JP3308955B2 - Data compression encoder - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

[0001] 1. Field of the Invention [0002] The present invention relates to a data compression encoding apparatus constituted by using, for example, a DSP (Digital Signal Processor).

[0002]

2. Description of the Related Art In general, as a method of reducing the transmission bit rate by compressing the number of bits of a supplied PCM signal,
For example, there is known a so-called filter selection type bit compression encoding system in which a filter that can obtain the highest compression ratio in a block unit for each of a plurality of samples is selected from a plurality of filters prepared in advance.

[0003] In such a filter selection type bit compression encoding system, for example, an input signal having 16 blocks as one block is supplied to each block. The input signal for each block is a plurality of filters, for example, a zero-order filter that outputs straight PCM, a first-order difference filter that outputs a first-order difference, and a second output that outputs a second-order difference.
It is supplied to each of the three filters of the next difference filter. Then, the maximum absolute value in the block is detected for each filter, the block data that has passed through the filter that minimizes the maximum absolute value in the block is selected, and re-quantized from, for example, 16 bits to 4 bits and output. . At the time of the requantization, the difference (quantization error) between the input and the output of the quantizer is fed back to the input side of the quantizer and added to the data newly supplied to the quantizer. A so-called noise shaping process is performed. At the time of the output, filter information indicating a filter through which the data of the recorded block has passed, range information indicating a range of the data, and the like are output together with the data, and the decoder side responds to the information. Data is reproduced.

[0004]

When the quantization error is added to new input data to the quantizer due to the noise shaping process and a carry occurs, a bit for requantization is generated. A so-called overflow may occur such that a value larger than the extraction range is input to the quantizer. When this overflow occurs, there is a problem that a direct current error component occurs at the time of reproduction, and good reproduction cannot be performed. Therefore, usually, the maximum absolute value in each block from the first-order differential filter and the second-order differential filter is used. The weight information is multiplied by a weighting coefficient (for example, 1.5), and the range information is obtained based on the weighted maximum absolute value.

[0005] However, even if such weighting is performed, an overflow may occur, and if data with the overflow is output, a problem such as deterioration of the SN ratio during reproduction on the decoding side may occur. is there.

As described above, the filter selection type bit compression encoding system outputs a primary or secondary difference value with respect to previous data based on, for example, the leading data in a block. When data is reproduced from the data, an error occurs in the reproduced data if the data reproduced at the arbitrary point is data that has passed through a primary or secondary filter. As described above, once an error occurs, the error propagates to the subsequent data, and a DC error component occurs, so that it is impossible to obtain a good reproduced sound. Therefore, there are many problems in using a filter having a high prediction gain in consideration of an error margin.

SUMMARY OF THE INVENTION The present invention has been made in view of the above-described problems, and can effectively prevent overflow during noise shaping. In addition, the present invention reduces a DC error component generated due to the propagation of the error. It is an object of the present invention to provide a data compression encoding device that can be eliminated.

[0008]

SUMMARY OF THE INVENTION In order to solve the above-mentioned problems, the present invention provides a plurality of filters including a mode for dividing an input signal into fixed samples and, for each block, directly outputting the input signal. A plurality of modes to be output through the data compression encoding apparatus, wherein a mode that provides an output signal having the highest compression ratio is selected. Weighting means for weighting at least a part of the value for each mode compared with the above, a quantization means for quantizing the filter output data of the selected mode, and a quantization error in the quantization means. Noise shaping means for returning to the quantizing means, and when the overflow is detected by the noise shaping means, Stop the input and output of the condensation coding changes to again the overflow weighting coefficients of the weighting means is characterized by performing a retry of processing data detected block.

Here, a counter is provided for counting the number of blocks and supplying a count reaching signal to the weighting means when the counted number reaches a predetermined value. The weighting means supplies the count reaching signal to the counter. Accordingly, the weighting coefficient may be changed so that the mode for directly outputting the input signal is forcibly selected.

[0010]

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a block diagram showing an embodiment of a data compression encoding apparatus according to the present invention. FIG. 1 is a functional block diagram showing in block form each function of a data compression encoding apparatus according to an embodiment of the present invention.

In FIG. 1, the audio data generation block 1 can stop reading at an arbitrary point.
The playback can be performed instantaneously from the stop point. For example, a CD (compact disc) player or a DA
It can be configured using a device that generates digital data such as T (Digital Audio Tape Recorder), a memory, and the like.

From the audio data generating block 1, reading is controlled by a control signal from a reading control circuit 22. For example, an audio PCM signal in which one sample is made into 16 bits (one word) and 16 samples are made into one block is generated. The audio PCM signal is stored in a buffer 3 via a switch 2 described later, and is also supplied to a zero-order filter 5, a first-order filter 6, a second-order filter 7, and a counter 8 in the predictor 4. . The encode filter has a configuration of a so-called difference filter. The audio PCM signal of the one block supplied to the zero-order filter 5 is used as a so-called straight PCM signal without any difference, and is used as a range detector 9. Supplied to

In the range detector 9, the maximum absolute value in one block of the supplied straight PCM signal is detected, and the maximum absolute value in the block is compared to a comparison (minimum value detection) block 14 and to the entire block. The data is supplied to the selector 15.

Next, the primary filter 6 supplied to the primary filter 6
The audio PCM signal of a block is
The next difference is obtained, and the first difference data is supplied to the range detector 10. The range detector 10 detects the maximum absolute value in the block from the primary difference data of one block. The maximum absolute value of the primary difference data in the block is supplied to the weighting unit 12, and the primary difference data of the one block is supplied to the selector 15. The maximum absolute value of the primary difference data supplied to the weighting unit 12 in the block is determined by a predetermined coefficient (for example, 1.5).
Is multiplied, that is, weighted and supplied to the comparison block 14.

Next, the first filter supplied to the secondary filter 7 is
The audio PCM signal of the block is 2
The next difference is obtained and supplied to the range detector 11. The range detector 11 detects the maximum absolute value in the block from the supplied secondary difference data of one block. The maximum absolute value of the secondary difference data in the block is supplied to the weighting unit 13, and the secondary difference data of the one block is supplied to the selector 15. And weighting device 1
The maximum absolute value in the block of the secondary difference data supplied to the block 3 is multiplied by a predetermined coefficient (for example, 1.5).

The comparison block 14 has a maximum absolute value in the block of the supplied straight PCM data, a maximum absolute value in the block of the weighted primary difference data, and a maximum absolute value in the block of the weighted secondary difference data. The absolute value is compared to detect the minimum value, the range of the detected minimum value is output as range information, and it is detected which of the three filters has passed the minimum value. This is output as filter information.

The range information is supplied to a quantizer 18 and a multiplexer 17, and the filter information is supplied to a selector 15 and a noise shaping circuit 19 and to the multiplexer 17.

The selector 15 selects an audio PCM signal through a filter specified by the filter information from the comparison block 14 by, for example, switching a switch, and selects the selected audio PCM signal of one block.
The M signal is supplied to the quantizer 18.

As described above, the selector 15 selects an optimum one from the audio PCM signals passed through the above three filters and supplies it to the quantizer 18, while the counter 8 outputs the supplied audio PCM signal. The number of blocks is counted, and when a predetermined count number (for example, several blocks to several tens of blocks) is reached, each of the weighters 12,
13 is supplied with a count reaching signal.

Each of the weighters 12, 13 increases the weighting coefficient, which was 1.5, for example, to a sufficiently large value (for example, 100, etc.) when the count reaching signal is supplied. As mentioned above,
The selector 15 selects the supplied three audio PCMs.
In order to select the block having the largest absolute value in the minimum block from the signals, the block of the straight PCM signal through the zero-order filter 5 is necessarily selected. After the counting, the block of the straight PCM signal can be forcibly arranged.

Here, assuming that the above-mentioned fixed interval (constant number of blocks) is N and the number of samples in one block is n (for example, n = 28 for CDI and n = 16 for APU), every nN samples must be obtained. 0-order filter 9 is guaranteed to be selected, in the case of reproducing the data BRR encoding from any point, even the longest nNt _s [sec]
It is possible to obtain a complete reproduced sound after (t _s = sampling interval [sec]). Specifically, when f _s = 32 kHz and one block is 16 samples, the time required for one block is about 0.5 [msec], and the straight PC is output every several tens of blocks.
Even if M signals are arranged, they can be arranged at intervals of about 20 to 30 [msec]. This interval is not a problem because it is not recognizable by human hearing.

[0022] Although the causing some errors until the block of the straight PCM signal is disposed, may have been the mask data during the nNt _s [sec] from the start .

According to the range information supplied from the comparison block 14, the 16-bit audio PCM signal thus selected and supplied to the quantizer 18 is initially "1" in the 16 bits as viewed from the upper bits. For example, 4 bits are taken out from the bits immediately after "", that is, so-called requantization is performed and supplied to the multiplexer 17 via the noise shaping circuit 19. The noise shaping circuit 19 converts the so-called quantization error, which is the error between the input and the output of the quantizer 18, into the comparison block 14.
Is fed back to the input of the quantizer 18 in accordance with the filter information from. That is, the data of the lower 4 bits extracted at the time of the requantization is fed back to the input of the quantizer 18 and added to the data newly supplied to the quantizer 18.

Here, as described above, the quantizer 18 outputs, for example, 4 bits counted from the first bit of "1" in the 16 bits as viewed from the upper bits, based on the range information. The added quantization error and new data cause a carry, so that "1" shifts to a bit higher than the extracted bit specified by the range information, so-called overflow may occur. .

When this overflow occurs, there is a problem that a DC error component occurs during reproduction, and good reproduction cannot be performed.
12 and 13, the maximum absolute value in the block through the primary filter and the secondary filter is weighted by multiplying the maximum absolute value by, for example, 1.5 , and data is extracted from, for example, one bit higher than the bit to be originally extracted. I have. However, even if the above weighting is performed, an overflow may occur.

Therefore, in the present embodiment, when the overflow occurs, the noise shaping circuit 19
Supplies the switch 2, the buffer 3, the mask circuit 21 and the read control circuit 22 with overflow information indicating the amount of overflow and the current filter number (0th, 1st or 2nd filter), and the audio PCM in which the overflow has occurred. Stop the signal output once.

When the overflow information is supplied, the multiplexer 17 stops outputting data, the read control circuit 22 stops reading data from the audio data generation block 1, and the switch 2 sets the selection terminal 2c to the selected terminal. Switching from the terminal 2a to the selected terminal 2b temporarily stops taking in data from the audio data generation block 1. As described above, the buffer 3 stores audio data for several blocks.
The predictor 4 calculates again the audio PCM signal of the block in which the overflow has occurred, based on the overflow information.
A so-called retry is supplied to the power supply.

When this retry is started, the mask circuit 2
1 knows the current filter number from the supplied overflow information, and supplies a coefficient increment signal to a weighter connected to this filter. The weighter to which the coefficient increment signal is supplied increments the coefficient value so that the coefficient value is 1.5, for example, to 1.6, for example. The re-supplied audio PC of the overflowed block
The M signal is supplied to each of the range detectors 9, 10, and 11 via each filter as described above, and the maximum absolute value in each block is detected.
Supplied to Of the detected maximum absolute values in the block, the maximum absolute value in the block of the straight PCM signal via the range detector 9 is used as it is in the comparison block 1.
4, the maximum absolute value of the primary difference data in the block via the range detector 10 is
The maximum absolute value of the secondary difference data in the block via the range detector 11 is supplied to the weighter 13. As described above, the coefficient value of one of the weighters that caused the overflow is incremented, and the maximum absolute value in each of the blocks is multiplied by the coefficient value and weighted and supplied to the comparison block 14. . The comparison block 14 compares the maximum absolute value in each block including the maximum absolute value in the block re-weighted by the incremented coefficient to detect the one having the minimum value, and based on the detection result, The range information is supplied to the quantizer 18. Based on this range information, the quantizer 18 performs requantization as described above. When the overflow is eliminated by performing the requantization, data from which the overflow has been eliminated by the requantization is output, and the mask circuit 21 calculates a coefficient value based on the incremented coefficient value of the weighting device. And the switch 2 switches the selection terminal 2c to the selected terminal 2a, and resumes capturing of new data.

Note that this retry is performed until the overflow is resolved. If an overflow occurs despite the retry, the weighting coefficient value is incremented again.

The re-quantized 4-bit data is supplied to the multiplexer 17.

The multiplexer 17 outputs the supplied data and the above-described range information and filter information.
These data and information are taken out via the output terminal 20.

As the output data taken out from the output terminal 20, for example, one block is as shown in FIG. 2, and 1-byte header information (compression parameter information or auxiliary information, etc.) RF and 8-byte Of data D _{A0 to} D _B3 . The header information RF includes 4-bit range information, 2-bit mode selection information or filter selection information, two 1-bit flag information, information LI indicating the presence or absence of a loop, and an end block (end) of a waveform. Block) or information EI indicating whether or not the block is a block. As described above, one sample of data is, for example, 16 bits to 4 bits.
The data D _{A0 to} D _B3 include 4-bit data D _{A0H to} D _{B3L for} 16 samples.

And such data and each information are:
For example, the data is recorded on a recording medium or the like, or transmitted directly to the decoding side for reproduction.

As is clear from the above description, the data compression encoding apparatus according to the present embodiment divides an input signal into blocks of, for example, every 16 samples, and forms 0-order filters 5, primary filters 6, and secondary filters 7 The counter 8 counts the number of blocks input when a filter that provides the highest compression ratio is selected and re-quantized to transmit an output signal. By setting the coefficients of the weighters 12 and 13 provided in the secondary filter 6 and the secondary filter 7 to predetermined large values,
Since the 0th-order filter can be forcibly selected and the straight PCM signal can be arranged, it is possible to reproduce data from an arbitrary point. This can be solved by arranging a straight PCM signal in the filter, so that a filter having a high prediction gain can be used.

Further, in a data compression encoding apparatus that changes the weighting parameter for each fixed block and indirectly controls the filter selection, the data according to the present invention can be obtained by only slightly changing such a system. It can be changed to a compression encoding device.

Further, in the data compression encoding apparatus of the present embodiment, the output and the input are stopped at the time when the overflow is detected in the noise shaping circuit 19, the coefficient values of the weighters 12 and 13 are incremented, and the data is again encoded. By performing the retry of processing the data of the block in which the overflow has been detected, the data in which the overflow has occurred is not supplied, so that the deterioration of the SN ratio at the time of reproduction can be prevented.

It should be noted that the coefficient value of the above-mentioned weighting device is merely an example, and various changes are possible, for example, the base coefficient is set to 1.4 and the increment at the time of retry is set to 1.6. Of course, it is not limited.

[0038]

According to the data compression encoding apparatus of the present invention, an input signal is divided into blocks for each fixed sample, and each block is output through a plurality of filters including a mode for directly outputting the input signal. In a data compression encoding device that selects a mode that provides an output signal having the highest compression rate among a plurality of modes, a comparison is performed to select any of the plurality of modes for each of the blocks. Weighting means for weighting at least a part of the value for each mode, quantization means for quantizing the filter output data of the selected mode, and quantization error in the quantization means And a noise shaping means for returning to the data compression code when overflow is detected by the noise shaping means. Input and output are stopped, the weighting coefficient value of the weighting means is changed, and the data of the block in which the overflow is detected is retried, so that the data having the overflow is not supplied. In addition, it is possible to prevent the SN ratio from deteriorating during reproduction.

Further, the mode in which the input signal is directly output is forcibly selected by setting the weighting value for mode selection to a predetermined value at intervals of a fixed number of blocks, thereby achieving a simple configuration. It is possible to clear a DC error component generated due to accumulation of errors generated in the encoding / decoding system, and to enable reproduction from substantially any point.

[Brief description of the drawings]

FIG. 1 is a functional block diagram showing functions of an embodiment of a data compression encoding apparatus according to the present invention in a block form.

FIG. 2 is a schematic diagram showing a format of output data of the data compression encoding apparatus according to the embodiment of the present invention;

[Explanation of symbols]

1 audio data generation block, 2 switches,
3 buffers, 4 predictors, 5 0th-order filters,
6 primary filter, 7 secondary filter, 8 counter, 9, 10, 11 range detector, 12, 1
3 weighter, 14 comparison (minimum value detection) block, 15 selector, 17 multiplexer, 1
8 Quantizer, 19 Noise shaping circuit, 2
0 output terminal, 21 mask circuit, 22 read control circuit

──────────────────────────────────────────────────続 き Continued on the front page (58) Field surveyed (Int. Cl. ⁷ , DB name) H03M 1/00-11/00

Claims (2)

(57) [Claims] An output having the highest compression ratio among a plurality of modes for outputting an input signal through a plurality of filters including a mode for directly outputting an input signal for each block, and for each block. In a data compression encoding apparatus that selects a mode in which a signal is obtained, a block of at least some of the plurality of modes is used.
Weight the maximum absolute value in the
Weighting means for performing weighting, and a mode in which weighting is not performed by the weighting means.
Maximum absolute value in the block and the value from the above weighting means
And the smallest one is detected.
Output the range as range information, and
Comparing means for selecting the supplied mode, the lens filter output data of the selected mode
And a noise shaping means for feeding back a quantization error in the quantization means to the quantization means, wherein the data is obtained when the overflow is detected in the noise shaping means. stop the input and output of the compression coding, data compression coding apparatus and performs a retry again the overflow by changing the weighting coefficient values processes the data of the detected block. 2. A counter for counting the number of blocks and supplying a count reaching signal to the weighting means when the count reaches a predetermined value, wherein the weighting means supplies the count reaching signal. in response, data compression coding apparatus according to claim 1, wherein changing the weighting coefficients such mode for outputting the input signal directly is forcibly selected.