JPH0693197B2 - Band-division type speech analysis / synthesis device - Google Patents

Description

TECHNICAL FIELD The present invention relates to a speech synthesizer used in a band division type speech analysis and synthesis apparatus.

(Conventional Technology) The Bell System Technical Journal has been used as a technology of this type.
l), 55 [8] (1976-10) (US) P.1069-1085 Band-division type speech analysis and synthesis method (Sub-Band Codin)
Also called g method, abbreviated as SBC method afterwards) is known. In this SBC system, as shown in FIG. 4, a plurality of frequency bands of an audio signal (usually 4 to 8) (in the figure,
It shows with and. ), And the output of each divided channel is encoded and decoded separately.

FIG. 5 shows the basic circuit configuration of this SBC system. 6 (A) to 6 (E) are diagrams for explaining the operation of the circuit of FIG. Hereinafter, FIG. 5 and FIG.
The operation of the SBC method will be described with reference to (E).

First, the operation of the analyzer is as follows. An analog audio signal input from a microphone (not shown) is input to a low-pass filter (not shown) to remove frequency components of 1/2 or more of a predetermined sampling frequency, and then an A / D converter An analog signal is converted into a digital signal S _(n) at a predetermined sampling frequency (not shown). Here, n is a sample number. The digitized input signal S _(n) is input to the bandpass filter 50, and a specific band component (here, W _1k −W _2k ) is input as shown in FIG. 6 (A).
Is extracted. Next, the output signal of the bandpass filter 50 is multiplied by a cosine wave (cos wave) having a frequency of W _1k shown in FIG. As shown in Fig. 6 (C) (0- _Wk )
Is shifted to the baseband of. An unnecessary frequency component of 2W _1k or more (for example, a component shown by a dotted line in FIG. 6C) generated at this time is removed by the low-pass filter 52. Since the signal r _{k (n)} obtained in this way requires only frequency components of W _k or lower, necessary and sufficient information can be maintained by sampling at a sampling frequency of 2W _k . Therefore perform downsampling drop the high sampling frequency than necessary by the downsampling unit 53 to 2W _k to the signals the downsampling encoded at the encoder 54, and transmits the encoded signal to the combiner.

Next, the signal sent from the analyzer is decoded by performing the processing which is completely opposite to that of the analyzer in the synthesizer. That is, after the encoded signal is decoded by the decoder 55, the interpolation unit 56 performs upsampling to restore the signal downsampled by the analyzer to the original sampling frequency. The output signal from the interpolator 56 is demodulated by being multiplied by the cos wave having the frequency W _1k shown in FIG. 6 (D) in the multiplier 57, and shown in FIG. 6 (E). After returning from the base band (0−W _k ) to the original frequency band (W _1k −W _2k ), the band pass filter 58 removes the components of the band other than (W _1k −W _2k ). Remove.

In this way, the combiner outputs the signal _{k (n)} .

The above series of processing is performed for each divided band (channel), and finally the outputs of all channels are added to obtain an output audio signal.

The above is the basic operation contents of the SBC method, but the circuit configuration of FIG. 5 is rarely directly implemented, and the bandpass filters 50 and 58 are not used to reduce the circuit amount. An SBC system with such a configuration has also been proposed.

Next, the operation of the circuit shown in FIG. 7 will be described.

First, in the analyzer, the digitized input signal S
_(n) is a complex signal e ^jωkn [where ω _k = (W _1k + W _2k ) /
2] is complex-modulated. This complex modulation is performed by cos modulation by the multiplier 61a (modulation wave is cos ω _kn ) and sine modulation by the multiplier 61b (modulation wave by sin ω _kn ). Multiplier 61a, the output of 61b is a low-pass filter 62a of the bandwidth _{(0-ω k / 2)} , are filtered are input to 62b. In this way, the low-pass filter 62a is the real part a _k of the complex signal _{a k (n) + jb k} (n) (n) is, from the low-pass filter 62b is a complex signal a _k of _{(n) + jb k (n} ) The imaginary part b _{k (n)} is output. Each of the signals a _{k (n)} and b _{k (n)} is down-sampled to a frequency W _k by the down-samplers 63a and 63b and then encoded by the encoder 64,
It is transmitted to the synthesizer side. In the combiner, the coded signal is decoded by the decoder 65 and then interpolators 66a and 66b.
After being returned to the original sampling frequency by the filter, and then filtered through low-pass filters 67a and 67b having a bandwidth (0−ω _k / 2), multiplication with a cos wave by the multiplier 68a and a sine wave by the multiplier 68a are performed. Demodulated by multiplication of
Further, the cos component and the sin component of the signal are added by the adder 69, and the signals of the divided band are combined.

The above series of processing is performed for each divided band (channel), and finally the outputs of all channels are added to obtain an output audio signal.

The above is the operation principle of the SBC system, but this system has the following features compared to the system that encodes the audio signal itself.

The quantization error of each channel is close to white noise and spreads over the entire frequency spectrum, but only the noise within the band of each channel falls to each channel.
Quantization noise can be reduced.

Also, the quantization error of each channel is related only to the signal within that frequency band, and low-frequency components are large like speech,
In the case of a signal having a small high frequency component, the error in the channel of the high frequency band is only a slight error when viewed from the whole signal.

Further, the high frequency component of the voice signal is mainly a noise component, and an error in this band has little effect on hearing.

Therefore, by setting the band division method and the number of quantization bits to be given to the signals of each channel in consideration of such properties, the amount of information is about half that of the method of directly encoding the audio signal. Can be realized. That is, if PCM audio sampled at 8 kHz is directly subjected to ADPCM coding, for example, an information amount of about 30 Kbit / sec is required.
With BC, it is possible to obtain a synthesized sound of almost the same quality in terms of hearing with an information amount of about 16 Kbit / sec.

By the way, the above-mentioned coding of the down-sampled signal includes adaptive quantization (APC).
M) and Adaptive Differential Quantization
generalization, ADPCM) is generally used, but the most commonly used adaptive quantization will be described below.

Adaptive quantization is a method of quantizing while changing the quantization width adaptively according to the level of the input signal, and the quantizing cycle is sufficiently short with respect to the amplitude change of the signal. This is an effective method when it is gentle.

FIGS. 8 (A) and 8 (B) are block diagrams for explaining the principle for realizing this adaptive PCM system, and the prediction adaptive system (forward adaptive system) as shown in FIG. 8 (A). : F
(eed Forward) and a forward value adaptive method (backward adaptive method: Feed Back) as shown in FIG. 8 (B). In the figure, 70 is a multiplier, 71 and 76 are gain adaptive control systems, 72 is a quantizer, 73 is an encoder, 74 is a decoder, and 75 is a divider.

In the prediction adaptive method, the signal y controlled by the gain adaptive control system 71 that adaptively controls the gain of the input signal x (n)
The quantizer 72 quantizes (n), the encoder 73 encodes this, and the encoded result c (n) is transmitted. Therefore, it is impossible to reproduce the input signal on the receiving side without separately transmitting the output G (n) of the gain adaptive control system 71, which is obtained by multiplying the input signal x (n) and adaptively controlling the gain. . On the other hand, in the previous value adaptive method, the gain is controlled only by the coding result of the immediately preceding input signal level. The coding result of the immediately preceding signal is directly known to the transmitting side, and can also be automatically detected as the received code value on the receiving side. Therefore, no information for adaptive control is needed.

Considering in this way, the output G of the gain adaptive control system 71
Since it is not necessary to transmit (n), the pre-value adaptive method seems to be superior, but the quantizer does not overload the input signal and the signal-to-noise ratio (S / N ratio) In the predictive adaptive method, in which the gain is adaptively controlled so that the deterioration of) is in the range that cannot be heard and detected, more appropriate quantization and coding are performed, and the quality is stable. Can be said.

Further, if a fixed sample is set as one frame and G (f) for adaptively controlling the gain for this frame is used as a substitute value of G (n) for each sample of the frame, the increment of the bit rate is increased for each sample. Distributed, without incurring a seemingly extreme bit rate increase,
It is possible to prevent the deterioration of the synthetic sound quality.

This method is often used to improve APCM and is relatively easy to implement, and is called so-called segmental APCM (SAPCM). Here, it is necessary to select the frame length so that the input x (n) does not change extremely.

Although the basic concept of the APCM coding method has been described above, a method of directly controlling the width of quantization is also used instead of controlling the gain of the input signal. This method will be described below by taking the prior value adaptation method as an example.

Δ (n) is the quantization width for the input (n-th sample value) to be quantized, and the quantization width Δ immediately before that is Δ (n).
When the level value of the quantization result of the input that has been quantized by (n-1) is L _n-1 , it is expressed as Δ (n) = Δ (n-1) M (| L _n-1 |). However, M (| L _n-1 |) is a coefficient given in advance in the table.

The principle of selecting this coefficient will be described. Level value L _n-1
The absolute value of is encoded by B bits (excluding the sign bit)
When there in 2 ^B-number of levels of the smaller half of the range (2 ^B-1 one level lower), the input signal is considered to be in the interval to a gradual change, small coefficient M from 1 By setting the quantization width Δ (n) to the immediately preceding quantization width Δ (n−
1) It is narrower than 1) so that finer quantization can follow finer changes in the input signal, and the occurrence of quantization noise is suppressed. Conversely | L _n-1 | when is in the range of 2 ^B pieces of the larger half in level (2 ^B-1 one level upper) is the interval in which the input signal is changing drastically Considering that there is, the coefficient M is made larger than 1 to further widen the quantization width so that it is possible to follow a large and fast change of the waveform, and the occurrence of overload noise is prevented.

When using the method of controlling the width of this quantization in SAPCM (segmental APCM: predictive adaptive coding using frame processing), for example, the maximum amplitude value of the data in the frame is calculated,
This value is 2 ^B (B is the number of quantization bits excluding the sign bit)
The value divided by is the quantization width. However, in practice, this quantization width needs to be expressed by a limited number of bits and sent to the receiving side, so (maximum amplitude value) / (2 ^B ) is encoded once and transmitted to the receiving side at the same time. At the transmitting side, the decoding is performed again, and the input signal is quantized using the decoded result as the quantization width.

The APCM method described above is often used to encode a downsampled signal. As mentioned earlier, AP
The CM method is an effective method when the quantizing cycle is sufficiently short for the amplitude change of the signal and the relative amplitude change is gentle.

Therefore, it is already well known that it is effective when the sampling frequency is sufficiently high (for example, 12 kHz or more) with respect to the change in the amplitude of the signal such as a voice signal, and the sampled value changes only slowly, The down-sampled signal in the SBC does not always have a gradual change in amplitude like a voice signal, and there are extreme cases where the difference in amplitude between adjacent samples is about 1/2 of the dynamic range.

In this way, using the pre-value adaptive method for encoding a signal whose amplitude change is not always gentle is likely to cause an increase in quantization noise and S / N deterioration, and the preferred method is No, but when the amplitude of the signal was examined over several samples, it was found that there was a clear correlation, rather than being completely uncorrelated and random, so it was determined from the maximum amplitude in the frame. It seems reasonable to use the SAPCM method, which quantizes by the quantization width.

Next, a conventional SAPCM method for optimizing the quantization width will be described. FIG. 9 is a block diagram showing the outline of an apparatus for implementing the conventional SAPCM method.

The signal input from the input terminal 81 is a horizontal component (also called a cos component) or a vertical component (also called a sin component) of the output of the frequency division channel of the SBC.

One frame of the input signal is once stored in the input buffer memory 82. The quantization width control unit 83 determines the quantization width for the input signal for one frame stored in the input buffer memory. This quantization width is obtained by, for example, obtaining the maximum value of the input data amplitude in one frame, and dividing this maximum value by the maximum value that can be represented by (the number of bits given to the transmission of the encoding result of the input signal) -1 bit. . The output of the quantization width control unit 83 is encoded by the quantization width encoder 84 with the number of bits given to the quantization width transmission, and sent to the receiving side. On the other hand, the encoded quantization width is decoded by the quantization width decoder 85 and sent to the input signal encoder 86. Input signal encoder
At 86, by using the quantization width sent from the quantization width decoder 85, the input signal for one frame in the input buffer memory is encoded and sent to the receiving side. On the reception side, the quantization width decoder 87 decodes the encoded quantization width sent from the transmission side, and sends it to the reception signal decoder 88. The reception signal decoder 88 decodes the reception signal, which is the result of encoding the input signal by using the quantization width sent from the quantization width decoder 87, for one frame and outputs the result to the output terminal 89. Send to.

(Problems to be Solved by the Invention) In the conventional SAPCM technology for optimizing the quantization width in this way, the SAPCM processing is performed for one frequency division channel.
Since there is one for each of the horizontal and vertical components, AP
Although there is an improvement in quality compared to CM, the increase in bit rate (that is, the increase in bit amount) that occurs due to quantization width transmission becomes a considerable amount when each channel is added together.

An object of the present invention is to solve such a problem and reduce the number of bits for transmitting a quantization width when quantizing and transmitting a signal downsampled in SBC using SAPCM technology, It is an object of the present invention to provide a band-division type voice analysis / synthesis device that obtains a high-quality synthesized voice with an information amount.

(Means for Solving the Problems) In order to achieve this object, the present invention employs the following means on the transmitting side and the receiving side to perform the quantization width determination process.

First, on the transmission side (or analysis side), an amplitude maximum value detection unit that determines the maximum amplitude value of the horizontal component and the maximum amplitude value of the vertical component of each channel for each frame, and the maximum amplitude value of the horizontal component A quantization width coding unit that codes the maximum amplitude value of the vertical component to form a corresponding quantization width code, and a quantization that compares the sizes of the quantization width code of the horizontal component and the quantization width code of the vertical component. A result determining unit; a difference calculating unit that calculates a difference between the quantization width code of the horizontal component and the quantization width code of the vertical component; a difference encoder that encodes the difference to form a difference code; From a quantization width decoding unit that decodes the larger quantization width code obtained by, a difference decoder that decodes the difference code, and from the larger quantization width code obtained by the comparison Subtract the difference obtained by the decoding A subtracter, a quantization width decoding unit that determines a smaller quantization width according to the result obtained from the subtractor, and an input signal code that performs APCM encoding of the input signal by these two quantization widths. With a chemical device. Also, on the receiving side (or on the combining side), the quantization width decoding unit that decodes the larger coded quantization width code sent from the transmission side, and sent from the transmission side. A difference decoder for decoding the difference code, a subtracter for subtracting the difference obtained by the decoding from the larger quantization width code sent from the transmission side, and a result obtained by the subtractor Correspondingly, the quantization width decoding unit that determines the smaller quantization width, and the AP sent from the transmission side with the two quantization widths
A received signal decoder for decoding the CM code.

(Operation) As described above, the present invention uses the SAPCM method when coding a downsampled signal in a band-division type speech analysis and synthesis method that uses complex modulation for modulation, and as a result, in addition to the coded data, The configuration is provided with a means for increasing the efficiency of bit allocation when quantizing the quantization width that must be transmitted to the synthesis side and reducing the transmission bit rate while maintaining the quality of synthesized speech.

Focusing on one of the frequency-divided channels having a plurality of SBCs, the signal to be encoded has two components, a horizontal component and a vertical component, for one downsampling point. There is a strong correlation. Furthermore, if the concept of the frame is introduced,
This becomes more apparent. For example, the average amplitude of the horizontal component and the average amplitude of the vertical component in the frame often show almost similar values, and the same tendency is observed in the maximum amplitude. The present invention utilizes this property, for example, as is conventionally done, the quantization width obtained from the maximum amplitude of the horizontal component in the frame and the quantization width obtained from the maximum amplitude of the vertical component are necessary and sufficient. Encoding by allocating the number of bits,
Instead of transmitting, compare the quantization width obtained from the maximum amplitude of the horizontal component with the quantization width obtained from the maximum amplitude of the vertical component, and allocate the necessary and sufficient number of bits to the larger one as before. The smaller one is not transmitted, and the difference between the two is encoded by assigning a bit number smaller than the number of bits assigned to encode and transmit the quantization width conventionally obtained from the maximum amplitude. It is possible to reduce the number of bits required for the quantization width transmission in SAPCM, and to reduce the transmission bit rate. Comparing the quantization width of the horizontal component and the quantization width of the vertical component with the difference between the two, the latter is considerably smaller than the former. Therefore, the number of bits required to quantize this difference is about half the number of bits allocated for transmission of the quantization width.

(Embodiment) An embodiment of the band division type speech analysis and synthesis apparatus of the present invention will be described below with reference to the drawings.

The configuration of the device of the present invention will be described while explaining its operation.

FIG. 1 is a diagram showing an embodiment of the present invention. This is a block diagram of one frequency division channel of a portion for encoding, transmitting and decoding the downsampled signal of the band division type voice analysis and synthesis apparatus using the dual modulation for the modulation of the input signal by the APCM method. Therefore, the signals input from the input terminals 11 and 12 respectively correspond to the horizontal and vertical components of the downsampled signal of the frequency division channel of the SBC.

Transmission side First, the operation of the transmission side will be explained. One frame of the signals input from the input terminals 11 and 12 is once stored in the respective input buffer memories 13 and 14. The input signal for one frame stored in the input buffer memories 13 and 14 is sent to the absolute value calculation units 15 and 16, and these absolute value calculation units
Reference numerals 15 and 16 calculate absolute amplitude values of the input signals for one frame.

Based on the absolute amplitude value of the input signal calculated and output by the absolute value calculation units 15 and 16, the maximum amplitude value detection unit 17 and
Reference numeral 18 denotes a maximum amplitude value of each of the horizontal component and the vertical component, and the result is the quantization width encoder of the quantization width encoding units 19 and 20.
Send to 19a and 20a respectively.

In the quantization width coding units 19 and 20, the maximum amplitude detection unit 17 and
The quantization width is encoded based on each of the maximum amplitude value of the input of the horizontal component and the maximum amplitude value of the input of the vertical component sent from 18. In this embodiment, this encoding is performed by referring to the quantization width encoding table ROMs 19b and 20b. As shown in FIG. 2, the quantization width coding tables ROM19b and 20b show the maximum value quantization level logarithmically distributed to the dynamic range of the maximum amplitude value of the input of the horizontal component and the vertical component. Are stored in ascending order. In the quantization width encoders 19a and 20a, the numerical values sent from the amplitude maximum value detection units 17 and 18 are sequentially compared with the numerical values in the quantization width encoding tables ROM19b and 20b, (maximum value quantization level). _{When k-1} <(amplitude maximum value detection unit output value) ≦ (maximum value quantization level) _k is satisfied, the quantization width code k is output as the quantization result of the amplitude maximum value detection unit output value. This quantization width code is input to the quantization result determination unit 21 and the difference calculation unit 22.

The quantization result determination unit 21 compares the two quantization results of the input horizontal and vertical components, that is, the quantization width code,
The larger quantization result is sent to the quantization width decoding unit 23, and the smaller quantization result is calculated based on the larger quantization result, so that it is also sent to the subtractor 24. On the other hand, the larger quantization result is transmitted through the transmission line 31 to the quantization width decoder 40 of the reception side quantization width decoding unit 40.
It is also sent to a and the subtractor 41.

In addition, the quantization result determination unit 21 determines a signal indicating which of the two input signals is larger by the differential encoder 25.
Send to. The difference calculator 22 calculates the difference between the two input quantization width codes, and sends the result to the difference encoder 25.
In the differential encoder 25, the number that can be represented by (the number of bits allocated for differential encoding-1) bits is limited to the numerical value sent from the differential calculator 22 as necessary, and the remaining With 1 bit, for example, which of the quantization result of the maximum amplitude value of the input signal of the horizontal component and the quantization result of the maximum amplitude value of the input signal of the vertical component sent from the quantization result determination unit 21 is larger, Expressed as 0 when (maximum input of horizontal component) ≥ (maximum input of vertical component) and 1 when (maximum input of horizontal component) <(maximum input of vertical component) In addition, the differential code is also output as the encoding result. The output of this difference code is the difference decoder 26.
On the other hand, it is also sent to the differential decoder 46 on the receiving side through the transmission path 32.

The quantization width decoding unit 23 decodes the quantization width code sent from the quantization result determination unit 21. Decoding is performed by referring to the quantization width decoding table ROM23b. As shown in FIG. 3, the quantization width decoding table ROM 23b stores the quantization width corresponding to each maximum value quantization level. In the quantization width decoder 23a, this table ROM23
The quantization width is generated and output by referring to b. This output is the APCM of the component containing the signal with the maximum amplitude in the frame, that is, using the larger quantization width.
In order to perform encoding, the output is input to the multiplexer 27.

The differential decoder 26 decodes two pieces of information from the differential code sent from the differential encoder 25. One of them is the magnitude relationship between the quantization result of the maximum amplitude value of the input signal of the horizontal component and the quantization result of the maximum amplitude value of the input signal of the vertical component, and the other is the amplitude relationship between the input signals of both components. It is the difference of the quantization result of the maximum value. As described above, the former can be represented by 1 bit and can be easily decoded. This decoding result is sent to the multiplexer 27.

On the other hand, the latter has undergone limiter processing by the difference calculation unit 22, and is immediately sent to the subtractor 24 as an integer value.

In the subtractor 24, from the quantization result (quantization width code) of the amplitude maximum value of the larger input signal sent from the quantization result determination unit 21 as a result of comparison, it is sent from the differential decoder 26. , The difference between the quantization results of the maximum amplitude values of the input signals of both components is reduced. The result of this subtractor is the quantization width decoding unit 28
Send to.

The quantization width decoding unit 28 decodes the quantization width in the same procedure as the quantization width decoding unit 23. That is, by referring to the quantization width decoding table ROM 28b, the quantization width value corresponding to the value input to the quantization width decoder 28a is read as the smaller quantization width and sent to the multiplexer 27. The quantization width decoding table ROM28b is configured similarly to the quantization width decoding table ROM23b.

Next, the multiplexer 27 will be described. The multiplexer 27 receives the two quantization widths sent from the two quantization width decoders 23a and 28a on the basis of which of the two quantization widths sent from the differential decoder 26 is larger. Input signal encoder for horizontal component by switching the encoding width 29
And the vertical component to the input signal encoder 30.

The input signal encoders 29 and 30 use the quantization width sent from the multiplexer 27 to input the input buffer memory 13
And the input signal for one frame in 14 is APCM encoded by the number of bits assigned to the APCM encoding of the input signal, and received signal decoders 43 and 44 on the receiving side through transmission lines 33 and 34.
Send to each.

Receiving Side Next, the operation of the receiving side will be described. The quantization width decoding unit 40 decodes the encoded quantization width sent from the quantization result determination unit 21 and outputs it. The decoding procedure is similar to that of the quantization width decoding unit 23, and is performed by referring to the quantization width decoding table ROM 40b. The output of the quantization width decoding unit 40 is input to the multiplexer 45 for use in APCM decoding of the component including the signal having the maximum amplitude in the frame.

In the differential decoder 42, the quantization result of the maximum amplitude value of the input signal of the horizontal component in the differential code sent from the differential encoder 25, and the quantization result of the maximum amplitude value of the input signal of the vertical component Is sent to the subtractor 41, and the bit information expressing which of the two quantization widths is larger is sent to the multiplexer 45.

In the subtractor 41, from the quantization result (quantization width code) of the amplitude maximum value of the larger input signal sent from the quantization result judging unit 21 on the transmission side, it is sent from the differential decoder 42. The difference between the quantization results of the maximum amplitude values of the input signals of both components is reduced. The result of the subtractor 41 is the quantization width decoding unit 46.
Send to. In the quantization width decoding unit 46, the quantization width decoding unit 40
Decoding of the quantization width is performed in the same procedure as. That is, by referring to the quantization width decoding table ROM46b, the quantization width decoding device
The quantization width value corresponding to the value input to 46a is read as the smaller quantization width and sent to the multiplexer 45. The quantization width decoding table ROM46b has the same configuration as the quantization width decoding table ROM23b.

Next, the multiplexer 45 will be described. The multiplexer 45 receives the two quantization widths sent from the two quantization width decoders 40a and 46a, based on the information indicating which of the two quantization widths sent from the differential decoder 42 is larger. Horizontal component received signal decoder 43
And to the vertical component received signal decoder 44, respectively.

The reception signal decoders 43 and 44 use the quantization width sent from the multiplexer 45, and the transmission line 33 and
The APCM code sent through 34 is decoded and sent to the output terminals 47 and 48.

The present invention is not limited to the embodiments described above, and many modifications and changes can be made without departing from the scope of the present invention. For example, the example in which the quantization width encoding unit is configured by the quantization width encoder and the quantization width encoding table ROM has been described, but the configuration is not limited to this. Also, the quantization width decoding unit is not limited to the above-described embodiment. Further, each constituent component shown by a block in FIG. 1 can be easily configured by hardware and / or software by using a conventional electronic technique.

Further, in the above-described embodiment, the description has been made according to the communication system, but by storing the code sent from the transmission side to the reception side through the transmission lines 31 to 34 in FIG. 1 in the semiconductor memory, for example, It is possible to record high-quality voice in a semiconductor memory at a low bit rate.

(Effect of the Invention) As described in detail above, according to the present invention, the SAPC
When the SBC down-sampled signal is encoded by the M method and transmitted or stored in a memory, it is possible to reduce the number of bits allocated to the transmission and storage of the quantization width, so the bit rate of transmission and storage is reduced. Is possible. Furthermore, if the reduced number of bits is assigned to the coding of the downsampled signal according to the present invention, it is expected that the quality of the voice to be transmitted and stored is improved.

[Brief description of drawings]

FIG. 1 is a block diagram showing an embodiment of a band division type speech analysis / synthesis device of the present invention, and FIG. 2 is a quantization width coding table used in the device of FIG.
ROM explanatory diagram, FIG. 3 is a quantization width decoding table used in the apparatus of FIG.
ROM explanatory diagram, FIG. 4 is an explanatory diagram of the band division type voice analysis and synthesis system (SBC system), FIG. 5 is a conventional basic block configuration diagram of the SBC system, and FIG. 6 is the SBC system of FIG. 7 is a block diagram for explaining the principle of the adaptive quantization (APCM) system, and FIG. 9 is a quantization width diagram. FIG. 9 is a block diagram for explaining the principle of the SAPCM method for optimization. 11, 12 …… Input terminals 13, 14 …… Input buffer memory 15, 16 …… Absolute value calculator 17, 18 …… Amplitude maximum value detector 19, 20 …… Quantization width encoder 19a, 20a …… Quantization width encoders 19b, 20b ... Quantization width encoding table ROM 21 ... Quantization result determination unit 22 ... Difference calculation unit 23, 28, 40, 46 ... Quantization width decoding unit 23a, 28a , 40a, 46a ... Quantization width decoder 23b, 28b, 40b, 46b ... Quantization width decoding table ROM 24, 41 ... Subtractor, 25 ... Differential encoder 26, 42 ... Differential decoding Input signal encoder 31, 32, 33, 34 ... Transmission line 43, 44 ... Received signal decoder 47, 48 ... Output terminal.

Claims (1)

[Claims] 1. A downsampled signal of each band
APCM encoding, transmission or storage and APCM using SAPCM method
In the band-division type speech analysis and synthesis device for decoding, the maximum amplitude of the horizontal component signal and the maximum amplitude of the vertical component signal of each channel are determined for each frame on the transmission side in order to perform the quantization width determination process. An amplitude maximum value detection unit, a quantization width encoding unit that encodes the amplitude maximum value of the horizontal component and the amplitude maximum value of the vertical component to form a corresponding quantization width code, and a quantization width code of the horizontal component And a quantization result determination unit that compares the magnitudes of the quantization width codes of the vertical components, a difference calculation unit that calculates the difference between the quantization width code of the horizontal component and the quantization width code of the vertical component, and the difference A differential encoder that encodes to form a differential code, a quantization width decoding unit that decodes the larger quantization width code obtained by the comparison, and a differential decoder that decodes the differential code And obtained by the comparison A subtracter that subtracts the difference obtained by the decoding from the larger quantization width code, and a quantization width decoding unit that determines the smaller quantization width corresponding to the result obtained from the subtractor, , And an input signal encoder that performs APCM encoding of the input signal with these two quantization widths, and at the receiving side, decodes the coded larger quantization width code sent from the transmitting side. Quantization width decoding unit, a differential decoder that decodes the differential code sent from the transmitting side, and a larger quantization width code sent from the transmitting side, which is obtained by the above decoding. A subtractor that subtracts the difference, a quantization width decoding unit that determines a smaller quantization width corresponding to the result obtained from the subtractor, and an AP sent from the transmission side with the two quantization widths.
A band-division type speech analysis / synthesis device comprising a reception signal decoder for decoding a CM code.