JPH07115403A - Circuit for encoding and decoding silent section information - Google Patents

Abstract

PURPOSE:To prevent the degradation in quality of a reproduced signal and to eliminate and unnecessary processing by selecting a minimum error pattern corresponding to the background noise at the time of silence and adding it to the last of a post pattern to send it. CONSTITUTION:At the time of silence, a frequency characteristic extractor 13 extracts the background noise pattern is accordance with a prescribed system and sends it to a minimum error pattern selecting circuit 14. This circuit 14 selects the minimum error pattern, whose error from the background noise pattern is minimum, from a preliminarily stored noise pattern group. A switch 17 adds the minimum error pattern to the last of the post pattern and sends it as encoded data for background noise generation. A noise characteristic convolutional operation circuit 24 performs the convolutional operation between the minimum error pattern sent following the post pattern and the output of a white noise generator 23 and outputs the result.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a silent section signal encoding and decoding circuit in a VOX (Voice Operated Transmission) system, which is a power saving technology for mobile communication devices, and particularly makes a speaker feel uncomfortable. In addition, the present invention relates to an encoding / decoding circuit for silent section information capable of reducing power consumption.

There is a strong demand for smaller size, lighter weight, and lower power consumption in mobile communication devices than in other devices, and technical development for realizing them is being advanced. Among them, as a method for reducing power consumption in voice communication, a VOX method has been put into practical use, in which the voice encoder is operated only when it is determined that there is sound, and the operation of the voice encoder is stopped when it is determined that there is no sound. There is.

In the VOX system, the coding circuit uses a post pattern that informs that the coding operation is stopped and becomes silent when the voice changes from voiced to silence, and a background noise generation signal accompanying it. A certain encoded data is transmitted, and a pre-pattern for notifying that the voice becomes voiced when the voice changes from the voiceless to the voiced is transmitted, and then the output of the voice encoder is transmitted. The decoding circuit generates a reproduced voice based on the received coded data when there is a voice, and generates background noise based on the coded data that accompanies the post pattern when there is no sound. This means that the decoding circuit is always operating, and it is difficult to realize low power consumption.

[0004]

2. Description of the Related Art FIG. 19 shows the configuration of a conventional encoding circuit. In FIG. 19, 10 is an analog-digital converter, 11 is a voice encoder, 12 is a voice detector, and 15 is a pre-
Pattern generator, 16 is a post pattern generator, 17
Is a switch, and 18 is a line encoder.

The input voice is digitized by an analog-to-digital converter and band compression encoded by a voice encoder. The digitized voice is also supplied to a voice detector, and the voice detector determines whether the voice is present or not. If it is determined that there is sound, the switch selects the output of the audio encoder according to the output signal of the audio detector and sends it.
If it is determined to be silent, the operation of the speech coder is stopped. Then, when the sound changes from sound to silence, the switch selects and outputs the output of the post pattern generator,
The decoding circuit is notified of the change to silence, and the encoded data for generating the background noise in the decoding circuit is transmitted. It should be noted that, during silence continuation, at regular intervals. post·
Sends coded data for pattern and background noise generation. On the other hand, when there is a change from silence to voice, the switch selects and sends the pre-pattern generated by the pre-pattern generator to notify the change to voice. Also, during times other than when the post pattern and encoded data are being output when there is no sound, the dummy pattern generator provided in the switch outputs a dummy pattern that is not used for decoding. There is. Then, regardless of whether there is sound or not, the signal to be transmitted is subjected to CRC coding or the like in the line encoder and converted into a code suitable for transmission.

FIG. 20 shows the configuration of a conventional decoding circuit (No. 1). In FIG. 20, 20 is a line decoder, 21
Is a voice decoder, 22 is a pre / post pattern detector,
26 is a digital / analog converter.

In the decoding circuit of FIG. 20, the line code received by the line decoder is converted into a band-compressed voice code and supplied to the pre / post pattern detector. When the pre / post pattern detector detects the pre pattern, the voice decoder generates a digitized voice signal from the voiced encoded data, and when the pre pattern is detected, the encoding accompanying the post pattern is performed. Generate background noise from the data. Then, the output of the audio decoder is supplied to the digital-analog converter to reproduce the original audio signal.

FIG. 21 shows the configuration of a conventional decoding circuit (No. 2). In FIG. 21, 20 is a line decoder, 21
Is a voice decoder, 22 is a pre / post pattern detector,
24 is a white noise generator, 25 is a switch, and 26 is a digital-analog converter.

In the decoding circuit of FIG. 21, the line code received by the line decoder is converted into a band-compressed voice code and supplied to the voice decoder and the pre / post pattern detector. Pre / post pattern detector
When the pattern is detected, the voice decoder operates, and the switcher selects the output of the voice decoder and supplies it to the digital-analog converter. When the pre / post pattern detector detects the post pattern, the voice decoder is stopped, and the switch selects the output of the white noise generator and outputs the white noise to the digital-analog converter.

In this way, when there is sound, the output of the voice decoder is converted into analog, and when there is no sound, the output of the white noise generator is converted into analog to generate reproduced sound.

[0011]

However, the conventional encoding circuit and decoding circuit have the following problems. The first is
When the decoding circuit of FIG. 20 is applied, it is necessary to send out the post pattern and the coded data each time when the coding circuit side continuously detects silence. The point is that the power consumption of both circuits increases.

Second, in the case of applying the decoding circuit of FIG. 21, when the coding circuit detects silence, the speech coder is stopped and the post pattern is transmitted, and the receiving side receives the post pattern. When the signal is detected, the speech decoder is stopped and the output of the white noise generator is selected, so power consumption can be reduced, but the frequency characteristics of the output of the white noise generator and the background noise around the speaker on the transmitting side match. Therefore, when switching between sound and silence, the listener recognizes the difference between the background noise included in the sound signal and the white noise selected when there is no sound, and feels uncomfortable.

Thirdly, as a matter of course, wireless communication is applied in mobile communication, but other radio waves are used in wireless communication.
Code errors due to interference such as automobile noise, and code errors due to the level drop when entering a tunnel or a building are likely to occur. These code errors are pre-pattern and post-code.
If it is included in the pattern, there is an error in determining whether there is sound or no sound, and the quality of the reproduced sound is significantly deteriorated.

It is an object of the present invention to provide a coding / decoding circuit for silent section information capable of preventing the deterioration of the quality of a reproduced signal and further reducing the power consumption by coping with such a problem.

[0015]

[Means for Solving the Problems] For the first problem,
Since the solution is brought about by applying the encoding circuit of FIG. 19 and the decoding circuit of FIG. 21 in combination, the means for solving the second and following problems will be examined.

There are two means to solve the second problem. The first is that in the encoding circuit, the frequency characteristics of the background noise are extracted for each frame, the noise pattern closest to this is selected and transmitted, and the noise pattern received by the decoding circuit and the output of the white noise generator are output. This is a method of performing a convolution operation with and for each frame, and converting the result into an analog signal to obtain a reproduced voice. The second is a method of transmitting coded data for background noise generation once in a plurality of frames and generating reproduced background noise that gradually changes from the coded data to white noise.

It is also possible to combine the following means with the first means. That is, in the encoding circuit, the power quantized value (power code) of the background noise is detected, the power code at time t and the power code at time (t-1) are compared, and there is a significant difference between the two. When there is a change, the post pattern and the coded data for generating the background noise are transmitted, and when there is no significant difference in the power code, neither is transmitted. Then, in the decoding circuit, the received signal for background noise generation is stored, and when there is a change in the same signal, it is replaced with the stored same signal and white noise is generated by the latest stored pattern of the same signal. The convolution operation with the pattern is performed, and the result is converted into an analog to obtain the reproduced voice.

For the third problem, all dummy data that is transmitted except when the encoded data for generating the post pattern and background noise is transmitted in the silent section is the data in which the CRC error occurs. Leave.

[0019]

In the first means for the second problem, the noise pattern closest to the frequency characteristic extracted on the encoding circuit side is selected and the pattern is transmitted, and the decoding circuit convolves the noise pattern with the white noise. Therefore, the frequency characteristic of its output is close to the frequency characteristic of background noise, and therefore,
In the reproduced voice, it becomes difficult to feel the background noise when switching between voiced and silent. In the second means, the signal reproduced from the coded data for background noise generation accompanying the received post pattern is set as C, and the signal reproduced from the white noise pattern prepared in advance on the decoding circuit side is set to W. The reproduced background noise is represented as R, and a positive number whose value is 0 at time 0 when the first post pattern is transmitted and whose value is 1 at time T when the next post pattern is transmitted is When m, a reproduction background noise R that satisfies the following equation is generated.

[0020]

[Equation 1]

Therefore, the reproduced background noise is equal to that of the reproduced background noise coded data when the post pattern is transmitted for the first time after changing to silence, and gradually changes with time, and at time T. Becomes equal to white noise, so that the encoded data continuously changes to white noise.
Therefore, it becomes difficult to feel a change in background noise when switching between voiced and silent. Then, when silence continues after time T, m is fixed to 1 and white noise of constant power is continuously generated as reproduction background noise.

The first means for the second problem is
The power consumption of the encoding circuit can be further reduced by combining the means that neither the post pattern nor the signal for generating the background noise is transmitted unless there is a significant change in the power code in the encoding circuit.

As for the third problem, if there is no error in the voice coded data in the voiced section, the CRC check result will be "0", and if there is no error in the dummy data, all will be C.
The RC check result becomes "1". Therefore, even if the pre / post pattern cannot be detected correctly due to a code error, if the received signal is CRC checked, the correct pre / post pattern can be obtained.
The time position of the post pattern can be detected.

[0024]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS FIG. 1 shows an embodiment of the present invention. In FIG. 1, 10 is an analog-digital converter, 11 is a voice encoder, 12 is a voice detector, 13 is a frequency characteristic extractor, and 14
Is a minimum error pattern selection circuit, 15 is a pre pattern generator, 16 is a post pattern generator, 17 is a switch,
A line encoder 18 constitutes an encoding circuit. Also, 20
Is a line decoder, 21 is a voice decoder, 22 is a pre / post pattern detector, 23 is a white noise generator, 24 is a noise characteristic convolution operation circuit, 25 is a switcher, 26 is a digital-analog converter, It constitutes a decoding circuit.

In the encoding circuit, the input voice is subjected to analog-to-digital conversion and then band-compressed by the voice encoder. At the same time, the digitally converted voice band signal is input to the voice detector, and when the voice detector determines that the voice is present, the output of the voice encoder is selected and transmitted by the switching device according to the output of the voice detector. When it is determined that there is no sound, the operation of the voice encoder is stopped. The switch selects and sends the pre-pattern when the sound changes from silence to sound, and the switch selects and sends the post pattern and the signal for background noise generation when the sound changes from sound to silence. However, in the encoding circuit of the present invention, the signal for background noise generation is determined as follows.

That is, the frequency characteristic of the digitally converted voice band signal (this is background noise) when it is determined to be silent is extracted and patterned according to a predetermined rule. This pattern is input to the minimum error pattern selection path, and the error with the frequency characteristic pattern of the extracted background noise is minimized from the noise pattern group (normally called a codebook) included in the minimum error pattern selection circuit. Then, a noise pattern (minimum error pattern) that satisfies the following is selected, and this is sent as encoded data for background noise generation after the post pattern.

It should be noted that both the data with voice and the data with voice are subjected to CRC encoding, etc. in a line encoder, converted into a code suitable for a transmission line, and transmitted. In the decoding circuit, the band-compressed voice signal is restored from the signal received by the line decoder, and this is input to the voice decoder to extend the band. At the same time, the band-compressed voice signal is input to the pre / post pattern detector. No pre / post pattern is detected when there is sound, and at this time the output of the decoder is selected by the switch and digital-analog converted to be reproduced voice. On the other hand, when there is no sound, a post pattern is detected, the voice decoder is stopped by this output, and the switch selects the output of the noise characteristic convolution operation circuit and supplies it to the digital-analog converter. In the noise characteristic convolution operation circuit, the convolution operation is performed on the minimum error pattern sent together with the post pattern and the white noise generated by the white noise generator. Therefore, the output of the noise characteristic convolution operation circuit is It approximates background noise. That is, when the post pattern is detected, the background noise is selected and reproduced by the digital-analog converter. Since the background noise of this reproduced background noise is reproduced by using the minimum error pattern selected by the encoding circuit, the background noise of the reproduced silent section has continuous frequency characteristics with the background noise of the voiced section. It becomes difficult for the hand to notice that the sound is switched to the sound.

FIG. 2 shows the configuration of the second embodiment of the decoding circuit of the present invention. In FIG. 2, 20 is a line decoder, 2
1 is a speech decoder, 22 is a pre / post pattern detector, 23 is a white noise generator, 24 is a noise characteristic convolution operation circuit, 25 is a switcher, 26 is a digital-analog converter, and 27 is a frequency characteristic extractor. 28 are minimum error pattern selection circuits. The encoding circuit facing the decoding circuit of this configuration has the same configuration as the encoding circuit shown in FIG.

The configuration of FIG. 2 differs from the configuration of the decoding circuit of FIG. 1 in the circuit for generating the noise pattern used in the noise characteristic convolution operation. In the case of FIG. 2, a frequency characteristic pattern is extracted from the output signal of the speech decoder, and a noise pattern having a minimum error with the extracted pattern is selected from the codebook included in the minimum error pattern selection circuit, The convolution operation of the selected noise pattern and the white noise pattern is performed. As can be readily understood by comparing this with the generation of the signal for background noise generation in FIG. 1, the noise pattern selected by the decoding circuit in FIG.
The operation of the decoding circuit shown in FIG. 2 is exactly the same as the operation of the decoding circuit shown in FIG.

FIG. 3 shows the configuration of a third embodiment of the decoding circuit of the present invention. In FIG. 3, 20 is a line decoder, 2
1 is a speech decoder, 22 is a pre / post pattern detector, 23 is a white noise generator, 24 is a noise characteristic convolution operation circuit, 25 is a switcher, 26 is a digital-analog converter, and 27 is a frequency characteristic extractor. , 28 is a minimum error pattern selection circuit, and 29 is a perceptual weighting circuit. The encoding circuit facing the decoding circuit of this configuration is the encoding circuit shown in FIG.

The configuration of FIG. 3 differs from the configuration of FIG. 2 in that a perceptual weighting circuit is newly connected to the output side of the frequency characteristic extractor, and the minimum error pattern approximating the output of the perceptual weighting circuit is the minimum error pattern. This is a point to be selected by the selection circuit. The perceptual weighting circuit adds the frequency characteristic of human hearing to the output of the frequency characteristic extractor, and has an advantage that background noise sounds more natural.

FIG. 4 is a fourth embodiment of the decoding circuit of the present invention. In FIG. 4, 20 is a line decoder, 21 is a voice decoder, 22 is a pre / post pattern detector, and 23.
Is a white noise generator, 30 is a first gain control circuit, 30a
Is a second gain control circuit, 31 is an adding circuit, and 26 is a digital-analog converter.

The configuration of FIG. 4 is premised on that silence lasts much longer than the frame period. The encoding circuit to which the decoding circuit having the configuration of FIG. 4 faces is an encoding circuit similar to the encoding circuit shown in FIG. 9, and the post pattern and the encoded data have a period considerably longer than the frame period while silence is continued. Send to T once.

The feature of the configuration of FIG. 4 is that when the post pattern is first detected, the reproduction background noise is not immediately formed by white noise, and the transmitter side associated with the post pattern when the encoding circuit determines that there is no sound. The coded data that is the background noise of is the reproduction background noise. Thereafter, the reproduced background noise obtained by the equation (1) is generated by the encoded data and the white noise, and the background noise on the transmitting side is gradually switched to the white noise.

This switching is performed by the first gain control circuit and the second gain control circuit of FIG. 4, the gain coefficient of the first gain control circuit is (1-m), and the gain control circuit of the second gain control circuit is The gain coefficient is m. This m takes a value 0 at the time 0 when the first post pattern is detected, takes a value 1 at the time T when the next post pattern is detected, and m is 1 when silence continues after the time T. Fixed to. Therefore, the playback background noise can be continuously switched from the background noise on the transmitting side to the white noise output by the white noise generator provided on the receiving side, and the listener can confirm that the voice has switched from voiced to silence. It is difficult to feel, and the playback background noise is continuous even after the time T during silence.

The value m for controlling this gain is 0 at time 0.
Then, if the value is 1 at time T and changes substantially uniformly during time T, there is no restriction on the way of change, but it is easiest to change linearly during time T. The m changing in this way is such that m is 0 ≦ t ≦ T, where T _F is the frame period, T is the period of the post pattern, k is the ratio of the period of the post pattern to the frame period, and t is time. It is given by the following formula.

[0037]

[Equation 2]

In reality, it is impossible to generate completely continuous m. Therefore, if m is calculated for each frame period T _F , values from 0 to 1 can be obtained in 1 / k steps. . Then, after the time T, m is fixed to 1.

FIG. 5 is a flow chart of background noise generation of the fourth embodiment of the decoding circuit of the present invention. The calculation of m will be described below with reference to the reference numerals in FIG. A. Set m to 0. B. A post pattern is detected, and when no post pattern is detected, the process returns to A. C. When the post pattern is detected, the equation (2) is calculated to obtain the reproduced background noise. D. The pre-pattern is detected, and when the pre-pattern is detected, the process returns to A. E. When the pre-pattern is not detected, the frame period T _F is measured. F. It is determined whether the number of frame period measurements exceeds k, and k
When not exceeding, return to C. F. When it exceeds k, m is fixed at 1 and the process returns to C.

FIG. 6 is a diagram showing the transition of the background noise level thus generated. In FIG. 6, t = 0 is the time when the post pattern is first detected, and at this time, the background noise sent along with the post pattern is reproduced. After that, the reproduction background noise is calculated for each T _F by the expression (1), and the reproduction is continued by the expression (1) until t = T (the number of times T _F is measured is k). At t = T, m = 1, and the reproduced background noise is generated only by the white noise generated by the white noise generator of the decoding circuit. Then, at t> T, m = 1 is fixed, so that the reproduced background noise is fixed at t = T.

In the flowchart of FIG. 5, t
At m> T, m is fixed to 1 and the calculation of the reproduction background noise of the formula (1) is continued for each frame.
If the reproduced background noise at = T is stored in the adding circuit, the calculation by the equation (1) is not necessary at t> T.

FIG. 7 is a fifth embodiment of the decoding circuit of the present invention. In FIG. 7, 20 is a line decoder, 21 is a voice decoder, 22 is a pre / post pattern detector, and 23.
Is a white noise generator, 30 is a first gain control circuit, 30a
Is a second gain control circuit, 30b is a third gain control circuit,
30c is a fourth gain control circuit, 31 is a first adding circuit,
31a is a second addition circuit, 32 is a power calculation circuit, 33
Is a multiplication circuit, and 26 is a digital-analog converter. In the fourth embodiment of the present invention, m is 1 when t> T.
The background noise is reproduced by the white noise generated by the white noise generator of the decoding circuit fixed to, but in the fifth embodiment of the present invention, when the silence continues even at t> T,
The level of white noise is made asymptotic to the level of coded data sent along with the post pattern. That is, t> T
In the above, the output of the second gain control circuit, the power P _C of the encoded data and the power P _{W of the} white noise are calculated in the power calculation circuit, and the square root of the ratio P _C / P _W of the two is calculated to obtain the white noise. By the output of the generator multiplied by
A reproduction background noise R ′ given by the following equation is generated.

[0043]

[Equation 3]

Here, n is 0 when t ≦ T, and T <t <2.
It is a number that gradually changes from 0 to 1 at T and takes a value of 1 when t ≧ 2T. Further, R _T is the reproduction background noise R according to the equation (1) at t = T and is equal to W.

FIG. 8 is a flow chart of background noise generation in the fifth embodiment of the decoding circuit of the present invention. A method of generating background noise in the fifth embodiment of the decoding circuit of the present invention will be described below with reference to the reference numerals of FIG. A. Set m and n to 0. B. A post pattern is detected, and when no post pattern is detected, the process returns to A. C. When the post pattern is detected, the equation (2) is calculated to obtain the reproduced background noise. D. The pre-pattern is detected, and when the pre-pattern is detected, the process returns to A. E. When the pre-pattern is not detected, the frame period T _F is measured. F. It is determined whether the number of frame period measurements exceeds k, and k
If it does not exceed, return to C. G. When k is exceeded, m is fixed to 1 and the reproduction background noise R ′ is generated by the equation (3). H. The pre-pattern is detected, and when the pre-pattern is detected, the process returns to A. I. When the pre-pattern is not detected, the frame period T _F is measured starting from t = T. J. It is determined whether the number of frame period measurements exceeds k, and k
If it does not exceed, return to G. K. When the number of frame cycle measurements exceeds k, n is set to 1
Fix to and return to G.

FIG. 9 is a diagram showing the transition of the background noise level thus generated. In FIG. 9, t = 0 is the time when the post pattern is first detected, and at this time, the background noise sent along with the post pattern is reproduced. After that, the reproduction background noise is calculated for each T _F by the expression (1), and the reproduction is continued by the expression (1) until t = T (the number of times T _F is measured is k). At t = T, m = 1 is fixed, and thereafter, reproduction background noise is generated by the expression (3). Therefore (1)
The background noise according to the formula and the formula (3) becomes continuous at t = T, and thereafter, the level of the white noise generated by the formula (3) is the level of the background noise on the transmitting side transmitted along with the post pattern. , And becomes equal to the level of the background noise on the transmitting side which is transmitted along with the post pattern at t = 2T. After that, since n is fixed at 1, the reproduction background noise level is kept constant.

The feature of the fifth embodiment of the decoding circuit of the present invention is that the background noise does not change significantly, so that when silence changes continuously to 2T or more, it changes from silence to speech. The difference in the level of background noise due to is reduced, and the listener does not feel the change.

In the flowchart of FIG. 8, t
At> 2T, n is fixed to 1 and the calculation of the reproduced background noise is continued for each frame by the formula (3).
If the reproduced background noise at t = 2T is stored in the second addition circuit, the calculation by the equation (3) is not necessary at t> 2T.

Further, in the configuration of FIG. 7, the third gain control circuit is provided on the output side of the first addition circuit to add with the output of the fourth gain control circuit. If the control circuit is provided on the output side of the second gain control circuit and the outputs of the first gain control circuit, the third gain control circuit, and the fourth gain control circuit are added, the formula (3) is obtained. In addition to obtaining background noise close to the reproduced background noise obtained in step 1, one adder circuit can be omitted.

FIG. 10 shows a sixth embodiment of the decoding circuit according to the present invention. In FIG. 10, 20 is a line decoder, 21
Is a voice decoder, 22 is a pre / post pattern detector,
23 is a white noise generator, 30 is a first gain control circuit, 3
0a is a second gain control circuit, 30b is a third gain control circuit, 30c is a fourth gain control circuit, 31 is a first addition circuit, 31a is a second addition circuit, and 26 is a digital-analog converter. In a fifth embodiment of the invention, where t>
When silence continues even at T, the level of white noise is made asymptotic to the level of encoded data sent in association with the post pattern. In the sixth embodiment of the present invention, t> At T, the white noise is asymptotic to the coded data sent along with the post pattern.
That is, the reproduction background noise R ″ given by the following equation is generated.

[0051]

[Equation 4]

Here, n is 0 when t ≦ T, and T <t <2.
It is a number that gradually changes from 0 to 1 at T and takes a value of 1 when t ≧ 2T. Further, R _T is the reproduction background noise R according to the equation (2) at t = T and is equal to W.

FIG. 11 is a flowchart of background noise generation in the sixth embodiment of the decoding circuit of the present invention. A background noise generating method in the fifth embodiment of the decoding circuit of the present invention will be described below with reference to the reference numerals of FIG. A. Set m and n to 0. B. A post pattern is detected, and when no post pattern is detected, the process returns to A. C. When the post pattern is detected, the equation (2) is calculated to obtain the reproduced background noise. D. The pre-pattern is detected, and when the pre-pattern is detected, the process returns to A. E. When the pre-pattern is not detected, the frame period T _F is measured. F. It is determined whether the number of frame period measurements exceeds k, and k
If it does not exceed, return to C. G. When it exceeds k, m is fixed to 1 and the reproduction background noise R ″ is generated by the equation (4). H. When the pre-pattern is detected and the pre-pattern is detected (Yes), it is set to A. Return I. When no pre-pattern is detected (No), t
The frame period T _F is measured starting from = T. J. It is determined whether the number of frame period measurements exceeds k, and k
When it does not exceed (No), the process returns to G. K. When the number of frame cycle measurements exceeds k (Yes)
, Fix n at 1 and return to G.

FIG. 12 is a diagram showing the transition of the background noise level thus generated. In FIG. 12, t = 0 is the time when the post pattern is first detected, and at this time, the background noise sent along with the post pattern is reproduced. After that, the reproduction background noise is obtained by the equation (2) for each T _F , and the reproduction is continued by the equation (2) until t = T (the number of times T _F is measured is k). At t = T, m = 1 is fixed, and thereafter, reproduction background noise is generated by the equation (4). Therefore, the background noise according to the equations (1) and (4) is continuous at t = T, and thereafter, the reproduced background noise generated according to the equation (4) is the background of the transmitting side transmitted along with the post pattern. It asymptotically approaches the noise, and becomes equal to the background noise on the transmitting side that is sent along with the post pattern at t = 2T. After that, since n is fixed at 1, the reproduction background noise level is kept constant.

The feature of the sixth embodiment of the decoding circuit of the present invention is that the background noise does not change significantly, so when silence changes continuously to 2T or more, it changes from silence to speech. In addition to being able to reduce the level difference of the background noise associated with, the frequency characteristics are also similar to the background noise sent along with the post pattern, so that the listener does not feel the switching.

In the flow chart of FIG. 11,
When t> 2T, n is fixed to 1 and the calculation of the reproduction background noise is continued for each frame by the formula (4). However, the reproduction background noise at t = 2T is stored in the second addition circuit. If t> 2T, then (4)
Eliminates the need for formula calculations.

FIG. 10 also shows a seventh embodiment of the decoding circuit of the present invention. 13 is a flowchart of background noise generation of the seventh embodiment of the decoding circuit of the present invention, and FIG. 14 is a transition of background noise level of the seventh embodiment of the decoding circuit of the present invention.

As can be seen from FIGS. 13 and 14, the operation of the seventh embodiment of the present invention differs from the operation of the sixth embodiment of the present invention in that t = 0 (T = 0 ( Received at t = pT (p is a positive integer greater than or equal to 2) instead of using the encoded data received accompanying the post pattern at the time of the first post pattern reception) to generate background noise. The point is that background noise is generated using encoded data.

Although the background noise does not change significantly on the transmitting side, a slight change occurs. Therefore, updating the playback background noise every time the post pattern is received causes the playback background noise to be the same as the playback background noise when there is no sound. There is an advantage that the difference from the background noise when returning to the sound is less likely to be felt.

In the seventh embodiment of the present invention as well, t
It is also possible to store the reproduced background noise obtained by substituting the coded data received accompanying the post pattern at> 2T so that the reproduced background noise is not calculated in the frame period.

FIG. 15 shows the configuration of the second embodiment of the encoding circuit of the present invention. In FIG. 15, 10 is an analog-
Digital converter, 11 is a voice encoder, 12 is a voice detector, 15 is a pre-pattern generator, 16 is a post-pattern generator, 17 is a switcher, 18 is a line encoder, 19 is a power code detector Is. The configuration of FIG. 5 differs from the configuration of FIG. 1 in that a power code detector is newly provided and the output of this power code detector causes
The point is to control the transmission of signals for pattern and background noise generation.

The power code detector detects the power of the output of the speech coder immediately after the output of the post pattern, encodes it, and stores it. After that, post patterns are generated at regular intervals while detecting silence, but each time a post pattern is generated, the power code is detected from the output of the speech encoder and stored as the detected power code. The output power codes are compared, and when there is no significant difference, the signal transmission for post pattern and background noise generation is stopped. This reduces the opportunity for the speech coder to operate, thus saving transmission power.

Any of the decoding circuits already described can be opposed to the encoding circuit of FIG. However, in this case, if the encoding circuit does not detect a change in the power code, the transmission of the post pattern and the signal for generating the background noise is stopped. It is necessary to add the function of storing encoded data for.

FIG. 16 shows the configuration of the third embodiment of the encoding circuit of the present invention. In FIG. 16, 10 is an analog-
A digital converter, 11 is a voice encoder, 12 is a voice detector, 13 is a frequency characteristic extractor, 14 is a minimum error pattern selector, 15 is a pre-pattern generator, and 16 is a post-pattern generator.
Pattern generator, 17 is a switch, 18 is a line encoder, 1
9 is a power code detector. The configuration of FIG. 6 is different from that of FIG. 5 in that the frequency characteristic of a digitally converted voice band signal in the absence of sound is extracted, and the noise pattern having the smallest error from the extracted frequency characteristic pattern is minimized. This is the point of selecting from the codebook of the error pattern selector and transmitting the selected pattern as encoded data for generating background noise. Then, when the power code detected from the output of the speech coder does not change, the transmission of the pattern selected from the codebook is stopped. The configuration of the decoding circuit used in opposition to the encoding circuit of FIG. 16 may be any of FIG. 1, FIG. 2 and FIG. 3 already described. Also, in this case, if the encoding circuit does not detect the change in the power code, the transmission of the signal for generating the post pattern and the background noise is stopped, so that the latest background noise is supplied to the speech decoder. It is necessary to add the function of storing encoded data for generation.

Now, the coding circuit sends dummy data, which is not used in the decoding circuit, from the switcher except when the signal for generating the post pattern and the background noise is sent in the silent section of the voice. There is. CRC error does not occur if voice coded data is transmitted without error (CR
Since C = 0), a CRC error always occurs in the dummy data if there is no error during transmission (C
If it is formed such that RC = 1), even if the pre / post pattern cannot be recognized due to a code error, the decoding circuit performs a CRC check on the received signal to switch between sound and silence. The time position of can be detected.

FIG. 17 shows the configuration of an eighth embodiment of the decoding circuit of the present invention based on the above principle. In FIG. 17, 20 is a line decoder, 21 is a voice decoder, 22 is a pre / post pattern detector, 23 is a white noise detector, and 2
4 is a noise characteristic convolution operation circuit, 25 is a switcher, 26
Is a digital / analog converter, and 30 is a pre / post pattern time position determiner. The feature of the configuration of FIG. 17 is that
This is to provide a pre / post pattern time position determiner capable of determining the time position of the pre / post pattern even if a code error occurs in the transmission line. 17 is provided in the decoding circuit having the configuration shown in FIG. 1, the pre / post pattern time position determination circuit can be provided similarly in the configurations shown in FIGS.

FIG. 18 is a flow chart of the pre / post pattern time position determiner. The pre / post pattern time position determiner uses the fact that even if there is an error in the predetermined bit length, it does not continue for a long time if the error is within the range in which communication can be continued, and If it is received, a CRC error will always occur only in the dummy data, and in other cases, the CRC error will not occur and the CRC check of the code to be transmitted is performed to determine whether or not it is the dummy data. The operation will be described below step by step. A. Check the contents of the register that stores the voiced and silent states. B. If there is sound, an error counter that counts the number of CRC check errors is reset. C. A CRC check is performed to determine whether the check result is incorrect. When there is no error, that is, when CRC = 0 (No), the procedure returns to B. D. When the CRC check result is incorrect, that is, CRC = 1
In the case of (Yes), the number of errors is counted by the error counter. E. It is determined whether or not the counting result of the error counter has reached a predetermined number N. If it has not reached N (No), the procedure returns to C. F. When the count result of the error counter reaches N (Ye
In s), it was determined that the post pattern was lost,
It is assumed that it is a silent section. G. The contents of the register storing the voiced and silent states are rewritten and the process returns to A. H. On the other hand, if it is confirmed from the contents of the register storing the silent state that there is a silent period, the no error counter is reset. I. A CRC check is performed to determine whether the check result is correct. When it is not correct, that is, when CRC = 1 (No), the process returns to H. J. When the CRC check result is correct, that is, CRC = 0
In the case of (Yes), the correct number is counted by the no-error counter. K. It is determined whether or not the count result of the no-error counter has reached a predetermined number M. If it has not reached M (No), the process returns to I. L. When the count result of the no-error counter reaches the predetermined number M (Yes), it is determined that the pre-pattern has been lost, and it is determined that it is a voiced section. G. The contents of the register storing the voiced and silent states are rewritten and the process returns to A.

In the configuration of the encoding circuit, from the conventional configuration to the configuration of the present invention, the line decoder is installed at the output of the encoding circuit, but the line encoder is the switch. It is also possible to install it between the audio encoder and the audio encoder. In this case, since the line encoder can be stopped when there is no sound, the power consumption can be further reduced. However, in this case, the post pattern and the pre pattern
It is necessary to convert the pattern into a pattern that has been subjected to CRC calculation in advance.

Further, in the above, the switch in the encoding circuit and the decoding circuit is not described. Since it is simple, the switching device in the encoding circuit of FIG. 1 will be described as an example without using the drawing.

The voice detector detects sound and silence and outputs a signal. For example, it is 1 when there is sound and 0 when there is no sound.
And In response to this, the switching device outputs the output of the voice encoder when the voice continues, the post pattern when the voice changes from the voice to the silence and the post pattern when the voice changes from the voice to the voice. Select and send the pattern. Therefore, by monitoring the output of the voice detector and the output of the voice detector one cycle before, it is possible to determine which of the above states is present. That is, it can be seen that 11 indicates the continuation of the sound, 10 indicates the change to the silence, 00 indicates the continuation of the sound, and 01 indicates the change to the sound. Therefore, the output of the voice detector and the output of the voice detector one cycle before are the logical product circuit (a) which does not logically invert the input and the logical product circuit (b) which logically invert the second input. ), And a logical product circuit (designated as c) that logically inverts the first and second inputs, and a logical product circuit (designated as d) that logically reverses the first input, and when the output of a is 1. The output of the speech encoder is selected, the post pattern is selected when the outputs of b and c are 1, and the pre pattern is selected when the output of d is 1. The outputs of the speech encoder, pre-pattern generator, and post-pattern generator are not single bit, so for example the output of the speech detector is latched for one period. The output of the minimum error pattern selection circuit may be selected by a signal obtained by delaying the signal for selecting the post pattern by the duration of the post pattern.

[0071]

As described above, according to the present invention, it is possible to realize a coding / decoding circuit for silent section information capable of preventing deterioration of quality of a reproduced signal and deleting unnecessary processing, and mobile communication. It is possible to contribute to improving the call performance of the device and reducing the power consumption.

[Brief description of drawings]

FIG. 1 is an embodiment of the present invention.

FIG. 2 is a second embodiment of the decoding circuit of the present invention.

FIG. 3 shows a third embodiment of the decoding circuit of the present invention.

FIG. 4 is a fourth embodiment of the decoding circuit of the present invention.

FIG. 5 is a flowchart of background noise generation of the fourth embodiment of the decoding circuit of the present invention.

FIG. 6 is a background noise level transition of a fourth embodiment of the decoding circuit of the present invention.

FIG. 7 is a fifth embodiment of the decoding circuit of the present invention.

FIG. 8 is a flowchart of background noise generation of the fifth embodiment of the decoding circuit of the present invention.

FIG. 9 is a transition of the background noise level of the fifth embodiment of the decoding circuit of the present invention.

FIG. 10 is a sixth embodiment of the decoding circuit of the present invention.

FIG. 11 is a flowchart of background noise generation of the sixth embodiment of the decoding circuit of the present invention.

FIG. 12 is a transition of the background noise level of the sixth embodiment of the decoding circuit of the present invention.

FIG. 13 is a flowchart of background noise generation of the seventh embodiment of the decoding circuit of the present invention.

FIG. 14 is a background noise level transition of the seventh embodiment of the decoding circuit of the present invention.

FIG. 15 is a second embodiment of the encoding circuit of the present invention.

FIG. 16 shows a third embodiment of the encoding circuit of the present invention.

FIG. 17 is an eighth embodiment of the decoding circuit of the present invention.

FIG. 18 is a flowchart of a pre / post pattern time position determiner.

FIG. 19 is a conventional encoding circuit.

FIG. 20 shows a conventional decoding circuit (No. 1).

FIG. 21 shows a conventional decoding circuit (2).

[Explanation of symbols]

 10 analog-digital converter 11 speech encoder 12 speech detector 13 frequency characteristic extractor 14 minimum error pattern selection circuit. 15 pre-pattern generator 16 post-pattern generator 17 switcher 18 line encoder 20 line decoder 21 voice decoder 22 pre / post pattern detector 23 white noise generator 24 noise characteristic convolution operation circuit 25 switcher 26 Digital / Analog converter

Claims (9)

[Claims] 1. A method for detecting voiced and non-voiced speech, outputting the output of the voice encoder when the voice is present, and stopping the voice encoder when there is no voice to generate post pattern and encoded data for background noise reproduction. A coding circuit and a decoding circuit for silent section information in the VOX system that reproduces background noise when a post pattern is detected on the receiving side and reproduces the output of a voice decoder when there is sound. Is a frequency characteristic extractor (13) for extracting and patterning the frequency characteristic of the analog-digital converted voice band signal.
And a minimum error pattern selection circuit (14) that has a noise pattern group and selects and outputs a noise pattern that is closest to the output pattern of the frequency characteristic extractor, and outputs the minimum error as encoded data when there is no sound. The decoding circuit is an encoding circuit that outputs the output of the pattern selection circuit, and the decoding circuit includes a noise characteristic convolution operation circuit (24) that performs a convolution operation of the received minimum error pattern and the output pattern of the white noise generator. A coding and decoding circuit for silent section information, which is a decoding circuit for reproducing the output of the noise characteristic convolution operation circuit when receiving a pattern. 2. The encoding / decoding circuit for silent section information according to claim 1, wherein the decoding circuit extracts a frequency characteristic of an output of the speech decoder, and patterns the frequency characteristic. A minimum error pattern selection circuit that has a noise pattern group and that selects and outputs a noise pattern that is closest to the output pattern of the frequency characteristic extractor, and the output of the minimum error pattern selection circuit and the white noise generator A coding and decoding circuit for silent period information, which is a decoding circuit for performing a convolution operation on the output and the noise characteristic convolution operation circuit. 3. The encoding / decoding circuit for silent period information according to claim 1, wherein the decoding circuit extracts a frequency characteristic of a speech decoder output signal and patterns it, A hearing weighting circuit for convoluting the average frequency characteristic pattern of the human auditory sense into the output pattern of the frequency characteristic extractor, and a noise pattern group having a noise pattern group, and selecting a noise pattern closest to the output pattern of the hearing weighting circuit And a minimum error pattern selection circuit for outputting the output of the minimum error pattern selection circuit and the output of the white noise generator is a decoding circuit for performing a convolution operation in the noise characteristic convolution operation circuit. Encoding and decoding circuit for section information. 4. A method for detecting voiced and non-voiced voice, outputting the output of the voice encoder when the voice is present, and stopping the voice encoder when there is no voice to generate post pattern and encoded data for background noise reproduction. A coding circuit and a decoding circuit for silent section information in the VOX system that reproduces background noise when a post pattern is detected on the receiving side and reproduces the output of a voice decoder when there is sound. Is a coding circuit that transmits the post pattern and the coded data in a cycle that is a multiple of the frame cycle, and the decoding circuit sets m to be 0 before the first post pattern is received, When the number is between 0 and 1 which becomes 1 after the second post pattern is received, the gain coefficient for controlling the output of the speech decoder is (1-m). And the first gain control circuit A second gain control circuit having a gain coefficient of m for controlling gain with respect to the output of the white noise generator; and a first addition circuit for adding outputs of both gain control circuits, Is a decoding circuit for reproducing the output of the adding circuit of 1., and a coding and decoding circuit for silent period information. 5. The encoding and decoding circuit for silent interval information according to claim 4, wherein the decoding circuit has a value of 0 before receiving the second post pattern, The gain coefficient for the output of the first adder circuit is (1-n) when the number is between 0 and 1 which becomes 1 after the reception of the th post pattern. ) Is a third gain control circuit, a power operation circuit that obtains the power P _C of encoded data and the power P _{W of} white noise, and outputs the square root of P _C / P _W , and the output of the power operation circuit. A multiplication circuit that outputs the product of white noise, and a gain coefficient n that controls the gain of the output of the multiplication circuit.
And a second adder circuit for adding the output of the fourth gain control circuit and the output of the third gain control circuit, and reproducing the output of the second adder circuit. And a decoding circuit for silent period information. 6. The encoding and decoding circuit for silent interval information according to claim 4, wherein the decoding circuit has a value of 0 before receiving the second post pattern, The gain coefficient for the output of the first adder circuit is (1-n) when the number is between 0 and 1 which becomes 1 after the reception of the th post pattern. ) Is a third gain control circuit, a fifth gain control circuit having a gain coefficient of n for performing gain control on the output of the speech decoding circuit, an output of the fifth gain control circuit and the A third adder circuit for adding the output of the third gain control circuit, and a decoding circuit for reproducing the output of the third adder circuit. Circuit. 7. Detecting voiced and non-voiced voices, sending the output of the voice encoder when there is voice, and stopping the voice encoder when there is no voice to generate post pattern and encoded data for background noise reproduction. A coding circuit and a decoding circuit for silent section information in the VOX system that reproduces background noise when a post pattern is detected on the receiving side and reproduces the output of a voice decoder when there is sound. In addition, a power code detector for detecting and storing a power code from the output signal of the voice encoder, and storing the latest power code when the power code changes, the power code An encoding and decoding circuit for silent period information, which outputs a post pattern and encoded data when there is a change in. 8. The encoding / decoding circuit for silent section information according to claim 7, wherein the encoding circuit extracts a frequency characteristic from an analog-digital converted signal to form a frequency characteristic extraction pattern. And a noise pattern group, and a minimum error pattern selection circuit that selects and outputs a noise pattern that most closely approximates the output pattern of the frequency characteristic extractor. A coding and decoding circuit for silent period information, which is characterized by transmitting a pattern and coded data. 9. Detecting voiced and silent voices, outputs the output of the voice encoder when there is voice, stops the voice encoder when there is no voice, and outputs post pattern and encoded data for background noise reproduction. A coding circuit and a decoding circuit for silent section information in the VOX system which reproduces background noise when a post pattern is detected on the receiving side and reproduces the output of a voice decoder when there is sound. Pre- and post-preparation for the decoding circuit, adding the CRC check bit calculated by the CRC calculation rule different from that of the receiving side to only the dummy pattern generated by the dummy pattern generator included in In the pattern time position determination circuit, 1 and 0 of the CRC check result of the received signal are counted, and whether the dummy pattern or the non-dummy pattern is detected is determined by the counted result. Encoding and decoding circuit of the silent segment information, characterized by a constant.