JP3580906B2 - Voice decoding device - Google Patents

Description

[0001]
[Industrial applications]
The present invention relates to a speech decoding device in speech coded communication, and more particularly to a speech decoding device in which an adaptive difference PCM (ADPCM) method is applied to a coding method.
[0002]
[Prior art]
When voice communication is being performed, it is said that the time rate at which one of the callers is speaking is about 35%.
2. Description of the Related Art In recent years, the range of personal communication, which is communication performed by individuals, has been expanding. Here, voice communication using a portable terminal is mainly performed. First, cordlessness is one of the items required for such a mobile terminal. Secondly, a battery is used for portability and it is necessary to withstand use for a long time, so that reduction in circuit power consumption is required.
[0003]
As a conventional method for reducing the circuit power consumption, there is a method in which a transmission circuit is operated only when a voice is being uttered while paying attention to a voice utterance time rate, and the circuit is in a rest state during other transmission times. In order to realize such a technique, it is only necessary to provide a voice detection function on the transmission side and add a discontinuous transmission device that transmits only a sound section. In that case, the problem is on the receiving side. That is, on the receiving side, the reproduced sound is intermittent, resulting in a very unpleasant sound. This is because background noise is superimposed on the sound when the sound is being transmitted, but is not transmitted when there is no sound. That is, it is known that background noise is modulated and transmitted only when there is an audio signal.
As a method of solving such a problem, a method of generating a pseudo background noise similar to the background noise on the transmission side in a silent section in which no audio signal is transmitted on the reception side is known.
[0004]
Although the present invention is intended for a decoder on the receiving side, the decoder on the receiving side has the same components as the encoder on the transmitting side, so the outline of the ADPCM encoder on the transmitting side will be described. .
FIG. 5 is a block diagram of the transmission side encoder. In the figure, reference numeral 51 denotes a uniform PCM converter for converting a 64 kb / s μ-law PCM input signal into a linear 13-bit PCM.
Reference numeral 52 denotes a subtractor 52 that subtracts the prediction signal j output from the adaptive predictor 53 from the output of the uniform PCM converter 51 to obtain a difference signal k. The difference signal k is quantized by the adaptive quantizer 54, and 32 kb / s voice data is transmitted to the transmission path as an output of the ADPCM voice coder.
On the other hand, the adaptive inverse quantizer 56 performs adaptive inverse quantization on the output audio data of 32 kb / s and outputs a quantized difference signal m. The adder 55 adds the quantized difference signal m and the prediction signal j and outputs a reproduced signal n.
The adaptive predictor 53 calculates a prediction coefficient and uses it to generate and output a prediction signal j from the quantized difference signal m and the reproduction signal n.
[0005]
The prediction coefficient calculated by the adaptive predictor 53 to generate the prediction signal j is a coefficient for predicting a sample value at a certain point in time using two adjacent past sample values, and the value represents a timbre. For example, the prediction coefficients have different occurrence distributions in the case of a speech signal and in the case of background noise, and even with the same background noise, a slope such as white noise having flat frequency characteristics and pink noise which is generally well known is used. The distribution of occurrence is different even from noise having
As an example, FIG. 6 is a waveform example diagram of an input signal and a prediction coefficient, and shows a temporal change of a1 and a2 and a speech of the prediction coefficient calculated by the adaptive predictor 53. It can be seen that the values of the prediction coefficients are different between the section with speech in sections A, B and C and the section with only background noise in sections D and E. Here, the background noise superimposed on the voice is, for example, noise of an actual air conditioner (air conditioner) having an inclined frequency characteristic.
Both the prior art and the present invention focus on this prediction coefficient as a means for generating pseudo background noise.
An example will be described.
[0006]
FIG. 4 is a block diagram of a conventional decoding device.
In FIG. 4, reference numeral 1 denotes an antenna. Reference numeral 2 denotes a reception demodulator for receiving and demodulating a modulated wave obtained by multiplexing and modulating ADPCM coded data with a voice detection flag of a voice detector for judging the presence / absence of voice from an encoder on the transmission side. Reference numeral 3 denotes a demultiplexer that separates the demodulated signal a into ADPCM encoded data c and a voice detection flag b. Reference numeral 4 denotes a controller which receives a voice detection flag b from the demultiplexer 3 and outputs a control signal for performing control to be described later to the power holder, the ADPCM decoder 5, and the prediction coefficient holder 6. Numeral 5 receives the control signal d from the controller 4, generates random ADPCM data for pseudo background noise, transmits and receives the prediction coefficient data to and from the prediction coefficient holder 6, and outputs the ADPCM encoded data from the demultiplexer 3. ADPCM decoder for decoding c. Numeral 6 calculates the average value of the prediction coefficient g of the ADPCM decoder 5 for each frame, updates the storage of the average value in the voiced section by the control signal e from the controller 4, and suspends the update in the silent section. It is a prediction coefficient retainer. 8 is a speaker.
[0007]
The operation will be described with reference to FIGS.
First, in FIG. 4, a modulated wave obtained by multiplexing and modulating a speech detection flag and ADPCM encoded data is received by an antenna 1, received and demodulated by a reception demodulator 2, and a demodulated signal a is sent to a demultiplexer 3. The speech detection flag referred to here indicates a result of detection of a portion of the input speech of the encoder on the transmitting side that has speech (speech section) and a section that does not (speech section) by the speech detector.
The demultiplexer 3 separates the signal into a voice detection flag b and ADPCM encoded data c. In this case, as an example, the ADPCM encoded data c is set to 5 msec as one frame. The voice detection flag b is sent to the controller 4 every 5 msec.
Upon receiving the voice detection flag b, the controller 4 outputs control signals d and e indicating either a sound section or a silent section.
The ADPCM encoder 5 internally generates random data within a range that can be taken by the ADPCM encoded data c in order to generate pseudo background noise because the modulated signal is interrupted in a silent section, and stores the data in a prediction coefficient. The decoding is performed using the prediction coefficient f given from the unit 6.
[0008]
The prediction coefficient holding unit 6 extracts a prediction coefficient g in the ADPCM decoder 5 in order to add spectrum information of actual noise to the pseudo background noise generated by the ADPCM decoder 5 when transmission is interrupted, and The average value is obtained and updated and held. When the control signal e changes from the voiced section to the voiceless section, the update is stopped while retaining the average value of the prediction coefficient of the immediately preceding frame, that is, the last frame of the voiced section.
The ADPCM decoder 5 stops the update, decodes the internally generated random data using the held prediction coefficient f, and outputs pseudo background noise.
[0009]
The speech detector on the encoder side has a function generally called hangover, which determines a voiced section for about several hundred msec when changing from a voiced section to a silent section, in order to prevent speech ending. Since it is provided, the prediction coefficient value of the last frame of the sound section becomes the value of the background noise. With this effect, even if decoding is performed using random data, noise having a timbre similar to the background noise superimposed on the actual speech of the encoder side is generated.
[0010]
FIG. 7 is a waveform example illustrating the prediction coefficient for the pseudo background noise, and shows a case where there is no disturbance. In the figure, (A) shows an input speech signal (speech + background noise) of an encoder, and (B) shows a temporal change of a prediction coefficient a1 as an example among prediction coefficients corresponding thereto.
Assuming that speech is ideally detected as in the speech detection flag (C) in the speech detection processing on the encoder side, the encoder intermittently transmits only the audio data of the audio signal (D). (E) is a prediction coefficient a1 for the audio signal (D).
In the prior art, in order to insert the pseudo background noise into the sections a, b, and c as in the audio signal (D), the frame immediately before the change from the voiced section to the silent section, here, held in the prediction coefficient holding unit 6 The random data of the sections a, b, and c are decoded using the average value (F) of the prediction coefficients a1 of t, t-1, and t-2, respectively, and pseudo background noise is output. That is, the prediction coefficient used for decoding by the decoder 5 is like the prediction coefficient a1 (G).
In this state, the timbre of the pseudo-background noise in the silent sections a, b, and c is the prediction coefficient value of the actual background noise having substantially the same value in the sections a, b, and c as the prediction coefficient a1 (F). , A pseudo background noise which is continuous without any discomfort.
[0011]
[Problems to be solved by the invention]
Next, a problem of the related art will be described with reference to FIG. FIG. 8 is an example of a waveform of a prediction coefficient when there is a disturbance.
When a suddenly changing disturbance or the like is added to the last background noise frame (t-1) of the sound section, for example, when an abnormal sound enters the microphone on the encoder side or when an error occurs on the transmission path Etc., or when the received voice detection flag indicates an erroneous determination result, for example, when a voice detector on the encoder side makes a determination error and determines that there is a silent section during a voiced section, or transmission When an error occurs in the voice detection flag on the road or the like, as shown by the input voice signal (A), the prediction coefficient (B), and the voice detection flag (C), the last frame t-1 of the voiced section is , Input speech (A) and prediction coefficient a1 (B) are in different levels from the background noise.
In the conventional method, the pseudo background noise of the subsequent silent section is generated using only the average value of the prediction coefficients of the last frame of the sound section. That is, since the average value of the Z portion of the prediction coefficient a1 (E) is used as the prediction coefficient of the section b, the section a and the section c are considerably different from each other, such as the difference x and y of the prediction coefficient a1 (F). Therefore, the tone color of the pseudo background noise in the section b is different, and in the sections a, b, and c, there is a disadvantage that the continuity is unnatural as shown in (G).
[0012]
SUMMARY OF THE INVENTION An object of the present invention is to reduce the adverse effect on pseudo-background noise generation due to short-time intermittent reception, which is a problem of the prior art, and to make the last part of a voiced section unvoiced in the ADPCM speech coding system. It is an object of the present invention to provide a speech decoding apparatus in which reproduced pseudo background noise does not cause a sense of incongruity even if there is a disturbance immediately preceding the level or background noise, which has a large difference in timbre.
[0013]
[Means for Solving the Problems]
The speech decoding apparatus according to the present invention decodes the adaptive difference PCM coded signal using a prediction coefficient in a voiced section of the received adaptive difference PCM coded signal and outputs the decoded signal as a reproduced voice. An adaptive difference PCM decoder for decoding internally generated random codes and outputting pseudo-background noise as reproduction noise in a silent section, and a prediction coefficient holder for obtaining and updating an average value of the prediction coefficients for each frame In the audio decoding device provided with
A prediction coefficient memory for sequentially updating and storing the average value of the prediction coefficient for each frame obtained by the prediction coefficient holder over a plurality of times in the past, and an average value of a plurality of prediction coefficients stored in the prediction coefficient memory over a plurality of times in the past And an auxiliary prediction coefficient holding unit for calculating a prediction coefficient of a silent section for the adaptive difference PCM decoder.

[0014]
Further, the speech decoding apparatus of the present invention decodes the adaptive difference PCM coded signal using a prediction coefficient in a voiced section of the adaptive difference PCM coded signal and outputs the decoded signal as a reproduced voice, and furthermore, decodes the prediction coefficient in a silent section. An adaptive difference PCM decoder for decoding internally generated random codes and outputting pseudo-background noise as reproduction noise in a silent section, and a prediction coefficient holder for obtaining and updating an average value of the prediction coefficients for each frame In the audio decoding device provided with
A first multiplier for multiplying the average of the prediction coefficients obtained for each frame by the prediction coefficient holder by a positive number less than 0.5 as a first multiplication value, and an input value from the first multiplier An adder for adding the other input value to the prediction coefficient to be given to the prediction coefficient holder, a prediction coefficient memory for updating and storing the output of the adder, and an output value of the prediction coefficient memory. A second multiplier configured to multiply a value obtained by subtracting the first multiplied value from 1 as a second multiplied value and use the multiplied value as the other input value of the adder.
[0015]
【Example】
FIG. 1 is a block diagram showing an embodiment of the present invention. In FIG. 1, 1 is an antenna, 2 is a receiving demodulator for receiving and demodulating a modulated wave obtained by multiplex-modulating a voice detection flag and ADPCM coded data, and 3 is a receiving demodulation signal a for converting ADPCM coded data c and a voice detection flag b. The demultiplexer 4 receives the audio detection flag b and outputs the control signals d and e. The controller 5 and the ADPCM decoder 5 are the same as those of the conventional circuit.
6 is a predictive coefficient retainer, which extracts a predictive coefficient g which is an internal variable of the ADPCM decoder 5, calculates an average value for each frame, stores and updates the average value, and stops updating the silent section for a sound section. I do. Reference numeral 7 denotes an auxiliary prediction coefficient holder added in the present invention, and reference numeral 8 denotes a speaker.
[0016]
〔motion〕
The operation is the same as that of the prior art except for the operations of the prediction coefficient holder 6 and the auxiliary prediction coefficient holder 7, and therefore, the points different from the prior art will be described in detail.
The prediction coefficient holding unit 6 extracts a prediction coefficient g calculated in the ADPCM decoder 5 in accordance with the control signal e, calculates an average value for each frame, performs holding and updating, and updates the speech section from the speech section to the silent section. , The updating is stopped, and the value i of the last frame is output to the auxiliary prediction coefficient holding unit 7.
Each time the value i of the last frame calculated by the prediction coefficient holder 6 is input, the auxiliary prediction coefficient holder 7 uses the prediction coefficient value immediately before the change from the past voiced section to the silent section. A process is performed on the prediction coefficient value for generating the pseudo background noise to reduce the influence of the problem of the related art. For example, weighting the data of the last frame in the past to suppress the influence of the current data, or performing a process of weighting the past data by averaging the data of the past several frames to perform prediction coefficient holding The processed value h is output to 6. The prediction coefficient holder 6 outputs this value h as f in a silent section, and the ADPCM decoder 5 generates pseudo background noise using this value f.
[0017]
FIG. 2 is a block diagram showing a first embodiment of the auxiliary prediction coefficient holding unit 7 which is a main part of the present invention. In FIG. 2, reference numeral 21 denotes an average calculator, and reference numeral 22 denotes a prediction coefficient memory. The average value calculator 21 calculates the average value of the prediction coefficients of the final frame in the memory 22 for a plurality of times, and outputs the average value h to the prediction coefficient holder 6. The prediction coefficient memory 22 sequentially stores and updates past prediction coefficient values for a plurality of final frames, for example, five frames from t to t-4.
[0018]
〔motion〕
When a change from a voiced section to a silent section is detected, the auxiliary prediction coefficient holding unit 7 extracts the prediction coefficient average value i of the last frame from the prediction coefficient holding unit 6. The memory 22 updates and stores the value in the memory each time the prediction coefficient average value of the last frame is input. Here, t is the value of i that is currently input, t-1 is the value that was input one time in the past, and similarly, a total of five values up to t-4 are stored, and a new value i is input. Each time, each value is sequentially updated to the past memory. Next, the average calculator 21 calculates the average of the five values output from the memory 22 and supplies the average h to the prediction coefficient holder 6. The prediction coefficient holder 6 gives the value h to the decoder 5 as the value f.
As a result of this processing, the prediction coefficient of the area b in FIG. 8F becomes as indicated by a broken line l, and in the area c, the prediction coefficient smoothly approaches a normal value as indicated by the broken line k, and is affected by Z as shown in FIG. Also, the conventional difference x becomes m and y becomes n, and the sense of incongruity is extremely reduced without any extreme difference between the regions a, b, and c.
[0019]
Next, FIG. 3 is a block diagram showing a second embodiment of the auxiliary prediction coefficient holding unit 7 which is a main part of the present invention. In FIG. 3, 31 is an adder, 32 and 33 are multipliers, and 34 is a prediction coefficient memory. In the multiplier 32, a multiplication value α of 0 <α <0.5 is set, and the weight of the prediction coefficient average value i of the currently input last frame is lightened. The multiplier 33 is set with a multiplication value β that satisfies 1−α = β, and acts to increase the weight of the previous prediction coefficient average value h output from the adder 31. For example, when a multiplication value of α = 0.05 and β = 1−α = 0.95 is set, the adder 31 calculates 5% of the current input value i and 95% of the output value h from the previous adder 31. % Is added and output. The prediction coefficient memory 34 sequentially stores and updates the output value h of the adder 31 and provides the stored value p to the multiplier 33 to be multiplied by the multiplied value β to be used as one input of the next adder 31.
[0020]
When the interval changes from a voiced section to a silent section, the auxiliary prediction coefficient holder 7 extracts the prediction coefficient value i of the last frame from the prediction coefficient holder 6. The following equation (1) is calculated in the adder 31, the multipliers 32 and 33.
(Equation 1)
h = p × 0.95 + i × 0.05 (1)
That is, h is a prediction coefficient value including the previous value of i, and is not greatly affected by the currently input value of i. This means that even if the value of i at a certain point changes abruptly, it does not significantly affect the value of h that is output, and settles to an appropriate value. The prediction coefficient memory 34 updates and stores the new h. Here, the initial value is set to 0 (settles at an appropriate value in a short time).
By this processing, an effect is obtained that a rapid change of the present value is suppressed by the past value as in the first embodiment of FIG.
[0021]
【The invention's effect】
As described in detail above, by generating the pseudo background noise using the prediction coefficient of the last frame in the past, even when a signal or the like that changes rapidly for a short time enters, the influence of the signal can be reduced. Since it is possible to generate pseudo background noise having almost the same spectrum information as background noise superimposed on voice without receiving it, the sense of discomfort given to the listener is reduced.
[Brief description of the drawings]
FIG. 1 is a configuration example diagram of a speech decoding device according to an embodiment of the present invention.
FIG. 2 is a configuration example diagram showing a first embodiment of the auxiliary prediction coefficient holder of FIG. 1 of the present invention.
FIG. 3 is a configuration example diagram showing a second embodiment of the auxiliary prediction coefficient holder of FIG. 1 of the present invention.
FIG. 4 is a diagram illustrating a configuration example of a conventional speech decoding device.
FIG. 5 is a block diagram of a transmission-side encoder using the ADPCM encoding method.
FIG. 6 is a waveform example of an input signal and a prediction signal.
FIG. 7 is a waveform example illustrating a prediction coefficient for pseudo background noise.
FIG. 8 is a waveform example illustrating a prediction coefficient for pseudo background noise when there is disturbance;
[Explanation of symbols]
REFERENCE SIGNS LIST 1 antenna 2 reception demodulator 3 demultiplexer 4 controller 5 ADPCM decoder 6 prediction coefficient holder 7 auxiliary prediction coefficient holder 8 speaker a demodulated signal b audio detection flag c ADPCM coded signal e control signal f pseudo loss noise Predicted coefficient value g Predicted coefficient 21 Average value calculator 22 Predicted coefficient memory 31 Adder 32 Multiplier 33 Multiplier 34 Predicted coefficient memory 51 Uniform PCM converter 52 Subtractor 53 Adaptive predictor 54 Adaptive quantizer 55 Adder 56 adaptive inverse quantizer j prediction signal m quantization difference signal n reproduced signal

Claims (2)

In the voiced section of the received adaptive differential PCM coded signal, the adaptive differential PCM coded signal is decoded using a prediction coefficient and output as a reproduced voice. An audio decoding apparatus comprising: an adaptive difference PCM decoder that decodes a code and outputs a pseudo background noise as reproduction noise in a silent section; and a prediction coefficient holder that obtains an average value of the prediction coefficients for each frame and updates and holds the average.
A prediction coefficient memory for sequentially updating and storing the average value of the prediction coefficient for each frame obtained by the prediction coefficient holder over a plurality of past times; and an average value of a plurality of prediction coefficients stored over the past several times in the prediction coefficient memory. And a supplementary prediction coefficient holding unit for obtaining a prediction coefficient of a silent section for the adaptive difference PCM decoder. In the voiced section of the received adaptive differential PCM coded signal, the adaptive differential PCM coded signal is decoded using a prediction coefficient and output as a reproduced voice. An audio decoding apparatus comprising: an adaptive difference PCM decoder that decodes a code and outputs a pseudo background noise as reproduction noise in a silent section; and a prediction coefficient holder that obtains an average value of the prediction coefficients for each frame and updates and holds the average.
A first multiplier for multiplying the average of the prediction coefficients obtained for each frame by the prediction coefficient holder by a positive number less than 0.5 as a first multiplication value, and an input value from the first multiplier An adder that adds the other input value to the prediction coefficient to be given to the prediction coefficient holding unit, a prediction coefficient memory that updates and outputs an output of the adder, and an output value of the prediction coefficient memory. A second multiplier that multiplies a value obtained by subtracting the first multiplied value from 1 as a second multiplied value and uses the multiplied value as the other input value of the adder.

Images