JP2001249699A - Sound compression device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To improve sound quality by almost averaging the bit distribution number of each group in which sub-bands are divided, in layer-II mode of MPEG1 audio. SOLUTION: A scale factor(SF) is generated by normalizing audio data divided by a sub-band analysis part 2 in every plural sub-bands by a scaling part 3. A scale factor(SF) selection information calculating part 4 makes SF common in 3-frame coding units for each sub-band. A bit distribution table selection part 9 selects a bit distribution table corresponding to the number of the common SF from a bit distribution table part 8, and a bit distribution allocation part 10 allocates the bit distribution number of this bit distribution table, so that the sub-bands are divided into groups of different frequency bands and the bit distribution number of each group is almost averaged, and a quantization part 5 quantizes the data in coded units for each normalized sub-band according to this new bit distribution table.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

The present invention relates to a DVD (Digital)
The present invention relates to an audio compression device used for a Versatile Disc (Digital Video) camera, a DV (Digital Video) camera, an electronic still camera, a telephone, and the like.

[0002]

2. Description of the Related Art Conventionally, as a compression method for audio data,
For example, ISO / IEC standard method (ISO / IEC 11172-3: 1993 (E))
MPEG1 audio which is compression-encoded in conformity with the standard is known. The present invention realizes highly efficient compression of digital audio data by utilizing human psychoacoustic characteristics.

FIG. 6 is a block diagram showing an audio compression apparatus based on this standard system, wherein 1 is an audio compression apparatus, 2 is a subband analysis section, 3 is a scaling section, 4 is a scale factor selection information calculation section, 7 Denotes an psychoacoustic processing unit, 5 denotes a quantization unit, and 6 denotes a bit stream generation unit. Here, the encoding mode is layer II.

In FIG. 1, digital audio data A _IN output from an A / D conversion circuit (not shown) is supplied to a sub-band analyzer 2. In this sub-band analyzer 2,
The digital audio data A _IN is digitally filtered in units of 12 samples, thereby dividing it into 32 sub-bands and outputting audio data of 32 different sub-bands. In MPEG1 audio, this 12 samples / 32 subbands (ie, 12 × 32 = 3)
(84 samples) is regarded as one frame, but in the case of the layer II mode, three consecutive frames are encoded collectively for each subband. The three-frame unit is the smallest unit AAU (Audio Access Unit) that can be decoded independently. However, in this case, since audio data is encoded, it is referred to as an audio encoding unit for convenience.

[0005] Audio data of 32 sub-bands is supplied to a scaling unit 3. The scaling unit 3 normalizes the waveform of the audio data of each sub-band so that the sample of the maximum amplitude value has the same specified amplitude for each frame (12 samples).
The scaling factor (dB) for returning the normalized audio data to the original waveform at the time of decoding is represented by a scale factor (Scale F).
actor). In MPEG audio, according to the above-mentioned ISO / IEC standard method, the scale factor has a value of 0.
Index consisting of 63 6-bits (Inde)
x), these indexes have +6 dB to -1
A scale factor of about 2 dB is associated with a range of 18 dB, and the scale factor is represented by a number called a 6-bit index. Therefore, in the following, such an index is referred to as a scale factor number and is referred to as an SF number. The smaller the SF number, the larger the scale factor. Scaling unit 3
Is used to normalize the audio data of one frame unit for each subband as described above.
The most suitable scale factor among the scale factors defined in the ISO / IEC standard system is represented by an SF number, and this is used as the scale factor of this subband.

[0006] The thus-normalized audio data of each sub-band is supplied to the quantization unit 5, and the scale factor is supplied to the scale factor selection information calculation unit 4.

The scale factor selection information calculation section 4
Detects, for each sub-band, a change in scale factor in an audio coding unit of three consecutive frames, which is an audio coding unit, by using an SF number. This is to make the scale factor common among the frames to be processed. For example, now, S frames of three frames forming an audio coding unit
Assuming that the F numbers are SF1, SF2, and SF3 in order, and the amount of change from SF1 to SF2 is within a specified range, in this audio coding unit, the scale factor of SF2 is not used, and the two scales of SF1 and SF3 are used. The factor will be used. Here, “used” means that the encoded scale factor is added to the audio stream A _ST and output, and is used for decoding the audio stream A _ST .
In this case, the scale factor of SF1 is
It is used in place of SF2, and performs commonality of assigning the scale factor of SF1 instead of the scale factor of SF2 to a frame in which the scale factor of SF1 is originally decoded and reproduced using the scale factor of SF2. Things. In order to indicate such a state of commonality, the scaling unit 3 further includes scale factor selection information (SCFSI: SCALEFACTOR SELECTION IMFORMA) indicating how the scale factors are standardized.
TION) is formed. This scale factor selection information SCF
The SI is also supplied to the quantization unit 5 together with the common scale factor (SF).

[0008] Even for a sub-band of an audio coding unit in which the amount of change between the three scale factors is large and the scale factors are not shared, the 3-bit scale factor selection information SCFSI indicating the same is also provided. It is formed.

The scale factor (SF number) and the scale factor selection information SCFSI represent the scale factor actually used in the subband, and this is represented by a bit stream generation unit described later. 6.

As described above, the scale factor (SF number) and the scale factor selection information subjected to the common processing
The SCFSI is supplied to the bit stream generator 6.

On the other hand, the psychoacoustic processing unit 7 uses the psychoacoustics of a human from the input digital audio data A _IN for each three consecutive audio coding units.
A bit allocation table (allocation: Al) for allocating the number of bits for each subband when performing quantization in the next quantization unit 5
location). Therefore, even in the same subband,
The number of bits allocated to each audio coding unit may be different. In the case of allocating the number of bits using such a human auditory psychology, the minimum audible limit during silence and a masking threshold determined by this and the sample of the input digital audio data A _IN (by actually listening to the original sound) The number of quantization bits of the sub-band to be quantized by the quantization unit 5 is distributed from the minimum perceptible level of unnecessary sound such as noise when the sub-band is quantized (hereinafter, distributed to the sub-bands in this manner). The number of quantization bits is referred to as a bit allocation number). In this case, the quantization step is determined so that the level of the quantization noise is smaller than the masking threshold. As a result, sound quality degradation due to quantization noise can be prevented. Digital audio data A _IN of an input audio coding unit In a sub-band smaller than the masking threshold, the number of quantization bits to be allocated is set to 0 (allocation = 0). This bit allocation table is supplied to the quantization unit 5 and the bit stream generation unit 6.

The quantization unit 5 quantizes and encodes audio data of each subband for each audio encoding unit. This quantization is performed for each sample of 3 frames / 32 sub-bands in a predetermined order, and the bit allocation number of each sample is allocated to the sub-band to which it belongs according to the bit allocation table formed by the psychoacoustic processor 7. Is the number of quantized bits. The audio data of the audio coding unit including the quantized samples is supplied to the bit stream generation unit 6.

The bit stream generator 6 adds the audio data quantized by the quantizer 5 to the psychoacoustic processor 7.
, A scale factor selection information calculation section 4 and a scale factor selection information SCFSI, and a header including information such as a synchronization word, a layer designation, a bit rate, and a sampling frequency, and an error check. A code (optional) is added, and output as an audio stream _AST . In addition, for the sub-band where the bit allocation number is 0,
The scale factor (SF number) is omitted.

In such MPEG audio, the psychoacoustic processing unit removes audio information that is low in level and hardly audible to human ears, or audio information that is difficult to hear due to high-level audio information in a nearby frequency band. 7, the bit allocation for each sub-band is performed, so that the code amount of the audio data can be greatly reduced, and within a narrow bandwidth, a low-level voice is masked by a high-level voice and becomes inaudible. The use of such a technique (sub-band division) makes it possible to eliminate the influence of quantization noise generated at the time of encoding, thereby enabling highly efficient compression.

By the way, in such a voice compression device, the bit allocation processing (processing for creating a bit allocation table) for each sub-band in the psychoacoustic processing apparatus 7 requires an enormous amount of calculation, and the creation processing takes an extremely long time. However, since the digital audio data A _IN cannot be compressed in real time by software using a microcomputer, a dedicated LSI for audio is required. Japanese Patent Laid-Open No.
This is proposed in Japanese Patent Publication No. 1-195995.

This is not to create a bit allocation table from input audio data using the psychoacoustic characteristics as in the above-described audio compression apparatus, but to use one or more bit allocation tables in advance based on the sampling frequency and bit rate. A table is created and used. When a plurality of bit allocation tables are used, these bit allocation tables differ in the total number of bits allocated to each subband, and the allocation to the subbands is performed by humans based on the minimum audible limit described above. Many bits are allocated to subbands that are easy to detect. Therefore, more bits are allocated to the lower band subband. Further, based on the total number of scale factors shared by the scale factor selection information calculation unit (see FIG. 6), an optimal bit distribution table is selected and used.

According to this, at a commonly used bit rate (for example, about 128 kbps), the number of bits allocated to each subband increases, so that the compression process can be performed effectively and Sound quality is also improved.

Japanese Patent Application No. 10-253,642 discloses that
There is also an example in which bit allocation is calculated according to the size of a subband, and the sound quality is further improved.

[0019]

Problems to be Solved by the Invention Japanese Patent Laid-Open No. 11-19 / 1999
In the prior art of No. 5995, as described above, there is an effect that the reproduction process of the bit allocation table for each sub-band is not required and the sound quality is improved, but on the other hand, the sampling frequency is limited to CD or MD. Used 44.
In the case of 1 kHz, the D
At 48 kHz, which is the audio sampling frequency used in the VD camera, the user cannot help but expect higher sound quality. Therefore, it is necessary to further improve the sound quality.

In the case of using the bit allocation table prepared in advance as described above, if the bit allocation is performed for all the sub-bands at one time, the bias of the bit allocation is inevitably shifted to the lower band side sub-band. And the number of bits allocated to the high frequency side is reduced, so that improvement in sound quality may not be expected.

An object of the present invention has been made in view of the above points, and an object of the present invention is to provide an audio compression apparatus capable of surely achieving further improvement in sound quality while using a bit allocation table. Is to do.

[0022]

In order to achieve the above object, the present invention provides a method of processing digital audio data for each frame of a fixed number of samples, dividing the data into a plurality of sub-bands, and encoding a predetermined number of frames. Normalize the audio data of each sub-band for each unit to form a scale factor for each, and divide the plurality of sub-bands into a plurality of groups having different bands for each coding unit, and Are allocated to the groups such that the total number of allocated bits among the groups is substantially averaged,
In addition, a bit allocation table is created by allocating the number of allocated bits to sub-bands according to the scale factor for each group, and the quantization bit number of the number of allocated bits in the bit allocation table is used for each encoding unit. The quantized audio data of the band is quantized, and a header, a scale factor, and the like are added to the quantized audio data of a coding unit to form an audio stream.

Also, the present invention processes digital audio data for each frame of a fixed number of samples and divides the data into a plurality of subbands, and normalizes the audio data of each subband for each frame. A scaling unit that uses a scaling factor for normalization as a scale factor, and a plurality of consecutive frames as a coding unit, and a scale factor in the coding unit for each subband is shared between frames, and A scale factor selection information calculation unit for generating scale factor selection information representing a factor commonality state, and a bit distribution table unit storing a plurality of bit distribution tables in which the number of bits for each subband is allocated in a specific manner in advance And the bit allocation table corresponding to the scale factor selection information for each coding unit. A bit allocation table selecting unit for selecting from the bit allocation table unit, and the subband divided by the subband analyzing unit are divided into a plurality of different band groups, and the subband belongs to each coding unit. A group determining unit for determining a group; and a bit allocation table for each of the coding units, based on a determination result of the group determining unit, such that the distribution of the number of quantization bits among the groups is substantially averaged. Allocating the bit allocation number in the bit allocation table selected by the section to the sub-band,
A bit allocation allocating unit that creates a new bit allocation table; and a quantization unit that quantizes audio data of each of the subbands with the number of bits allocated by the new bit allocation table for each coding unit. The new bit allocation table, the common scale factor, the scale factor selection information, and the like are added to the audio data of the subband of the coding unit quantized by the quantization unit to generate an audio stream. And a bit stream generation unit.

As described above, the total number of bit allocations among the groups is substantially averaged, and the number of bits allocated in the subband is determined in each group according to the scale factor. The sound quality of the sound obtained by decoding the audio stream is improved without being biased to bands. In addition, since the bit allocation table used for quantization is created by allocating the number of bit allocations of the bit allocation table created in advance, the bit allocation table used for quantization is obtained almost in real time with a small amount of calculation. be able to.

Further, according to the present invention, if the number of subbands divided from digital audio data is 32, one frame is 12 samples, and three consecutive coding units are used, By using the common scale factor, the code amount of the scale factor can be reduced.

[0026]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing an embodiment of an audio compression apparatus according to the present invention, wherein 8 is a bit allocation table section, 9 is a bit allocation table selection section, 10 is a bit allocation distribution section, and 11 is a band determination section. 6, parts corresponding to those in FIG. In this case, too, the mode is changed to Layer II, as in the previous examples.
And

In the figure, a plurality of bit allocation tables having different sums of the number of allocated bits are set in the bit allocation table section 8. Bit distribution table selector 9
Determines the number of scale factor scalefactor selection information SCFSI that the scale factor selection information calculating section 4 is calculated as described above based on is added to the audio stream A _ST (SF number), depending on the number The bit allocation table of the total number of allocated bits is selected from the bit allocation table section 8.

The band determining unit 11 determines a subband group. In this embodiment, 32 subbands subdivided by the subband analysis unit 2 are divided into a plurality of groups (subband groups) having different bands, and the band determination unit 11 allocates the number of allocated bits. This is to determine and specify which subband group the subband to be processed belongs to.

The bit allocation allocating unit 10 determines the number of bits allocated in the bit allocation table selected by the bit allocation table selecting unit 9 on the basis of the scale factor calculated by the The distribution is performed for each sub-band of the band group, and a new bit distribution table is created.

The quantizing section 5 includes a bit distribution allocating section 10
Then, the audio data normalized by the scaling unit 3 is quantized for each audio coding unit in the same manner as in the conventional audio compression apparatus shown in FIG. In addition, the bit stream generation unit 6 may use a CRC (Cyclic Redundancy Che
ck), an error detection word is generated, and the audio data for each audio coding unit quantized by the quantization unit 5 is
Header, error detection word, bit allocation table formed by bit allocation distribution unit 10, scale factor (SF number) output from scale factor selection information detection unit 4
And scale factor selection information SCFSI and the like, and output as an audio stream _AST .

FIG. 2 is a diagram showing a specific example of a bit distribution table stored in the bit distribution table section 8 in advance.

In FIG. 3, three bit distribution tables A, B, and C are shown. These are for the case of layer II having a sampling frequency of 48 kHz and a bit rate of 128 kbps. In this case, the subbands to which the quantization bits are allocated are the MPEG audio standard ISO / IEC 11172-3. : According to 1993 (E), 27 sub-bands out of 32 sub-bands, and the remaining
The 31st subband (referred to as subbands 27 to 31).
The same applies to the following). Here, for the sake of simplicity, the remaining subbands 0 to
26 of 27 subbands, 1 of subbands 0 to 14
The five subbands will be described. For the subbands 0 to 26, the total number of bits allocated in the bit allocation table A is 50, the total number of bits allocated in the bit allocation table B is 51, and the total number of bits allocated in the bit allocation table C is 52. Further, the maximum value of the bit allocation based on psychological perception is about 7 bits at 128 kbps per channel.

The bit distribution tables stored in the bit distribution table section 8 are bit distribution tables A, B, and C.
As shown by, the bit allocation number is arranged such that the bit allocation number increases in descending order of the scale factor. In such a bit distribution table,
A large number of quantization bits are allocated to the low-frequency side sub-band and become biased. For example, in a band easily perceived by the human ear such as 1 kHz to 5 kHz, the number of allocated bits decreases, The sound quality of the decoded speech will be impaired.

Therefore, in this embodiment, when one bit allocation table is selected from the bit allocation table section 8 by the bit allocation table selection section 9, the bit allocation distribution section 10 causes the scale factor obtained by the scaling section 3 to be obtained. (SF number) and the number of bits allocated to the subband are re-allocated based on the determination result of the band determination unit 11,
This is to create a new bit distribution table.

The distribution of the number of bits to be allocated is as follows.
The distribution of the number of bits over the sub-band is approximately averaged. For this purpose, in this embodiment, all the sub-bands are divided into a plurality of groups (sub-band groups) having different bands, and the total number of bits allocated to each sub-band group is substantially averaged. As an example, all the sub-bands are divided into three sub-band groups of a low band, a middle band, and a high band. The bit distribution table A shown in FIG.
In B and C, the leftmost subband is the lowest subband of the 15 subbands, and the rightmost subband is also the highest subband of the 15 subbands.
Subbands are subbands of a lowband subband group, the middle five subbands are subbands of a middle subband group, and the right five subbands are subbands of a high subband group. .

If the bit allocation table C in FIG. 2 is selected by the bit allocation table selecting section 9, the bit allocation allocating section 10 determines that the total number of bits allocated to each of the three subband groups is three. Assign to each sub-band group so that they are almost averaged.
Here, the bit allocation number 7, 6, 5, 4, 3 in the bit allocation table C is assigned to the low band subband group (total 25).
Bit), the bit allocation number 6, 4, 2, 2, 1 (total 15 bits) for the middle band sub-band group, and the bit allocation number 5, 3, also for the high band sub-band group.
2, 1, and 1 (total of 12 bits) are respectively allocated. This is averaged between the sub-band groups as compared with the case where the bit distribution table C in FIG. 2 is similarly grouped (29 bits in the low band, 16 bits in the middle band, 7 bits in the high band). However, rather than completely averaging the number of allocated bits among the low band, the middle band, and the high band, it is better to increase the number of allocated bits in the lower band from the relationship of the amount of information. Thus, the total number of allocated bits becomes smaller in the order of the middle band and the high band.

The scale factor obtained by the scaling unit 3 is also sent to the bit allocation distribution unit 10 and stored by being divided for each subband group.

The band judging section 11 designates a low band. As a result, the bit allocation distribution unit 10 compares the scale factors of the five subbands in the low frequency subband group with the SF numbers. Now, the subbands in the lowband subband group are sequentially assigned to subbands 0, 1,
2, 3, 4 and these SF numbers (0), (1),
Assuming that (2), (3), and (4) are in the order of (0) <(1) <(2) <(3) <(4), first, low SF number is lower than subband 0. The highest allocated bit number 7 allocated to the subband group of the area is allocated, the second highest allocated bit number 6 is assigned to the subband 1 having the second lowest SF number, and the third lowest allocated number is next. The sub-band 2 of the SF number has the third largest bit allocation number 5,
The fifth largest bit allocation number 4 is allocated to the subband 3 having the lowest SF number, and the smallest bit allocation number 3 is allocated to the subband 4 having the highest SF number.

After the distribution of the number of allocated bits in the low-frequency sub-band group is completed in this way, the band determination unit 11 next specifies the middle frequency band. As a result, the bit allocation distribution unit 10 selects 5
The SF numbers of the scale factors of the subbands are compared. Now, the subbands in the subband group in the middle band are referred to as subbands 5, 6, 7, 8, 9 in order from the low band side,
These SF numbers (5), (6), (7), (8),
Assuming that (9) is in the order of (5) <(6) <(7) <(8) <(9), subband 5 having the smallest SF number is first allocated to the subband group in the middle band. Then, the largest bit allocation number 6 is assigned, the second largest bit allocation number 6 is assigned to the subband 6 having the second lowest SF number, and the third is assigned to the subband 7 having the third lowest SF number. The 5th largest bit allocation number is 5
The fourth largest bit allocation number 4 is allocated to the subband 8 having the lowest SF number, and the smallest bit allocation number 3 is allocated to the subband 9 having the highest SF number.

After the distribution of the bit allocation numbers in the middle band sub-band group is completed in this way, the band determination unit 11 next specifies the high band. As a result, the bit allocation distribution unit 10 selects 5
The SF numbers of the scale factors of the subbands are compared. Now, the subbands in the highband subband group are sequentially arranged from the lowband side to subbands 10, 11, 12, 13,.
14 and these SF numbers (10), (11), (1)
2), (13), and (14) are in the order of (10) <(11) <(12) <(13) <(14). Are assigned the largest bit allocation number of 5 assigned to the subband groups of
The sub-band 11 with the lowest SF number has the second largest bit allocation number 3, the sub-band 12 with the third lowest SF number has the third largest bit allocation number 2, and the fourth next. The fourth largest bit allocation number 1 is allocated to the sub-band 13 having the lowest SF number, and the smallest bit allocation number 1 is allocated to the sub-band 14 having the highest SF number.

Thus, a new bit distribution table as shown in FIG. 3A is created. In this new bit allocation table, as described above, the number of bits allocated to the low band sub-band group is 25, the number of bits allocated to the middle band sub-band group is 15, and the number of bits allocated to the low band is 15 The number is 12, which is averaged as compared with the bit allocation shown in FIG. 2 (low frequency: 29 bits, middle frequency: 16 bits, low frequency: 7 bits). This new bit allocation table is used for quantization of each subband in the quantization unit 5, and is added to the audio stream _AST by the bit stream generation unit 6.

The method of assigning the number of allocated bits is merely an example, and other methods may be used. For example, in the above example, the distribution is performed for the low band sub-band group, and when this is completed, the distribution is performed for the middle band sub-band group, and finally, the distribution is performed for the high band sub-band group. However, one subband may be sorted in the order of the low band subband group, the middle band subband group, and the high band subband group.

FIG. 3B shows a bit distribution table obtained by this distribution method. In this case, the SF numbers of the sub-bands in the low-band, middle-band, and high-band sub-band groups are set as described above. First, the band determination unit 11 specifies a low band. As a result, the bit allocation distribution unit 10 first compares the SF numbers of the five subbands in the low band subband group, and maps the subband 0 having the smallest SF number in the low band subband group. The largest bit allocation number 7 in the bit allocation table C of 2 is assigned. Next, the band determination unit 11 specifies a middle band. As a result, the bit allocation distribution unit 10 selects 5
The SF numbers of the subbands are compared, and the subband 5 having the lowest SF number in the subband group in the middle band is assigned the second largest bit allocation number 6 in the bit allocation table C of FIG. Next, the band determination unit 11 specifies a high band. As a result, the bit allocation distribution unit 10 compares the SF numbers of the five subbands in the high band subband group and determines the S number in the high band subband group.
The third largest bit allocation number 6 in the bit allocation table C of FIG. 2 is allocated to the subband 10 with the smallest F number. Next, the band determination unit 11 specifies a low band. Thereby, the bit allocation allocating unit 10 assigns the largest bit allocation among the remaining bit allocation numbers in the bit allocation table C of FIG. 2 to the subband 1 having the second lowest SF number in the low band subband group. Equation 5 is assigned. Next, the band determination unit 11 specifies a middle band. As a result, the bit allocation distribution unit 10 assigns the subband 6 having the second lowest SF number in the subband group in the middle band to the second largest among the remaining bit allocation numbers in the bit allocation table C in FIG. A bit allocation number of 5 is assigned. Next, the band determination unit 11 specifies a high band. As a result, the sub-band 11 having the second lowest SF number in the high-band sub-band group
Is assigned the third largest bit allocation number 4 among the remaining bit allocation numbers in the bit allocation table C of FIG.

Similarly, the bit allocation number 4 is assigned to the subband 2, the bit allocation number 3 is assigned to the subband 7, the bit allocation number 2 is assigned to the subband 12, the bit allocation number 3 is assigned to the subband 3, and 8, bit allocation number 1 for subband 13, bit allocation number 2 for subband 4, bit allocation number 1 for subband 9, subband 1
4, bit allocation numbers 1 are sequentially assigned.

In this way, a new bit distribution table as shown in FIG. 3B is obtained. In this new bit allocation table, the number of bits allocated to the low band sub-band group is 22, the number of bits allocated to the middle band sub-band group is 17, and the number of bits allocated to the low band is 13 2 (low frequency: 29 bits, middle frequency: 16 bits, low frequency: 7 bits) shown in FIG.

FIG. 4 is a flow chart showing the operation when the bit distribution distribution unit 10 in FIG. 1 creates the bit distribution table shown in FIG.

In the figure, first, the low band sub-band group is designated by the band judgment section 11 (step 10).
0), whereby the operation from step 104 is executed as follows.

That is, in the case of Layer II, encoding is performed for each audio encoding unit (three consecutive frames), and a scale factor (SF number) is set for each of these three frames. Thus, for each subband, the scale factor of the maximum value (SF number of the minimum value) among the scale factors for the three frames of the audio coding unit is detected, and this is set as the scale factor of the subband (step 105). . As a result, five subbands in the lowband subband group (FIG. 3)
In the case of (b), the scale factor of each of the subbands 0 to 4) is determined. Then, the order of these five sub-bands 0 to 4 is set in the descending order of the scale factor (step 105).

Next, the scale factor selection information calculation section 4
The number of scale factors in the low-frequency subband group is calculated based on the scale factor selection information SCFSI from, and a bit allocation table selecting unit 9 generates a bit allocation table corresponding to the number of scale factors by the bit allocation table selecting unit 9. , And the number of bits allocated in the bit allocation table is adjusted so that the total number of bits allocated in the low band, middle band, and high band sub-band groups is substantially averaged. Classification is performed for each band group (step 106). Assuming that the read bit allocation table is the bit allocation table C shown in FIG. 2, the bit allocation number of the bit allocation table C is allocated to the low band, middle band, and high band sub-band groups as described above.

Then, the low band sub-band groups are allocated to the low band sub-band groups in accordance with the order set in step 105, so that the larger the number of allocated bits is, the faster the sub-band is. A bit allocation number is assigned to each subband (step 107). When the bit distribution table shown in FIG. 2 is selected from the bit distribution table section 8, FIG.
As described in (b), the bit allocation table is allocated to subbands 0 to 4.

When the process of allocating the number of allocated bits in the low band subband group is completed (step 10).
8) Next, the middle band sub-band group is designated by the band decision unit 11, and the designation is decided (step 102), whereby the same processing of steps 104 to 107 is performed for the middle band sub-band group. Is performed,
Bit allocation numbers are assigned to subbands 5 to 9. When the process of allocating the number of allocated bits in the middle band sub-band group is completed (step 108), the band judging unit 11 next specifies the high band sub-band group, and the designation is determined ( Step 10
3) The same processing of steps 104 to 107 is performed for the high band sub-band group, and the sub-band 10
The number of allocated bits is assigned to .about.14. When the allocation of the bit allocation numbers to all the subbands is completed, a new bit allocation table as shown in FIG. 3B is obtained.

When the allocation bit allocation table shown in FIG. 3B is created, the sub-band group is changed every time the number of allocated bits is allocated to one sub-band. If it is determined in step 100 that the subband group is a low band subband group, the operations of steps 104 to 107 are performed for one subband, and the next middle band subband group is specified in step 108. From step 104. Hereinafter, steps 104-1 in one subband
Each time the operation of step 07 is performed, steps 100, 10
Determinations are made in the order of 2, 103, and the routine proceeds to step 104. However, in this case, the detection of the scale factor of the maximum value and the ordering of the subbands in steps 104 and 105 are performed for those other than those to which the number of allocated bits has already been assigned. Step 106 is performed at the time of assigning the number of allocated bits to the first subband. At this time, the number of allocated bits remaining after being allocated to each subband group is performed.

FIG. 5 is a block diagram showing a specific example of a system using the audio compression apparatus according to the present invention. Here, a DVD camera is shown.
1 is a lens, 22 is an image sensor, 23 is a camera signal processing unit, 24 is an MPEG CODEC unit, 25 is a microphone (hereinafter referred to as a microphone), 26 is an A / D converter, 27
, An audio buffer; 28, a microcomputer; 29, an MPEG audio unit using the audio compression apparatus according to the present invention; 30, a multiplexing unit; and 31, a recording medium.

In FIG. 1, the DVD camera 20 encodes a video signal in accordance with the MPEG1 or 2 standard, encodes an audio signal in accordance with the MPEG audio standard, and multiplexes and outputs these. Or a built-in recording medium (for example, a large-capacity recording medium such as an optical disk such as a DVD-RAM or a magnetic tape) 31.

The subject image incident through the lens 21 is converted into an electric signal by the image pickup device 22. This electric signal is processed by the camera signal processing unit 23, converted into a video signal, and compression-encoded by the MPEG CODEC unit 24 to generate a video stream. This video stream is supplied to the microcomputer 28. An audio signal output from a microphone 25 for inputting audio information is converted into digital audio data by an A / D converter 26, and then temporarily stored in an audio buffer 27.
Read by.

The microcomputer 28 has an MPEG audio unit 29 and a multiplexing unit 30 each of which is the above-described audio compression device according to the present invention, and converts digital audio data read from the audio buffer 27 as described above. Then, the audio stream is formed by compression encoding in accordance with the MPEG audio standard in the MPEG audio unit 29, and the audio stream and the video stream fetched from the MPEG CODEC unit 24 are multiplexed in the multiplexing unit 30. . The multiplexed signal thus obtained can be output to the outside or recorded on the built-in recording medium 31.

Although the embodiments of the present invention have been described above, the present invention is not limited to only such embodiments.

For example, in the above embodiment, all the sub-bands are allocated to three sub-band groups of low band, middle band, and high band. However, the present invention is not limited to this. The number of bits allocated to each of the subband groups is, of course, substantially equal, and averaged bit allocation is performed.

In the above-described embodiment, the mode is the layer II in which three frames are used as an audio coding unit.
It may be a layer I in which one frame is an audio coding unit. However, in this case, since the scale factor is not shared, the scale factor selection information calculation unit 4 is not used, and since the number of scale factors in the coding unit is constant, one bit Using a distribution table, this is processed by the bit distribution allocating unit 10 to create a new bit distribution table in which the number of bits allocated to each subband is determined according to the scale factor of each subband group at that time. You may.

[0060]

As described above, according to the present invention,
The bias of the number of allocated bits to a specific sub-band can be alleviated, and compression encoding of audio data with further improved sound quality can be performed.

According to the present invention, since a bit distribution table prepared in advance can be used, a bit distribution table making use of the auditory truth characteristic which occupies most of the operation amount of the compression encoding processing of audio data. Becomes unnecessary, and almost real-time compression encoding processing becomes possible.

[Brief description of the drawings]

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

FIG. 2 is a diagram showing a specific example of a bit distribution table stored in a bit distribution table unit in FIG.

FIG. 3 is a specific example of a bit distribution table formed by bit distribution in FIG. 1;

FIG. 4 is a flowchart illustrating a specific example of the operation of the embodiment illustrated in FIG. 1;

FIG. 5 is a block diagram showing a specific example of a system using the present invention.

FIG. 6 is a block diagram illustrating an example of a conventional audio compression device.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Speech compression apparatus 2 Subband analysis part 3 Scaling part 4 Scale factor selection information calculation part 5 Quantization part 6 Bit stream generation part 8 Bit distribution table part 9 Bit distribution table selection part 10 Bit distribution distribution part 11 Band judgment part

Continued on the front page F term (reference) 5D045 DA20 5J064 AA01 BC01 BC06 BD03 9A001 BB02 EE05 FF05 GG05 HH15 JJ71 KK43

Claims (10)

[Claims] 1. An audio compression apparatus which divides digital audio data into a plurality of sub-bands and quantizes each sub-band to generate a compression-encoded audio stream. And a scale factor that is a scaling factor for normalization is generated. The plurality of sub-bands are divided into a plurality of groups having different bands, and a predetermined number of allocated bits is assigned to each group. In such a way that the sum of the bit allocations at
Creating a bit allocation table for the sub-band, quantizing the audio data of the sub-band for each coding unit with the number of bits allocated based on the bit allocation table, and quantizing the audio data for each coding unit. The audio compression apparatus characterized in that the audio stream is generated by adding the scale factor and the bit distribution table to the audio stream. 2. A sub-band analysis unit that processes digital audio data for each frame of a fixed number of samples and divides the digital audio data into a plurality of sub-bands, and normalizes audio data of each sub-band for each frame. And a scaling unit that uses a scaling factor as a scale factor for a plurality of consecutive frames as a coding unit, and for each of the subbands, a scale factor in the coding unit is shared between frames, and a common scale factor is used. A scale factor selection information calculation unit for generating scale factor selection information representing a state; a bit allocation table unit in which a plurality of bit allocation tables in which the number of bits for each subband is allocated in a specific manner in advance; For each unit, the bit allocation table according to the scale factor selection information is stored in the bit allocation table. A bit allocation table selecting unit that selects from the table unit, is divided into groups of the sub-band analysis unit in divided the subbands multiple bands different for each unit the coding,
A group determining unit that determines a group to which the subband belongs; and, for each of the coding units, based on a determination result of the group determining unit, the distribution of the number of quantization bits among the groups is substantially averaged. A bit allocation allocating unit for allocating the number of bits allocated in the bit allocation table selected by the bit allocation table unit to the sub-band and creating a new bit allocation table; A quantizer for quantizing each audio data by the number of bits allocated by the new bit allocation table; and a new bit for the audio data of the subband of the coding unit quantized by the quantizer. A bit stream for generating an audio stream by adding an allocation table, the common scale factor, and the scale factor selection information. Audio compression apparatus characterized by comprising a generating unit. 3. The bit allocation table according to claim 2, wherein the plurality of bit allocation tables stored in the bit allocation table section have a bit allocation number allocated to all of the available subbands in the coding unit. An audio compression device characterized by the following. 4. The bit allocation table according to claim 2, wherein the plurality of bit allocation tables stored in the bit allocation table unit have a bit allocation number smaller than the number of subbands available in the coding unit. An audio compression device comprising: 5. The bit allocation table selection unit according to claim 2, wherein the bit distribution table selection unit calculates the number of the common scale factors based on the scale factor selection information generated by the scale factor selection information calculation unit. An audio compression device, wherein an optimal bit distribution table is selected according to the number. 6. The sub-band analysis unit according to claim 2, wherein the sub-band analysis unit divides the digital audio data into 32 sub-bands for each coding unit, An audio compression apparatus comprising: dividing a band into a plurality of groups; and determining to which group the subband being processed belongs. 7. The bit allocation allocating unit according to claim 2, wherein the bit allocation allocating unit allocates, for each of the groups, each of the sub-bands based on the scale factor generated by the scaling unit. An audio compression device characterized by forming. 8. The method according to claim 2, wherein the bit allocation allocation unit allocates, for each group, a larger number of bit allocations for the subbands having the larger scale factor generated by the scaling unit. Audio compression device. 9. The method according to claim 2, wherein the bit stream generation unit performs a CRC as needed.
An audio compression device characterized by adding an error detection word using (Cyclic Redundancy Check). 10. The audio compression device according to claim 2, wherein the bit stream generation unit generates the audio stream to which a header and an error detection word are added.