JPH0844393A - Processing device for audio encoded data - Google Patents

Abstract

PURPOSE:To prevent sound quality from being degraded by reducing the effect of errors while applying error correction processing to respective band signals encoded while being divided into plural bands. CONSTITUTION:This device is, for example, a device applied with a band division encoding system and is constituted of an encoding processing part 1 encoding audio data while dividing them into plural frequency bands, a bit stream processing part 2 performing an assembling processing to a bit stream and a disassembling processing from a bit stream and a decoding processing part 3 decoding bits into audio data. The inverse bit allocation processing part 31 of the decoding processing part 3 obtains a quantization level from quantization level information recorded on a bit stream state. An interpolation means 32 rearranges the order of data in respective sub-band blocks divided into odd number-th and even number-th sample data from the inverse bit allocation processing part 31 and performs error correction to respective sub-band blocks and an inverse quantization processing part 33 applys inverse quantization processing to the result in accordance with the quantization level.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a device for processing audio coded data which performs frequency decomposition on digital audio data to decode coded audio data and corrects errors.

[0002]

2. Description of the Related Art Generally, there are various high-efficiency coding methods for reducing the data amount of digital audio data, for example, a block frequency band division method, so-called transform coding (transform coding), , A non-blocking frequency band division method, so-called band division coding (subband coding: SBC), and the like.

These high-efficiency coding techniques reduce redundant information by utilizing the property of uneven distribution in the frequency axis direction.

For example, as shown in FIG. 11, in the data processing by the band division encoding method, as shown in FIG. 11, an encoding processing unit 100 for dividing and encoding in a frequency band, a bit stream assembling process, and a bit stream are provided. Is performed by a device including a bitstream processing unit 200 that performs the decomposing process from the above and a decoding processing unit 300 that decodes the audio data.

The encoding processing unit 100 divides the input audio data of a plurality of channels into M frequency bands for each channel and encodes them.

At the time of encoding, the bit stream processing unit 200 rearranges the band signals of a plurality of channels encoded by the encoding processing unit 100 into a bit stream format and assembles them into one data.

Here, for example, the division number 32 and the code length 32
As shown in FIG. 12, the data in the bitstream state generated by the 0-sample band-division coding method is used as a word-length data part WL for recording information indicating the quantization level when quantizing each band signal. Information indicating the maximum value of data in each band signal, a scale factor data section SF recording a so-called scale factor, and a sign data section SIGN recording information indicating the polarity of each band signal,
It is composed of a subband block unit DATA for recording 32 band signals (hereinafter referred to as subband blocks).

The subband blocks DATA (0) to DATA (31) of the subband block unit DATA are
Each of the 10 sample data D (0) to D (9) constitutes one subband block.

Further, the word length data section WL,
The scale factor data part SF and the sign data part SIGN include the sub-band block DATA (0).
~ Word length data W corresponding to DATA (31)
L (0) to WL (31), scale factor data SF
(0) -SF (31), sign data SIGN (0)-
Record each SIGN (31).

The subband blocks DATA (0) to
Sample data D (0) to D (9) of DATA (31)
Since the quantization level differs depending on the frequency band, the length is usually not fixed, but here it is assumed that the total amount of information is constant. The information recorded in the word length data section WL, the scale factor data section SF, etc. is assumed to be the same for each subband block.

Therefore, the bit stream processing section 20 is
0 assembles each band signal into a bit stream state according to the above-mentioned format at the time of encoding. Further, at the time of decoding, the above-mentioned bit stream state is decomposed and distributed to the data strings of each subband block.

The decoding processing unit 300 decodes each sub-band block decomposed by the bit stream processing unit 200 into audio data.

A system configuration of an audio coded data processing device to which the band division coding method as described above is applied will be described with reference to FIG. 13. The coding processing unit 100 shown in FIG. An analysis filter bank 101 that divides the frequency band and a quantizer 102 that performs a quantization process on each of the divided subband blocks are provided.
Also, the bit stream processing unit 2 shown in FIG.
00 has a memory 201 for recording the data assembled in the bitstream state.

Further, the decoding processing section 30 shown in FIG.
0 includes an inverse quantizer 301 that performs inverse quantization processing on each subband block, and a synthesis filter bank 302 that synthesizes each subband block subjected to inverse requantization processing.

The audio coded data processing apparatus will be described in detail. The audio coded data processing apparatus is shown in FIG.
0 is a division processing unit 110 that divides into M frequency bands by the analysis filter bank 101 shown in FIG.
A bit allocation processing unit 120 that assigns a quantization level to each subband block, and a quantization processing unit 130 that performs quantization processing according to the quantization level by the quantizer 102 shown in FIG. Has been done.

The bit stream processing section 200 is also provided.
Is an assembly processing unit 210 that assembles each subband block that has been quantized at the time of encoding into data in a bitstream state, and decomposes the data in a bitstream state at the time of decoding into a data string of each subband block. And a distribution processing unit 220 for distributing.

Further, the decoding processing unit 300 has an inverse bit allocation processing unit 310 for obtaining the quantization level from the quantization level information indicating the quantization level recorded on the bit stream state, and FIG. An inverse quantizer 301 performs an inverse quantization process on each subband block according to the quantization level by the inverse quantizer 301.
And a synthesis processing unit 330 that synthesizes each sub-band block that has been subjected to inverse quantization processing by the synthesis filter bank 302 shown in FIG. 13 and an interpolation processing unit 340 that corrects errors in the synthesized audio data. It is configured.

Further, the audio coded data processing apparatus includes a timing control unit 40 for controlling a frame cycle in the coding / decoding processing by pulse oscillation, for example.
0, the audio coded data processing device and ECC
The system control unit 500 controls input / output data between blocks and digital devices.

First, the processing at the time of encoding will be described.

The division processing unit 110 divides the audio data, which is a wideband signal of the sampling frequency fs, into M equal sub-bands of the sampling frequency fs / M. The bit allocation processing unit 120 allocates the quantization level so that the total data amount is constant according to the data amount of each subband block divided by the division processing unit 110.

The quantization processing unit 130 performs a quantization process on the data of each subband block at the quantization level assigned by the bit allocation processing unit 120.

The assembly processing unit 210 is the quantization processing unit 1 described above.
Each sub-band block that has been quantized by 30 is assembled into a bit stream state together with the quantized level information indicating the quantized level according to a predetermined format, and recorded in the memory 201 shown in FIG.

Therefore, the data assembled in the bit stream state is added to the ECC (Error Check) on the reproducing side (not shown) for adding information for error correction.
ing and Correcting) block, and then recorded on a recording medium or transmitted.

Here, in the ECC block, error flag information is added to the data that cannot be error-corrected and output. Generally, a Reed-Solomon code, a product code, or the like as a block correction code is used as the error correction code. In the Reed-Solomon code, for example, as shown in FIG. 15, 1 byte (= 8
Error flag information is added in bit units. For example, as shown in the figure, error flag information F5, F6,
When F7 and F8 stand in byte units, in the case of a subband block DATA (k _a ) to which 1 bit of sample data is allocated 8 bits, sample data D
(3), D (4), D (5), and D (6) indicate an error.

Further, in the case of the sub-band block DATA (k _b ) in which 1 bit of sample data is assigned 6 bits, the sample data D (4), D (5), D
(6), D (7), D (8), and D (9) are in error, and in the case of the subband block DATA (k _c ) in which 1 bit of sample data is assigned 9 bits,
Sample data D (2), D (3), D (4), D
(5) and D (6) indicate an error.

In the above product code, as shown in FIG. 16, in the state in which a plurality of data streams in the form of bit streams are combined, one sync signal unit is formed, and error correction parity is applied to the sync block data. The information IP is added. Also, for these synchronized block data,
Parity information OP is added for the burst error to form a product code block ECB.

Therefore, with respect to the data which has been error-corrected and which cannot be error-corrected, the ECC block outputs the data in the bit stream state to which the error flag information is added as described above. It

Next, the processing at the time of decoding will be described.

The distribution processing section 220 receives the error-corrected data in the bit stream state from the ECC block. Here, regarding the data that cannot be error-corrected in the ECC block,
The error flag information as described above is added to the data in the bit stream state and input.

The distribution processing section 220 decomposes the data in the bitstream state and distributes it to the data string of each subband block.

The inverse bit allocation processing unit 310 is
The quantization level for each subband block is obtained from the quantization level information recorded in the bitstream data.

The inverse quantization processing unit 320 performs inverse quantization processing on each subband block according to the quantization level obtained by the inverse bit allocation processing unit 310.

The synthesis processing section 330 synthesizes each sub-band block subjected to the inverse quantization processing by the inverse quantization processing section 320 and outputs it as audio data.

The interpolation processing unit 340 determines which sample data of the audio data output from the synthesis processing unit 330 the data to which the error flag information is added corresponds to and corrects the error.

For example, in the ECC block, when the error flag information is added by the product code, the error flag information is added in block units. Therefore, the interpolation processing unit 340 performs error correction and error detection in block units.

Alternatively, when the error flag information is added by the Reed-Solomon code, the error flag information is
It is added in 1-byte units. Here, in each band signal, since the quantization level is different, the range in which the error influences differs depending on how many bits are allocated to one sample data, but the error flag information is recorded in each subband block. Therefore, it is possible to accurately determine which sample data the error flag information corresponds to.

In this case, in the error correction, the sample data immediately before the error flag information is set is held, and the sample data at the end of the error is the average of the next correct sample value and the held value. The error is corrected by taking

For example, as shown in FIG. 17A, an error flag is added to the sample data D (3), D (4), D (5), D (6) of the kth subband block DATA (k). If the information stands, sample data D
(3) holds sample data D (2), sample data D (4) holds sample data D (3), and sample data D (5) holds sample data D (4). Then, for the sample data D (6) at the end of the error, the value of the sample data D (5) and the value of the sample data D (7) are averaged.

Further, as shown in FIG. 17 (b),
When error flag information is set in the sample data D (6), D (7), D (8), D (9) of the subband block DATA (k), the sample data D
(6) holds sample data D (5), sample data D (7) holds sample data D (6), and sample data D (8) holds sample data D (9). Then, for the sample data D (6) at the end of the error, the value of the sample data D (8) and the sample data D of the next subband block DATA (k + 1)
Take the average with the value of (0).

As shown in FIG. 17C, the error flag information is added to the sample data D (0), D (1), D (2), D (3) of the subband block DATA (k). If is standing, sample data D (0)
Holds the sample data D (9) of the previous subband block DATA (k-1),
(1) is the sample data D (0)
(2) holds the sample data D (3). Then, regarding the sample data D (4) at the end of the error,
Value of sample data D (3) and sample data D (5)
Take the average with the value of.

The error correction method as described above is used as an error correction method for uncoded audio data, for example, in a D2 video tape recorder.

[0042]

However, conventionally, as described above, it has been determined which audio data sample data the data for which the error flag information is set corresponds to, and the error correction is performed after the decoding process. For this reason,
When an error is detected in one sample data in one subband block, the effect of the error is, for example,
The data in the bit stream state shown in FIG. 12 has spread to 32 sample data, that is, the number of divided frequency bands. That is, 3 after decryption
It was supposed to set an error flag to 2 sample data.

When the error correction is performed on the sample data of such long audio data, the influence of the error cannot be reduced and it is perceived as an abnormal sound, so that the deterioration of the sound quality cannot be prevented.

Therefore, the present invention has been made in view of the conventional circumstances as described above, and has the following objects.

That is, an object of the present invention is to provide an audio coded data processing device capable of preventing the deterioration of sound quality by reducing the influence of an error.

..

In order to solve the above problems, an audio coded data processing apparatus according to the present invention decodes audio data coded by dividing into a plurality of frequency bands and interpolating means. A device for processing audio coded data for performing error correction processing according to, wherein the interpolation means performs error correction processing on each of the coded band signals.

Further, in the audio coded data processing apparatus according to the present invention, the interpolating means replaces the band signal containing the error sample data with the band signal immediately before the band signal, and each coded band. It is characterized in that the signal is subjected to error correction processing. Further, in the audio coded data processing device according to the present invention, the band signal is composed of a plurality of predetermined sample data, and the interpolating means rearranges the order of the sample data and codes each band. It is characterized in that the signal is subjected to error correction processing.

Further, in the audio coded data processing device according to the present invention, the interpolation means arranges the order of the sample data in each band signal divided into the odd-numbered sample data and the even-numbered sample data. Instead, error correction processing is performed on each of the coded band signals.

Further, in the audio coded data processing apparatus according to the present invention, the interpolation means, when the sample data at the first position of the band signal is in error, samples the data at the first position. An error correction process is performed on each coded band signal by substituting it with zero.

Further, in the audio coded data processing apparatus according to the present invention, the interpolation means, when the sample data at the last position of the band signal is an error, outputs the sample data at the last position. It is characterized in that the sample data immediately before the sample data at the last position is replaced with the sample data and error correction processing is applied to each of the coded band signals.

Further, in the audio coded data processing apparatus according to the present invention, the interpolating means, when the sample data at the middle position of the band signal is an error, samples the middle position sample data. An error correction process is performed on each of the coded band signals by substituting the average value of the sample data immediately before and the sample data immediately after the sample data at the middle position.

Further, in the audio coded data processing apparatus according to the present invention, the interpolation means replaces the sample data at the middle position with the sample data immediately before when the sample data immediately after the error is an error. Then, error correction processing is performed on each of the coded band signals.

[0053]

In the audio coded data processing apparatus according to the present invention, the interpolation means performs error correction on each of the coded band signals of the audio data coded by being divided into a plurality of frequency bands.

Further, in the audio coded data processing apparatus according to the present invention, the interpolation means replaces the band signal containing the error sample data with the band signal immediately before the band signal, and each coded band. Perform error correction processing on the signal.

Further, in the audio coded data processing apparatus according to the present invention, the interpolating means rearranges the order of the sample data of each band signal composed of a predetermined plurality of sample data and codes the coded data. Error correction processing is performed on each band signal.

Further, in the audio coded data processing apparatus according to the present invention, the interpolation means arranges the order of the sample data in each band signal divided into the odd-numbered sample data and the even-numbered sample data. Instead, an error correction process is performed on each of the coded band signals.

Further, in the audio coded data processing apparatus according to the present invention, the interpolating means may perform the following processing if the sample data at the first position of the band signal is in error.
The sample data at the first position is replaced with zero, and error correction processing is performed on the encoded band signals.

Further, in the audio coded data processing apparatus according to the present invention, the interpolating means may perform the following processing if the sample data at the last position of the band signal is erroneous.
The sample data at the last position is replaced with the sample data immediately before the sample data at the last position, and error correction processing is performed on the encoded band signals.

Further, in the audio coded data processing apparatus according to the present invention, the interpolating means may be arranged so that when the sample data at the middle position of the band signal has an error,
The sample data at the midway position is replaced with the average value of the sample data immediately before and the sample data immediately after the sample data at the midway position, and error correction processing is performed on each of the coded band signals.

Further, in the audio coded data processing apparatus according to the present invention, the interpolation means replaces the sample data at the middle position with the sample data immediately before when the sample data immediately after the above is an error. Then, error correction processing is performed on each of the encoded band signals.

[0061]

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

The audio encoded data processing apparatus of the present invention shown in FIG. 1 is, for example, one to which a band division encoding method is applied, and includes an encoding processing unit 1 for dividing and encoding into a plurality of frequency bands. , A bitstream processing unit 2 that performs assembly processing into a bitstream and decomposition processing from the bitstream, and a decoding processing unit 3 that decodes into audio data.

The decoding processing section 3 is provided with a correction means 32 for correcting an error in each band signal which has been encoded by the encoding processing section 1.

That is, the encoding processing unit 1 includes, for example, a division processing unit 11 that divides into 32 frequency bands by an analysis filter, a bit allocation processing unit 12 that assigns a quantization level to each band signal, The quantization processing unit 13 performs a quantization process according to the quantization level.

Further, the bit stream processing section 2 is
An assembly processing unit 21 that assembles each band signal that has been quantized at the time of encoding into data in a bit stream state,
It is composed of a distribution processing unit 22 which decomposes data in a bit stream state at the time of decoding and distributes the data into a data string of each band signal.

Further, the decoding processing section 3 obtains the quantization level from the quantization level information indicating the quantization level recorded on the bit stream state, and the inverse bit allocation processing section 31 according to the quantization level. An inverse quantization processing unit 32 that performs inverse quantization processing on each band signal, the interpolating unit 32, and a synthesis processing unit 33 that synthesizes each band signal subjected to the inverse quantization process by a synthesis filter. There is.

Further, in this audio coded data processing apparatus, for example, a timing control section 4 for controlling a frame cycle at the time of coding / decoding processing by pulse oscillation.
And a system controller 5 for controlling input / output data between the audio coded data processing device and the ECC block or digital device.

First, the encoding process will be described.

The division processing section 11 divides audio data, which is a wideband signal having a sampling frequency fs, into 32 equal sampling frequencies fs / 3 by an analysis filter.
Divide into 2 subbands.

The bit allocation processing unit 12 allocates the quantization level so that the total data amount becomes a constant amount according to the data amount of each band signal divided by the division processing unit 11.

The quantization processing unit 13 applies the quantization processing to the data of each band signal at the quantization level assigned by the bit allocation processing unit 12.

The assembly processing unit 21 is the quantization processing unit 13 described above.
Each band signal that has been quantized by 0 is assembled into a bit stream state together with quantization level information indicating the above-mentioned quantization level according to a predetermined format, and recorded in a memory (not shown). Here, the data in the bit stream state will be specifically described.

For example, as shown in FIG. 2, the data in the bit stream state generated by the band division coding method with the number of divisions of 32 and the code length of 320 samples has the quantization level at the time of quantizing each band signal. A word length data section WL for recording information, a maximum value of data in each band signal, a scale factor data section SF for recording a so-called scale factor, and a sign for recording information indicating the polarity of each band signal. It is composed of a data part SIGN and a sub-band block part DATA for recording 32 band signals.

Each band signal (hereinafter referred to as a subband block) of the subband block unit DATA constitutes one subband block with 10 sample data D (0) to D (9). The respective sample data are even-numbered sample data D (0), D
Odd sample data D (1), D (3), D (5), (2), D (4), D (6), D (8) are placed in front.
D (7) and D (9) are recorded in the latter stage.

The word length data section WL,
The scale factor data part SF and the sign data part SIGN include 32 sub-band blocks DATA.
Word length data WL (0) to WL (31) corresponding to (0) to DATA (31), scale factor data SF (0) to SF (31), sign data SIGN.
(0) to SIGN (31) are recorded.

It should be noted that the sample data D (0) to D (9) of each subband block are not usually defined in length because the quantization level differs depending on the frequency band, but here the total amount of information is Is constant. Also,
The information recorded in the word length data section WL, the scale factor data section SF, etc. is assumed to be the same for each subband block.

Therefore, the data assembled into the bitstream state as shown in FIG. 2 by the assembling unit 21 has ECC (Error) on the reproducing side (not shown) to which information for the next error correction is added. Check
ing and Correcting) block, and then recorded on a recording medium or transmitted.

In the ECC block, error flag information is added to the data that cannot be error-corrected and the data is output.

That is, as shown in FIG. 3, corresponding to the word length data WL (0) to WL (31) of the word length data section WL shown in FIG. 2, the word length data WL (0) to WL (0) to Error flag information EWL (0) to EWL (31) of WL (31) is added.

Further, the scale factor data SF (0) to S (S) of the scale factor data section SF shown in FIG.
F (31) and the signature data section S shown in FIG.
IGN sign data SIGN (0) to SIGN (3
Corresponding to 1), the scale factor data SF
Error flag information ESF (0) of (0) to SF (31)
~ ESF (31) and the signature data SIGN
Error flag information ESIG of (0) to SIGN (31)
N (0) to ESIGN (31) are added.

Further, each sub-band block DATA (0) to DATA (0) of the sub-band block unit DATA shown in FIG.
Odd sample data D (1) of DATA (31)
~ D (9) and even-numbered sample data D (0)
To D (8), error flag information ed (1) to e
d (9) and ed (0) to ed (0) are added.

The error flag information as described above is "H".
When IGH = 1 ", it indicates that the data corresponding to the error flag information is an error.

Therefore, with respect to the data which has been error-corrected and which has not been error-corrected, the ECC block outputs the data in the bitstream state to which the above-mentioned error flag information is added. .

Next, the processing at the time of decoding will be described.

The data in the bit stream state from the ECC block is input to the distribution processing section 22.

The distribution processing unit 22 decomposes the data in the bit stream state from the ECC block and distributes it to the data string of each subband block.

The inverse bit allocation processing section 31 obtains the quantization level for each subband block from the quantization level information recorded in the data string of each subband block.

In the correcting means 32, the sample data in each sub-band block is converted into the odd-numbered sample data D (1), D (3), D (5), D as shown in FIG.
(7), D (9) and even-numbered sample data D
Each sub-band block divided into (0), D (2), D (4), D (6) and D (8) is input.

Then, the correction means 32 causes the odd-numbered sample data D (1), D (3), D (5), D.
(7), D (9) and even-numbered sample data D
The order of the sample data in each subband block divided into (0), D (2), D (4), D (6), and D (8) is the original order, that is, the sample data is generated. Order D (0), D (1), D (2), D (3) ..., D
The sub-band blocks are rearranged into (9) to perform error correction on each sub-band block.

The correction means 32 will be specifically described below.

For example, as shown in FIG. 4A, the word length data W of the subband block DATA (k _a ) (k _a : indicates the subband block index).
L (k _a ), scale factor data SF (k _a ), and sign data SIGN (k _a ) are all normal,
Error flag information ed (6), ed (8), ed (1), of the sample data D (6), D (8), D (1), D (3) of the subband block DATA (k _a ). ed
When (3) is standing (= "1"), that is, when the sample data D (6), D (8), D (1), and D (3) are present, first, as described above. , The order of the sample data D0 to D (9) divided into the odd number and the even number is rearranged. For the sample data D (1), the immediately preceding sample data D (0) is held, and the next correct sample data D (2) and the held value are averaged. For the sample data D (3), similarly to the case of the sample data D (1), the value held by holding the immediately preceding sample data D (2) and the next correct sample data D (4) And take the average. Similarly, for the sample data D (6), the immediately preceding sample data D (5) is held, and the next correct sample data D (70 and the held value are averaged to obtain the sample data D ( Similarly for 8), hold the immediately preceding sample data D (7),
The next correct sample data D (9) and the held value are averaged.

As described above, by rearranging the data in the subband block divided into the odd-numbered sample data and the even-numbered sample data as described above, for example, one error flag information is plural. It is possible to prevent a series of errors that occur when the sample is spread over.

For example, as shown in FIG. 4B, the order D (0), D (1), D in which the sample data was generated.
(2), D (3), ..., D (9) are arranged in the sub-band block (k _b ) (k _b : index of the sub-band block). Error flag information ed (3), e at the same position as _a )
When d (4), ed (5), and ed (6) are standing (= "1"), that is, sample data D (3), D
If (4), D (5), and D (6) are errors, if errors are corrected to these, the errors will be continuous.

However, the sub-band block DATA
When the data in each subband block input in the state of (k _a ) is rearranged as described above and error correction is performed, the error becomes a dispersed form, and the influence of the sound of the error is detected. It gets harder.

In the above example, when the sample data D (j) in the middle of the subband block (k _a ) (j: indicates the index of the sample data) is in error, the sample data D (j) is generated. Is replaced with the average value of the sample data D (j-1) immediately before the sample data D (j) and the sample data D (j + 1) immediately after the sample data D (j). If the sample data D (j + 1) immediately after () is an error, the sample data D (j) is replaced with the sample data D (j-1) immediately before the sample data D (j).

Also, the subband block DATA
(K _a ), word length data WL (k _a ), scale factor data SF (k _a ), and sign data SI.
Although it is assumed that GN (k _a ) is all normal, for example, as shown in FIG.
Word length data WL (7) corresponding to TA (7)
Is an error, the above subband block D
All sub-band block DATA after ATA (7)
(8) to DATA (31) are regarded as errors, and the subband blocks DATA (7) to DATA (3) in error are considered.
Replace 1) with the previous subband block. Therefore, the sub-band blocks DATA (7) to DATA
All (31) will be replaced with the subband block DATA (6). Alternatively, as shown in FIG. 5B, the sub-band blocks DATA (6), DATA
Sign data SIGN (6), SIG corresponding to (7)
If N (7) is in error, the sub-band blocks DATA (6) and DATA (7) are regarded as errors, and the sub-band blocks DATA (6) and DATA are data.
Replace (7) with the previous subband block. Therefore, the sub-band blocks DATA (6), DATA
(7) will be replaced with the subband block DATA (5).

Also, when the scale factor data is in error, it is replaced with the preceding subband block, as in the case where the word length data is in error. That is, in this case, as shown in FIG. 6, when the subband block DATA (N) is in error, the subband block DATA (N) is the previous subband block DATA (N-1). , And thus the subband block DATA (N-1) is repeated.

Further, as shown in FIG. 5C, when the subband block DATA (19) and the subband block DATA (20) are in error, the subband block as described above is used. Correct the error in. Here, for example, the subband block DATA (1
If the number of errors of the sample data D (0) to D (9) in 9) is greater than or equal to the set threshold value, the sample data D (0) in the subband block DATA (19). -D (31) are all regarded as an error, and the subband block DATA (19) is replaced with the immediately preceding subband block DAT in the same manner as in the case of FIG. 5B.
Replace with A (18).

When the error correction is performed as described above, if the number of times the data of the previous subband block is repeated exceeds the set threshold value, the reproduced sound is set to be silent. To do.

The sample data rearrangement process as described above is performed, for example, in the recording area M01 of the word length data WL (0) to WL (31) as shown in FIG.
And scale factor data SF (0) to SF (31)
Recording area M02 and subband block DATA
This is performed by the memory address at the time of signal processing using a memory composed of the recording area M03 of (0) to DATA (31).

An error flag EF corresponding to each data and a repeat flag EBAND indicating that the data of the preceding subband block is repeated are added to each data recorded in each of the recording areas M01, M02, M03.

In the recording area M03, the sub-band blocks DATA (0) to DATA (31) are
The subband blocks DATA (0) to DATA (3
Polarity information SIGN (0) to SIGN indicating the polarity of 1)
Data multiplied by (31) ODATA (0) to ODA
TA (31) is recorded.

Further, for example, the repeat flag EBAND
When (k) is "1", the data of the subband block DATA (k) corresponding to the repeat flag EBAND (k) which is "1" is not written. As a result, the sub-band block DATA
Subband block DATA (k-
The data of 1) will be repeated.

Correction means 3 for correcting errors as described above
The error correction processing step of No. 2 will be specifically described with reference to the flowcharts shown in FIGS. 8 and 9.

First, as shown in FIG. 8, the repeat flag EB corresponding to the subband block DATA (k).
AND (k) and the error processing flag WFG are initialized (EBAND (k) = 0, WFG = 0) (step 1).

Next, it is determined whether or not the error flag information EWL (k) of the word length data WL (k) or the error processing flag WFG is set (step 2).

If the error flag information EWL (k) or the error processing flag WFG is set (EWL
(K) = 1 or WFG = 1), that is, the word length data WL (k) is in error, or the sub-band block DATA is currently in check symmetry.
Subband block DATA (x) before (k)
When (0 ≦ x <k) has an error, the repeat flag EBAND (k) indicating that the data of the previous subband block DATA (k−1) is repeated is set to “1” (EBAND ( k) = 1) (step 3).

Next, the scale factor data SF (k)
It is determined whether or not the error flag information ESF (k) is set (step 4).

When the error flag information ESF (k) is set (ESF (k) = 1), that is, when the scale factor data SF (k) is in error,
The repeat flag EBAND (k) indicating that the data of the previous subband block DATA (k-1) is repeated is set to "1" (EBAND (k) = 1) (step 5).

Next, it is judged whether or not the error flag information ESIGN (k) of the sign data SIGN (k) is set (step 6).

When the error flag information ESIGN (k) is set (ESSIGN (k) = 1), that is, when the sign data SF (k) is in error, the previous subband block DATA ( The repeat flag EBAND (k) indicating that the data of (k-1) is repeated is set to "1" (EBAND (k) = 1) (step 7).

The processing of steps 1 to 7 shown in FIG. 8 is performed by using 32 subband blocks DATA (0) to DATA (DA).
This is performed for TA (31) (0 ≦ k ≦ 31).

Next, as shown in FIG. 8, the count judgment flag CFG is cleared to zero (CFG = 0) (step 8).

Next, one subband block DATA
Zero clear (Ev = 0) of the error sample data number Ev indicating the error number of the sample data D (0) to D (9) in (k) is performed (step 9).

Then, the error sample data number Ev
, The error flag information ed (k, j) of the j-th sample data D (j) of the subband block DATA (k).
Are added (Ev = Ev + ed (k, j)) (step 10).

All the sample data D in the sub-band block DATA (k) are processed by the processing in step 10 above.
This is performed for (0) to D (9) (0 ≦ j ≦ 9).

Next, the error sample data count Ev
Is the upper limit value E of the number of errors of the sample data D (0) to D (9) in one subband block DATA (k)
It is determined whether or not it is VMAX or more (step 1
1).

When the error sample data number Ev is not equal to or larger than the upper limit value EVMAX (Ev <EVMA
X), it is determined whether or not the repeat flag EBAND (k) is set (step 12).

When the repeat flag EBAND (k) is not set (EBAND (k) ≠ 1), that is,
If none of the word length data WL (k), the scale factor data SF (k), and the sign data SIGN (k) is in error, the error flag information ed (k, j) (k: Subband block index, j: sample data index)
It is determined whether or not is standing (step 13).

When the error flag information ed (k, j) is not set (ed (k, j) ≠ 1), that is, when the sample data (j) is not an error, the sample data D ( The error correction for j) is not performed (D (j) = D (j)) (step 14), and the process proceeds to the check for the next sample data D (j + 1).

When the error flag information ed (k, j) is set (ed (k, j) = 1), that is, when the sample data (j) is an error, the sample data D (j) ) Is the first sample data D (j = 0)
Or not (step 15).

When the sample data D (j) is the first sample data D (j = 0), FIG. 10 (a)
As shown in, the sample data D (j =
0) is replaced with “0” (D (j) = 0) (step 1
6).

Alternatively, as a result of the judgment processing in step 15, when the sample data D (j) is not the first sample data D (j = 0), the sample data D (j) is the last sample data D (j). It is determined whether the sample data is D (j = 9) (step 17).

When the sample data D (j) is the last sample data D (j = 9), FIG. 10 (b)
As shown in, the sample data D (j =
9) is replaced with the sample data D (j-1) immediately preceding the sample data D (j) (D (j) = D (j-
1)) (step 18).

Alternatively, as a result of the determination processing in step 17, the sample data D (j) is not the final sample data D (j = 9), that is, the sample data in the middle (1≤j≤ If it is 8), it is determined whether or not the error flag information ed (j + 1) of the sample data D (j + 1) immediately after the sample data D (j) is set (step 19).

When the error flag information ed (j + 1) is not set (ed (j + 1) = 0), that is, the sample data D one after the sample data D (j).
When (j + 1) is not an error, the above-mentioned FIG.
As shown in, the average value of the sample data D (j) ((D (j j-1)
+ D (j + 1)) / 2) (step 20).

Alternatively, as a result of the judgment processing in the step 19, the error flag information ed (j + 1) is set (ed (j + 1) = 1), that is, one of the sample data D (j). When the subsequent sample data D (j + 1) is in error, the sample data D (j) is replaced with the previous sample data D (j-1), as in the case of FIG. 10B. D (j) = D (j-1)) (step 21).

Sample data D for which error correction has been performed by the processing of steps 13 to 21 described above.
(J), that is, j of the subband block DATA (k)
Sign data SIGN corresponding to the sample data D (k, j) is added to the th sample data D (k, j).
Multiply by (k, j) (D (k, j) = D (k, j) × S
IGN (k, j)) (step 17).

The sample data D (k, j) multiplied by the sign data SIGN (k, j) is recorded, for example, in the memory shown in FIG.

The above-mentioned steps 13 to 21 are performed on all the sample data D (0) to D (9) of the subband block DATA (k) (0≤j≤.
9). Then, the process returns to step 9 and the process for the next subband block DATA (k + 1) is performed.

On the other hand, as a result of the judgment processing in step 11, the error sample data number Ev is equal to the upper limit value EVM.
If it is equal to or larger than AX (Ev ≧ EVMAX), the repeat flag EBAND (k) indicating that the data of the previous subband block DATA (k−1) is repeated is set to “1” (EBAND (k) = 1). (Step 23).
Next, the sub-band block DAT currently being processed
It is determined whether the index k of A (k) is larger than the minimum threshold value EBMIN of the subband block number in which the error occurred (step 24).

As a result of the judgment processing in step 12,
When the repeat flag EBAND (k) is set (E
BAND (k) ≠ 1), that is, word length data W
The determination process of step 24 is also performed when any of L (k), scale factor data SF (k), and sign data SIGN (k) is in error.

The index k corresponds to the threshold value EB.
If it is larger than MIN (EBMIN <k), it is determined whether or not the count determination flag CFG is set (step 25).

When the count judgment flag CFG is not set (CFG = 0), that is, the number of times the data of the previous subband block DATA (k-1) is repeated is less than or equal to the set threshold value. In this case, the count flag ECNT is cleared to zero (ECNT = 0) (step 26).

Further, the mute flag is set to OFF so that the reproduced sound is not silent (step 27).

When the count judgment flag CFG is set (CFG = 1), that is, when the number of times the data of the previous subband block is repeated is less than or equal to the set threshold value, the process goes to step 9. Returning to this, processing is performed on the next subband block DATA (k + 1).

Similarly, after the process of step 27, the process returns to step 9 and the next subband block DAT is executed.
The process for A (k + 1) is performed.

When the index k is less than or equal to the threshold value EBMIN (EBMIN ≧ k) as a result of the judgment processing in step 24, the count judgment flag CFG.
It is determined whether or not is standing (step 28).

When the count judgment flag CFG is not set (CFG = 0), that is, the number of times the data of the previous subband block DATA (k-1) is repeated is less than or equal to the set threshold value. In this case, the count flag ECNT indicating the number of times the data of the previous subband block DATA (k−1) is repeated at the time of error is counted up (ECNT = ECNT + 1), and the count determination flag CFG is set to “1” (CFG). = 1) (step 29).

Then, the count value of the count flag ECNT indicates that the previous subband block DAT at the time of error.
It is determined whether the number of times A (k-1) is repeated is larger than the upper limit value EHMAX (step 30).

The count value of the count flag ECNT is larger than the upper limit value EHMAX (EHMAX <
In the case of (ECNT), the mute flag is set to ON so that the reproduced sound becomes silent (step 31).

Then, the process returns to step 9, and the process for the next subband block DATA (k + 1) is performed.

If the count determination flag CFG is set as a result of the determination processing in step 28 (C
FG = 1) or, as a result of the judgment processing in step 30, if the count value of the count flag ECNT is less than or equal to the upper limit value EHMAX (EHMAX ≧ ECNT), the process similarly returns to step 9 and the next sub-band. Block D
Perform processing on ATA (k + 1).

The above steps 9 to 31 are performed by sub-band blocks DATA (0) to DATA (3
Do 1).

Here, for example, continuous errors at a very high frequency are originally hard to be perceptually perceived, so in this case, the upper limit of the number of errors of the sample data in one subband block in the above step 11 is used. Value EVMAX, upper limit value EH of the number of times the previous subband block is repeated at the time of error in step 24
MAX, threshold value EBMIN of the minimum number of the sub-band block that caused the error in step 30 above
Etc. to change the error data to “0” or to turn on the mute flag so that the reproduced sound becomes silent. In this way, error correction processing is performed according to the frequency of the subband block in which the error has occurred.

It should be noted that the upper limit value EVM shown in FIG.
AX, EHMAX, and the threshold value EBMIN are set by the system control unit 5 shown in FIG. 1, for example.
The setting range of each value is 0 ≦ EVMAX ≦ 11, 0 ≦ E
HMAX ≦ 15 and 0 ≦ EBMIN ≦ 31. Also,
The range of the value of the error sample data number Ev is 0 ≦ E
v ≦ 10, and the count value range of the count flag ECNT is 0 ≦ ECNT ≦ 15.

Interpolation means 3 for performing the error correction as described above
The operation of the audio coded data processing apparatus including the second embodiment will be described.

At the time of encoding, the division processing section 11 divides the audio data of a plurality of channels from digital equipment such as an editing machine and a mixer into a plurality of frequency bands for each channel by, for example, an analysis filter, and divides the divided data. The respective band signals, that is, the sub-band blocks, are supplied to the bit allocation processing unit 12.

The bit allocation processing section 12 is
A quantization level is assigned to each subband block of a plurality of channels from the division processing unit 11, and the quantization level of each subband block is supplied to the quantization processing unit 13.

The quantization processing unit 13 performs a quantization process on each band signal at the quantization level of each subband block from the bit allocation processing unit 12. And
The quantized sub-band blocks and the quantization level information are supplied to the assembly processing unit 21.

The assembling processor 21 assembles the sub-band blocks of a plurality of channels which have been quantized by the quantization processor 13 and the quantization level information into a bit stream according to a predetermined format, and illustrates them. Not recording in memory.

At this time, the order of the sample data in each subband block is divided into odd-numbered sample data and even-numbered sample data and recorded in the memory.

Then, the bit stream state data recorded in the memory is an ECC (Error Check) (not shown) to which information for error correction is added.
ing and Correcting) block, and then recorded on a recording medium or transmitted.

Next, at the time of decoding, first, the data in the bit stream state, which has been error-corrected and to which the error flag information is added, is distributed from the ECC block.
Is input to

The distribution processing section 22 decomposes the data in the bit stream state from the ECC block and distributes it as a data string of each subband block. And
The decomposed subband blocks are supplied to the inverse bit allocation processing unit 31. The inverse bit allocation processing unit 31 obtains the quantization level of each subband block from the quantization level information recorded in the data string of each subband block from the distribution processing unit 22. Then, each sub-band block for which the quantization level is obtained is supplied to the interpolation means 32.

The interpolating means 32 rearranges the order of the data in each sub-band block divided into the odd-numbered sample data and the even-numbered sample data from the inverse bit allocation processing section 31 and rearranges them. Perform error correction on subband blocks. Then, the error-corrected subband blocks are supplied to the inverse quantization processing unit 33.

The inverse quantization processing unit 33 inversely quantizes each error-corrected subband block from the interpolation unit 32 according to the quantization level obtained by the inverse bit allocation processing unit 31. Apply processing. Then, each subband block subjected to the inverse quantization is supplied to the synthesis processing unit 34.

The synthesizing section 34 synthesizes the respective subband blocks of the respective channels which have been inversely quantized by the inverse quantizing section 33 by the synthesizing filter,
Output as audio data to digital devices such as editors and mixers.

Here, the timing control section 4 controls the frame period at the time of the above-mentioned encoding / decoding processing by the pulse oscillator. Further, the system control unit 5 controls input / output data between the ECC block and the digital device, and sets a threshold value for error correction of the interpolation unit 5.

As described above, each subband block is 1
The error correction is performed by rearranging each subband block in which the data in each subband block is divided into an odd number and an even number in a state of 0 sample and 1 subband block, in the order in which the sample data occurs. By doing so, the error does not become a continuous state, so that the influence of the error can be reduced and it becomes hard to be perceptually perceived. Therefore, deterioration of sound quality can be prevented.

For example, when a sample in which error flag information is set is data-interpolated from correct sample values before and after the sample, an error from the original data becomes an error.
When the data is manipulated by the synthesizing unit 34, it is mixed with the correct sample data, so the influence is about 1/32.
It is reduced to, and it becomes difficult to be perceived as an error.

Further, when the sample value of the subband block in which an error has occurred is small, or when the frequency region represented by the subband block is originally hard to be perceptually perceived, the influence of the error becomes even less perceptible. .

In the audio coded data processing apparatus shown in FIG. 1, the audio data coded by the band division coding method is processed, but the frequency is linearly processed and divided. It can also be applied to audio data encoded by all other encoding methods.

[0164]

In the audio coded data processing apparatus according to the present invention, the interpolation means performs error correction on each of the coded band signals of the audio data coded by being divided into a plurality of frequency bands. . As a result, the influence of the error can be reduced, and thus the sound quality deterioration can be prevented.

Further, in the audio coded data processing device according to the present invention, the interpolating means replaces the band signal containing the error sample data with the band signal immediately preceding the band signal, and codes each band. Perform error correction processing on the signal. As a result, the influence of the error can be further reduced, so that the sound quality deterioration can be prevented.

Further, in the audio coded data processing apparatus according to the present invention, the interpolating means rearranges the order of the sample data of each band signal composed of a predetermined plurality of sample data and codes the coded data. Error correction processing is performed on each band signal. As a result, a series of errors does not occur, and the influence of the errors can be further reduced, so that the sound quality deterioration can be prevented.

Further, in the audio coded data processing device according to the present invention, the interpolation means arranges the order of the sample data in each band signal divided into the odd-numbered sample data and the even-numbered sample data. Instead, an error correction process is performed on each of the coded band signals. As a result, a series of errors does not occur, and the influence of the errors can be further reduced, so that the sound quality deterioration can be prevented.

Further, in the audio coded data processing apparatus according to the present invention, the interpolating means may be arranged so that when the sample data at the first position of the band signal is erroneous,
The sample data at the first position is replaced with zero, and error correction processing is performed on the encoded band signals. As a result, a series of errors does not occur, and the influence of the errors can be further reduced, so that the sound quality deterioration can be prevented.

Further, in the audio coded data processing apparatus according to the present invention, the interpolation means, when the sample data at the last position of the band signal is erroneous,
The sample data at the last position is replaced with the sample data immediately before the sample data at the last position, and error correction processing is performed on the encoded band signals. As a result, a series of errors does not occur, and the influence of the errors can be further reduced, so that the sound quality deterioration can be prevented.

Further, in the audio coded data processing apparatus according to the present invention, the interpolating means, when the sample data in the middle position of the band signal is erroneous,
The sample data at the midway position is replaced with the average value of the sample data immediately before and the sample data immediately after the sample data at the midway position, and error correction processing is performed on each of the coded band signals. As a result, a series of errors does not occur, and the influence of the errors can be further reduced, so that the sound quality deterioration can be prevented.

Further, in the audio coded data processing apparatus according to the present invention, the interpolation means replaces the sample data at the middle position with the sample data immediately before when the sample data immediately after is an error. Then, error correction processing is performed on each of the encoded band signals. As a result, a series of errors does not occur, and the influence of the errors can be further reduced, so that the sound quality deterioration can be prevented.

[Brief description of drawings]

FIG. 1 is a diagram showing a configuration of a device for processing audio coded data according to an embodiment of the present invention.

FIG. 2 is a diagram showing a data structure assembled into a bit stream state to be input to the audio encoded data processing device.

FIG. 3 is a diagram for explaining error flag information added to the data assembled in the bitstream state.

FIG. 4 is a diagram for explaining a process of rearranging the order of data in a band signal and performing error correction.

FIG. 5 is a diagram showing a relationship between a position of error flag information and an error position of each band signal.

FIG. 6 is a diagram for explaining a case where an error band signal is replaced with an immediately preceding band signal.

FIG. 7 is a diagram showing a structure of memory data.

FIG. 8 shows word length data, scale factor data,
It is a flowchart which shows the check process of signature data.

FIG. 9 is a flowchart showing band signal check processing and error correction processing in an error correction processing step in the interpolation means.

FIG. 10 is a diagram illustrating an error correction process.

FIG. 11 is a diagram showing a configuration of a band division encoding method.

FIG. 12 is a diagram showing a data structure assembled in a bitstream state.

FIG. 13 is a diagram showing a system configuration of a device for processing audio coded data to which the conventional band division coding method is applied.

FIG. 14 is a diagram showing a configuration of a device for processing the audio encoded data.

FIG. 15 is a diagram for explaining an addition state of error flag information in a Reed-Solomon code.

FIG. 16 is a diagram for explaining an addition state of error flag information in a product code.

FIG. 17 is a diagram for explaining conventional data correction.

[Explanation of symbols]

 1 Encoding Processor 2 Bit Stream Processor 3 Decoding Processor 4 Timing Controller 5 System Controller 11 Division Processor 12 Bit Allocation Processor 12 13 Quantization Processor 13 21 Assembly Processor 22 Distribution Processor 31 Reverse Bit Allocation Processing unit 32 Interpolating means 33 Inverse quantization processing unit 34 Synthesis processing unit 34

Continuation of the front page (51) Int.Cl. ⁶ Identification number Office reference number FI Technical display location G11B 20/18 570 C 8940-5D 574 D 8940-5D H03M 7/30 A 9382-5K 13/00 8730-5K

Claims (8)

[Claims] 1. An audio coded data processing device for decoding audio data, which is divided into a plurality of frequency bands and coded, and for performing error correction processing by an interpolating means, wherein the interpolating means is the encoded data. An apparatus for processing audio coded data, wherein error correction processing is performed on each band signal. 2. The interpolation means replaces a band signal containing error sample data with a band signal immediately before the band signal and performs error correction processing on each of the coded band signals. Item 1. A device for processing audio coded data according to Item 1. 3. The band signal is made up of a plurality of predetermined sample data, and the interpolation means rearranges the order of the sample data and performs error correction processing on each of the coded band signals. An apparatus for processing audio encoded data according to claim 1. 4. The interpolation means rearranges the order of sample data in each band signal divided into odd-numbered sample data and even-numbered sample data, and corrects errors in each coded band signal. 4. The audio coded data processing apparatus according to claim 3, wherein processing is performed. 5. The interpolation means replaces the sample data at the first position with zero when the sample data at the first position of the band signal is erroneous, and encodes each band signal. 4. The audio coded data processing apparatus according to claim 3, wherein the error correction processing is performed on. 6. The interpolation means converts the sample data at the last position into the sample data immediately before the sample data at the last position when the sample data at the last position of the band signal is in error. 4. The audio coded data processing apparatus according to claim 3, wherein the coded band signals that are replaced are subjected to error correction processing. 7. The interpolating means, when the sample data at the midway position of the band signal is in error, sets the sample data at the midway position as sample data immediately before the sample data at the midway position. 4. The audio coded data processing apparatus according to claim 3, wherein the coded band signals are replaced with an average value of the immediately following sample data and error correction processing is performed on the coded band signals. 8. The interpolation means replaces the sample data at the middle position with the sample data immediately before and corrects the error in each of the encoded band signals when the sample data immediately after the error is an error. 8. The audio coded data processing device according to claim 7, wherein the device performs processing.