JPH07288476A - Coded signal decoding method/device - Google Patents

Abstract

PURPOSE:To eliminate such data that are extracted only when an error is detected by confirming only true error flag of the byte that includes the highest rank bit of the data which are extracted by the adverse quantization processing. CONSTITUTION:The coded signal received from an input terminal 7 is inputted to a demultiplexer part 8 and separated into the scale factor, the assigned bit number and the spectrum data. An adverse quantization part 9 performs the adverse quantization processing of the data based on the assigned bit number and the spectrum data and outputs the data to an adverse normalization part 10. The part 10 performs the adverse normalization of the data based on the scale factor and the output of the part 9 and converts the data into the spectrum data. An IMDCT part 11 applies the othogonal conversion to the spectrum data that undergone the adverse normalization through the IMDCT processing and turns the data into the data on a time area to output them to an output terminal 12 as the PCM digital audio signals that include the frequency components of DC-20kHz, for example.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a coded signal decoding method, for example, bit-compressing input data obtained by digitizing a continuous signal and recording it on a disc-shaped recording medium, and then performing the bit compression from the disc-shaped recording medium. The present invention is applied to a disc recording / reproducing apparatus that reproduces reproduced data and bit-expands it to obtain reproduced data.

[0002]

2. Description of the Related Art In recent years, with the progress and development of high-efficiency coding technology, the technology for high-efficiency coding of digital signals such as voice and images is about to expand its application field.
In a method of highly efficient coding of a signal such as audio or voice, for example, band division coding (sub-band coding) in which an audio signal on the time axis is divided into a plurality of frequency bands by a filter for coding : SBC)
Etc. Moreover, the signal on the time axis is converted into a signal on the frequency axis (orthogonal conversion) and divided into a plurality of frequency bands,
Examples include coding for each band, transform coding, and the like.

A high-efficiency coding method combining the above-described band-division coding and transform coding has also been considered. In this case, for example, the above-mentioned band-division coding is used to divide into several bands. After that, the signal for each band is orthogonally converted into a signal on the frequency axis, and the coding is performed for each band subjected to the orthogonal conversion.

Here, as a filter for the above band division, there is, for example, a QMF filter, and 1976 REC
rochier, Digital coding of speech in subbands, BellS
yst.Tech. J. Vol.55, No.8 1976. Also ICASSP 83, BOSTON Polyphase Quadrature filters-A
A new subband coding techniqe, Joseph H. Rothweiler, describes an equal bandwidth filter partitioning technique. Next, as the above-mentioned orthogonal transform, for example, the input audio signal is divided into blocks in a predetermined time unit (frame), and the modified DCT transform (MDCT) or the like is performed for each block to transform the time axis into the frequency axis. There are things. About MDCT ICASSP 1987 Subband
/ Transform Coding Using Filter Bank Designs Based
on TimeDomain Aliasing Cancellation, JPPrincen,
ABBradley, Univ. Of SurreyRoyal Melbourne Inst.
of Tech.

Further, as a frequency division width for quantizing each frequency component obtained by frequency band division, for example, band division utilizing human psychoacoustic characteristics or the like is performed, and when encoding data for each band. Is encoded by adaptive bit allocation for each band. For example, when the data obtained by the MDCT processing is encoded by the bit allocation, the MDCT data for each block is encoded by the adaptive allocation bit number.

[0006] created by such an encoding method,
FIG. 7 shows an example of encoded data of an audio signal. Figure 7
In general, 22 is scale information (scale factor) corresponding to the exponent part in the case of floating point notation. Reference numeral 23 denotes the number of bits assigned to each block, which is assigned bit information (word length) corresponding to the number of digits of the mantissa part in the floating point notation. 24
Is the spectral component data of each block, which is spectral information (spectral data) corresponding to the data in the mantissa part in the floating-point notation. The scale information 22 and the allocated bit number information 23 divide the input spectrum data into a plurality of bands (blocks) (for example, 5).
It is divided into two blocks, and the block with the lowest frequency is designated as block 0, block 1, ..., Block n. ), One is assigned to each block.

FIG. 8 is an enlarged view of the scale factor 22, the word length 23 and the spectrum data 24 of FIG. 7, which are scale factors 22 (1) to 22 (3) and blocks 0 to 22 of block 0 to block n, respectively. The word lengths 23 (1) to 23 (3) of the block n and the spectrum data 24 (1) to 24 (3) of the block 0 to the block n are obtained.

Further, in the encoded data of FIG. 7, in order to know whether or not an error has occurred in the data, as shown in FIG. 4, a data stream as shown in FIG. It is divided into 15 (a) with 1 byte as a unit. Error flag 15 (b) indicating whether each byte contains an error.
Will be attached. When the error flag is 1, the byte data 15 (a) includes an error, and when the error flag is 0, the byte data does not include an error.

FIG. 2 shows the configuration of a conventional coded signal decoding apparatus to which the coded signal as described above is input. In FIG. 2, 7 is an input terminal of the encoded signal decoding device, 8 is a demultiplexer section, 9 is an inverse quantization section, 10 is an inverse normalization section, 11 is an IMDCT section, and 12 is a PCM audio data output terminal. .

Next, the operation of the above conventional example will be described. In FIG. 2, the encoded signal from the input terminal 7 is input to the demultiplexer unit 8, and the scale factor (for example, 22 (1) to 22 (3) in FIG. 8) and the number of allocated bits (for example, in FIG.
23 (1) to 23 (3)), spectral data (for example, in FIG.
24 (1) to 24 (3)).

The dequantization unit 9 dequantizes the data from the allocated number of bits and the spectrum data, and outputs the data to the denormalization unit 10.

The denormalization unit 10 denormalizes the data from the scale factor and the output of the dequantization unit 9, and converts the data into spectrum data as shown in FIG. In FIG.
D1 to D8 represent the intensity of each frequency band of the spectrum data, and are output as the spectrum data (data on the frequency axis) denormalized in this way.

The IMDCT unit 11 orthogonally transforms the denormalized spectrum data by IMDCT processing into data in the time domain, and as shown in FIG.
It is output to the output terminal 12 as a PCM digital audio signal including a frequency component of DC to 20 KHz.

As described above, the conventional coded signal decoding apparatus described above can decode a digital audio signal from highly efficient coded data.

[0015]

However, in the above-mentioned conventional coded signal decoding apparatus, when the error occurs in the coded data as shown in FIG. 7, the above inverse quantization is performed. Therefore, there is a problem that the encoded signal cannot be decoded and the PCM audio data to be originally decoded is muted or noise is generated.

Now, with reference to FIG. 3, description will be given of a case where the spectrum data is extracted from the data stream. In FIG. 3, 13 (1) to 13 (4) are obtained by dividing the input data stream into byte units. 13
(1a) to 13 (4a) are error flags (1 indicates that an error is included, and 0 indicates that an error is not included) that indicates whether or not each byte includes an error. Further, 14 (1) to 14 (8) are data extracted from the above data stream with a word length of 2 to 16 bits, and Byte (n-1) 13 (2) and Byte are stored in a buffer for data extraction. When 2 bytes of data of (n) are input from the data stream, there are 8 patterns of byte striding when extracting word length data of 2 to 16 bits from this buffer. ) And 14 (5), some data spans 3 bytes.

When such data is extracted, it is checked whether or not there is an error in each byte including the extracted data, and if there is an error, the corresponding data is set to 0. However, there are 8 cases, and when extracting one data, it is necessary to check an error flag of up to 3 bytes, which requires a large-scale hardware and is difficult to realize. It was

The present invention solves such a conventional problem, and the deterioration of the sound quality is minimized even when a data error occurs in the high-efficiency coded signal with the hardware of the same scale as the conventional device. (EN) Provided is an excellent coded signal decoding device capable of decoding an audio signal by performing inverse quantization of data with a limit.

[0019]

In order to achieve the above object, the present invention provides only the error flag of the byte containing the most significant bit of the extracted data in the dequantization processing performed in the dequantization section. It is confirmed that the extracted data is set to 0 only when the most significant byte contains an error.

[0020]

Therefore, according to the present invention, by checking whether or not there is an error in the byte including the most significant bit of the data to be dequantized, the data to be dequantized can be set to 0 when there is an error. it can.

Even if there is an error in a byte other than the most significant byte of the data to be extracted, the most significant data is not an error. For example, if the extracted data is 8 bits and the upper byte contains 7 bits, Only the lower 1-bit data becomes uncertain and the data resolution is 8
This is equivalent to just dropping from 7 bits to 7 bits, and it has an effect of being able to interpolate into data close to the information of the original data.

Further, since the presence / absence of the error is confirmed only in the byte including the most significant bit of the data to be extracted, the number of times of confirming the error flag and the case of the error pattern are reduced, and the processing of the inverse quantizer is performed. Can be realized with simple hardware, and there is an effect that deterioration of sound quality can be minimized without significantly increasing the processing load for performing interpolation processing when there is an error in data.

[0023]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS FIG. 1 shows an embodiment of the present invention and is a flow chart showing the inverse quantization process performed by the inverse quantization section 9 of FIG. That is, the configuration of the coded signal decoding apparatus in the embodiment of the present invention is the same as the conventional configuration shown in FIG. 2, but the method of processing in the dequantization unit 9 is particularly characterized. It is a thing. Before describing the inverse quantization process in FIG. 1, first, the data input to the inverse quantization unit 9 will be described.

FIG. 5 shows the presence / absence of an error in each byte of the spectrum data and the spectrum data stream input to the inverse quantizer 9 of FIG. 2, and shows how the number of allocated bits is extracted. In FIG. 5, 16 (1) to 16 (4)
Is a byte obtained by dividing the spectrum data stream (here, the unit for error detection is a byte).
16 (1a) to 16 (4a) are error flags indicating whether or not the above byte contains an error. Here, if the error flag is 1, the relevant byte contains an error, and if the error flag is 0, it indicates that the relevant byte has no error. Therefore, byte A (16 (1)), byte B (16
(2)), byte D (16 (4)) is a byte having no error, and byte C is a byte containing an error.

DATA (n-1) (17 (1)), DATA (n) (17 (2)),
DATA (n + 1) (17 (3)) is the data obtained by extracting the number of allocated bits from the spectrum data stream, and 18 is DATA (n) (1
This is the original value when there is no error in the part of byte C (16 (3)) of 7 (2)). Next, the manner in which the error-free DATA (n) (17 (2)) is dequantized in the most significant bit (MSB) of the data will be described with reference to the flowchart of FIG. Byte B (16 including the MSB of DATA (n)
Since there is no error in (2)) (step 1), it is confirmed whether or not there is data for the number of extracted bits (8 bits in this example) in the bit extraction buffer (step 2a).

However, since 8-bit data cannot be extracted from byte B, the next byte C (16 (3)) is input from the spectrum data stream (step 3).
a). Further, it is confirmed whether the data for the number of extracted bits is input to the bit extraction buffer (step 4
a).

Here, since 13-bit data is in the buffer, 5-bit data from byte B and 3-bit data from byte C are taken out, and DATA (n) is dequantized (step 5).

In this case, since the byte B containing the MSB of DATA (n) has no error, even if the byte C, which is the lower bit part of DATA (n), has an error, it is dequantized as it is.

Further, the manner in which the data DATA (n + 1) having an error in the MSB of the data is inversely quantized will be described with reference to the flowchart of FIG. It is confirmed whether or not an error is included in the byte C including the MSB of DATA (n + 1) (step 1). It is confirmed whether or not data for the number of extraction bits (8 bits in this case) is in the bit extraction buffer (step 2b). Now, since there are only 3 bits left in the buffer, the next byte D from the spectral data stream is read into the buffer.

Since the bit extraction buffer has 13-bit data, the process proceeds to the next process, but DATA (n + 1) is
Since the MSB contains an error, the data for the extracted number of bits is dequantized with 0 (step 6).

The dequantization process is performed in this way, and FIG. 6 shows the data DATA (n) (17) extracted in FIG.
(2)) represents the difference in the dequantized value due to the difference in processing method when dequantizing.

The true value of DATA (n) cannot be decoded because there is an error in the part included in the byte C on the least significant bit (LSB) side. 19 indicates the true inverse quantization value when DATA (n) has no error, and 20 indicates DA
Only the byte B including the MSB of TA (n) is error-checked, and the value is dequantized by the method shown in the flowchart of FIG. 1. 21 indicates that the byte C including the LSB of DATA (n) has an error. Therefore, DATA (n) is regarded as an error and represents the value when all extracted bits are interpolated by 0.

As can be seen from this example, the dequantized value of the original 19 is 65, 20 is 71 when only the block B is error-checked, 21 is 0 when interpolated, and 21 is 0.
By taking advantage of the byte containing the MSB of the data to be extracted, it is possible to make the value close to the true value even if there is an error in the data on the LSB side. In this example, originally 8-bit data only needs to fall to 5-bit accuracy. This means that the larger the number of bits included in the byte on the MSB side, the closer the value can be interpolated to the true value.

[0034]

As is apparent from the above description, the present invention has the following advantages.
It has the following effects. (1) Only the error of the byte containing the most significant bit of the spectrum data is checked, and if there is no error in that byte, the data containing the most significant bit is utilized and the data is dequantized. Compared with the method of clearing the spectrum to be converted as an error, the difference from the original data is small, and in particular, the larger the number of bits of the byte including the upper bits, the closer to the original value. (2) When the data is inversely quantized, the presence or absence of an error in the data to be extracted from the spectrum data is confirmed by checking only the byte including the most significant bit of the extraction block, so that the data to be extracted is plural. Inverse quantization can be done by checking the error flag only once even if it spans bytes, which simplifies the processing of the inverse quantizer and reduces the hardware scale. The advantage is that the encoded signal decoding device can be realized at a processing speed.

[Brief description of drawings]

FIG. 1 is a flowchart of an inverse quantization process performed by an inverse quantization unit of an encoded signal decoding device according to an embodiment of the present invention.

FIG. 2 is a block diagram showing a schematic configuration of a coded signal decoding device according to an embodiment of the present invention and a conventional example.

FIG. 3 is an error pattern diagram of data that occurs during inverse quantization.

FIG. 4 is an explanatory diagram in which a data stream input to the device is divided into byte units and an error detection flag is added to each byte.

FIG. 5 is an explanatory diagram of data to be dequantized when an error occurs in a part of the spectral data stream according to the embodiment of the present invention.

FIG. 6 is a diagram showing the difference in value due to the difference in processing of the dequantized data DATA (n) in FIG. 5 above.

FIG. 7 is an explanatory diagram of an encoded signal input to this device.

FIG. 8 is an enlarged view of the encoded signal of FIG.

FIG. 9 is a spectrum diagram showing the intensity of spectrum data after denormalization of the coded signal decoding device.

FIG. 10 is a waveform diagram showing PCM audio data output from the IMDCT unit of the encoded signal decoding device.

[Explanation of symbols]

 7 Input Terminal 8 Demultiplexer Section 9 Inverse Quantization Section 10 Inverse Normalization Section 11 IMDCT Section 12 Output Terminal

─────────────────────────────────────────────────── ─── Continuation of the front page (51) Int.Cl. ⁶ Identification code Internal reference number FI Technical display location G11B 20/18 560 G 8940-5D 570 C 8940-5D H03M 7/30 A 8842-5J

Claims (2)

[Claims] 1. A data stream divided into blocks and including an error flag indicating the presence / absence of an error in the block in the block unit is different from a break in the block unit including the error flag from the data stream. When extracting data in units, only the error flag of the block that contains the most significant bit of the data to be extracted is detected, and if the error flag indicates that there is no error, the extracted data is valid Method. 2. A system for decoding bit-compressed data, wherein a plurality of data encoded in an error detection variable length format are concatenated to form a bit stream, which is an error detection block having a fixed length. A decoding device for dividing a signal recorded / reproduced or transmitted by adding a correction code by dividing into, a means for detecting an error in data to be decoded using the correction code, and an interpolation means for interpolating erroneous data. If the data to be decoded spans multiple error detection blocks, only the error detection block that contains the most significant bit of the data is checked, and only if the error detection block has an error. A coded signal decoding device that interpolates and decodes the decoded data.