitude of a plurality of frequency components obtained by subdividing the input digital signal on the frequency axis. Before the quantization of each frequency component, determine the spread of the weight of the frequency component to a high band, the tonality based on the change in the size of each frequency component, and the required bit amount at the time of quantization of each frequency component. , According to this function, the gain of the frequency component within the bandwidth which is symmetric with respect to the folding frequency and close to the folding frequency is reduced, so that the input digital signal in which the isolated spectrum component stands sparsely in the whole band In addition to adaptive gain attenuation, it is possible to reduce quantization noise while minimizing signal deterioration due to high-frequency gain limitation.

[Brief description of drawings]

FIG. 1 is a block circuit diagram showing the overall configuration of a high efficiency coding apparatus according to an embodiment of the present invention.

FIG. 2 is a block circuit diagram showing a schematic configuration of a mapping circuit.

FIG. 3 is a block circuit diagram showing a more specific configuration of a mapping circuit.

FIG. 4 is a block circuit diagram showing a configuration of a high frequency gain attenuation circuit.

FIG. 5 is a diagram showing an example of a distribution of spectral components.

FIG. 6 is a diagram showing a high-frequency gain attenuation characteristic.

FIG. 7 is a diagram showing an example of widening a high band as an example of an adaptive high band gain attenuation characteristic.

FIG. 8 is a diagram showing an example of increasing a gain attenuation amount in a high range as an example of an adaptive high range gain attenuation characteristic.

FIG. 9 is a block circuit diagram showing a configuration of a control circuit.

FIG. 10 is a flowchart of a tonality calculation.

FIG. 11 is a flowchart of energy calculation and equivalent bandwidth calculation.

FIG. 12 is a flowchart of a psychoacoustic model calculation and a required bit amount calculation.

FIG. 13 is a flowchart of gain control signal selection.

FIG. 14 is a normalization flowchart.

FIG. 15 is a flowchart of quantization.

FIG. 16 is a flowchart of inverse quantization.

FIG. 17 is a flowchart of inverse normalization.

FIG. 18 is a block circuit diagram showing a schematic configuration of an inverse mapping circuit.

FIG. 19 is a block circuit diagram showing a more specific configuration of the inverse mapping circuit.

[Explanation of symbols]

 1 High Efficiency Coding Device 2 Decoding Device 102 Mapping Circuit 103 High Band Gain Attenuation Circuit 104 Normalization Circuit 105 Quantization Circuit 107 Transmission or Recording / Reproducing Unit 109 Inverse Quantization Circuit 110 Inverse Normalization Circuit 111 Inverse Mapping Circuit 113 Control Circuit

Claims (54)

[Claims] 1. A signal encoding method for encoding an input digital signal, wherein the input digital signal is decomposed into a plurality of frequency components subdivided on a frequency axis, and a function related to the magnitude of each frequency component is obtained. A signal encoding method characterized by attenuating a gain of a frequency component within a bandwidth which is symmetric with respect to the folding frequency and close to the folding frequency according to the above function, and quantizes each frequency component. 2. The signal encoding method according to claim 1, wherein the larger the function is, the wider the bandwidth for attenuating the gain of the frequency component is. 3. The signal coding method according to claim 1, wherein the larger the function is, the larger the amount of reducing the gain of the frequency component is. 4. The signal coding method according to claim 2, wherein the function is a spread of a weight of the magnitude of each frequency component on the frequency axis to a high band. 5. The signal coding method according to claim 2, wherein the function is a tonality based on a change in magnitude of each frequency component. 6. The signal encoding method according to claim 2, wherein the function is a required bit amount at the time of quantization of each frequency component. 7. The tonality based on the spread of the weight of the magnitude of each frequency component on the frequency axis to a high band and the tonality based on the change of the magnitude of each frequency component on the frequency axis. 2. The signal encoding method as described in 2. 8. The function is the spread of the weight of the size of each frequency component on the frequency axis to a high band, and the required bit amount at the time of quantization of each frequency component. 2. The signal encoding method as described in 2. 9. The signal encoding method according to claim 2, wherein the function is a tonality based on a change in the magnitude of each frequency component and a required bit amount at the time of quantization of each frequency component. 10. The function is the spread of the weight of the magnitude of each frequency component on the frequency axis to a high band, the tonality based on the change of the magnitude of each frequency component, and the quantization time of each frequency component. 3. The signal encoding method according to claim 2, wherein the requested bit amount is 11. The signal encoding method according to claim 3, wherein the function is a spread of a weight of the magnitude of each frequency component on the frequency axis to a high band. 12. The signal encoding method according to claim 3, wherein the function is a tonality based on a change in magnitude of each frequency component. 13. The signal encoding method according to claim 3, wherein the function is a required bit amount at the time of quantization of each frequency component. 14. The tonality based on the spread of the weight of the magnitude of each frequency component on the frequency axis to a high band and the tonality based on the change of the magnitude of each frequency component on the frequency axis. 3. The signal encoding method according to item 3. 15. The function is the spread of the weight of the magnitude of each frequency component on the frequency axis to a high band, and the required bit amount at the time of quantization of each frequency component. 3. The signal encoding method according to item 3. 16. The signal encoding method according to claim 3, wherein the function is a tonality based on a change in the magnitude of each frequency component and a required bit amount at the time of quantization of each frequency component. 17. The function is the spread of the weight of the magnitude of each frequency component on the frequency axis to a high band, the tonality based on the change of the magnitude of each frequency component, and the quantization time of each frequency component. 4. The signal encoding method according to claim 3, wherein the requested bit amount is 18. The signal encoding method according to claim 1, wherein the amount of attenuation of the gain of the frequency component is approximately 6 dB / octave to 12 dB / octave. 19. A signal encoding device for encoding an input digital signal, which relates to a decomposing means for decomposing the input digital signal into a plurality of frequency components subdivided on a frequency axis, and a magnitude of each frequency component. A function calculating means for obtaining a function, a gain attenuating means symmetric with respect to the folding frequency and attenuating the gain of a frequency component within a bandwidth close to the folding frequency, and a quantizer for quantizing each frequency component. And a signal coding apparatus. 20. The signal coding apparatus according to claim 19, wherein the gain attenuating unit widens a bandwidth for attenuating the gain of the frequency component as the function increases. 21. The signal encoding apparatus according to claim 19, wherein the gain attenuator increases the amount of reducing the gain of the frequency component as the function increases. 22. The signal coding apparatus according to claim 20, wherein the function is a spread of a weight of the magnitude of each frequency component on the frequency axis to a high band. 23. The signal coding apparatus according to claim 20, wherein the function is a tonality based on a change in magnitude of each frequency component. 24. The signal encoding apparatus according to claim 20, wherein the function is a required bit amount at the time of quantization of each frequency component. 25. The tonality based on the spread of the weight of the magnitude of each frequency component to a high band on the frequency axis and the change in the magnitude of each frequency component on the frequency axis. 20. The signal encoding device according to 20. 26. The function is the spread of the weight of the size of each frequency component on the frequency axis to a high band, and the required bit amount at the time of quantization of each frequency component. 20. The signal encoding device according to 20. 27. The signal coding apparatus according to claim 20, wherein the function is a tonality based on a change in magnitude of each frequency component and a required bit amount at the time of quantization of each frequency component. 28. The function is the spread of the weight of the magnitude of each frequency component on the frequency axis to a high band, the tonality based on the change of the magnitude of each frequency component, and the quantization time of each frequency component. 21. The signal encoding apparatus according to claim 20, wherein the requested bit amount is 29. The signal coding apparatus according to claim 21, wherein the function is a spread of a weight of the magnitude of each frequency component on the frequency axis to a high band. 30. The signal coding apparatus according to claim 21, wherein the function is a tonality based on a change in magnitude of each frequency component. 31. The signal encoding apparatus according to claim 21, wherein the function is a required bit amount at the time of quantization of each frequency component. 32. The tonality based on the spread of the weight of the magnitude of each frequency component on the frequency axis to a high band and the tonality based on the change of the magnitude of each frequency component on the frequency axis. 21. The signal encoding device according to 21. 33. The function is the spread of the weight of the magnitude of each frequency component on the frequency axis to a high band, and the required bit amount at the time of quantization of each frequency component. 21. The signal encoding device according to 21. 34. The signal encoding apparatus according to claim 21, wherein the function is a tonality based on a change in magnitude of each frequency component and a required bit amount at the time of quantization of each frequency component. 35. The above-mentioned function is the spread of the weight of the magnitude of each frequency component on the frequency axis to a high band, the tonality based on the change of the magnitude of each frequency component, and the quantization time of each frequency component. 22. The signal encoding device according to claim 21, wherein the requested bit amount is 36. The signal encoding apparatus according to claim 19, wherein the amount of attenuation of the gain of the frequency component is approximately 6 dB / octave to 12 dB / octave. 37. A bandwidth that is obtained by decomposing a digital signal into a plurality of frequency components subdivided on a frequency axis and is symmetric with respect to the folding frequency and close to the folding frequency according to a function related to the magnitude of each frequency component. A recording medium characterized in that the encoded information obtained by attenuating the gain of the frequency components inside and quantizing each frequency component is recorded. 38. The recording medium according to claim 37, wherein the larger the function, the wider the bandwidth for attenuating the gain of the frequency component. 39. The recording medium according to claim 37, wherein the larger the function, the larger the amount of reducing the gain of the frequency component. 40. The recording medium according to claim 38, wherein the function is a spread of a weight of the magnitude of each frequency component on the frequency axis to a high band. 41. The recording medium according to claim 38, wherein the function is a tonality based on a change in the magnitude of each frequency component. 42. The recording medium according to claim 38, wherein the function is a required bit amount at the time of quantization of each frequency component. 43. The function is a tonality based on a spread of a weight of the magnitude of each frequency component on a frequency axis to a high band and a change in the magnitude of each frequency component. 38. A recording medium according to item 38. 44. The function is the spread of the weight of the size of each frequency component on the frequency axis to a high band, and the required bit amount at the time of quantization of each frequency component. 38. A recording medium according to item 38. 45. The recording medium according to claim 38, wherein the function is a tonality based on a change in magnitude of each frequency component and a required bit amount at the time of quantization of each frequency component. 46. The above-mentioned function is the spread of the weight of the magnitude of each frequency component on the frequency axis to a high band, the tonality based on the change of the magnitude of each frequency component, and the quantization time of each frequency component. 39. The recording medium according to claim 38, wherein the requested bit amount is 47. The recording medium according to claim 39, wherein the function is a spread of the weight of the magnitude of each frequency component on the frequency axis to a high band. 48. The recording medium according to claim 39, wherein the function is a tonality based on a change in the magnitude of each frequency component. 49. The recording medium according to claim 39, wherein the function is a required bit amount at the time of quantization of each frequency component. 50. The tonality based on the spread of the weight of the magnitude of each frequency component on the frequency axis to the high band and the tonality based on the change of the magnitude of each frequency component on the frequency axis. 39. The recording medium according to 39. 51. The function is a spread of a weight of a size of each frequency component on a frequency axis to a high band, and a required bit amount at the time of quantization of each frequency component. 39. The recording medium according to 39. 52. The recording medium according to claim 39, wherein the function is a tonality based on a change in magnitude of each frequency component and a required bit amount at the time of quantization of each frequency component. 53. The function is such that the weight of the magnitude of each frequency component on the frequency axis spreads to a high band, the tonality based on the change of the magnitude of each frequency component, and the quantization time of each frequency component. 40. The recording medium according to claim 39, wherein the requested bit amount is 54. The recording medium according to claim 37, wherein the amount of attenuation of the gain of the frequency component is approximately 6 dB / octave to 12 dB / octave.