JP2762938B2 - Audio coding device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a speech coding apparatus, and more particularly to a speech coding apparatus used in a digital mobile communication type mobile telephone system.

[0002]

2. Description of the Related Art As shown in FIG. 2, a speech encoding apparatus used in a conventional digital mobile communication type mobile telephone system has a power quantizer 2 for inputting a speech signal and outputting a power index and a power value. And input the audio signal, LP
C (Linear Predictive Coefficient) and LSP (Line
An LPC analyzer 5 that outputs a spectrum pair, a pitch analyzer 6 that receives an audio signal, performs a linear prediction inverse filter on the audio signal, and outputs a plurality of lag candidates from the autocorrelation of the LPC prediction residual signal. The voice signal and the prediction residual gain output from the LSP quantizer 4 are input, and voice / non-voice detection of voice is performed.
A sound / silence detector 3 that outputs OX MODE information (referred to as “VOX mode information”) is provided in parallel.

[0003] The speech coding apparatus inputs and quantizes the LSP output from the LPC analyzer 5, converts the quantized LSP into a quantized LPC, and calculates the LSP index, the quantized LPC, and the prediction error gain. An LSP quantizer 4 for outputting, and a filter coefficient for inputting LPC and calculating a filter coefficient for performing auditory weighting, further inputting an audio signal, and outputting an auditory weighting audio signal 7
A perceptual weighting / synthesizing filter 9 that inputs a quantized LPC and a zero signal and outputs a zero-input response signal, and a subtractor 10 that subtracts the zero-input response signal from the output signal of the perceptual weighting / combining filter 9. .

Further, the above-mentioned speech coding apparatus comprises a subtractor 1
0, a quantized LPC, and a lag candidate, and a cross-correlation between a signal obtained by weighting and synthesizing each lag candidate and an audio signal obtained by subtracting a zero-input response output signal from an auditory-weighted audio signal. An adaptive codebook searcher 11 that finds an adaptive codebook signal that maximizes this and outputs an adaptive codebook index and an adaptive codevector signal
And an output signal of the subtracter 10, a quantized LPC, and an adaptive code vector signal, and a signal obtained by subjecting each noise code vector to an adaptive code vector signal and orthogonalization processing to perform weighting synthesis, and a perceptually weighted speech A cross-correlation with a speech signal obtained by subtracting a zero-input response output signal from a signal is searched, a noise code vector signal which maximizes the cross-correlation is searched, and a random code book searcher 12 which outputs a random code book index and a random code vector signal And the output signal of the subtracter 10, the adaptive code vector signal, the noise code vector signal, the power value, and the prediction residual gain, and the output signal of the subtracter 10, the noise code vector signal obtained by adding the gain, and the adaptive vector A gain codebook searcher 13 that searches for a gain codebook that minimizes an error from the sum of the signals and outputs a gain codebook index;
A VOX unique word generator 14 for outputting unique pattern data of OX (Voice Operated Transmitter), a power index, an adaptive codebook index, a noise codebook index, a gain codebook index, and LS
The data transmission unit 15 receives a P index, a VOX unique word, and VOX mode information, converts the input data into a predetermined format, and outputs encoded data.

[0005] When the uplink voice is a silent section, the mobile station transmits a unique word indicating stop and start of the transmission of the encoded index, stops the output of the power amplifier, and suppresses power consumption. The presence / absence determination of the upstream voice is performed by the presence / absence detector 3.

The sound / non-speech detector 3 has a power value of the audio signal equal to or less than a predetermined threshold "1" (for example, -45 dBm), or a prediction residual gain equal to or more than a predetermined threshold "2" (for example, 0.4).
In the case of (1), a silent state is set. If this condition is not satisfied, a sound state is set.

The sound / silence detector 3 sets a state “0” when the sound state is continuous, a state “1” when the state changes from the sound state to the silence state, and further sets a state “1” when the silence state continues. , The number of states increases to state “2”, state “3”,..., State “25”. When the state changes from the silent state to the sound state, the state is set to “−1”.

The sound / silence detector 3 determines the number of states as VO
Output as X mode information.

If the state following the state "25" is also a silent state, the state returns to the state "1". This is to transmit information on the ambient noise of the mobile station to the base station at certain intervals.

[0010] When the VOX mode information is in the state "0" or in the state "2 or more", the data transmitting section 15 performs coding index (power index, adaptive codebook index, noise codebook index, gain codebook index, LSP index) in a data format defined in advance.

When the VOX mode information is in the state "1", the data transmitting section 15 outputs a unique word ("postamble") for notifying the stop of the transmission of the encoded index in a predetermined data format. And
In the case of the state "-1", a unique word (referred to as "preamble") for transmitting the start of encoding index transmission is output in a data format defined in advance.

The decoder of the base station monitors the VOX unique word to determine whether the mobile station is transmitting the coded index or the coded index transmission in the silent section is stopped. Generate background noise. Background noise is generated with a coding index following a periodically transmitted postamble as an initial value.

[0013]

In the conventional speech coding apparatus, a unique word is transmitted when the upstream speech is silent, and the transmission of the coding index to the base station is stopped.
However, when the silent state continues, the coding index is periodically transmitted to the base station as information on the ambient noise.

[0014] The speech coding apparatus of the base station generates background noise using the information of the ambient noise at the base.

Therefore, the background noise that changes periodically is
There is a problem that unlike a sound obtained by encoding and decoding ordinary noise, the sound becomes very uncomfortable.

SUMMARY OF THE INVENTION The present invention has been made in view of the above-mentioned problems, and an object of the present invention is to provide a voice coder used in a digital mobile communication type mobile telephone system, and a mobile station VOX (or VOX). VAD: Voice Activi
It is an object of the present invention to provide a speech encoding device that reduces discomfort caused by a periodic tone change of background noise transmitted when silence is detected in ty Detect processing.

[0017]

In order to achieve the above object, a speech coding apparatus according to the present invention receives a speech signal, calculates and quantizes a power value of the speech signal, and quantizes the power value with a power index. A power quantizer for outputting a power value, and an input of the audio signal, the LPC (Linear Predictive Coef)
ficient) is calculated and output, and LS is calculated from the LPC.
An LPC analyzer that converts and outputs a P (Line Spectrum Pair), a pitch analyzer that receives the audio signal, performs a linear prediction inverse filter, and outputs a plurality of lag candidates from the autocorrelation of the LPC prediction residual signal. A speech / silence detector which inputs the speech signal and a prediction residual gain output from an LSP quantizer, detects presence / absence of speech / silence, and outputs VOX mode information indicating a speech state or a silence state. , And LSP are input and quantized to obtain a quantized L
The L which converts an SP into a quantized LPC and outputs an LSP index, the quantized LPC, and the prediction error gain
An SP quantizer, a perceptual weighting filter for inputting the LPC and calculating a filter coefficient for performing perceptual weighting, and receiving the voice signal and outputting a perceptually weighted voice signal; A perceptual weighting synthesis filter that inputs a signal and outputs a zero-input response signal, a subtractor that subtracts the zero-input response signal from an output signal of the perceptual weighting filter, an output signal of the subtractor, the quantized LPC, The lag candidates are input, and a cross-correlation between a signal obtained by weighting and synthesizing each lag candidate and an audio signal obtained by subtracting a zero-input response output signal from an auditory-weighted audio signal is determined. An adaptive codebook searcher for searching a code vector signal and outputting an adaptive codebook index and an adaptive code vector signal; an output signal of the subtractor and the quantity The adaptive LPC and the adaptive code vector signal are input, the adaptive code vector signal is subjected to orthogonalization processing to each noise code vector and weighted and synthesized, and the zero input response output is obtained from the auditory weighted speech signal. A noise codebook searcher that calculates a cross-correlation with a speech signal obtained by subtracting the signal, searches for a noise code vector signal that maximizes the cross-correlation, and outputs the noise codebook index and the noise code vector signal, and the subtractor The output signal of the adaptive code vector signal, the noise code vector signal, the power value and the prediction residual gain are input, the output signal of the subtracter, the noise code vector signal obtained by adding gain, and the adaptive A gain codebook searcher that searches for a gain codebook that minimizes an error from the sum of the code vector signals and outputs a gain codebook index; And VOX unique word generator for outputting a unique pattern data of X, and the power index, and the adaptive codebook index, and the random codebook index, and the gain codebook index, and the LSP index, the VOX
A VOX unique word output from a unique word generator and the VOX mode information, and a data transmission unit that converts the VOX mode information into a predetermined format and outputs encoded data. It is.

In the speech coding apparatus of the present invention, it is preferable that the speech coding apparatus include an LPF (low-pass filter) that receives the speech signal and suppresses and outputs a high-frequency component. One of an audio signal and an output signal of the LPF is switched and input based on VOX mode information output from the sound / silence detector, and an auditory weighted audio signal is output.

In the speech coding apparatus according to the present invention, preferably, the power quantizer inputs VOX mode information output from the voiced / silent detector, and the VOX mode information is in a silent state. Is characterized by outputting a power index obtained from a power value averaged over a long period of time.

In the speech coding apparatus of the present invention, preferably, the LPC analyzer inputs VOX mode information output from the voiced / silent detector, and the VOX mode information is in a silent state. , LPC to a unique value, and outputs the same. The LSP quantizer inputs VOX mode information output from the voiced / silent detector,
When the OX mode information is in a silent state, a quantized LSP index and a quantized LPC when the value is a unique value of the LPC are output.

In the speech coding apparatus of the present invention, preferably, the adaptive codebook searcher inputs VOX mode information output from a voiced / silence detector, and the VOX mode information is in a silent state. Wherein the adaptive codebook index is controlled to a unique value and output.

Further, in the speech coding apparatus according to the present invention, preferably, the adaptive codebook searcher inputs VOX mode information output from the voiced / silence detector, and the VOX mode information is in a silent state. In the middle case, the adaptive codebook index is controlled to a unique value, and control is performed so that search processing is not performed.

In a preferred embodiment, the speech encoding apparatus according to the present invention includes an LPF (low-pass filter) for inputting a speech signal and attenuating a predetermined high frequency component, and an LPF for the speech signal and the speech signal. And a perceptual weighting filter that switches and inputs one of the outputs according to VOX mode information output from the voiced / silent detector and outputs a perceptually weighted audio signal, and a VOX mode information output from the voiced / silent detector. When the VOX mode information is in a silent state, a power quantizer that outputs a power index obtained from a power value averaged for a long time and a VOX mode information that is output from the voiced / silent detector. type in,
When the VOX mode information is in a silent state, an LPC analyzer that controls and outputs LPC to a unique value and VOX mode information output from the sound / silence detector are input, and the VOX mode information is In a silent state, a quantized LSP index and a quantized L when LPC is set to a unique value
An LSP quantizer that outputs a PC, a pitch analyzer that receives the audio signal, performs a linear prediction inverse filter, and outputs a plurality of lag candidates from the autocorrelation of the LPC prediction residual signal, and the voiced / silent detection Codebook searcher which inputs VOX mode information output from the device and controls the adaptive codebook index to a unique value when the VOX mode information is in a silent state, so as not to perform a search process. A random codebook searcher that searches for a random codebook signal and outputs a random codebook index and a random code vector signal;
A gain codebook searcher for outputting a gain codebook index, a VOX unique word generator for outputting VOX unique pattern data, the power index, the adaptive codebook index, the noise codebook index, and the gain Data transmission for inputting a codebook index, the LSP index, a VOX unique word output from the VOX unique word generator, and the VOX mode information, converting the VOX mode information into a predetermined format, and outputting encoded data. And at least a part.

[0024]

The principle of the present invention is as follows.
LPF (Low Pass F) that suppresses and outputs high frequency components
ilter), one of the audio signal and the output signal of the LPF is switched and input according to the VOX mode information output from the soundless / silent detector, and an auditory weighting filter for outputting an auditory weighting audio signal, and the VOX mode information are input. , VOX
When the mode information is in a silent state, a power quantizer that outputs a power index obtained from a power value averaged over a long period of time and VOX mode information are input, and VOX mode information is input.
When the mode information is in the silent state, the LPC analyzer controls and outputs the LPC to a unique value, and when the VOX mode information is in the silent state, the LPC analyzer uses the LPC as the unique value. An LSP quantizer for outputting an index and a quantized LPC, and VOX mode information,
When the OX mode information is in a silent state, an adaptive codebook searcher that controls the adaptive codebook index to a unique value so as not to perform a search process is provided, and is output from a decoder of the base station. This is to reduce the discomfort caused by the periodic tone change of the background noise.

Further, according to the present invention, applying an LPF to the audio signal input to the auditory weighting filter can reduce background noise in an adaptive spectral postfilter used to enhance the auditory quality of synthesized speech. Suppress frequency emphasis.

[0026]

Embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram showing the configuration of an embodiment of the speech encoding apparatus according to the present invention.

Referring to FIG. 1, an audio signal is input from input terminal 1. The power quantizer 2 receives an audio signal from the input terminal 1, finds the power value, searches the power codebook prepared in advance by learning for the index of the best approximated quantized power value, and Output as an index.

Since the power quantizer 2 performs the same operation as the decoder, the power value quantized based on the searched power index (hereinafter, this quantized power value is referred to as “R
MS ”).

The power quantizer 2 outputs the VOX MODE information (“VO_MODE”), which is input particularly from the sound / silence detector 3.
When the “X mode information” is in a silent state, a power value averaged over a long period of time in a silent section is obtained, an index of a quantized power value that is best approximated from a power codebook is searched, and the power is calculated. Output as an index.

By taking an average over a long period of time,
The power level of the background noise generated by the base station decoder is
Suppresses sudden level fluctuation due to sudden noise.

The sound / silence detector 3 receives a speech signal from the input terminal 1, inputs a prediction residual gain output from the LSP quantizer 4, performs speech / silence detection of the speech, and It outputs VOX mode information indicating a sound state or a silence state.

The VOX mode information is defined as a state "0" when the sound state is continuous, a state "1" when the state changes from the sound state to the silent state, and a state "1" when the silent state continues. When the number of states increases to "2", state "3",..., State "25", and the state following the state "25" is a silent state, the state returns to the state "1" and the state changes from the silent state to the sound state. If so, the state "-1" is output.

The sound / non-speech state of the sound / silence detector 3 is determined, for example, when the power value of the audio signal is equal to or less than a predetermined threshold “1” (for example, −45 dBm) or when the prediction residual gain is equal to or smaller than the predetermined threshold “2” If it is equal to or more than 0.2 (for example, 0.2), the sound state is set. If this condition is not satisfied, the sound state is set.

The LPC analyzer 5 inputs a voice signal from the input terminal 1, finds an LPC by an autocorrelation method or a covariance method, and further changes parameters from LPC to LSP.
Outputs LPC and LSP.

In particular, when the VOX mode information input from the sound / silence detector 3 is in a silent state, the LPC analyzer 5 sets the LPC to a unique value (for example, the transfer function of a synthesis filter using the LPC is 1) is output. Since the LSP calculates the prediction residual gain, it does not replace the output data.

The pitch analyzer 6 receives a speech signal from the input terminal 1, performs a linear prediction inverse filter, and outputs a plurality of lag candidates from the autocorrelation of the LPC prediction difference signal.

The LSP quantizer 4 inputs and quantizes the LSP output from the LPC analyzer 5, converts the quantized LSP into a quantized LPC, and outputs an LSP index, a quantized LPC, and a prediction error gain. .

The hearing weighting filter 7 receives the LPC output from the LPC analyzer 5 and calculates a hearing weighting filter coefficient.

Further, when the VOX mode information input from the sound / silence detector 3 is in a silence state, the auditory weighting filter 7 is an LPF (low band) for blocking and attenuating a predetermined high frequency component of the audio signal. An output signal passing the audio signal through a pass filter 8 and, in the case of a sound section, an audio signal from the input terminal 1 are switched and input via a switch 17,
An auditory weighted audio signal is output.

The auditory weighting synthesis filter 9 calculates the quantization L
A PC and a zero signal are input, and a zero input response signal is output. The subtractor 10 subtracts the quiescent response signal from the output signal of the auditory weighting filter 7.

The adaptive codebook search unit 11 receives the output signal of the subtracter 10, the quantized LPC, and a plurality of lag candidates, and weights and synthesizes each lag candidate, and obtains a perceptually weighted speech signal. A cross-correlation with a speech signal obtained by subtracting the quiescent response output signal is obtained, an adaptive code vector signal that maximizes the correlation is searched, and an adaptive codebook index and an adaptive code vector signal are output.

The adaptive codebook searcher 11 uses a unique value (for example, an index not using the adaptive codebook) as an adaptive codebook index, particularly when the VOX mode information input from the sound / silence detector 3 is in a silent state. ) Is output.

In this case, since the adaptive codebook searcher 11 sends out a fixed index, the search processing can be omitted, and the amount of calculation and power consumption can be reduced.

The noise codebook search unit 12 receives the output signal of the subtractor 10, the quantized LPC and the adaptive code vector signal, and performs an adaptive code vector signal and orthogonalization processing on each noise code vector to perform weighting synthesis. Cross-correlation between the obtained signal and the audio signal obtained by subtracting the zero input response output signal from the auditory weighted audio signal, searching for a noise code vector signal that maximizes this, a noise codebook index and a noise code vector signal Is output.

The gain codebook searcher 13 receives the output signal of the subtractor 10, the adaptive code vector signal, the noise code vector signal, the power value, and the prediction residual gain, and multiplies the output signal of the subtracter 10 by the gain. A gain codebook that minimizes an error between the obtained random code vector signal and the added value of the adaptive code vector signal is searched, and a gain codebook index is output.

The VOX unique word generator 14 outputs the VO
The unique pattern data of X is output.

The VOX unique word includes a postamble for notifying the transmission of the encoded index and a preamble for notifying the start of the encoded index transmission.

The data transmitting section 15 receives a power index, an adaptive codebook index, a noise codebook index, a gain codebook index, an LSP index, a VOX unique word, and VOX mode information, and Convert to the format and output the encoded data.

When the VOX mode information is in the state “0” or “state” 2 or more, the data transmitting unit 15 sets the coding index (power index, adaptive codebook index,
Noise codebook index, gain codebook index, L
SP index).

The data transmitting section 15 outputs a postamble when the VOX mode information is in the state "1", and outputs a preamble when the VOX mode information is in the state "-1".

As described above, according to the present embodiment, when the VOX mode information indicates a silent state, the audio signal input to the auditory weighting filter 7 is subjected to the LPF 8 to improve the auditory quality of the synthesized voice. In the adaptive spectrum postfilter to be performed, the frequency emphasis of background noise is suppressed, and the power quantizer 2 averages the power value in a silent period for a long time, thereby generating the signal at the decoder of the base station. The power level of the background noise suppresses a sudden level fluctuation caused by sudden noise.

In the present embodiment, the adaptive codebook searcher 11 outputs a fixed index, which has the effect of suppressing the periodic change of the background noise.

Furthermore, in this embodiment, the adaptive codebook searcher 11 can output a fixed index, thereby omitting the search processing, greatly reducing the amount of calculation and reducing power consumption. Becomes

Although the present invention has been described with reference to the above embodiments, the present invention is not limited only to the above embodiments.
It includes various aspects according to the principles of the present invention.

[0056]

As described above, the present invention (claim 1)
According to the above, by performing the state control in the voiced state and the silent state based on the VOX mode information output from the voiced / silence detector, the periodic tone change of the background noise output from the decoder of the base station is performed. This has the effect of reducing the sense of discomfort that occurs.

According to the present invention (claim 2), when the VOX mode information indicates a silent state, the audio signal input to the auditory weighting filter is subjected to LPF to improve the auditory quality of synthesized speech. In the adaptive spectrum post-filter used in (1), frequency enhancement of background noise is suppressed.

Further, according to the present invention (claim 3), the power quantizer averages the power value in a silent section for a long time to thereby obtain the power of the background noise generated by the decoder of the base station. The level has an effect of suppressing a sudden level change caused by sudden noise.

According to the present invention (claim 4), a search process can be omitted by outputting a fixed index, and the amount of calculation can be greatly reduced and power consumption can be reduced.

Further, according to the present invention (claim 5), in the adaptive codebook searcher, by outputting a fixed index, there is an effect that a periodic change of background noise is suppressed.

Furthermore, according to the present invention (claim 6), in the adaptive codebook searcher, the search processing can be omitted by outputting a fixed index. Power can be reduced.

According to the present invention (claim 7), each of the above effects can be achieved synergistically.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating a configuration of an embodiment of a speech encoding device according to the present invention.

FIG. 2 is a block diagram illustrating a configuration of a conventional speech encoding device.

[Explanation of symbols]

 Reference Signs List 1 input terminal 2 power quantizer 3 voiced silence detector 4 LSP quantizer 5 LPC analyzer 6 pitch analyzer 7 auditory weighting filter 8 LPF 9 auditory weighting synthesis filter 10 subtractor 11 adaptive codebook searcher 12 noise code Book searcher 13 Gain codebook searcher 14 VOX unique word generator 15 Data transmission unit 16 Output terminal 17 Switch

Claims (6)

(57) [Claims] A power quantizer that receives an audio signal, calculates and quantizes a power value of the audio signal, and outputs a power index and a quantized power value; Linear Predictive
Coefficient) is calculated and output, and L is converted from the LPC to an LSP (Line Spectrum Pair) and output.
A PC analyzer, and inputting the audio signal, performing a linear prediction inverse filter,
A pitch analyzer that outputs a plurality of lag candidates from the autocorrelation of a PC prediction residual signal, and a prediction residual gain that is output from the voice signal and an LSP quantizer to detect presence / absence of voice and silence. , A voiced / silent detector that outputs VOX mode information indicating a voiced state or a voiceless state, and a voice / silence detector in a parallel form. And the quantization L
A PC and the LSP quantizer for outputting the prediction error gain; a hearing device for inputting the LPC to calculate a filter coefficient for performing auditory weighting, inputting the audio signal, and outputting an auditory weighted audio signal A weighting filter, an auditory weighting synthesis filter that inputs the quantized LPC and the zero signal, and outputs a zero input response signal, a subtractor that subtracts the zero input response signal from an output signal of the auditory weighting filter, Of the quantizer output signal, the quantized LPC and the lag candidate, and a signal obtained by weighting and synthesizing each lag candidate, and an audio signal obtained by subtracting a zero-input response output signal from an auditory-weighted audio signal. An adaptive code that obtains a correlation, searches for an adaptive code vector signal that maximizes the correlation, and outputs an adaptive codebook index and an adaptive code vector signal. A book searcher, a signal obtained by inputting the output signal of the subtractor, the quantized LPC, and the adaptive code vector signal, performing a weighting synthesis by performing an orthogonalization process with the adaptive code vector signal for each noise code vector, and Finding the cross-correlation with the audio signal obtained by subtracting the quiescent response output signal from the auditory weighted audio signal, searching for a noise code vector signal that maximizes the cross-correlation, the noise codebook index and the noise code vector A noise codebook searcher that outputs a signal, an output signal of the subtractor, the adaptive code vector signal, the noise code vector signal, the power value, and the prediction residual gain, and an output signal of the subtractor. Searching for a gain codebook in which an error between the noise code vector signal obtained by adding the gain and the added value of the adaptive code vector signal is minimized. A gain codebook searcher for outputting a dex, a VOX unique word generator for outputting VOX unique pattern data, the power index, the adaptive codebook index, the noise codebook index, and the gain codebook index. , The LSP index, and the VOX
A VOX unique word output from a unique word generator and the VOX mode information are input and converted into a predetermined format, and a data transmitting unit that outputs encoded data ; Output with suppressed frequency components
And LPF (low pass filter) that, wherein the perceptual weighting filter, the audio signal and the LP
Either one of the output signals of F
Switching input based on the output VOX mode information,
A speech coding apparatus for outputting a sense-weighted speech signal . 2. The sound quantizer according to claim 2, wherein said power quantizer is said sound / silence detector.
Input the VOX mode information output from the
If the load information is in a silent state, the average
Output the power index calculated from the measured power value
The speech encoding device according to claim 1, wherein: 3. The sound / silence detector according to claim 1 , wherein
Input the VOX mode information output from the
When the card information is in the silent state, the LPC
The LSP quantizer outputs the output from the sound / silence detector.
Is input, and the VOX mode information is
In the case of silence, if the
Output the quantized LSP index and the quantized LPC
The speech encoding device according to claim 1, wherein 4. The voice codebook searcher according to claim 1, wherein said adaptive codebook searcher is a soundless / silence detector.
Input the VOX mode information output from the
When the code information is in a silent state, the adaptive codebook
The index is controlled to a unique value and output
The speech encoding device according to claim 1, wherein 5. The voice codebook searcher according to claim 1 , wherein
The VOX mode information output from the output device is input, and the VO
When the X mode information is in a silent state, the adaptive code
Performs search processing by controlling the issue book index to a unique value
5. The sound according to claim 4, wherein the sound is controlled so as not to exist.
Voice coding device. 6. An audio signal is inputted and a predetermined high frequency component is generated.
LPF (low-pass filter) to be attenuated , and either the audio signal or the LPF output of the audio signal
One is the VOX mode information output from the sound / silence detector.
Switch and input auditory weighted audio signal
An auditory weighting filter and VOX mode information output from the sound / silence detector.
If the VOX mode information is silent,
Output the power index calculated from the averaged power value.
Power quantizer and VOX mode information output from the sound / silence detector.
Input, and when the VOX mode information is in a silent state,
An LPC analyzer that controls and outputs a PC to a unique value, and VOX mode information that is output from the sound / silence detector.
Type, when the VOX mode information is in silence, L
Quantized LSP index when PC is set to a unique value
And an LSP quantizer that outputs a quantized LPC and the audio signal , performs a linear prediction inverse filter,
Output multiple lag candidates from autocorrelation of PC prediction residual signal
And a VOX mode information output from the sound / silence detector.
If the VOX mode information is in a silent state,
The codebook index is controlled to a unique value, and search processing is performed.
An adaptive codebook searcher controlled not to perform the search and a random codebook vector signal
Codebook searcher that outputs noise and random code vector signals
When the gain codebook search outputting a gain codebook index
And VOX unique which outputs VOX unique pattern data
A word generator, the power index, and the adaptive codebook index
And the noise codebook index and the gain codebook index.
Index, the LSP index, and the VOX
VOX unique word output from unique word generator
Mode and the VOX mode information, and
Output encoded data after converting to
And a data transmission unit .