JPH08211899A - Method and device for encoding voice - Google Patents

Abstract

PURPOSE: To prevent a band to which no bit of a high frequency band is allocat ed from existing by allocating the number of bits of an irreducible minimum for the specified band. CONSTITUTION: The number of bits (information amount) allocated to the whole bands are set in an initial value zero bit, and SMRs (signal masking ratio) of respective bands are calculated. NMRs (noise masking ratio) are calculated from the SMRs and SNRs(signal noise ratio) of respective bands. Together with that, SARs(signal minimum audible limit ratio) are calculated. The band having the maximum NMR among respective bands is detected, and the band is defined No. 16. The No. 16 is made a NMRB(maximum noise masking ratio band), and whether or not the NMR of the No. 16 arrives at a minimum allocation threshold value is judged, and when smaller than the minimum allocation threshold value, bit allocation is ended. Then, by allocating the number of lowest bits even for the band of a small dynamic range by using auditory sense, the bands to which no bit are allocated are reduced.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a highly efficient digital signal coding apparatus, and more particularly to bit allocation.

[0002]

2. Description of the Related Art In the processing of a conventional highly efficient coding method, an input signal is band-divided, orthogonal transformation is applied to each band, and the auditory characteristic is used to make a masking threshold (hereinafter referred to as a masking threshold). Value)) is calculated. Then, a method is used in which the masking threshold value is compared with the signal level in each band, and bits are assigned to each band so that the bits fit in the coding format according to the comparison result.
This encoding method will be described with reference to FIG. Figure 6
FIG. 4 is a schematic block diagram showing a conventional encoding device.

In FIG. 6, a band division unit 601 divides an input signal into bands either uniformly or non-uniformly.
The band division is performed by orthogonal transform such as filter bank or discrete cosine transform that specifically divides the input signal into bands using a plurality of filters.

Further, a masking threshold value calculation unit 602
Performs frequency analysis on the input signal by Fast Fourier Transform (FFT) or the like, and calculates a masking threshold value by utilizing auditory characteristics. This masking threshold value is compared with the signal of each band output from the band division unit 601 in the bit allocation unit 603, and the amount of information assigned to each band so as to fit in the encoding format (hereinafter referred to as the number of bits).
To decide. This is called bit allocation.

The requantization unit 604 requantizes the band-divided signal based on the number of bits determined by the bit allocation unit 603. The requantized signal is sent to the formatting unit 605 and the bit allocation unit 6
It is output together with the bit allocation information and so on in 03 in a predetermined format.

[0006] As a typical one used as a format of such an encoding process, ISO / IEC DIS 11
There is MPEG audio described in 172. The bit allocation processing in the masking threshold value calculation unit 602 and the bit allocation unit 603 of the above encoding device will be described with reference to FIG. FIG. 7 is a flowchart showing a conventional bit allocation process.

Hereinafter, the signal masking ratio (Signal
Masking Ratio) to SMR, Signal Noise Ratio (Signal Noise Ratio) to SNR,
Noise masking ratio (Noise Masking R)
NMR), the maximum noise masking ratio band (Noise Masking Ratio Ban)
Let d) be NMRB.

In the initial value setting step 701, the number of bits assigned to all bands is set to an initial value (usually 0), and the signal masking ratio SMR is calculated from the ratio of the signal level of each band to the masking threshold value.

Noise masking ratio calculation step 702
Is a signal noise ratio SNR that is the ratio of the signal level derived from the number of bits assigned to each band and the quantization noise level (noise level), and subtracts the SNR from SMR to obtain the noise level and masking threshold. The noise masking ratio NMR which is the ratio of

The maximum noise masking ratio band detection step 703 is a band NMR having the largest NMR among all bands.
B is detected.

Noise masking ratio comparison step 704
Determines whether the NMR of NMRB has reached the minimum allocation threshold for allocating information, and if not, terminates bit allocation, and if so, performs the processing of the next noise masking ratio update step 705.

Noise masking ratio update step 705
Increases the number of bits of NMRB by one step (1 bit),
The SNR derived from the increased number of bits is subtracted from the SMR to obtain the NMR, and the NMR is updated as the NMR of the band.

The necessary information amount calculating step 706 calculates the required number of bits as the number of bits increases.

The information amount comparing step 707 compares the required bit number with the usable bit number. If the required number of bits is less than or equal to the number of usable bits, the number of bits,
The NMR is updated and the process returns to the maximum noise masking ratio band detection step 703. If the required number of bits is larger than the usable number of bits, the next update cancel step 708 is performed.

Update cancellation step 708 reverts to updating the number of bits and sets NMRB NMR to the minimum allocation threshold. Then, the process returns to the maximum noise masking ratio band detection step 703.

Bit allocation is performed by the above processing loop. This process will be described with reference to FIG. FIG.
FIG. 6 is a diagram schematically showing a conventional information allocation amount of bit allocation.

In FIG. 8, the band-divided signal of each band is compared with the masking threshold value calculated by the masking threshold value calculation unit 602.

Each SMR is calculated from the result of comparison between the signal level of each band and the masking threshold. The value calculated by this signal level and the masking threshold value becomes an audible threshold value of the signal that is not affected by the masking effect.

Since an initial value of 0 bits is assigned as the initial number of bits, the SNR of each band is 0, and the signal level and noise level are equivalent. Table 1 shows the relationship between the number of bits and SNR.

[0020]

[Table 1]

FIG. 9 is a schematic diagram showing the relationship between the number of bits and SNR. FIG. 9A shows a case where 2 bits are allocated, and FIG. 9B shows a case where 6 bits are allocated.

As shown in FIG. 9A, in the case of 2 bits, the quantization noise level is close to the signal level,
The quantization noise level becomes a value above the masking threshold value, and the quantization noise is audibly recognized. By allocating bits, the SNR increases.
For example, as shown in FIG. 9B, when 6 bits are allocated, the quantization noise level becomes a value below the masking threshold value, and the quantization noise is masked, which makes it difficult to perceptually recognize.

SNR is subtracted from SMR in each band to calculate NMR. In each band, the band in which NMR is maximum is detected, and the number of bits is preferentially assigned from that band.

Then, when the number of usable bits is allocated, the bit allocation process ends.

[0025]

In the above bit allocation, the number of bits is preferentially allocated from the band requiring a large dynamic range based on the masking threshold value calculation result. FIG. 10 is a schematic diagram showing a bit allocation calculation result in the conventional technique.

As shown in FIG. 10, since the number of bits is preferentially assigned from the band requiring a large dynamic range, there is a band in which the number of bits is not assigned in the high frequency band. When transmitting a transmission system that has a band where the number of bits is not allocated,
The sound reproduced from the signal is distorted as an overtone of the musical instrument, for example, and is detected as high-frequency noise synchronized with the musical instrument, which adversely affects the hearing.

In view of the above, it is an object of the present invention to reduce distortion in a high frequency band by using an adaptive bit allocation method in a coding method and a coding apparatus for allocating bits.

[0028]

A speech coding method according to the present invention divides an input signal into a plurality of bands and allocates an information amount based on a ratio between a signal level and a masking level of each band. In the speech coding method having the information amount allocating means for deciding, the information amount allocating means comprises a signal minimum audibility ratio detecting step for obtaining a ratio between a signal level and a minimum audible level for each band, and a signal minimum audible limit. Minimum signal minimum audibility for determining the ratio of the minimum signal level and minimum audible level to be transmitted based on the ratio of at least one band among the ratios of the signal level and the minimum audible level obtained in the ratio detection step. In the limit ratio detection step and the minimum signal minimum audible limit ratio detection step, the ratio of the signal level and the minimum audible level that are equal to or higher than the ratio of the minimum signal level and the minimum audible level is A band extraction step of extracting a frequency band,
The allocation amount comparison step that compares the information amount of the band extracted in the band extraction step with the minimum allocation amount of the information amount in each band, and the minimum allocation amount that is less than or equal to the minimum allocation amount among the bands compared in the allocation amount comparison step Is characterized by having a minimum allocation step of allocating.

In the voice coding method according to the present invention, the input signal is divided into a plurality of bands, and the information amount allocation means determines the allocation of the information amount based on the ratio between the signal level and the masking level of each band. In the speech coding apparatus having, the information amount assigning means obtains the signal minimum audible ratio detecting means for obtaining the ratio between the signal level and the minimum audible level for each band, and the signal minimum audible ratio detecting means. A minimum signal minimum audibility ratio detecting means for obtaining a ratio between the minimum signal level and the minimum audible level to be transmitted based on the ratio of at least one frequency band among the ratios of the signal level and the minimum audible level. Band extraction means for extracting a band having a ratio of the signal level and the minimum audible level that is equal to or higher than the ratio between the minimum signal level and the minimum audible level of the minimum signal minimum audible ratio detection means; When the amount of information is less than or equal to the minimum amount of allocation among the amounts of information of the bands extracted by the output unit and the minimum amount of allocation of information in each band, and the bands compared by the amount of comparison unit , And has a minimum allocation means for allocating the minimum allocation amount.

Further, in the speech coding apparatus, of the ratio of the signal level obtained by the signal minimum audibility ratio detecting means and the minimum audible level, the minimum signal level and the minimum signal level for transmitting at least one band ratio. It is characterized in that it is a ratio of audible limits.

Further, in the speech coding apparatus, the signal level and the noise level assigned to the band from the ratio of at least one band among the ratios of the signal level and the minimum audible level obtained by the signal minimum audible ratio detecting means. It is characterized in that the value obtained by subtracting the ratio of is the ratio of the minimum signal level to be transmitted and the minimum audible level.

[0032]

According to the present invention, the ratio between the noise level and the masking level and the ratio between the signal level and the minimum audible level are obtained, and the band having the maximum ratio between the noise level and the masking level is detected. , Allocate bits to that band. The ratio of the signal level of the band and the minimum audible level is used as the threshold value of the ratio of the audible signal level and the minimum audible level to be transmitted as it is, and the threshold value of the ratio of the signal level and the minimum audible level. By allocating the required minimum number of bits to the band having the ratio of the above signal level and the minimum audible level, it is possible to prevent the existence of the band in which the bit is not allocated in the high frequency range.

Further, the ratio between the noise level and the masking level and the ratio between the signal level and the minimum audible level are obtained, and a bit is assigned to a certain band according to the ratio between the noise level and the masking level. The ratio of the signal level and the noise level is calculated, and the ratio of the signal level and the noise level is subtracted from the ratio of the signal level and the minimum audible level of the band to determine the audible signal level and the minimum audible level. Obtain the ratio threshold and assign the minimum required number of bits to a band with a signal level-minimum audible level ratio greater than the audible signal level-minimum audible level ratio threshold. As a result, it is possible to prevent the existence of a band in which bits are not allocated in a high frequency range.

[0034]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Embodiments of the present invention are shown in FIGS.
A description will be given based on FIG. FIG. 1 is a flowchart showing an embodiment according to the present invention. FIG. 2 is a schematic explanatory diagram of the masking effect. FIG. 3 is a schematic explanatory diagram of the minimum audible limit.

In FIG. 1, 101 is an initial value setting step, 102 is a noise masking ratio and signal minimum audible ratio calculating step, 103 is a maximum noise masking ratio band detecting step, 104 is a noise masking ratio comparing step, and 105 is a first step. Noise masking ratio update step, 1
06 is a signal minimum audibility ratio threshold calculation step, 107
Is a signal minimum audibility ratio comparing step, 108 is a minimum information allocation amount comparing step, 109 is a second noise masking ratio updating step, 110 is a band detecting step, 111 is a necessary information amount calculating step, 112 is an information amount comparing step, 1
Reference numeral 13 is an update cancellation step.

The signal minimum audibility ratio detection step is a noise masking ratio and signal minimum audibility ratio calculation step 102, the minimum signal minimum audibility ratio detection step is a signal minimum audibility ratio threshold calculation step 106, and the band The extraction step is a signal minimum audibility ratio comparison step 107,
The allocation comparison step is the minimum information allocation detection step 10
8 and the lowest allocation step is the second noise masking ratio updating step 109.

The processing of each step in the bit allocation processing method of the present invention will be described. Hereinafter, the signal minimum audible ratio (Signal Threshold
of Audio Ratio) to SAR.

In the initial value setting step 101, the number of bits assigned to all bands is set to an initial value (usually 0), and the signal masking ratio SMR is calculated from the ratio of the signal level of each band to the masking threshold value.

Here, masking means that, as shown in FIG. 2, for example, the sound of murmuring can be heard in a quiet environment, but it may not be heard in a storm. In other words, when there are two or more sounds, it means that one sound interferes with the other and cannot be heard. The masking threshold is
It refers to the level of masking in each divided band.

A noise masking ratio and signal minimum audibility ratio calculating step (hereinafter referred to as NMR and SAR calculating step) 102 is a signal noise which is a ratio of a signal derived from information allocated from the SMR to each band and a quantization noise. The noise masking ratio NMR of each band is obtained by subtracting the ratio SNR. At the same time, the ratio SAR between the signal level of each band and the minimum audible limit level is calculated.

Here, the minimum audible limit refers to, for example, the minimum level of sound pressure that can be detected by the auditory sense during silence, and has a characteristic as shown in FIG. Generally, in quiet time, it is related to the limit of noise, and the minimum audible limit varies depending on the frequency of the sound (the pitch of the sound).

The maximum noise masking ratio band detection step (hereinafter referred to as NMRB detection step) 103 selects a band having the maximum NMR among all bands.

Noise masking ratio comparison step (hereinafter N
This is called MR comparison step. ) 104 is NM of NMRB
It is for determining whether R has reached a minimum allocation threshold for allocating information.

Here, the minimum allocation threshold indicates the minimum NMR threshold for allocating the number of bits. In other words, NMR smaller than the masking threshold
Even if a large number of bits are assigned to the band having
There is no point in allocating bits. This is a threshold value for providing a threshold value for NMR and for not performing the bit number allocation processing in a band smaller than the threshold value for NMR. Generally, this minimum allocation threshold value uses a minimum value (for example, minus infinity).

The first noise masking ratio updating step (hereinafter referred to as the first NMR updating step) 105 is an NMR.
The number of bits of B is increased by one step (1 bit), and the SNR derived from the increased number of bits is subtracted from SMR to update the NMR. In SNR, when the number of bits is increased by 1 bit, the quantization noise decreases,
The SNR for that band increases.

Signal minimum audibility ratio threshold calculation step (hereinafter referred to as SAR threshold calculation step) 106.
Is the minimum signal minimum audible ratio threshold (hereinafter referred to as SAR threshold) that should be transmitted from the SAR in NMRB.
Is to seek.

Here, the SAR threshold is calculated firstly by using the SAR obtained by NMRB as the SAR threshold, and secondly by the SAR obtained by NMRB by SNR by bit allocation of the band. Is subtracted, and the value obtained thereby is used as the SAR threshold value.

Of these two methods, the first method is easy to process, but in this embodiment, the case of using the second method will be described. The second method is less perceptually important but less transmitted than the first method, so that bit allocation which is good for hearing is performed. In this embodiment, a method will be described in which the SNR obtained by NMRB is subtracted from the SNR obtained by bit allocation in the band, and the value obtained by the subtraction is used as the SAR threshold.

The signal minimum audibility ratio comparison step 107 is
SAR of each band and SAR threshold are compared, and the SAR
Is for detecting a band that is equal to or higher than the SAR threshold.

The minimum information allocation amount comparison step 108 is S
The number of bits in a band having a SAR equal to or larger than the AR threshold is compared with the minimum number of allocated bits.

Here, the minimum number of allocated bits is defined in advance, for example, No. 0-No. 31 of 3
When there is a band of No. 2, No. 0-No. Bands up to 7 are 3 bits, No. 8 to No. The band up to 31 is 2 bits.

The second noise masking ratio updating step (hereinafter referred to as the second NMR updating step) 109 is performed by the SAR.
The number of bits in the band having the SAR less than the threshold is changed to the minimum number of allocated bits, and the SNR derived from the changed number of bits is subtracted from the SMR to update the NMR of the band.

The band detection step 110 detects whether or not bits have been allocated to all bands.

The necessary information amount calculation step 111 calculates the required number of bits for the updated information allocation amount.

The information amount comparing step 112 compares the required bit number with the usable bit number.

The update cancel step 113 returns the update of the number of bits to the original value and sets the NMR of NMRB to the minimum allocation threshold value.

The bit allocation processing operation in each of the above steps will be described. Initial value setting step 101
Sets the number of bits to be allocated to all bands to an initial value (usually 0 bit) for an input signal band-divided by a certain bandwidth, and calculates the SMR from the ratio of the signal in each band to the masking threshold value. calculate.

Then, the NMR and SAR calculation step 1
02 subtracts the SNR derived from the number of bits allocated to each band from the SMR to calculate the NMR of each band. Simultaneously with the calculation of NMR, SAR, which is the ratio between the signal level of each band and the minimum audible level, is also calculated.

In the NMRB detection step 103, the band having the maximum NMR is selected from the entire bands calculated by the NMR and the SAR calculation step 102.

Then, in the NMR comparison step 104, N
NM of NMRB selected in MRB detection step 103
R determines whether the minimum allocation threshold for allocating the number of bits is reached, and if the minimum allocation threshold is reached, the bit allocation ends, and if the minimum allocation threshold is not reached, the first allocation is finished. 1 NMR update step 105
To update the NMR.

In the first NMR updating unit 105, the number of bits of NMRB is increased by one step (1 bit), the SNR is derived from the increased number of bits, and the SNR is subtracted from the SMR to calculate the NMR, and a new NMR is calculated. Update as NMR.

Next, SAR threshold calculation step 106
Then, the minimum signal minimum audibility ratio threshold value (hereinafter referred to as SAR threshold value) to be transmitted from SAR in NMRB.
To calculate.

The SAR comparison step 107 compares the SAR threshold of each band with the SAR threshold of the NMRB. If the SAR is a band equal to or more than the SAR threshold, the process of the minimum information allocation amount comparison step 108 is performed, and the SAR is compared. Is less than the SAR threshold, the process of band detection step 110 is performed.

If SAR is greater than or equal to the SAR threshold, the minimum information allocation comparison step 108 is
The number of bits in the band that is equal to or larger than the AR threshold value is compared with the minimum number of allocated bits, and if the number of bits is less than the minimum allocated number of bits, the first NMR updating unit 109 performs the process so that If there is, the processing of band detection step 110 is performed.

The second NMR updating unit 109 changes the bit number of the band whose bit number is determined to be less than the minimum allocated bit number to the minimum allocated bit number, and then changes the SNR from the changed bit number.
Is derived, and SMR is subtracted from the SNR to calculate NMR. The NMR is updated as the NMR of the band. Then, the process of band detection step 110 is performed.

In the band detection step 110, the SAR comparison step 107, the minimum information allocation amount comparison step 108,
It is detected whether or not each processing of the second NMR updating unit 109 has been executed for all the bands selected in the SAR comparison step 107, and if there is a band that has not been processed yet, the SAR comparison step 107 Return to processing. S
When it is determined that the processing has been performed on all the bands selected in the AR comparison step 107, the necessary information amount calculation step 111 then determines the number of bits after the update,
Calculate the required number of bits.

The information amount comparing step 112 compares the required number of bits with the usable number of bits, and if the required number of bits is less than the usable number of bits, the number of bits and the NMR
Is accepted and the processing of the maximum noise masking ratio band detection unit 104 is performed. If the required number of bits is larger than the usable number of bits, the update cancel step 113 is performed on all the bands.

In the update cancel step 113, the update of the number of bits is returned to the original value, and the NMR of NMRB is set to the minimum allocation threshold value. Then, the maximum noise masking ratio band detection step 104 is performed.

The above processing is repeated to assign bits in all bands.

Specific processing will be described with reference to the schematic diagrams of FIGS. FIG. 4 is a diagram schematically showing the information allocation amount of bit allocation according to the present invention. FIG. 5 shows S
It is the figure which showed the AR threshold calculation result typically.

For example, the number of bits assigned to all bands is set to an initial value of 0 bits, and the SMR of each band is calculated. NMR is calculated from the SMR and the SNR of each band. At the same time, the SAR of each band is calculated.

The band having the maximum NMR is detected in each band, and that band is No. Assume that it is 16. This No.
No. 16 as NMRB. Determine if 16 NMR has reached the minimum assigned threshold. If the NMR is less than the minimum allocation threshold, terminate bit allocation.

NMRB No. If the 16 NMR has reached the minimum assigned threshold, No. The number of bits of 16 is increased by one step, the NMR is calculated from the SNR derived from the increased number of bits, and the NMR is calculated as No. 16 NMR is updated.

Next, the SAR of this band is obtained and set as the SAR threshold.

This No. No. 16 is compared with the SAR threshold of 16 and the SAR of each band. Detect bands with SAR greater than 16 SAR threshold. No. 16 SAR
The band having the SAR larger than the threshold is No. 1, N
o. 3, No. 6 ... etc.

First of all, For 1 band,
The allocated number of bits is compared with the minimum allocated bit number of the band, and if it is less than the minimum allocated bit number, it is changed to the minimum allocated bit number and the SN derived from the updated bit number
After obtaining the NMR from R and updating the NMR, the SAR of the next band and the SAR threshold value are compared.

No. If the number of bits in one band is equal to or more than the minimum number of allocated bits, the SAR of the next band is compared with the SAR threshold value.

When the detection of all the bands exceeding the SAR threshold is completed, No. If 16 is NMRB, the required number of bits is calculated for the updated number of bits.

The required number of bits and the usable number of bits are compared, and if the required number of bits is less than or equal to the usable number of bits,
The update of the bit number is accepted, and the NMRB detection process is performed again.

If the required number of bits is larger than the number of usable bits, the updating of the number of bits is canceled and the No. 16 NMR is set to the minimum allocation threshold value, and then NMRB detection processing is performed.

The process of detecting the NMRB is repeated, and finally, when the NMR of the NMRB falls below the minimum allocation threshold value, the process of bit allocation is terminated.

The high band has a relatively low signal level and is easily masked, so that the dynamic range is small. Therefore, the number of bits may not be assigned, but even in a band with a small dynamic range, by assigning the lowest number of bits by using the auditory characteristics, the band without bit assignment decreases, It is possible to prevent distortion that adversely affects hearing.

[0083]

According to the present invention, when bit allocation processing is performed on a signal within a predetermined number of bits, the audible adverse effect caused by the bit allocation is prevented and the sound quality is improved.

[Brief description of drawings]

FIG. 1 is a flow chart showing an example of the present invention.

FIG. 2 is a schematic explanatory view showing a masking effect.

FIG. 3 is a schematic explanatory diagram showing a minimum audible limit.

FIG. 4 is a diagram schematically showing an information allocation amount of bit allocation in the present invention.

FIG. 5 is a diagram schematically showing a SAR threshold value calculation result in the present invention.

FIG. 6 is a schematic block diagram showing a conventional encoding device.

FIG. 7 is a flowchart showing a conventional bit allocation process.

FIG. 8 is a diagram schematically showing an information allocation amount of a conventional bit allocation.

FIG. 9 is a schematic diagram showing the relationship between the number of bits and SNR.
(A) is a schematic diagram when 2 bits are allocated.
(B) is a schematic diagram when 6 bits are allocated.

FIG. 10 is a diagram schematically showing a conventional bit allocation calculation result.

[Explanation of symbols]

101 ... Initial value setting step 102 ... Noise masking ratio and signal minimum audibility ratio calculation step 103 ... Maximum noise masking ratio band detection step 104 ... Noise masking ratio comparison step 105 ... First noise Masking ratio updating step 106 ... Signal minimum audible ratio threshold calculating step 107 ... Signal minimum audible ratio comparing step 108 ... Minimum information allocation amount detecting step 109 ... Second noise masking ratio updating step 110 ... Band detection step 111 ... Necessary information amount calculation step 112 ... Information amount comparison step 113 ... Update cancellation step 601 ... Band division unit 602 ... Masking threshold value calculation unit 603. .... Bit allocation unit 604 ... Requantization unit 605 ... Adjusting unit 701 ... Initial value setting step 702 ... Noise masking ratio calculation step 703 ... Maximum noise masking ratio band detection step 704 ... Noise masking ratio comparison step 705 ... Noise masking ratio updating step 706. .. Necessary information amount calculation step 707 ... Information amount comparison step 708 ... Update cancellation step

Claims (4)

[Claims] 1. A speech coding method having information amount allocating means for dividing an input signal into a plurality of bands and deciding allocation of information amount based on a ratio of a signal level and a masking level in each band. The information amount allocating means includes a signal minimum audibility ratio detecting step for obtaining a ratio between a signal level and a minimum audible level for each band, and a signal level and a minimum audible level for the signal minimum audible ratio detecting step. A minimum signal minimum audibility ratio detecting step for obtaining a ratio between a minimum signal level and a minimum audible level to be transmitted based on a ratio of at least one band among the ratio of A band extraction step of extracting a band having a ratio of a signal level and a minimum audible level equal to or higher than a ratio of a minimum signal level and a minimum audible level in the step; When the amount of information in the bands compared in the band compared in the allocated amount comparison step and the amount of information of the band extracted in A speech coding method comprising: a minimum allocation step of allocating the minimum allocation amount. 2. A speech coding apparatus having an information amount allocating means for dividing an input signal into a plurality of bands and deciding allocation of an information amount based on a ratio between a signal level and a masking level in each band. The information amount allocating means includes a signal minimum audibility ratio detecting means for obtaining a ratio between the signal level and the minimum audible level for each band, and a signal level and a minimum audible level for the signal minimum audible ratio detecting means. And a minimum signal minimum audibility ratio detecting means for obtaining a ratio between a minimum signal level and a minimum audible level to be transmitted based on a ratio of at least one frequency band among the ratios of the minimum signal and the minimum audible ratio. A band extracting unit that extracts a band having a ratio of a signal level and a minimum audible level that is equal to or higher than a ratio of the minimum signal level and the minimum audible level of the detecting unit, and information of the band extracted by the band extracting unit. Amount and a minimum allocation amount of the information amount in each band, and a minimum allocation amount of the minimum allocation amount when the information amount is less than or equal to the minimum allocation amount in the bands compared by the allocation amount comparing unit. A speech coding apparatus having an assigning means. 3. The speech coding apparatus according to claim 2, wherein
It is characterized in that the ratio of at least one band of the ratio of the signal level and the minimum audible level obtained by the signal minimum audible ratio detecting means is set to the ratio of the minimum signal level to be transmitted and the minimum audible level. Speech coding device. 4. The speech coding apparatus according to claim 2, wherein
A value obtained by subtracting the ratio of the signal level and the noise level assigned to the band from the ratio of at least one band of the ratio of the signal level and the minimum audible level obtained by the signal minimum audible ratio detecting means is transmitted. A speech coding apparatus characterized by setting a ratio between a minimum signal level to be performed and a minimum audible level.