JPH0487422A - Bit rate compressing device - Google Patents

Abstract

PURPOSE:To execute the bit rate compression, and also, to improve the dynamic range in a high frequency by increasing the number of quantizing bits, when a filter of a straight PCM is selected from in plural filters. CONSTITUTION:The bit rate compressing device is provided with not only a filter (O-order predictive filter 2) for outputting straight PCM data but also a primary difference predictive filter 3 and secondary difference predictive filters 4, 5 being plural bit compression predictive filters of different frequency characteristics, and selects one of the predictive filters 2-5 in accordance with a characteristic of an input signal (PCM digital audio data) and performs a floating point processing, and thereafter, quantizes it by a quantizer 9 and outputs it. When the predictive filter 2 of a straight PCM is selected, the number of quantizing bits by the quantizer 9 is increased based on a switching control signal from an adaptive switching selection processing circuit 60. For instance, when the predictive filters 3-5 are selected, an output obtained by a round-off processing by the quantizer 9 is set as the upper 4-bits, and when the predictive filter 2 is selected, it is set as the upper 8-bits.

Description

Detailed Description of the Invention [Field of Industrial Application] The present invention relates to a bit rate compression device that compresses the bit rate of digital data. In a bit rate compression device that selects one of the following according to the input signal and quantizes the resulting output, when a straight PCM filter is selected, the number of bits is increased by increasing the number of quantization bits. The present invention provides a bit rate compression device that can perform rate compression and improve the dynamic range in high frequencies.

[Conventional technology]

For example, so-called CD-1, which handles conventional PCM (pulse coded) digital audio data.
, CD-ROM XA, etc. These CD-1
, CD-ROM XA, for example, the adaptive difference P
The bit rate of the PCM digital audio data is compressed using adaptive predictive coding (APC) such as CM (ADPCM).

That is, bit rate compression of this digital audio data using APC is performed by, for example, a bit rate compression device configured as shown in FIG. This fourth
The apparatus shown in the figure can process, for example, 28 samples (for example, 3
0.741 m5ec as 7.8 kHz sampling
) is supplied as a block of data, and the 16-bit data is compressed into, for example, 4 bits or 8 bits and output. In addition, the above CDi
, AP in CD-ROM XA formatter
In C, four prediction filters are used: a zero-order prediction filter 2, a first-order differential molecular measurement filter, and a second-order differential molecular measurement filter 4 and 5, each having a different frequency characteristic. 3.4゜5, the above human power P
The CM digital audio data is filtered and the characteristics of the input signal (PCM digital audio data) are
The output of each of these filters is selected depending on the characteristics (characteristics).

That is, in the device shown in FIG.
For each bronc mentioned above, 16 pint straight PCM
Digital audio data is supplied. This digital audio data is processed by each of the prediction filters 2,
Sent to 3.4.5. Each of the above prediction filters 2.3, 4
．． 5 are two one-sample delayers 23, 24,
33° 34.43, 44, 53.54 and two multipliers 25.26, 35, 36.45.46, 55.56,
2 adders 21.22 31,32,41.42,5
1.52. Here, in the prediction filter 2, the multiplication coefficients of the multipliers 25 and 26 are both 0, and therefore the prediction filter 2 operates as a zero-order (straight) prediction filter. In the above prediction filter 3, the multiplication coefficient of the multiplier 35 is -0,9375
The multiplication coefficient of the multiplier 36 is O, and the prediction filter 3 operates as a first-order difference measurement filter. In the prediction filter 4, the multiplication coefficient of the multiplier 45 is -1,
796875, and the multiplication coefficient of the multiplier 46 is 0.81.
25, and operates as a second-order difference molecular measurement filter. In the prediction filter 5, the multiplication coefficient of the multiplier 55 is -1, 5.
3125, and the multiplication coefficient of the multiplier 56 is 0.8593.
75, and operates as a second-order difference molecular measurement filter. As a result, the frequency characteristics of each of the prediction filters 2.3 and 4.5 become as shown in FIG. The characteristic curve SP in FIG. 5 indicates the filter characteristic of the zero-order prediction filter 2, and the characteristic curve ID in the figure indicates the characteristic of the first-order difference molecular measurement filter 3. a2DA represents the characteristic of the second-order difference molecular measurement filter 4, and a characteristic curve 2DB in the figure represents the filter characteristic of the second-order difference molecular measurement filter 5. Therefore, in FIG. 5, a high compression effect can be obtained by using a filter having the characteristic that the predicted gain, that is, the instantaneous S/N amount is the largest at each frequency. The frequency characteristics of each of these prediction filters 2.34.5 are determined by the formats of the CD-1 and CD-ROM XA. Also, this CC
In the D-1CDROXA format, it is not necessary to use all of the above prediction filters 2.3.4.5.

The output of each of these prediction filters 2, 3, 4.5 is transmitted to each selected terminal 6°, 6. ．． Sent to 6A6. This changeover selection switch is selected according to the characteristics of the input signal (PCM digital audio data), and the output (output of the prediction filter) selected by the changeover selection switch 6 is a so-called floating It is sent to the decimal point processing (block floating processing or block ranging processing) section. Here, the switching selection by the switching selection switch 6 is based on the characteristics of the input signal, for example, the maximum absolute value (hereinafter referred to as peak value) in the output block of each prediction filter, or the peak value multiplied by a coefficient. This is done based on the value given.

In the ranging process in the block floating processing section (block ranging processing section) to which the output of the changeover selection switch 6 is sent, the peak value is 16 bits.
Find the gain closest to the maximum positive value in the complement representation of This is done by rounding to, for example, 4 pinto or 8 pinto using a converter) 9.

At this time, so-called noise shaving is performed by feeding back the rounding noise (quantization error) generated at the end of quantization to the adder 7 via the so-called noise shaper 11.

That is, the 16-bit digital data sent from the ranging amplifier 8 to the quantizer 9 is divided by the quantizer 9 into the upper 4 bits and the lower 12 bits (or the upper 8 pins and the lower 8 pins). After being subjected to a so-called rounding process in which only the 4 bits (or the upper 8 bits) are extracted, they are outputted from the quantizer 9 via the output terminal 0. Further, the lower 2 bits (or lower 8 bits) are sent to the noise shaper 11. That is, 8 noise shapers 11
is a one-sample delayer 12.13 and a multiplier 14.15.
and an adder 16,
With this noise shaper 11, the above quantization error (rounding error)
The spectral distribution of the signal is corrected so as to have an audibly favorable shape, and then fed back to the adder 7.

[Problem to be solved by the invention]

In the conventional bit rate compression device described above, each prediction filter 2.3.4.5 has a different compression effect for each frequency band as shown in FIG. In the medium frequency band, a dynamic range (dynamic range when compressed digital audio data is reproduced) that is greater than the word length of the rounding process described above can be obtained.

However, for example, some sound sources (including harmonics) whose signal spectrum concentrates in the high range (for example, triangle 5
When digital audio data (violin, insect chirping, etc.) is supplied as an input signal, as shown in Figure 5, in the high range where straight processing by the O-order side filter 2 is effective, the dynamic signal is determined by the word length of the rounding process described above. You will only get a range. It is inevitable that the reproduced sound at this time will give an unnatural impression to the auditory sense.

In particular, in the case of 4-bit compression with a high compression ratio, the dynamic range in the high frequency range is limited to 24 dB (4X6=24, assuming 6 dB per bit), making the reproduced sound unnatural to the ear.

On the other hand, with 8-bit compression, which has a lower compression ratio than the 4-pin compression, a dynamic range of 48 dB (8X6-48) or more can be secured in the high range, so there is no audible problem with the reproduced sound. However, when 8-bit compressed data is recorded on a medium such as the CD-1 or CD-ROM te 1/2
There is a drawback.

Note that as other data compression methods such as the above-mentioned APC, there are also compression methods such as so-called SBC (bandwidth division coding). This SBC is said to have good reproducibility of high frequency signals despite its high compression ratio. In addition, among the current compression methods using SBC, various methods have been proposed that make heavy use of FFT (fast Fourier transform) processing.
It is realized using digital signal processing chips such as . However, if this digital signal processing chip is used to perform SBC data compression accompanied by the FFT processing, the frequently used FFT processing takes time and requires a large amount of memory space.

SUMMARY OF THE INVENTION The present invention has been devised in view of the above-mentioned circumstances, and an object of the present invention is to provide a bit rate compression device capable of compressing bit rates and improving dynamic range in high frequencies. The purpose is to

[Means to solve the problem]

The bit rate compression device of the present invention has been proposed to achieve the above-mentioned object, and has a plurality of bit compression predictive filters with different frequency characteristics including a filter that outputs straight PCM data, and In a bit rate compression device that selects one of the plurality of filters according to the above, performs floating point processing, quantizes and outputs the result, the straight PC is selected from among the plurality of filters.
When M filters are selected, the number of quantization focuses is increased.

[Effect]

According to the present invention, when a filter that outputs straight PCM data is selected, by increasing the number of quantization bits, the dynamic range of the reproduced signal (high frequency) of this straight PCM data is increased. .

[Example] Hereinafter, an example to which the present invention is applied will be described with reference to the drawings.

FIG. 1 shows a block diagram of a bit rate compression device according to an embodiment of the present invention. In FIG. 1, the same components as those in FIG. 4 described above are designated by the same reference numerals, and their explanation will be omitted.

The bit rate compression device shown in FIG. 1 includes a plurality of bit compression prediction filters having different frequency characteristics, including a filter (the zero-order side filter 2) that outputs straight PCM data. 2. First-order difference molecular measurement filter 3. Second-order difference molecular measurement filter 4.5,
Depending on the characteristics of the input signal (PCM digital audio data), one of the plurality of sub-measurement filters 2, 3, 4, 5 is selected and the floating point processing (the above-mentioned floating point processing or block ranging processing is performed). ), the bit rate compression device performs quantization by a quantizer (local quantizer) 9, and outputs the result. 2 is selected, the number of quantization bits in the quantizer 9 is increased based on the switching control signal from the adaptive switching selection processing circuit 60.

That is, in this embodiment, when the prediction filters 3, 4.5 other than the prediction filter 2 are selected, the output obtained by the rounding process in the quantizer 9 is the upper 4 bits. When the straight PCM prediction filter 2 is selected, the output obtained by the rounding process in the quantizer 9 is switched to the upper 8 bits. In other words, since the prediction filter 2 of the above-mentioned stray) PCM is selected in a higher frequency range than that shown in FIG. The dynamic range of high frequencies will increase.

Here, in this embodiment, in order to perform variable length encoding in which the number of bits in the quantizer 9 is switched between the upper 4 bits and the upper 8 bits, the following configuration is used. .

That is, the apparatus of this embodiment has each prediction filter 2.3.
4.5 outputs are supplied, and the quantizer 9 performs variable length bit switching control based on the outputs of each of these prediction filters, and the switching selection switch 6 is adapted to adapt to the characteristics of the input signal. The adaptive switching selection processing circuit 60 outputs a switching control signal for performing switching control.

This adaptive switching selection processing circuit 60 has, for example, two
Each prediction filter 2, 3 for 8 words (1 block)
Memories 62, 63, 6 for writing/reading the output of 4.5
4.65, and a comparison and selection circuit 66 that outputs the switching control signal based on data from each memory 6263, 64, and 65. The above comparison selection circuit 6
6, for example, the maximum absolute value (hereinafter referred to as peak value) in the output block of each prediction filter 2, 3, 4.5 from each memory 62.63, 64.65, or the value obtained by multiplying this peak value by a coefficient. Based on this, as shown in FIG. 5, the filter having the characteristic with the largest predicted range, that is, the instantaneous S/N amount, is selected for each band of the input signal (audio data).

Therefore, the changeover selection switch 6 is controlled in response to the changeover control signal from the adaptive changeover selection processing circuit 60, so that the output from the changeover selection switch 6 follows the characteristic curve F in FIG. The output is obtained by filtering with the filter characteristics shown. That is, the filter characteristic shown by the characteristic curve F shown in FIG. ,1
In the characteristic curve of the order difference molecular measurement filter 2, the characteristic curve of the zero-order prediction filter 2 is straight above about 5.2 kHz. Note that in the formats of CD-1 and CD-ROM XA, the prediction filters 2, 3, 4, .
In this embodiment, the output of the second-order difference molecular measurement filter 5 is not used.

At this time, the quantizer 9 uses the 0th-order predictive filter 2 which has the best characteristics in the high frequency range based on the switching control signal.
When the above-mentioned upper 8 bits are selected, the above-mentioned upper 8 bits are outputted, and when the middle and low band prediction filter 3.4 is selected, the above-mentioned upper 4 bits are outputted.

Note that the switching control signal from the adaptive switching selection processing circuit 60, that is, the selection information of each predictive filter and the quantizer 9
The information indicating the variable length bits is also output from the output terminal 19, and subsequent decoding is performed using this information.

Third, filter characteristic selection processing (filter word length selection processing) in the adaptive switching selection processing circuit 60;
A flowchart of a specific algorithm for variable length bit switching processing in the quantizer 9 is shown. Note that FIG. 3 describes an example in which the output of the second-order difference molecular measurement filter 5 is not used.

That is, in this flowchart, in step Sl, the peak value (straight PCM peak) in the output block of the zero-order prediction filter 2 stored in the memory 62 is
A comparison is made between the peak value (first-order difference peak) in the output block of the first-order difference molecular measurement filter 3 stored in the memory 63, and the first-order difference peak is straight PCM.
A determination is made whether it is greater than (or greater than or equal to) the peak. However, the stray at this time) PC
The M peak value may be set to, for example, 0.0 to suppress error propagation.
It is multiplied by a factor of 66. When the stray) PCM peak is small (Yes), the process proceeds to step S2, and in step S2, the quantizer 9 switches the ranging amplifier 8 (the zero-order Rounding processing is performed to output the upper 8 bits from the output of prediction filter 2). When this step S2 is completed, return to the start and process the next bronch. If the first difference peak is small (No) in the step S1, only the value of the first difference peak is determined in step S3. is extracted and sent to step S4. In step °S4, the 2
The peak in the output block of the differential molecular measurement filter 4 [(2
A comparison is made between the first-order difference peak) and the first-order difference peak value, and it is determined whether the first-order difference peak is larger than the second-order difference peak. However, the secondary difference peak value at this time is multiplied by a coefficient of 1.5, for example.

When the secondary difference peak is small (Yes), the process proceeds to step S5, and based on the switching control signal from the adaptive switching selection processing circuit 60, the quantizer 9
However, the ranging amplifier 8 (the above-mentioned second-order difference molecular measurement filter 4
Rounding processing is performed to output the upper 4 bits from the output of

If the first-order difference peak is small (No) in step S4, the quantizer 9 performs rounding processing to output the top four bits from the first-order difference measurement filter 3 in step S6. These steps S5. When S6 is completed, the process returns to the start and the next bronc process is performed.

As described above, in the bit rate compression device of this embodiment, the plurality of prediction filters 2.3, 4.5
When the straight PCM prediction filter 2 is selected, the number of quantization bits in the quantum vase 9 is increased to 8 bits, so that the
A dynamic range of 8dB or more can be secured, for example,
Even when digital audio data from some sound sources is supplied as an input signal, where the signal spectrum is concentrated in the high range (e.g., triangles, fiddler insects, etc. that contain harmonics), the reproduced sound is not audible. There is no problem. In addition, for example, in the low and medium frequency bands that constitute the basic spectrum of most sound sources, the prediction filter 3.4 (5
) is used to perform 4-bit high compression, so the recording time to the media can be improved. Furthermore, since the differential processing by the prediction filter 3.4 (5) in the low and mid-range is effective, the dynamic range (when playing compressed digital audio data) is greater than the word length of the rounding process described above. dynamic range). Furthermore, this embodiment is based on the conventional (7) CD-1
, CD-ROM XA, etc., and these formats can be maintained. It should be noted that as the algorithms of this conventional format evolve, it is also possible to improve the auditory and measurable characteristics. In addition, unlike the above-mentioned SBC (bandwidth division coding) and other methods, the bit rate compression device of this embodiment does not use processing means such as FFT (fast Fourier transform). D
SP, etc.) can be executed within the capabilities of the SP, etc. In other words, it is possible to reduce the cost by reducing the processing power of the digital signal processing chip.

〔Effect of the invention〕

The bit rate compression device of the present invention has a plurality of bit compression prediction filters with different frequency characteristics including a filter that outputs straight PCM data, and selects one of the plurality of filters according to an input signal. When a straight PCM filter is selected from among multiple filters for output after floating point processing and quantization, the bit rate compression ratio is increased by increasing the number of quantization bits. It is possible to maintain a high level and improve the dynamic range in the high range. Furthermore, since the compression ratio has been increased for the low and midrange frequencies, it is possible to improve the recording time on the media.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing a schematic configuration of a bit rate compression device according to an embodiment of the present invention, FIG. 2 is a characteristic diagram showing frequency characteristics of a filter in the device of this embodiment, FIG. 3 is a flowchart, and FIG. 4 is a conventional FIG. 5 is a block diagram showing a schematic configuration of a bit rate compression device, and FIG. 5 is a characteristic diagram showing frequency characteristics of each prediction filter. 2.3.4.5... Prediction filter

Claims (1)

[Claims] It has a plurality of bit compression prediction filters with different frequency characteristics including a filter that outputs straight PCM data, and performs floating point processing by selecting one of the plurality of filters according to the input signal. In the bit rate compression device that quantizes and outputs the bit rate compression device, the number of quantized bits is increased when the straight PCM filter is selected from among the plurality of filters. Rate compression device.