JPH0870285A - Voice decoder - Google Patents

Abstract

PURPOSE: To improve a problem of a pseudo background noise for a silence interval of receiver side causing unnatural sound quality when a sound detection flag is outputted by mistake for the silence interval from the sender side in the receiver side voice decoder for the adaptive differential PCM coding system communication. CONSTITUTION: An auxiliary controller 6 is provided between a controller 5 receiving a sound detection flag (b) separated from a reception demodulation signal (a) to provide an output of a signal (e) representing discrimination of sound/silence and a reset signal (d) at a change point of the sound/silence and a predict,ion coefficient storage device 7 extracting a prediction coefficient (g) from an ADPCM decoder 4 decoding a coded signal (c), calculating an average for each frame, storing and updating it and providing an output of a stored value (h) just before the change from the silence interval into the sound interval. The auxiliary controller 6 outputs continuously a flag (f) representing the silence interval for nearly 5 frames even when the input flag (e) indicates the change of the silence interval into the sound interval so that the sound quality of pseudo background noise generated and decoded by the decoder 4 for the silence interval is not disturbed.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a speech decoding device, and more particularly to an adaptive differential PCM (ADPCM) coding system.
The present invention relates to a voice decoding device that decodes a received signal that is compatible with the method.

[0002]

2. Description of the Related Art When voice communication is performed, it is said that the time rate during which either one is uttering is about 35%. In recent years, the range of personal communication, which is a communication mainly based on individuals, has been expanding. There, mainly voice communication using a portable terminal is used. As a matter required for such a mobile terminal, firstly, there is a cordless method. Secondly, it is required to reduce the circuit power consumption because a battery is used for convenience of carrying and it is required to endure long-term use.

As a method of reducing the circuit power consumption more than ever before, there is a method of paying attention to the utterance time rate of voice, operating the transmission circuit only when uttering, and putting the circuit in a rest state during the other transmission time. In order to realize such a technique, a voice detection function may be provided on the transmission side and a discontinuous transmission device may be added. In that case, the problem is on the receiving side. That is, the reproduced sound is intermittent on the receiving side, resulting in a very unpleasant sound. The reason for this is that background noise is superimposed on the voice when transmitting the voice, but when the voice is not transmitted, the background noise is not transmitted, and the background noise is modulated and transmitted only when the voice signal is continuous. It is known that this is because.

As a method of solving such a problem,
It is known that the receiving side produces pseudo noise similar to the background noise on the transmitting side while no voice signal is being transmitted. FIG. 3 is a block diagram of a conventional speech decoding apparatus for explaining the above example. In FIG. 3, 1 is an antenna,
Reference numeral 2 denotes a reception demodulator, which includes a voice detection flag indicating whether or not voice is detected by the voice detector of the transmission side encoder and AD.
A modulated wave that is modulated by being multiplexed with PCM encoded data is received and demodulated. A demultiplexer 3 demultiplexes the received and demodulated signal a into ADPCM encoded data c and a voice detection flag b. Reference numeral 5 denotes a controller, which receives a voice detection flag b from the demultiplexer 3 and receives a control signal (reset pulse) d.
And a control signal e indicating sound / silence is output.

Reference numeral 4 denotes an ADPCM decoder, the prediction coefficient of which is reset to "0" by the control signal d from the controller 5, and the ADPCM coded data c from the demultiplexer 3 is decoded in the voiced section to produce silence. In the section, the random ADPCM coded data internally generated for pseudo background noise is decoded by the prediction coefficient input from the prediction coefficient holder 7. 7
Is a prediction coefficient holder, which extracts a prediction coefficient which is an internal variable of the ADPCM decoder 4, calculates an average value for each frame, stores and updates it, and outputs a control signal e from the controller 5.
Therefore, in the silent section, the prediction coefficient immediately before that is retained while A
It is given to the DPCM decoder 4. Reference numeral 8 is a speaker.

FIG. 4 is a time chart showing the source sound on the transmitting side and the signal waveform of each part of the speech decoding apparatus. At the top of the figure is the waveform of the source sound on the transmission side, in which voice and background noise are superimposed. The second stage is a voice detection flag b sent from the transmitting side and separated by the demultiplexer 3, in which the voiced section is "1" and the silent section is "0". In this example, the voice is erroneously shown for a short period of time during the silent section. The third stage shows the reset signal d output from the controller 5. The x in the fourth row is a quadratic prediction coefficient a among the prediction coefficients for the source sound (voice + background noise).
It is a waveform which shows the time change of l (t). The next j is
It is a waveform showing the temporal change of the prediction coefficient al (t) of the ADPCM decoder 4, and the voiced section is the same as x, but the silent section is input from the prediction coefficient holder 7 and is a prediction coefficient for decoding the pseudo background noise. Shows the change over time.

The operation of the conventional device will be described with reference to FIGS. 3 and 4. The modulated wave in which the voice detection flag and the ADPCM coded data are multiplexed is received by the antenna 1, received and demodulated by the reception demodulator 2, and the demodulated signal a is sent to the demultiplexer 3. The voice detection flag mentioned here is a signal indicating the result of detection by the voice detector of a portion (voiced section) of the input voice of the transmission side encoder and a portion (voiceless section) of the input voice. In the demultiplexer 3, the voice detection flag b and the ADPCM encoded data c
To separate. In this case, the ADPCM coded data c is 5 ms
ec is one frame. The voice detection flag b is sent to the controller 5 every 5 msec.

Upon receiving the voice detection flag b, the controller 5 sends a control signal (reset) to the ADPCM decoder 4 at the time of transition from the voiced section to the voiceless section and at the transition from the voiceless section to the voiced section. Pulse) d is output. This control signal d is for initializing a predetermined variable such as a prediction coefficient in the ADPCM decoder 4, and here, the transmission side encoder at the time of transmission stop due to voice detection and at the time of transmission start In order to make the state of 1 and the state inside the receiving side decoder the same, a reset is applied at each transition. Without this reset, the internal states of the encoder and the decoder are different after transmission interruption due to voice detection, and the sound quality of the reproduced voice decoded by the decoder is deteriorated. The control signal d resets the ADPCM decoder 4. Further, the ADPCM decoder 4 internally generates random data within a range that the ADPCM coded data c can take in order to generate pseudo background noise by itself because the modulated signal is cut off in the silent section, and the data is used as a prediction coefficient. Decoding is performed using the prediction coefficient input from the holder 7.

The prediction coefficient holder 7 is used for ADPC when transmission is interrupted.
In order to add the spectrum information of the actual noise to the pseudo background noise generated by the M decoder 4, the prediction coefficient in the ADPCM decoder 4 is extracted, the average value is calculated for each frame, updated and held, and the control signal When a silent section is detected by e, the average value of the prediction coefficients of the last frame of the sound section is held and given to the pseudo background noise. With this effect, even if decoding is performed using random data, pseudo background noise having a tone color similar to the background noise superimposed on the actual speech on the encoder side is generated.

[0010]

As described above, x in FIG. 4 represents the temporal change of the second-order prediction coefficient a1 (t) of the prediction coefficients for the source sound (voice + background noise). When the voice detection flag is erroneous due to the determination error of the voice detector or the influence of fading on the transmission path, and the B section in the unvoiced section is processed as the voiced section like the voice detection flag b in FIG. Reset is applied at the end of section A of j. The prediction coefficient which has become “0” by this reset rises to a value for background noise in the B section. Here, if the B section becomes a silent section after an extremely short section (for example, 5 frames), in the next C section (silent section),
Since the pseudo background noise is generated by holding the value before the prediction coefficient rises, the noise becomes different in tone color from the actual background noise in the section C, and the listener feels uncomfortable.
That is, there is a drawback that the pseudo background noises in the section A and the section C in FIG.

It is an object of the present invention to provide a speech decoding apparatus which eliminates the unnaturalness of the pseudo background noise inserted in the reproduced speech signal on the receiving side, which is a problem of the prior art in the ADPCM speech coding system. is there.

[0012]

According to the speech decoding apparatus of the present invention, a coded speech signal subjected to adaptive differential PCM coding and a speech detection flag indicating one of a voiced section and a silent section of voice are multiplexed and modulated. A reception demodulator that receives and demodulates a signal transmitted only in a voiced section, a demultiplexer that demultiplexes the demodulated signal from the reception demodulator into the encoded voice signal and the voice detection flag, and the demultiplexer And a controller that outputs the voice detection flag as it is and outputs a reset signal each time the voice detection flag changes from a voiced section to a silent section or from a voiced section to a voiced section, and by the reset signal during the voiced section. The encoded voice signal from the demultiplexer is decoded by the reset prediction coefficient rising from 0, and the reproduced voice signal is output. During the silent period, the voiced segment immediately before the silent segment is output. An adaptive difference PCM decoder that decodes a random code internally generated by a prediction coefficient of the background noise and outputs pseudo background noise, and a voice detection flag from the controller are input, and the adaptive difference is applied during a sound section. PCM
The prediction coefficient is fetched from the decoder, the average value is calculated for each frame and updated and stored, and during the silent section, the update of the prediction coefficient average value is stopped and the predicted coefficient average value immediately before entering the stored silent section is stored. In a speech decoding device provided with a prediction coefficient holder which is given to the adaptive differential PCM decoder as a prediction coefficient of background noise, a received speech detection flag erroneously indicates a voiced section for a short time when there is no speech. In order to reduce the change in the sound quality of the pseudo background noise when
Between the controller and the prediction coefficient holder, the voice detection flag from the controller is output as it is during the voiced section continuation and the voiced section continuation, and when the voiced section is changed to the voiced section, In the adaptive differential PCM decoder, the time until the value of the prediction coefficient reset to 0 rises and becomes substantially equal to the value of the prediction coefficient of the silent section immediately before the change to the voiced section, the voice detection flag indicating the silent section. Is provided to the prediction coefficient holder, and an auxiliary controller for keeping the storage and updating of the prediction coefficient is provided.

[0013]

1 is a diagram showing an example of the configuration of a speech decoding apparatus according to the present invention. In FIG. 1, an antenna 1, a reception demodulator 2, a demultiplexer 3, an ADPCM decoder 4, a controller 5, a prediction coefficient holder 7, and a speaker 8 are the same as in the conventional case. 6 is a controller 5
It is an auxiliary controller that controls the operation of the prediction coefficient holder 7 by receiving a control signal from. Reference numeral 7 extracts the prediction coefficient g which is an internal variable of the ADPCM decoder 4, calculates an average value for each frame, and stores and updates the average value by the control signal f from the controller 5 and the auxiliary controller 6. And the value h is ADP
This is a prediction coefficient holder provided to the CM decoder 4.

Next, the operation of the embodiment of the present invention shown in FIG. 1 will be described. Other than the operations of the auxiliary controller 6 and the prediction coefficient holder 7, it is the same as the conventional technique, and therefore the difference from the conventional technique will be described with reference to FIG. When the auxiliary controller 6 receives the control signal e indicating the presence or absence of the sound according to the voice detection flag from the controller 5, the auxiliary controller 6 switches from the sound section to the sound section until the reset prediction coefficient rises. During the period, the holder update signal f for keeping the updating of the prediction coefficient holder 7 stopped, that is, the voice detection flag indicating the silent section is output to the prediction coefficient holder 7. Prediction coefficient holder 7
Performs update / stop operation according to the retainer update signal f.

FIG. 2 is a detailed block diagram of the auxiliary controller 6 which is a main part of the present invention. In the figure, 61 is a control signal e
In response, the counter controller updates and resets the voiced section counter 62. Reference numeral 62 is a voiced section counter that counts voiced sections. Reference numeral 63 is a comparator that compares the count value of the sound counter with the value of a predetermined constant and outputs the result to the holder update controller 64. Reference numeral 64 is a retainer update controller that outputs a retainer update signal f in response to the comparison result of the comparator 63.

Next, the operation of the auxiliary controller 6 will be described. The auxiliary controller 6 indicates during the voiced section and during the voiced section, except for a short time when the control signal e (voice detection flag) input from the controller 5 changes from the voiceless section to the voiced section. In the meantime, the cage update signal f corresponding to the voice detection flag e is output. Further, when the soundless section is changed to the soundable section, counting of the number of frames is started from the time when it is changed to the soundable section, and a signal indicating the silence section is continuously output for a short time, for example, within 5 frame sections, to the holder update signal. Output as f. Therefore, in the prediction coefficient holder 7 that receives the holder update signal f, 5
The memory update of the prediction coefficient remains suspended during the frame.

As the time period for which the retainer update signal f indicating the silent section is subsequently output, the adaptive differential PCM decoder 4 is used.
At the time when the value of the prediction coefficient reset to 0 rises and becomes substantially equal to the value of the prediction coefficient of the background noise in the silent section immediately before changing to the sound section, the time is set. The distribution and level of the prediction coefficient when encoding and decoding the background noise differ depending on the tone color of the background noise, and generally the value of the prediction coefficient of the background noise is not so large, so this setting time is 5 frames (25 msec). The degree is good, but when the value of the prediction coefficient is large due to the timbre of the background noise, the effect is good when it is set to 10 frames (50 msec). However, when setting a time longer than this, it is necessary to consider the effect of speech break in the voiced section.

The detailed circuit of FIG. 2 will be described below.
The counter controller 61 that receives the control signal e outputs the counter update signal k while the frame-based control signal e indicates the voiced section, and sequentially updates the voiced section counter 62 by one, and outputs the silent section. Is displayed, the counter reset signal m is output to reset the voiced section counter 62 to "0". Then, the sound section counter 62 outputs the counter value to the comparator 63 each time. In the comparator 63, until the value of the prediction coefficient reset to 0 in the adaptive difference PCM decoder 4 rises and becomes almost equal to the value of the prediction coefficient of the background noise in the silent section immediately before the change to the sound section. A numerical value indicating time is set as a comparison constant. For example, 5 is set for 5 frames (25 msec).
Therefore, in the comparator 63, the input counter value is 0 to 5
It is determined whether it is within the range or exceeds 5, and the comparison result is output to the holder update controller 64.

In accordance with the above comparison result, the cage update controller 64 outputs a cage update signal f indicating a voiced section when the result exceeds 5 to update and store the prediction coefficient of the prediction coefficient holder 7. If it is within the range of 0 to 5, the holder update signal f indicating the silent section is output to stop the update of the prediction coefficient of the prediction coefficient holder 7. That is, the prediction coefficient holder 7
Holds the prediction coefficient for 5 frames (25 msec) even during the silent section and when the silent section is changed to the sound section.
That is, since the retainer update signal f as shown in (f) of FIG. 4 is output, the update is stopped during the section B of the conventional j, and the prediction coefficient of the silent sections A and C is shown as h. Since the value of becomes the same value continuously, and the background noise does not change, annoying noise is eliminated.

[0020]

As described above in detail, even if a voice detection flag indicating an erroneous voiced section is received, the prediction coefficient held for a time period taking into account the rise time of the prediction coefficient due to reset is simulated. By giving noise, it is possible to solve the hearing problem.

[Brief description of drawings]

FIG. 1 is a diagram showing a configuration example of a speech decoder according to the present invention.

FIG. 2 is a diagram showing a configuration example of an auxiliary controller forming a main part of the present invention.

FIG. 3 is a diagram showing a configuration example of a conventional speech decoder.

FIG. 4 is a time chart showing a time change of each signal.

[Explanation of symbols]

 1 antenna 2 reception demodulator 3 demultiplexer 4 ADPCM decoder 5 controller 6 auxiliary controller 7 prediction coefficient holder 8 speaker a demodulation signal b voice detection flag c ADPCM coded signal d control signal e control signal f holder update Control signal g prediction coefficient h held prediction coefficient i prediction coefficient k counter calculation signal m counter reset signal 61 counter controller 62 voiced section counter 63 comparator 64 holder update controller

Claims (1)

[Claims] 1. A received signal is demodulated by receiving a signal in which a coded voice signal subjected to adaptive differential PCM coding and a voice detection flag indicating any one of a voiced section and a voiceless section of voice are multiplexed and transmitted only in a voiced section. A receiving demodulator, a demultiplexer that separates the demodulated signal from the receiving demodulator into the encoded voice signal and the voice detection flag, and the voice detection flag from the demultiplexer that is output as it is A controller that outputs a reset signal each time the detection flag changes from a voiced section to a silence section, and from a silence section to a voiced section, and the demultiplexing by a prediction coefficient rising from 0 reset by the reset signal during the voiced section. The encoded voice signal from the device is decoded and the reproduced voice signal is output.In the silent period, it is generated internally by the prediction coefficient of the background noise in the voiced period immediately before the silent period. An adaptive differential PCM decoder that decodes a random code and outputs pseudo background noise, and a voice detection flag from the controller are input, and the prediction coefficient is fetched from the adaptive differential PCM decoder during a sound period. The average value is calculated and updated and stored for each frame, the update of the prediction coefficient average value is stopped during a silent section, and the average value of the prediction coefficient immediately before entering the stored silent section is used as the prediction coefficient of the background noise. In a voice decoding device provided with a prediction coefficient holder for giving to a PCM decoder, a pseudo background noise when a received voice detection flag is an error flag indicating a voiced section by mistake for a short time when there is no voice In order to reduce the change in sound quality, the voice detection flag from the controller is output as it is between the controller and the prediction coefficient holder during the voiced section and the silent section. At the same time, when the silent section is changed to the voiced section, the value of the prediction coefficient reset to 0 in the adaptive difference PCM decoder rises to almost the same value as the prediction coefficient of the silent section immediately before the change to the voiced section. A speech decoding apparatus, characterized in that an auxiliary controller is provided for giving a speech detection flag indicating a silent section to the prediction coefficient holder for keeping the storage and updating of the prediction coefficient stopped until the time becomes equal.