JPH0969266A - Method and apparatus for correcting sound - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a method or an apparatus for correcting and reproducing audio signals or sound signals by means of a relatively easy means in the case of decoding codes containing errors, with a small degree of deterioration. SOLUTION: Error detection data of each frame which are detected by an error detector 111 are stored in an error detection data buffer 104 and a coding data buffer 102. While an error is detected in coding data, a data-discarding signal is fed to the buffers 102, 104 by a switching device 109 to continue discarding of the data. When an error comes not to be detected in the coding data, the switching device 109 switches a switch 108 to feed the coding data to a decoder 112. When a soundless section is detected in the coding data by a power operator 105 and a soundless section detector 106, the switching device 109 switches the switch 8 to the soundless side, so that the soundless coding data are sent to the decoder 112.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a decoding technology of a coding / decoding device for a wide band so-called audio signal or a telephone band voice signal, and is particularly characterized in data interpolation when an error occurs in coded data. A certain audio correction method and apparatus.

[0002]

2. Description of the Related Art PCM for audio signals or voice signals
A coding / decoding technique has been established for converting the digital signal into a digital signal and transmitting or accumulating the digital signal by compressing or compressing the data amount, and is used in each field. In this encoding / decoding technology, encoded data is generated through an encoder, and this encoded data is transmitted via a transmission line in the case of a transmission system and to a storage medium such as a disk or tape in the case of a storage system. To the decoder via.

By the way, an error may occur in encoded data due to various factors. In a communication system, the encoded data mainly occurs in a transmission line, and a storage system occurs in the encoded data when reading from a storage medium using magnetism or light. Sometimes. These errors can be completely repaired by adding an error correction code if the frequency of occurrence is below a certain level. However, when many mistakes occur, it is impossible to completely repair it. In addition, an error correction code may not be added to the coded data in the storage system because the possibility of an error occurring is small.

In order to know whether or not there is an error in the coded data, that is, in order to detect an error, it is possible to divide the coded data into frames and perform error detection for each frame using an error detection code. it can. Then, when an error is detected, a method of reproducing an audio signal or a voice signal by interpolating the frame is generally adopted.

As a method of the interpolation, a method of performing mute (silence suppression, volume reduction), repeat, and interpolation is known. Repeat, which is one of the interpolation methods, is a method of repeatedly using the parameter or PCM waveform of the immediately preceding frame when an error is detected in a certain frame, and is effective for the case of a stationary signal. is there. Here, the term "stationary" means that the signal is periodic or the waveform of the signal is noisy, but the time variation of the approximate shape of the spectrum is small. Interpolation, which is another method of interpolation, is a method of predicting a waveform or a parameter value of a frame in which an error is detected from a frame before and after an error is detected in a frame when a signal is detected. It is effective when is constant or non-stationary but changes gradually compared to the frame length.

In all of these methods, it is premised that the target audio signal or voice signal is stationary or quasi-stationary in nature. Therefore, when this interpolation method is applied to a non-stationary signal, an oscillating sound-like abnormal sound caused by repetition and an unnatural change in sound due to a failure in parameter prediction are caused, and subjective quality is deteriorated.

Alternatively, if a code is retransmitted when an error occurs, it is possible to reproduce without error, but the method is limited in its application because the delay tends to be large. .

[0008]

The technical difficulty is to try to estimate the signal of the lost frame while suppressing the deterioration of quality. It is an object of the present invention to provide a method or apparatus for correcting and reproducing an audio signal or a voice signal with a relatively easy means when decoding a code containing an error and with a small deterioration.

[0009]

According to a first aspect of the present invention, there is provided a voice correction method, wherein a digital signal obtained by decoding encoded data obtained by converting an audio signal or a voice signal into a digital signal and encoding the digital signal. A voice correction method used for reproducing a signal, wherein encoded data or a digital signal obtained by decoding encoded data is divided into frames, and the encoded data is subjected to error detection in each frame to obtain encoded data. Alternatively, the decoded digital signal and the error detection code detected by the error detection are stored for each frame for a plurality of frames. When an error is detected in the stored frame, the frame is discarded and the stored frame is stored. Of these, the error-free ones are output next, the time shift that occurs when the frame is discarded is interpolated, and the frame is interpolated Give over data or digital signals mentioned above decoding.

The audio correction apparatus according to the second aspect of the present invention reproduces a digital signal obtained by decoding encoded data obtained by converting an audio signal or an audio signal into a digital signal and encoding the digital signal. An audio correction device used, wherein an encoded data or a digital signal obtained by decoding the encoded data is divided into frames, and an error detector that detects an error in the encoded data in each frame, and the encoded data or the decoding A buffer for storing the converted digital signal and the error detection code detected by the error detector for a plurality of frames for each frame, and when an error is detected in the frame stored in the buffer, the frame is removed from the buffer. A switch for discarding and outputting the error-free frames stored in the buffer subsequently, and a frame Interpolating time lag that occurs when the discards arm, and interpolation means for providing a digital signal encoded data or decoded to the interpolation frame,
Having.

According to the invention of claim 3 of the present application, in addition to the configuration of claim 2, a voice correction device further has a data buffer for storing an error-free encoded data or a decoded digital signal. However, when the buffer is empty, the interpolation means selects and gives the encoded data or the decoded digital signal stored in the data buffer.

According to the invention of claim 4 of the present application, in addition to the structure of claim 2, the speech correction apparatus has a decoder for decoding the coded data in which the error is detected by the error detector,
The decoder is characterized by giving a digital signal obtained by decoding encoded data to a buffer.

According to the invention of claim 5 of the present application, in addition to the structure of claim 2, the speech correction apparatus has a decoder for decoding the coded data in which the frame is interpolated by the interpolating means. Is characterized by outputting a digital signal obtained by decoding encoded data.

According to a sixth aspect of the present invention, in addition to the configuration of the second aspect, the interpolating means compares the coded data or the decoded digital signal with a predetermined level in a silent section. It is characterized in that predetermined data is inserted into coded data or a decoded digital signal by the number of frames which are discarded when a silent section is detected, and frames are interpolated.

According to a seventh aspect of the present invention, in addition to the configuration of the sixth aspect, the interpolating means compares the coded data or the decoded digital signal with a predetermined level in a silent section. A silent section detector that detects whether or not, when the silent section is not detected, the coded data stored in the buffer or the decoded digital signal is selected, and when the silent section is detected, a predetermined number of discarded frames is determined. It is characterized in that it is configured to include a switch for selecting data.

In the voice correction device according to the invention of claim 8 of the present application, in addition to the configuration of claim 6 or 7, the predetermined data is
Either the same data as the silent section, the encoded data of the frame determined to be the silent section, or a specific sample value,
When inserting the predetermined data over a plurality of frames, it is characterized in that the predetermined data is repeatedly used, or the value of the predetermined data is changed to reduce the power and then repeatedly used.

According to the ninth aspect of the present invention, in addition to the configuration of the sixth aspect, the interpolating means has a predetermined number of frames of coded data or decoded digital signals for the number of discarded frames. It is characterized by interpolating a frame by expanding it.

According to the voice correction device of the present invention,
By performing error detection for each frame of encoded data or a digital signal obtained by decoding the encoded data,
If an error is detected, the frame is discarded, and an error-free one is subsequently output. The frame in which the error is detected in this way is not decoded or reproduced, but the sound of the frame having no error is immediately continued, so that the sound is not interrupted. Therefore, the occurrence of abnormal noise can be avoided without causing a failure in interpolation.

By outputting the error-free frames successively, there is a time lag with respect to the case where no error occurs, but as one method of the interpolation means,
By outputting predetermined data for the number of discarded frames, it is possible to perform compensation in a form close to natural. Also,
As another method of the interpolating means, the coded data or the decoded digital signal is expanded over a predetermined number of frames by the number of discarded frames (compensation is performed by expanding the frame length to the extent not perceivable). You can also

Further, by further including a data buffer for storing an error-free encoded data or a decoded digital signal, interpolation can be performed as in the conventional case even when the buffer becomes empty. Become.

[0021]

BEST MODE FOR CARRYING OUT THE INVENTION A voice correction device of the present invention is a device used on the decoding side of an audio / voice encoding / decoding device, which converts an audio signal or a voice signal into a digital signal, and outputs the digital signal. Is used in a system in which encoded data is generated by an encoding device in order to transmit or store the compressed or uncompressed data amount, and the encoded data is reproduced by a decoding device. Then, the decoding of the frame in which the error is detected is stopped and the frame without the latest error is immediately decoded, or the decoding is performed once including the frame in which the error is detected, and the error in the reproduced sound is detected. The part corresponding to the frame is discarded and the sound is output immediately after continuing the error-free frame part, and then a silent interval is detected and extended, or the frame length is extended, if necessary. It is characterized by doing.

Embodiments of the present invention will be described with reference to the drawings. FIG. 1 shows an audio correction apparatus according to a first embodiment of the present invention, and shows its function in a block diagram.

The error detector 111 is for detecting an error from the coded data input from the input terminal for decoding, and detects whether or not there is an error in the coded data by 1-bit error detection for each frame. Error detection information 103 that is a code
Output as

The encoded data buffer 102 is a buffer 102 for storing a plurality of frames of the encoded data 101 that has passed through the error detector 111, and the stored encoded data 101 is from the left to the right in the figure. It is designed to be transferred to and output. When this buffer is empty, a buffer empty signal is output, and when there is a data discard signal, the coded data transferred to the rightmost side in the figure and being output is discarded.

In the error detection information buffer 104, the error detection information 103 is stored in the buffer 104 from the input terminal 103 for each frame stored in the buffer 102. Then, like the data buffer 102, the error detection information 103 is transferred and output from the left side to the right side of the drawing.

The counter 107 is for counting the number of frames in which an error has been detected in the encoded data using the error detection information output from the error detection information data buffer 104.

The power calculator 105 is for calculating the level (power) of the encoded data output from the encoded data buffer 102 for each frame. Then, when the value of the counter 107 is not zero and the encoded data read from the buffer 102 is error-free, the power calculator 105 calculates the power P (n) of the frame from the encoded data.

The silent section detector 106 compares the power P (n) of the frame with a threshold level Pth to determine whether the frame is in a silent section. Then, the result is given to the switch 109.

Since the error-free coded data buffer 110 stores one frame of the coded data 101 output from the coded data buffer 102 that has no error using the error detection information and sequentially outputs it to the next stage. Buffer 1
02, and the stored data is output as interpolation data.

The switch 109 switches the switch 108 on the basis of three pieces of information, that is, the error detection information from the error detection information data buffer 104, the judgment result of the silent section detector 106, and the count value of the counter 107. Further, by giving a signal indicating data discard (data discard signal) to the buffers 102 and 104, the encoded data and error detection information to be output at that time are discarded, and the next encoded data from the buffers 102 and 104, Causes error detection information to be output.

Under the control of the switch 109, the switch 108 selects from the coded data (decoding data) from the coded data buffer 102, silence, and interpolated data, and supplies it to the decoder 112. The decoder 112 decodes the supplied encoded data and outputs it as an audio signal or a voice signal, and is equivalent to that used for decoding in ordinary data compression or digital modulation.

FIG. 2 is a flow chart showing the switching operation. Further, FIG. 3 shows an example of the state of processing. The operation will be described below using these.

First, the error detection of the encoded data input by the error detector 111 is performed, the error detection information of each frame is stored in the buffer 104 from the input terminal 103, and the encoded data is transmitted for each frame. Buffer 102
Stored in. Then, when the encoded data read from the buffer 102 has no error (reference numeral 201 in FIG. 2) and there is no error (reference numeral 202 in FIG. 2), the switch 109 is used.
Indicates that the state of the switch 108 is the decoder 1 for the encoded data.
Then, the decoded voice data is passed to the decoder 112 as it is so that there is no time lag so that it is given to the decoder 12 and the decoding is performed (reference numeral 206 in the figure).

For example, this processing is performed for Nos. 1 and 2 of the encoded data of FIG. 3, and the state of the switch 108 gives the encoded data to the decoder 112.

Next, when it is determined by the error detection information from the error detection information buffer 104 that an error has been detected in the encoded data from the encoded data buffer 102 (reference numeral 201 in FIG. 2), the buffer 102 is not empty. In this case (reference numeral 205 in the figure), the switch 109 gives a signal indicating data discard (data discard signal) to the buffers 102 and 104 so as to discard the encoded data of the frame, and the next encoded data, error The detection information is output (reference numeral 208 in the figure). If an error is also detected in the subsequent encoded data (reference numeral 201 in the figure), it is subsequently discarded (reference numeral 208 in the figure).

If the encoded data has no error (reference numeral 201 in the figure), the value of the counter 107 is not 0 (reference numeral 202 in the figure), and no silent section is detected (reference numeral 20 in the same figure).
3) Since it is not in a state to be inserted (reference numeral 204 in the figure), the switch 109 decodes to the decoder 112 as it is so that the state of the switch 108 gives the encoded data to the decoder 112. The audio data thus obtained is passed and decoded (206 in the figure).

For example, the third code of the coded data shown in FIG. 3 is processed when this error is detected, and the state of the switch 108 discards the coded data of the frame. Then, the process of the coded data No. 4 is performed when there is no error, and the state of the switch 108 gives the coded data to the decoder 112. Thus, sequentially error-free frames are sequentially applied to the decoder 112 via the switch 108. At this time, when an error is detected, the subsequent error-free frame is output and decoded earlier in time than when there is no error.

When the power calculator 105 and the silent interval detector 106 detect a silent interval in the encoded data (reference numeral 203 in the figure), the switch 109 is
The coded data of silence is supplied to the decoder 112 so that the state of the switch 108 is silenced, and the same data as the silent section is inserted as predetermined data (reference numeral 20 in the figure).
7). In addition, a specific sample value can be used as the predetermined data. It is possible to use the coded data determined to be the silent section, and it is also possible to repeatedly use it while reducing the power.

Then, the counter 107 reduces the count value of the number of frames in which the error is detected and inserts the data of the silent section until it becomes 0 (reference numeral 202 in the figure). Value is 0
At that point, the insertion of the silent frame is stopped, and the state of the switch 108 gives the encoded data to the decoder 112. In this way, by inserting data of a silent section over a plurality of frames, the silent section is equivalently extended, thereby correcting the time lag caused by discarding the encoded data.

For example, the number 5 of the coded data of FIG. 3 is the process when a silence section is detected, and the decoder 112 is used to decode the coded data of silence so that the state of the switch 108 becomes silenced. To the encoded data of 5
The silent section is extended by the turn and the inserted silent section. Then, when the count value of the counter 107 becomes 0, as shown in FIG.
The state of the switch 108 causes the encoded data to be given to the decoder 112, like the encoded data No. 6 of FIG.
After that, the process returns to the above-described switching between decoding and discarding of the encoded data based on the error detection information.

Further, when the error detection of the coded data continues for a long time, the coded data stored in the buffers 102 and 104 may be used up and become empty. In that case, the interpolation is performed using the error-free coded data buffer 110 as is conventionally done. In this case, an error is detected (reference numeral 201 in the figure), and the buffers 102, 10
4 becomes empty (reference numeral 205 in the figure), and the switch 10
The state of 8 gives the interpolation data to the decoder 112. Extrapolation is performed by repeatedly using the last error-free coded data stored in the buffer 110 while reducing the power.

As described above, in this embodiment, when an error is detected in coded data, the data is discarded and a silent section having the same length as the discarded data is inserted to extend the silent section. Data is being corrected. Even when error detection of coded data continues for a long time, extrapolation is performed by using the last error-free coded data. With this, even if there is an error in the encoded data, it is possible to realize a voice correction device that outputs a reproduced sound without causing an abnormal sound due to a failure of interpolation or the like and without interrupting the sound.

In this embodiment, the coded data is processed, but it is also possible to process the reproduced sound which has already been decoded. In addition, we mentioned a method of inserting silence frames by compensating for time lag, but another method is to use frames when it is determined to be a silence section, or to reduce power sequentially when repeatedly used. There is also.

FIG. 4 shows an audio correction apparatus according to the second embodiment of the present invention. This embodiment is an example in which encoded data is decoded and then decoded data is corrected. Is shown.

In the speech correction apparatus of this embodiment, first, the error detector 111 detects an error from the coded data input from the input terminal for decoding,
The decoder 112 decodes the encoded data and outputs the decoded data 301 as an audio signal or a voice signal.

Here, interpolation such as power may be performed at the time of decoding as is conventionally done. In this case, when backward prediction is used in encoding / decoding, even if there is no code error in the subsequent frame, the frame is affected by the error that has occurred before, so that the effect can be reduced.

A reproduced sound buffer 302 is provided at the output of the decoder 112, and an error detection information buffer 104 is provided at the output of the error detector 111. Then, in the same way as the first embodiment, in the reproduction sound buffer 302,
The decoded data 301 is stored for each frame for a plurality of frames, and the error detection information buffer 304 stores the error detection information 303 for each frame stored in the buffer 302.

The error-free coded data buffer 309 is a buffer for storing one frame of the decoded data 301 output from the reproduced sound buffer 302 that has no error from the error detection information, and is stored. The data is output as interpolation data.

The switch 307 is provided with the error detection information from the error detection information data buffer 304 and the counter 305.
The switch 108 is switched on the basis of the three pieces of information called the count value of.

The interpolation filter and decimation filter 308 (time axis expander) expands the decoded data 301 from the reproduced sound buffer 302 along its time axis and outputs it.
If the rate of expansion of the time axis is about 1%, it is difficult to perceive it. Therefore, in this embodiment, the decoded data 30 is, for example, 101/100 times as large as the discard of one frame.
100 frames in which the time axis of 1 is expanded are output.

The counter 305 is similar to the above-mentioned embodiment.
Using the error detection information output from the error detection information data buffer 304, the number of frames in which an error is detected in the encoded data is counted. Then, in this embodiment, the number of decompressed frames is counted together with the number of discarded frames from the state of the switch 307. In the case where the expansion rate of the time axis is as described above, when the expanded frame count reaches 100, the discarded frame count is decreased by one.

The switch 108, under the control of the switch 307, decodes data from the reproduced sound buffer 302, discards, outputs from the interpolation filter and thinning filter 308,
Select from the interpolation data and give to. The interpolation filter and decimation filter 308 decodes the coded data given and outputs it as an audio signal or a voice signal.

FIG. 5 is a flow chart showing the switching operation. Further, FIG. 6 shows an example of the state of processing. The switching operation will be described below using these.

In this embodiment, first, the decoded data 301 which is the decoded sound is stored in the buffer 302 for each frame, and the error detection information 303 of each frame is buffered regardless of the error detection of the decoded data 301. Store in 304.

Then, due to the error detection information from the error detection information buffer 104, no error is detected in the decoded data 301 from the reproduced sound buffer 302 (reference numeral 50 in FIG. 5).
1) If the number of error detections of the counter 305 is 0 (no error is detected in the previous data) (reference numeral 502 in the figure), the switch 108 causes the reproduced sound buffer 302 to operate.
The decoded data from is output as it is (reference numeral 503 in the figure).

For example, the No. 1 and No. 2 of the decoded data in FIG. 6 is this processing, and the state of the switch 108 is to output the decoded data.

Next, according to the error detection information from the error detection information buffer 304, the reproduced sound data buffer 302
When it is determined that an error has been detected in the decoded data from (No. 501 in FIG. 5), if the buffer 302 is not empty (No. 505 in FIG. 5), the switch 307 switches so as to discard the decoded data of the frame. 108, the data discard signal is sent to the buffers 302, 3
Then, the next decoded data and error detection information are output in accordance with No. 04 (reference numeral 507). If an error is also detected in the subsequent decoded data (reference numeral 501 in the figure), it is continuously discarded (reference numeral 507 in the figure). The discarded number is counted by the counter 305.

For example, the No. 1 and No. 2 of the decrypted data in FIG. 6 is this processing, and the state of the switch 108 is that the decrypted data is discarded.

If there is no error in the decoded data (reference numeral 501 in the figure), the value of the counter 305 is not 0 (reference numeral 502 in the figure). Therefore, the switch 307 determines that the state of the switch 108 is the interpolation filter and the thinning filter 308. Then, the time axis is expanded and the decoded data is given (reference numeral 5 in FIG.
04). At this time, when the count number of the decoded data that has been expanded on the time axis reaches 100, the count number of the discarded data is decremented by 1.

For example, after this error is detected at the coded data No. 3 in FIG. 3, the process becomes when the coded data No. 4 is error free, and the state of the switch 108 is time-axis expanded decoding. You will be given data. Although the time axis is expanded 101/100 times in the coded data No. 4 and after, the coded data No. 104 has the decoded data 30
No error is detected at 1 (reference numeral 501 in FIG. 5), the number of error detections of the counter 305 becomes 0, and the decoded data from the reproduction sound buffer 302 is directly output (reference numeral 503 in FIG. 5).

When the error detection of the decoded data continues for a long time, the buffers 302 and 3 are similar to the first embodiment.
When 04 becomes empty, interpolation is performed using the error-free coded data buffer 309 as is conventionally done. In this case, an error is detected (reference numeral 50 in FIG.
1), the buffers 102 and 104 become empty (reference numeral 505 in the figure), and the state of the switch 108 gives interpolation data. Extrapolation is performed by repeatedly using the last error-free decoded data stored in the buffer 110 while reducing the power.

As described above, in this embodiment, the frame of which the error is detected is discarded, and the waveform of the output frame is expanded on the time axis to compensate for the time lag caused by immediately outputting the error-free frame. This is an example of a case of performing. In this method, as compared with the first embodiment, a silence section is required for compensating for the time shift in the first embodiment. However, in the case of an audio signal in which this silence section does not exist for a long time, etc. Can also be applied.

In this embodiment, the processing is performed on the decoded reproduced sound, but it may be processed at the stage of the encoded audio data before decoding. In that case, FIG.
As shown in, the decoder 112 may not be provided in the first stage of the audio correction apparatus, but may be provided in the latter stage of the switch 701 controlled by the switch 307.
In this case, the decoder 112 performs normal decoding, and the time-axis decompressor 308 performs processing such as time-axis decompression of the reproduced sound and outputting. Then, the output is switched using the switch 108 interlocking with the switch 701 as in the above-described embodiment. Even in this case, the same processing as that in the above embodiment can be performed.

As a method of extending the time axis, if the extension length is 1 to several samples, the last sample of the frame is repeated once, and the sample at a specific position is repeatedly used. The amount of calculation can be significantly reduced as compared with the method using a thinning filter or an interpolation filter.

By combining the first and second embodiments, more various audio signals or voice signals can be dealt with.

As described above, the frame in which the error is detected is discarded,
Although I have mainly described the method of correcting the time shift immediately by using error-free frames, when the time shift does not matter because the function only reproduces sound like a CD player. Does not need to add such a correction mechanism.

The present invention is applicable to any audio / voice signal coding / decoding apparatus as long as the buffer capacity allows, that is, the delay due to buffer data storage is allowed. It is possible to

[0068]

As described above, according to the present invention, when decoding coded voice data having a code error, a frame containing a code error is discarded and immediately an error-free frame data is output. By reproducing the sound by doing so, it is possible to realize the sound correction method and the apparatus for suppressing the generation of the abnormal sound due to the failure of the interpolation and outputting the reproduced sound without the sound being interrupted.

[Brief description of drawings]

FIG. 1 is a block diagram showing a first embodiment of the present invention.

FIG. 2 is a flowchart showing the operation of the switch of the first embodiment.

FIG. 3 is a diagram showing how processing is performed in the first embodiment.

FIG. 4 is a block diagram showing a second embodiment of the present invention.

FIG. 5 is a flowchart showing the operation of the switch of the second embodiment.

FIG. 6 is a diagram showing a processing state of a second embodiment.

FIG. 7 is a block diagram (partially) showing another processing method of the first embodiment.

[Description of Codes] 101 coded data, 102 coded data buffer, 103 error detection information, 104 error detection information buffer 105 power calculator, 106 silence interval detector 107 counter, 108 switch, 109 switcher 110 interpolation Error-free coded data buffer for use 301 coded data, 302 coded data buffer, 303 error detection information, 304 error detection information buffer 305 counter, 308 time axis expansion means, 309
Error-free coded data buffer 701 switch for interpolation

─────────────────────────────────────────────────── ─── Continuation of the front page (51) Int.Cl. ⁶ Identification code Office reference number FI Technical display location H04L 1/00 H04L 1/00 Z

Claims (9)

[Claims] 1. A voice correction method for reproducing a digital signal obtained by decoding coded data obtained by converting an audio signal or a voice signal into a digital signal and encoding the digital signal. The digital signal obtained by decoding the coded data is divided into frames, and the coded data is subjected to error detection for each frame, and the coded data or the decoded digital signal and the error detected by the error detection. The detection code and a plurality of frames for each frame are stored, when an error is detected in the stored frame, the frame is discarded, and an error-free one of the stored frames is subsequently output, By interpolating the temporal shift that occurs when a frame is discarded, the coded data or the interpolated frame is decoded. Voice correction method characterized by providing a Ijitaru signal. 2. A voice correction device used for reproducing a digital signal obtained by decoding encoded data obtained by converting an audio signal or a voice signal into a digital signal and encoding the digital signal, A coded data or a digital signal obtained by decoding the coded data is divided for each frame, an error detector that detects an error in the coded data for each frame, and the coded data or the decoded digital signal, A buffer for storing a plurality of frames of the error detection code detected by the error detector for each frame, and when an error is detected in the frame stored in the buffer, the frame is discarded from the buffer. A switch for successively outputting an error-free one of the stored frames from the buffer, An audio correction device comprising: an interpolating unit that interpolates a temporal shift generated when the frame is discarded and provides the coded data obtained by interpolating the frame or the decoded digital signal. 3. A data buffer for storing an error-free one of the encoded data or the decoded digital signal, wherein the interpolation means, when the buffer is empty, 3. The audio correction apparatus according to claim 2, wherein the encoded data or the decoded digital signal stored in a buffer is selected and given. 4. A decoder for decoding the coded data, the error of which is detected by the error detector, wherein the decoder decodes the coded data to obtain a digital signal. 3. The method according to claim 2, wherein
The audio correction device described. 5. A decoder for decoding the coded data obtained by interpolating the frame by the interpolation means, wherein the decoder outputs a digital signal obtained by decoding the coded data. The audio correction device according to claim 2, wherein 6. The interpolation means detects whether the coded data or the decoded digital signal is in a silent section by comparing with a predetermined level, and when the silent section is detected, the discarded frame is discarded. 3. The audio correction apparatus according to claim 2, wherein a predetermined number of predetermined data are inserted into the encoded data or the decoded digital signal to interpolate the frame. 7. The silent section detector compares the coded data or the decoded digital signal with a predetermined level to detect a silent section, and when the silent section is not detected. A switch for selecting the coded data or the decoded digital signal stored in the buffer and selecting the predetermined data by the number of the discarded frames when the silent section is detected. The audio correction device according to claim 6, wherein the audio correction device is provided. 8. The predetermined data is the same data as a silent section, encoded data of a frame determined to be a silent section,
It is one of specific sample values, and when inserting the predetermined data over a plurality of frames, the predetermined data is repeatedly used, or the value of the predetermined data is changed to reduce the power. The voice correction device according to claim 6 or 7, which is used repeatedly. 9. The interpolation means expands the coded data or the decoded digital signal by a predetermined number of frames by the number of the discarded frames to interpolate the frames. The voice correction device according to claim 2.