JPH04162100A - Multiple pulse type voice coding-decoding device - Google Patents

Abstract

PURPOSE:To prevent the omission of an essential pulse and improve sound quality by computing the maximum value and minimum value of the separation between adjacent pulses by analysis frames, and changing a separation expressing range. CONSTITUTION:A pulse separation maximum value detector 9 retrieves the maximum separation of multiple pulses and a pulse separation minimum value detector 10 retrieves the minimum separation of the multiple pulses so as to output the pulse separation maximum value 11 and the pulse separation minimum value 12 to a pulse separation difference computing element 13. The computing element 13 outputs separation difference 17, which is the difference between the maximum value and minimum value of the multiple pulses, so as to be supplied to a quantize bit distribution label output element 14 to determine/output a quantize bit distribution label 18 distributed to an assigned bit in the quantization of the amplitude and separation of the multiple pulses. A multiple pulse quantizer 15 outputs quantized pulse amplitude-separation 20. Three kinds of quantization input, that is, the label 18, a quantized LPC coefficient 19 and the amplitude-separation 20 are appropriately combined into a multiplexed signal to be sent out to the decoding side. On the decoding side, the multiplexing separation of the multiplexed signal is performed and decoded by a decoder 24. A digital input voice signal is reproduced and D/A converted so as to be outputted as analog synthesized voice 30.

Description

[Detailed Description of the Invention] [Field of Industrial Application] The present invention relates to a multi-pulse speech encoding/decoding device,
In particular, the present invention relates to a multi-pulse audio encoding/decoding device that encodes and decodes an input audio signal using multi-pulses as audio source information and transmits the input audio signal.

[Conventional technology]

A multi-pulse audio encoding/decoding device analyzes an input audio signal, extracts spectrum envelope information and sound source information that constitute audio information using an analysis device, and decodes this audio information via a transmission path. The input audio signal is transmitted by transmitting it to the other side and synthesizing the audio.

The spectral envelope information represents spectral distribution information of the vocal tract system that generates the input speech signal, and includes a number of LPG coefficients corresponding to the analysis order obtained by ordinary linear predictive coding (LPG) analysis.

For example, parameters such as the α parameter are used.

Further, the sound source information indicates the fine structure of the spectral envelope, and is what is called a residual signal obtained by removing the spectral envelope information from the input audio signal.

This sound source information can also be expressed by a plurality of impulse sequences, ie, multipulses, with a degree of freedom in amplitude and position. The multi-pulse audio encoding/decoding device calculates the spectrum envelope information and the sound source information, quantizes them based on predetermined quantization bit allocation, and transmits them from the encoding side to the decoding side.

[Invention or problem to be solved]

The conventional multi-pulse type speech encoding/decoding device described above has the following problems.

In a conventional multi-pulse type speech encoding/decoding device, the position information of multi-pulses used as sound source information is usually expressed by using a fixed number of quantization bits between each pulse to express the interval between adjacent pulses. Therefore, if the interval between adjacent pulses exceeds the maximum interval that can be expressed with a given number of quantization bits, a commonly used method is to set a pseudo pulse (dummy pulse) with zero amplitude. .

In this case, since the number of pulses per two analysis frames is fixed, it is necessary to discard pulses with small amplitude absolute values as the number of pulses increases due to the setting of dummy pulses.

If the bit rate is high and the number of multi-pulses that can be set within one analysis frame is sufficiently large, even if some pulses are discarded due to the setting of dummy pulses, no significant deterioration in sound quality will occur. This is because the number of set multi-pulses is sufficiently large, so the absolute value of the amplitude of the discarded pulses is small, and the effect on sound quality is small.

On the other hand, when the bit rate is low and the number of multipulses allocated to one analysis frame is small, discarding pulses by setting dummy pulses becomes a fatal problem. This is because because the number of multi-pulses that are set is small, the absolute value of the amplitude of the discarded pulses is large, which greatly affects the sound quality. Therefore, when the bit rate is low, it is necessary to prevent pulses from being discarded by setting dummy pulses. However, since the bit rate is low, the number of bits allocated to the interval between adjacent multi-pulses cannot be made large enough to eliminate the need for setting dummy pulses.

The purpose of the present invention is to eliminate the above-mentioned drawbacks, calculate the minimum and maximum values of the intervals between adjacent pulses for each analysis frame, and prevent the omission of 5 important pulses by changing the 9-interval expression range. An object of the present invention is to provide a multi-pulse audio encoding/decoding device that can improve audio quality.

[Means to solve the problem]

The device of the present invention sends an input audio signal encoded using multipulses as sound source information and linear prediction coefficients as spectral envelope information from the encoding side to the decoding side via a transmission path, and the decoding side In a multi-pulse audio encoding/decoding device that reproduces an input audio signal by decoding encoded multi-pulses and encoded linear prediction coefficients provided by It has a configuration in which the number of quantization bits is allocated to the amplitude and position information within a fixed number of bits in accordance with the difference between the maximum value and the minimum value of adjacent multi-pulse intervals.

Further, the device of the present invention is configured to preset information regarding allocation of quantization bit numbers to the amplitude and position information of the encoded multi-pulse based on the difference between the maximum value and the minimum value of the adjacent multi-pulse intervals. There is a configuration in which the information is represented by a label, abbreviated coded, and transmitted.

〔Example〕

Next, nine aspects of the present invention will be described with reference to the drawings.

FIG. 1 is a block diagram showing the configuration of one embodiment of the present invention. The embodiment shown in FIG. 1 consists of an encoding side and a decoding side that face each other via a transmission path 100. The encoding side includes a waveform memory 2 that stores an input speech waveform, and an LPG analyzer that extracts LPG coefficients. a multipulse analyzer 6 for extracting multipulses, an LPG coefficient quantizer 16 for quantizing LPG coefficients, and a pulse interval maximum value detector 9 for detecting the maximum and minimum intervals of the multipulses. A pulse interval minimum value detector 10, a pulse interval difference calculator 13 that calculates the interval difference between the maximum value and the minimum value of the multipulse interval, and a pulse interval difference calculator 13 that calculates the amplitude and interval (
a quantization bit allocation label output device 14 that sets a quantization bit allocation label that determines the allocation of allocated bits in quantization of position) information; a multipulse quantizer 15 that quantizes multipulses; The multiplexer 21 performs multiplexing.

The decoding side includes a multiplexer 22 that demultiplexes multiplexed signals provided from the encoding side.

LPG coefficient decoder 23 that decodes encoded LPG coefficients
and a pulse decoder 2 that decodes the encoded multi-pulse.
4, and a quantization bit allocation label decoder 2 that decodes the quantization bit allocation label and supplies it to the pulse decoder 24.
5, and a digital filter type LPC that synthesizes and outputs an input audio signal using LPG coefficients and multipulses,
and a synthesis filter 26.

Next, the operation of the three embodiments will be explained.

An input speech waveform 1 cut out at the analysis frame length is supplied and stored in a waveform memory 2. The input speech waveform 3 read from the waveform memory 2 is supplied to the LPC analyzer 4 . The LPC analyzer 4 calculates LPG coefficients of a predetermined order by LPG analysis using a known analysis method. The calculated LPC coefficient 5 and the input audio waveform 3 output from the waveform memory 2 are supplied to a multipulse analyzer 6, where multipulse analysis is performed by a known method using correlation processing to form a multipulse train. The amplitude and pulse interval of each pulse are calculated. The multi-pulse pulse interval 8 calculated by the multi-pulse analyzer 6 is supplied to a pulse interval maximum value detector 9 and a pulse interval minimum value detector 10.

The pulse interval maximum value detector 9 searches for the maximum interval among the pulse intervals 8 of the input multi-pulse, and outputs the pulse interval maximum value 11 to the pulse interval difference calculator 13. The minimum pulse interval detector 10 searches for the minimum interval among the pulse intervals 8 of the input multi-pulse, and outputs the minimum pulse interval 12 to the pulse interval difference calculator 13 .

The pulse interval difference calculator 13 calculates an interval difference 17, which is the difference between the maximum value and the minimum value of the multi-pulse pulse intervals for each analysis frame, from the maximum pulse interval value 11 and the minimum pulse interval value 12 for each input analysis frame. This interval difference 17 is supplied to the quantization bit allocation label output device 14.

The quantization bit allocation label output device 14 determines and outputs a quantization focus allocation label 18 that specifies the allocation of allocated bits in quantizing the amplitude and interval of the multi-pulse in accordance with the magnitude of the input interval difference 17. . Is the total number of quantization bits of a multi-pulse set to a certain value, or is there a degree of freedom as to whether to allocate this fixed number of bits to amplitude information and interval information of a multi-pulse? Therefore, in an analysis frame in which the interval difference between each pulse of a multi-pulse is large, the number of bits allocated to 9-interval information is given preferential treatment than in the case of an analysis frame in which the interval difference is small, and in the opposite case, the number of allocated bits is reduced. Correspondingly, the number of bits allocated to amplitude information is also adjusted in accordance with the magnitude of interval information to effectively allocate bits allocated to multi-pulse quantization.

The quantization bit allocation label 18 is a label as a representative code that specifies the allocation of the number of allocated bits in the quantization of multi-pulse amplitude and interval information determined in accordance with the magnitude of the multi-pulse interval difference 17. , this quantization biy) allocation label 18 specifies each time the amplitude of the multipulse and the allocation of the number of bits in the interval quantization in m-child processing with the analysis frame code in the multipulse quantizer 15.

The multi-pulse quantizer 15 quantizes the multi-pulse according to the allocated bit number allocation specified by the quantization bit allocation label 18 received every nine analysis frames, and outputs the quantized pulse amplitude/interval 20.

The quantization bit allocation label 8 sends the quantization LPG coefficient 19 and the quantization pulse amplitude ψ interval 20 to the multiplexer 2j.
is output with The multiplexer 21
The two quantized inputs are appropriately combined to form a multiplexed signal, and the multiplexed signal is sent from the encoding side to the decoding side via the transmission path 100.

On the decoding side, a demultiplexer 22 demultiplexes and quantizes the sent multiplexed signal, and the coefficients are L
The PG coefficient decoder 23 has a quantization pulse amplitude 0 amount interval 20
is supplied to the pulse decoder 24, and the quantization bit allocation label 18 is supplied to the quantization bit allocation label decoder 25, and decoded by these decoders. in this case,
The decoded quantized bit allocation label 28 is supplied to the pulse decoder 24, and the pulse decoder 24 outputs a multipulse 29 similar to the multipulse extracted by the multipulse analyzer 6 on the encoding side. Output is off.

The LPC synthesis filter 26 uses the LPG coefficient 27 as a filter coefficient, is driven by inputting the multi-pulse 29, reproduces the digital input audio signal, and then converts it into a D/A.
After conversion, analog synthesized speech 30 is output.

In this way, missing multi-pulses can be prevented and sound quality can be improved.

〔Effect of the invention〕

As explained above, according to the present invention, the minimum and maximum values of the intervals between adjacent pulses of a multi-pulse are determined on the encoding side, and the allocated bit allocation in interval quantization is made variable according to the interval difference. This has the effect of preventing pulse loss and improving sound quality.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing the configuration of an embodiment of the present invention. 1... Input audio waveform, 2... Waveform memory, 3...
Input audio waveform, 4...LPG analyzer, 5...LPG
coefficient. 6...Multi-pulse analyzer, 8...Pulse interval, 9
...Pulse interval maximum value detector, 10...Pulse interval minimum value detector, 11...Pulse interval maximum value, 12...
- Pulse interval minimum value, 13... Pulse interval difference calculator, 14... Quantization bit allocation label output device, 15...
・Multi-pulse quantizer, 16...LPC quantizer,
17... Interval difference, 18... Quantization bit label,
19... Quantized LPG coefficient, 20... Quantized pulse amplitude-interval, 21... Multiplexer, 22... Demultiplexer, 23... LPC coefficient decoder, 24
... Pulse decoder, 25... Quantization bit allocation label decoder, 26... LPC synthesis filter, 27.
・・LPC coefficient, 28・・Bit allocation label, 29・
...Multipulse, 30...Synthesized voice, 100...
transmission line.

Claims (1)

[Claims] 1. An input audio signal encoded using multipulses as sound source information and linear prediction coefficients as spectral envelope information is sent from the encoding side to the decoding side via a transmission path, and the decoding side encodes the input audio signal. In a multi-pulse audio encoding/decoding device that decodes encoded multi-pulses and encoded linear prediction coefficients provided from the encoder side to reproduce an input audio signal, the encoder should be supplied from the encoder side to the decoder side. A multi-pulse audio code characterized in that the number of quantization bits for multi-pulse amplitude and position information is allocated within a fixed number of bits in accordance with the difference between the maximum value and the minimum value of adjacent multi-pulse intervals. decoding device. 2. Information regarding the distribution of the number of quantization bits to the amplitude and position information of the encoded multi-pulse is shortened by representing it with a preset label based on the difference between the maximum value and the minimum value of the adjacent multi-pulse interval. 2. The multi-pulse speech encoding/decoding device according to claim 1, wherein the multi-pulse speech encoding/decoding device encodes and transmits the speech.