JPH11195995A - Audio video compander - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an audio video compander that uses a microcomputer to apply compression to audio data in real time by means of software where the processing of bit arrangement occupying about half of the compression processing is formed in a table beforehand so as to reduce considerably processing. SOLUTION: The compander consists of a sub band analysis section 101 that processes data sampled at a prescribed interval by a prescribed number of samples and that divides data into 32 bands, a scaling section 102 that normalizes a maximum amplitude of an output of the sub band analysis section to obtain a scale factor, a scale factor selection information calculating section 103 that sets a common scale factor when a change magnitude of the scale factors is small, a bit arrangement selecting section 104 that selects an optimal bit arrangement table among bit arrangement tables, a bit arrangement table section 105 where quantized bits number of each sub band is formed in pluralities of tables, and an audio stream generating section 107.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a video / audio compression / expansion apparatus for a video camera, an electronic still camera, a telephone and the like.

[0002]

2. Description of the Related Art As an audio compression / decompression device, for example, ISO / IE
C 11172-3: 1993 (E) (MPEG1 audio) (hereinafter, this is a known example).

[0003] MPEG1 Audio is a LAYER
1, LAYER2, and LAYER3. LAYER1 to LAYER2 and LAYER
As the number becomes 3, the compression / decompression process becomes complicated, but the sound quality and the compression become high.

FIG. 5 is a block diagram of LAYER1, and FIG.
An audio stream of ER1 is shown in FIG. 6, a block diagram of LAYER2 is shown in FIG. 7, and an audio stream of LAYER2 is shown in FIG.

In FIG. 5, reference numeral 501 denotes a sub-band analyzer which divides input audio data into 32 bands, 502
Is a scaling unit that uses the maximum amplitude of data divided into 32 sub-bands as a scale factor, 503 is an psychoacoustic processing unit that calculates the bit allocation of each band using the psychoacoustic input audio data, and 504 is A quantizing unit that quantizes the data processed by the subband analyzing unit 501 using the scale factor obtained by the scaling unit 502 and the bit allocation calculated by the psychoacoustic processing unit 503, and a quantizing unit 505 illustrated in FIG. And an audio stream generation unit that generates an audio stream of LAYER1.

In FIG. 7, reference numeral 701 denotes a sub-band analyzer for dividing input audio data into 32 bands;
Is a psychoacoustic processing unit that calculates the bit allocation of each band using the psychoacoustics of the input voice data, and 703 calculates the scale factor that is the maximum amplitude for every 12 consecutive samples of each subband, and calculates the amount of change. When the number is small, the scale factor selection information calculation unit 704, which is common, uses the bit allocation calculated by the psychoacoustic processing unit 702 and the scale factor calculated by the scale factor selection information calculation unit 703 to generate a sub-band analysis unit 701. A quantization unit 705 that quantizes the processed data includes an audio stream generation unit that generates the LAYER2 audio stream shown in FIG.

[0007]

In the above-mentioned known example, an enormous amount of calculation is required for a series of audio compression processing. Because of the enormous amount of computation, audio data cannot be compressed and processed in real time by software using a microcomputer, so a dedicated LSI for audio was required.

[0008]

In order to achieve the above object, the present invention provides a psychoacoustic processing unit 503 in FIG. 5 and a psychoacoustic processing unit 703 in FIG.
By reducing the amount of calculation to the maximum, compression processing can be realized in real time by software using a microcomputer.

[0009]

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

First, the operation of the embodiment of the present invention will be described.

FIG. 4 is a system block diagram of one embodiment of the present invention. In the figure, 401 is a lens, 402
Denotes an image sensor, 403 denotes a camera signal processing unit which converts an image signal obtained from the image sensor 402 into a video signal for photographing on a television or the like, 404 denotes a microphone, and 405 converts audio collected by the microphone 404 to digital. A / D converter 406, a video processor for compressing the video signal of the camera signal processor 103 to generate a video stream, and 407 temporarily store the audio data converted to digital by the A / D converter 405. An audio buffer 408 is an MPEG CODEC unit composed of a video processing unit 406 and an audio buffer 407, and 409 is an MPEG Audio unit 41 for compressing audio data in the audio buffer 407 and generating an audio stream.
A multiplexing unit 0 synchronizes the video stream generated by the video processing unit 406 with an audio stream generated by the MPEG audio unit 409 to generate one system stream. 411 captures and compresses video data and audio data. An MPEG camera 412 for generating a system stream is a recording medium for recording the system stream generated by the MPEG camera 411.

An image is condensed by a lens 401 and the image sensor 4
The video signal is converted into a video signal by a camera signal processing 403, and is compressed into a video stream compliant with MPEG by a video processing unit 406. On the other hand, audio
The sound is collected by the microphone 404, converted to digital by the A / D converter 405,
The audio unit 409 compresses the audio stream into an audio stream compliant with MPEG Audio. The multiplexing unit 411 synchronizes the video stream and the audio stream and records the video stream and the audio stream on the recording medium 412 in a format conforming to the MPEG system stream.

Next, the MPEG Audio unit 409 shown in FIG.
1, 8, 12, and 1 of the first embodiment of the processing in FIG.
3, FIG. 14, FIG. 15, FIG. 16, FIG. 17, FIG.

First, FIG. 1 will be described. 101 is A
The audio data processed by the / D conversion unit 405 is
, A scaling unit that uses the maximum amplitude of the data divided into 32 subbands as a scale factor, and a scale factor selection information calculation unit that is common when the change amount of the scale factor is small. , 104 is a bit allocation selection section for selecting an optimal bit allocation table from the bit allocation table section, 105 is a bit allocation table section in which the number of quantization bits of each subband is previously stored as a plurality of tables,
Reference numeral 106 denotes a scale factor calculated by the scaling unit 102 and the scale factor selection calculation unit 104, and the bit allocation table unit 10 selected by the bit allocation selection unit 104.
5, a quantization unit 107 for quantizing the data processed in the subband analysis 101, a header 107, a bit allocation,
A stream generation unit that generates an audio stream of LAYER2 shown in FIG. 8 using SCFSI (SCALEFACTOR SELECTION IMFORMATION, scale factor selection information), scale factor, and quantized data.

In the sub-band analysis 101, the band is divided into 32 sub-bands. Since this division divides the entire area equally into 32, when the sampling frequency is 32 KHz, the division is in units of 500 Hz. The scaling unit 102 calculates a maximum value for each band from the output data of the subband analysis 101, and sets a scale factor, which is a coefficient for normalizing for each subband, to three consecutive blocks in which 12 blocks are used as one block for each subband. Then, three scale factors are obtained, and a suitable index is obtained from the table of FIG. The scale factor selection information calculation unit 103 calculates the three scale factors sc calculated by the scaling unit 102.
dscf1 = scf1-scf2 dscf2 = scf2-scf3... (Equation 1) is obtained from f1, scf2, and scf3, and the class of dscf1 and dscf2 is shown in FIG.
From the value of class, and performs compression processing only with the scale factor shown in FIG. Therefore, the scale factor is compressed from a minimum of 30 to a maximum of 90.

For example, scf1 = 20, scf1 = 2
2, when scf3 = 18, dscf is obtained from (Equation 1).
1 = −2, dscf2 = 4, and from FIG.
ss1 = 2 and class2 = 5. From FIG. 17, the scale factors used for quantization in block 1, block 2, and block 3 are scf1, scf1, and scf, respectively.
1 is set to SCFSI which is 3 and scf2 is not used. In this way, when the change amount of the scale factor is small, the compression efficiency is increased by sharing the scale factor.

Next, a specific example of the bit allocation table 105 is shown in FIGS. 12, 13 and 14. FIG. In this case, the bit allocation table is selected based on the number of scale factors. Conventionally, this bit allocation is calculated for each audio frame based on psychological psychology, and this calculation occupies about half of the audio compression processing, which hinders real-time compression processing using a microcomputer. In the present invention, the part for calculating the bit distribution is deleted to make a table in advance.

In LAYER2, the maximum number of sub-bands that can be used according to the sampling frequency and bit rate, the value of bit allocation, and the corresponding quantization precision are determined in a table. FIG. 16 shows a quantization accuracy table with a sampling frequency of 32 KHz and a bit rate of 128 Kbps. In this case, up to 30 subbands out of 32 subbands can be used. In the case of LAYER2, when the quantization precision is 3, 5, or 9, quantization is performed by group coding in which three samples are combined into one and quantized.

(1) is a table used when all the scale factors are compressed, and (11) is a table used when the scale factors are compressed to the minimum. A method of creating a table used when the scale factor of (11) is compressed to the minimum will be described below.

LAYER2, sampling frequency 32K
One audio frame length 801 in Hz, bit rate 128 Kbps, and mona is 1 audio frame length = 128 [Kbps] × 1152 [samples] / 32 [KHz] = 4608 [bit]. 8, the header 802 is 32 bits, the bit allocation 803 is 94 bits in total of nbal of FIG. 16, the SCFSI is 2 bits × 30 subbands = 60 bits, and the scale factor 804 is 6 bits × 30 subbands × 1 = 180 bits. Therefore, the number of bits that can be allocated to the sample 805 is: number of bits of sample = 4608-32-94-60-180 = 4242 [bit]. First, the sub-bands 23 to 30, which are relatively high-frequency sub-bands that are hard to detect by humans, have a quantization accuracy of 3
(In the case of a quantization accuracy of 3, three samples are quantized with 5 bits). Therefore, the number of bits of the subbands 23 to 30 is 5 bits × 7 subbands × 12 = 420 bits. Therefore, the number of bits that can be allocated to subband 0 to subband 22 is 4242-420 = 3822 bits, and LAYE
In R2, since the data of 12 samples for each subband is processed as one block in three blocks, the number of assignable bits is divided by 36 to obtain 106. This is allocated to subbands 0 to 22.

Next, a method of creating a table used when the scale factor (1) is compressed at the maximum will be described.

LAYER2, sampling frequency 32K
Hz, a bit rate of 128 Kbps, and one audio frame length in monaural are also 4608 bits.
Among the 608 bits, the header 802 is 32 bits, the bit allocation 803 is 94 bits in total of nbal in FIG. 16, the SCFSI is 2 bits × 30 subbands = 60 bits, and the scale factor 804 is 6 bits × 30 subbands × 3 = 540 bits. Therefore, sample 8
The number of bits that can be assigned to 05 is: number of bits of sample = 4608-32-94-60-540 = 3882 [bit]. As in the case of (11), the subbands 23 to 30 are group-coded with a quantization accuracy of 3 (in the case of the quantization accuracy of 3, three samples are quantized by 5 bits in total). Therefore, the number of bits of subbands 23 to 30 is 5 bits × 7 subbands × 12 =
420 bits. Therefore, subband 0 to subband 22
Are 3882−420 = 346.
2 bits, and in LAYER2, 1 for each subband
In order to process two blocks of data as one block and process three blocks at a time, the number of bits that can be allocated is 36
Dividing by gives 96. This is allocated to subbands 0 to 22.

According to the present invention, a bit allocation table in units of 1 bit from 106 bits to 96 bits is created.
2, 11 tables as shown in FIGS. Minimum audible limit for distribution of bit allocation (minimum level of sound that can be detected by hearing and depends on sound frequency)
It is sorted based on. That is, more bits are allocated to sub-bands of a band that is easily detected by humans.

The bit allocation table to be selected by the bit allocation selection unit 104 is selected from the eleven tables based on the total number of scale factors calculated by the scale factor selection information calculation unit 103.

For the selection of the bit distribution table, it is conceivable to select an optimum table according to the input one (for example, the type of music).

The tables in FIGS. 12, 13 and 14 do not need to be allocated to all of the maximum number of usable sub-bands. For example, quantization bits are allocated to sub-bands 0 to 23 and sub-bands 24 to 29 are allocated. Is 0
A bit distribution table to which "" is assigned can also be created.

Next, the MPEG Audio unit 409 shown in FIG.
2, 8, 11, and 15 show a second embodiment of the processing in FIG.
This will be described with reference to FIG.

Reference numeral 201 denotes a subband analysis unit that divides the audio data processed by the A / D conversion unit 405 into 32 bands, 202 denotes a scaling unit that uses the maximum amplitude of the data divided into 32 subbands as a scale factor, 20
Numeral 3 denotes a scale factor selection information calculation unit for sharing when the amount of change in the scale factor is small, 204 denotes a bit distribution table unit in which the number of quantization bits of each subband is previously stored as a plurality of tables, and 205 denotes a scaling unit 202 and a scale unit. Scale factor calculated by factor selection calculation section 203 and bit distribution table section 204
The quantization unit 206 quantizes the data processed in the sub-band analysis 201 by using
This is a stream generation unit that generates a LAYER2 audio stream shown in FIG. 8 using CFSI, scale factor, and quantized data.

In the sub-band analysis 201, the band is divided into 32 sub-bands. Since this division divides the entire area equally into 32, when the sampling frequency is 32 KHz, the division is in units of 500 Hz. The scaling unit 202 calculates a maximum value for each band from the output data of the sub-band analysis 201, and sets a scale factor, which is a coefficient for normalizing for each sub-band, to three consecutive blocks in which each sub-band has 12 samples as one block. Then, three scale factors are obtained, and a suitable index is obtained from the table of FIG. The scale factor selection information calculation unit 203 obtains a scale factor used in quantization from the three scale factors calculated by the scaling unit 202. The bit distribution table 204 is a table in advance.

FIG. 11 shows a bit distribution table of one specific example. In the second embodiment, this bit distribution table is reduced to one. By using only one bit allocation table, the amount of calculation is further reduced, but a change in the scale factor causes a scale factor that is not actually compressed. However, since there is only one bit allocation table, bits cannot be allocated efficiently and The sound quality is worse than in the first embodiment.

The table in FIG. 11 does not need to be assigned to all of the maximum number of usable sub-bands. For example, quantization bits are assigned to sub-bands 0 to 23 and bits assigned to 0 are assigned to sub-bands 24 to 29. Distribution tables can also be created. MPEG Audio section 4
3, FIG. 6, FIG. 9, FIG.
Explanation will be made using 0. FIG. 3 is a block diagram of LAYER1 of the present invention.

Reference numeral 301 denotes a sub-band analysis unit which divides the audio data processed by the A / D conversion unit 405 into 32 bands, and 302 denotes a bit distribution which determines how many bits each of the 32 sub-bands is to be quantized. Table, 3
Reference numeral 03 denotes a scaling unit that uses the maximum amplitude of data divided into 32 subbands as a scale factor. Reference numeral 304 denotes a subband analysis 301 using the scale factor calculated by the scaling unit 103 and the bit allocation table 302.
The quantization unit 305 quantizes the data processed in step S 305. The scale factor and the quantized data are represented by L shown in FIG.
A stream generation unit that generates an AYER1 audio stream.

In the sub-band analysis 301, the band is divided into 32 sub-bands. This division equally divides the entire area into 32, and when the sampling frequency is 32 KHz, the division is made in units of 500 Hz similarly to LAYER2. The scaling unit 302 obtains a maximum value for each band from the output data of the subband analysis 301 and obtains a scale factor that is a coefficient for normalizing each subband. FIG. 9 shows a specific example of the bit allocation table 303.

This bit distribution table indicates how many bits are quantized for each subband. LAY
The actual number of quantization bits in ER1 is based on the standard (ISO / ISO
According to IEC11172-3: 1993 (E) (MPEG1 audio), the actual number of quantization bits = the value of the bit distribution table + 1 + 1 (Equation 2). However, when the value of the bit distribution table is 0, the actual number of quantization bits is 0.

Conventionally, this bit allocation is calculated for each audio frame based on psychological psychology, and this calculation occupies about half of the audio compression processing, which hinders real-time compression processing using a microcomputer. I was In the present invention, the part for calculating the bit distribution is deleted to make a table in advance. A specific example of creating the bit distribution table of FIG. 9 will be described below. LAYER1,
Sampling frequency 32KHz, bit rate 128K
bps, one audio frame length 601 in monaural
Is as follows: 1 audio frame length = 128 [Kbps] × 384 [samples] / 32 [KHz] = 1536 [bit]. In FIG. 9, the bit allocation is determined so as to use all 32 subbands per frame.

Next, a method of creating a bit distribution table will be described with reference to FIGS.

8, the header 602 has 32 bits, the bit allocation 603 has 4 bits × 32 subbands = 128 bits, and the scale factor 604 has 6 bits.
Bits × 32 subbands = 192 bits. Therefore, the number of bits that can be assigned to the sample 605 is: number of bits of sample = 1536-32-128-192 = 1184 [bit]. In LAYER1, in order to process data of 12 samples for each subband, the number of bits of the sample is divided by 12 and the number of bits allocated to the sample is set to 98 bits. The total number of bits in the bit allocation table is calculated by subtracting 32 from the expression 1 of the actual quantization bit number, and is 66.
This is distributed to 32 sub-bands. FIG. 9 shows a state of a table allocated to 32 sub-bands (sub-band 0 to sub-band 31). According to FIG. 9, since the bit allocation in subband 0 is 6, the actual number of quantization bits is quantized by 7 bits according to the above (Equation 2). This distribution is based on the minimum audible limit (the minimum level of sound that can be detected by hearing and depends on the frequency of the sound).

Next, a second specific example of the bit distribution table will be described with reference to FIGS. FIG. 10 is a bit allocation table using only 24 subbands out of 32 subbands. First, a total sum of bit allocation when using 24 subbands is obtained. Specifications are LAYER1, sampling frequency 32KHz, bit rate 128Kbps,
It is monaural. One audio frame length is 1536 bits as in the first specific example.

When 24 subbands are used, the total number of bits of the audio stream of LAYER1 in FIG.
Of the 536 bits, the header 602 is 32 bits, and the bit allocation 603 is 4 bits × 32 subbands = 12
The 8-bit scale factor 604 is 6 bits × 24 subbands = 144 bits. Therefore, sample 60
The number of bits that can be assigned to 5 is the number of bits of the sample = 1536-32-128-144 = 1232 [bits]. As in the first specific example, in LAYER1, in order to process data of 12 samples for each subband, the number of bits of the sample is divided by 12 and the number of bits allocated to the sample is set to 102 bits. From the above equation 1 of the actual quantization bit number, the total number of bits in the bit distribution table is 2
Subtract 4 to get 78. This is allocated to 24 sub-bands. FIG. 10 shows a state of a table allocated to 24 sub-bands (sub-band 0 to sub-band 23).

According to FIG. 10, since the bit allocation in subband 0 is 8, the actual number of quantization bits is quantized by 9 bits according to the above (Equation 1). This distribution is based on the minimum audible limit (the minimum level of sound that can be detected by hearing and depends on the frequency of the sound). Since the bit allocation in the subbands 24 to 31 is 0, the number of quantization bits in these subbands is 0.

That is, by using only the 24 sub-bands, the audio band becomes up to 12 KHz, but the number of quantization bits allocated to each sub-band increases. The sub-band analysis 101 and the scaling unit 10
2. Since the quantization unit 104 and the bit stream generation unit 105 need to calculate only 24 subbands out of 32 subbands, the amount of calculation can be further reduced.

[0042]

According to the present invention, the amount of calculation for audio compression processing can be reduced by previously setting a bit allocation obtained for each frame as a table.

As a result, since audio compression processing can be performed by software using a microcomputer, a dedicated LSI for audio is not required.

[Brief description of the drawings]

FIG. 1 is a block diagram showing an embodiment of a video / audio compression / decompression device according to the present invention.

FIG. 2 is a block diagram showing an embodiment of a video / audio compression / decompression device according to the present invention.

FIG. 3 is a block diagram showing an embodiment of a video / audio compression / decompression device according to the present invention.

FIG. 4 is a system block diagram showing an embodiment of a video / audio compression / decompression device according to the present invention.

FIG. 5 is a block diagram of a conventional example of LAYER1.

FIG. 6 is a diagram illustrating an audio stream of LAYER1.

FIG. 7 is a block diagram of a conventional example of LAYER2.

FIG. 8 is a diagram illustrating an audio stream of LAYER2.

FIG. 9 is a bit distribution table used in FIG. 1;

FIG. 10 is a bit distribution table used in FIG. 1;

FIG. 11 is a bit distribution table used in FIG. 2;

FIG. 12 is a bit distribution table used in FIG. 3;

FIG. 13 is a bit distribution table used in FIG. 3;

FIG. 14 is a bit distribution table used in FIG. 3;

FIG. 15 is a scale factor table.

FIG. 16 is a diagram used in a scale factor selection information calculation unit.

FIG. 17 is a diagram used in a scale factor selection information calculation unit.

FIG. 18 is a diagram illustrating quantization accuracy in a quantization unit.

[Explanation of symbols]

 101: Subband analysis unit 102: Scaling unit 103: Scale factor selection information calculation unit 104: Bit allocation selection unit 105: Bit allocation table unit 106: Quantization unit 107 -Bit stream generation unit 108 ... MPEG Audio processing unit

 ──────────────────────────────────────────────────続 き Continuing on the front page (72) Inventor Fumishige Yagi 1410 Inada, Hitachinaka-shi, Ibaraki Pref. Hitachi, Ltd. Visual Information Media Division

Claims (12)

[Claims] 1. A sub-band analyzer for processing data sampled at a fixed interval with a fixed number of samples and dividing the data into 32 bands, a scaling unit for normalizing the maximum amplitude of the output of the sub-band analyzer and using the scale factor as a scale factor, When the change amount of the scale factor is small, a common scale factor selection information calculation unit, a bit allocation selection unit that selects an optimum bit allocation table from the bit allocation table unit, and a plurality of quantization bits in each subband in advance , A quantization unit for quantizing the data processed by the sub-band analysis unit using a selected bit allocation table and a scale factor, a header, a bit allocation, a SCFSI (SCALEFACTO)
An audio / video compression / decompression device comprising: an audio stream generation unit that generates an audio stream from R SELECTION IMFORMATION, scale factor selection information), a scale factor of each subband, and quantized data. 2. The video / audio compression / expansion apparatus according to claim 1, wherein the bit allocation table section is not calculated for each frame, but is fixed as a table in advance. 3. The video / audio compression / expansion apparatus according to claim 1, wherein the bit allocation table section is a plurality of tables in which bit allocation is allocated to all of the maximum usable subbands. 4. The video / audio compression / expansion apparatus according to claim 1, wherein the bit allocation table section allocates bit allocation to several sub-bands among the maximum usable sub-bands. 5. A sub-band analyzer for processing data sampled at a fixed interval with a fixed number of samples and dividing the data into 32 bands, a scaling unit for normalizing the maximum amplitude of the output of the sub-band analyzer and using the scale factor as a scale factor, When the change amount of the scale factor is small, a scale factor selection information calculation unit to be used in common, a bit allocation table unit in which the number of quantization bits of each subband is previously stored as a table, and the subband analysis using the bit allocation table and the scale factor. A quantization unit for quantizing the data processed by the unit, a header, bit allocation, SCFSI, a scale factor of each subband, and an audio stream generation unit for generating an audio stream from the quantized data. Video / audio compression / decompression device. 6. A video / audio compression / decompression apparatus according to claim 5, wherein the bit allocation table section is not calculated for each frame, but is fixed as a table in advance. 7. A video / audio compression / decompression apparatus according to claim 5, wherein the bit allocation table section allocates bit allocation to all of the maximum usable subbands. 8. A video / audio compression / decompression apparatus according to claim 5, wherein the bit allocation table section allocates bit allocation to several sub-bands among the maximum usable sub-bands. 9. A sub-band analyzer for processing data sampled at a fixed interval with a fixed number of samples and dividing the data into 32 bands, a scaling unit for normalizing the maximum amplitude of the output of the sub-band analyzer and using the scale factor as a scale factor, A bit allocation table unit that previously stores the number of quantization bits of each subband as a table, a quantization unit that quantizes the data processed by the subband analysis unit using the bit allocation table and the scale factor, a header, and a bit allocation And an audio stream generation unit for generating an audio stream from a scale factor of each sub-band and quantized data. 10. A video / audio compression / decompression apparatus according to claim 9, wherein the bit allocation table section is not calculated for each frame, but is fixed as a table in advance. 11. The video / audio compression / decompression device according to claim 9, wherein the bit allocation table section allocates bit allocation to all 32 subbands. 12. A video / audio compression / expansion apparatus according to claim 9, wherein the bit allocation table section allocates bit allocation to several sub-bands out of 32 sub-bands.