JP2003271199A - Audio signal encoding method and encoding device - Google Patents

Abstract

(57) Abstract: An audio signal encoding method capable of performing efficient encoding and reducing deterioration of detected sound quality even when the number of usable bits is small in audio encoding. It is an object to provide an encoding device. SOLUTION: In the encoding method of the audio signal,
A conversion procedure for converting a signal in the time domain into a signal in the frequency domain, a division procedure for dividing the frequency coefficient group converted in the conversion step into a plurality of bands, and a frequency expressed by a product of a gain and a quantization value. The above object is achieved by including a control procedure for controlling the gain or the quantization value in a coefficient value in a step-like manner and an encoding procedure for encoding the quantization value.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an audio signal encoding method and an encoding apparatus, and more particularly to an audio signal encoding method and an encoding apparatus for efficiently encoding an audio signal.

[0002]

2. Description of the Related Art Generally, AAC (Advanced Audio Codin
g) In transform coding of an audio signal typified by coding, a time domain signal is converted into a frequency domain DCT (Discrete
Cosine Transform) Convert to coefficient. Also, when encoding, the value of the DCT coefficient is expressed by the product of a value that controls the quantization accuracy called scale factor (hereinafter referred to as scale factor value) and the quantized value, and is encoded by a code word such as Huffman code. Is done. Here, the quantized value refers to a mantissa value when the DCT coefficient is represented in a floating point format, and the scale factor value represents a value corresponding to the exponent value. Further, the exponent part itself is called a gain. here,
If the above relationship is expressed by a simple mathematical expression, K = R × G ^S , where DCT coefficient is K, quantized value is R, scale factor value is S, and gain is G ^S.

Another audio encoding system is ISO / IEC JTC1 / which is an international standard organization.
ISO / IEC 1 standardized by SC29 / WG11
3818 (MPEG-2). Since this standard system only defines the interpretation of a coded bit stream (compressed data) and its decoding process, the process related to bit allocation in coding can be freely performed.

By the way, in encoding an audio signal, when the number of bits available for encoding is insufficient,
Bit allocation is performed in which quantization noise larger than that permitted perceptually is generated, so that quantization distortion in the entire frequency domain becomes large, and deterioration of coding may be detected.

Therefore, in the prior art, when the required number of quantization bits is insufficient, the scale factor value that determines the gain of each frequency coefficient group is changed by 1 to obtain an optimum value, and the code is coded based on that value. An encoding method for performing encoding has been proposed.

By the above method, the number of quantization bits required for encoding can be calculated and assigned.

[0007]

However, in the above-mentioned method of assigning quantized bits, the quantized value and the required number of bits are recalculated each time the scale factor value is changed by 1, so that the calculation amount is enormous. Become. Also,
This is one of the causes of sound quality deterioration because no consideration is given to efficient bit allocation when converting to codewords.

Up to now, the efficiency of the number of bits to be assigned to a large quantized value has not been taken into consideration by making the determination of the assignment based on the number of bits of the entire frame. Therefore, depending on the band, the sound quality may be deteriorated due to the excess or deficiency of the number of bits.

The present invention has been made in view of the above-mentioned problems, and it is possible to reduce the deterioration of the sound quality that is detected even when the number of bits that can be used is small in audio encoding by performing efficient encoding. An object of the present invention is to provide an audio signal encoding method and an audio signal encoding apparatus which can be performed.

[0010]

In order to solve the above problems, the present invention employs means for solving the problems having the following features.

According to a first aspect of the invention, in the audio signal encoding method, a plurality of conversion procedures for converting a time domain signal into a frequency domain signal and a plurality of frequency coefficient groups converted in the conversion step are provided. , A division procedure for dividing in the band, a control procedure for stepwise controlling the gain or the quantized value in the value of the frequency coefficient represented by the product of the gain and the quantized value, and encoding the quantized value. And an encoding procedure.

According to the first aspect of the invention, the number of bits can be reduced by controlling the gain or the quantized value stepwise. Also, the bits that have been reduced are
Since it can be assigned to a portion that is important for sound quality, it is possible to improve a relatively audible sound quality.

According to a second aspect of the present invention, the gain is set for each scale factor band, and encoding is performed so that the code word of the maximum frequency coefficient in the scale factor band has the maximum quantization value. It is characterized by performing.

According to the second aspect of the present invention, since the quantized value of the gain value is determined based on the code word of Huffman coding or the like, the length of the code word can be shortened and, as a result, the bit value can be reduced. The number can be reduced.

According to a third aspect of the present invention, the gain is set for each scale factor band, and the quantization value of the code word of the maximum frequency coefficient in the scale factor band is set to 1 or any integer. Is characterized by.

According to the third aspect of the present invention, it is possible to easily control the encoding by determining the sound quality for each scale factor band and setting the required quantization value.

According to a fourth aspect of the present invention, when MS stereo is applied to an audio signal, the energy of the M and S components in the scale factor band, or the magnitude of the maximum frequency coefficient, or the auditory entropy is used. , And different quantized values are used as the maximum frequency coefficients of the M component and the S component in the scale factor band.

According to the invention described in claim 4, the number of bits can be separately allocated to each of the M component and the S component by using the MS stereo, and the encoding with the reduced number of bits can be performed.

According to a fifth aspect of the present invention, in the audio signal encoding device, a plurality of converting means for converting a time domain signal into a frequency domain signal and a plurality of frequency coefficient groups converted in the converting step are provided. Division means for dividing in the band of, the control means for controlling the gain or the quantized value in the value of the frequency coefficient represented by the product of the gain and the quantized value in steps, and encoding the quantized value. And an encoding unit.

According to the fifth aspect of the invention, the number of bits can be reduced by controlling the gain or the quantized value stepwise. Also, the bits that have been reduced are
Since it can be assigned to a portion that is important for sound quality, it is possible to improve a relatively audible sound quality.

According to a sixth aspect of the present invention, the control means sets the gain for each scale factor band, and the encoding means has a maximum code word of a frequency coefficient within the scale factor band. The encoding is performed so that the quantized value of

According to the invention of claim 6, since the quantized value of the gain value is determined based on the code word of Huffman coding or the like, the length of the code word can be shortened, and as a result, the bit value can be reduced. The number can be reduced.

In a seventh aspect of the present invention, the control means sets the gain for each scale factor band, and the encoding means quantizes a code word having a maximum frequency coefficient in the scale factor band. The characteristic value is 1 or an arbitrary integer.

According to the seventh aspect of the invention, it is possible to easily control the coding by determining the sound quality for each scale factor band and setting the required quantization value.

In the invention described in claim 8, when MS stereo is applied to an audio signal, the energy of the M component and the S component in the scale factor band, or the magnitude of the maximum frequency coefficient, or the auditory entropy is used. , And different quantized values are used as the maximum frequency coefficients of the M component and the S component in the scale factor band.

According to the invention described in claim 8, the number of bits can be separately allocated to each of the M component and the S component by using the MS stereo, and the encoding with the reduced number of bits can be performed.

According to a ninth aspect of the present invention, one of the plurality of encoding means is selected based on a plurality of encoding means having different encoding methods and the required number of bit rates for each scale factor band. It is characterized by having an evaluation / selection means for evaluating / selecting the encoding means.

According to the ninth aspect of the invention, by using the encoding means capable of reducing the number of bits to a minimum,
Encoding can be performed efficiently.

[0029]

BEST MODE FOR CARRYING OUT THE INVENTION The present invention uses a DCT transformation to
Controls the quantization accuracy of the frequency coefficients output from the 9 scale factor bands to reduce the number of bits, efficiently encodes using Huffman code, and detects even when the number of bits that can be used in audio encoding is small. The main purpose is to reduce the deterioration of the sound quality.

When converting the quantized value of each band frequency group into a code word represented by a Huffman code, a table of code words different depending on the maximum quantized value in the frequency band (band) or Utilizes the feature of using mathematical formulas. For example, AAC has 11 Huffman codeword tables, but the maximum quantized value of each table is a discrete value (for example, 16, 12, 7,
4, 2, 1). By assigning the number of bits in consideration of this maximum value, efficient encoding is performed.

Here, an example of the 11 kinds of Huffman code word tables described above is shown in FIG. The table shown in FIG. 1 has 0 to 80 (81 types) of "index",
It is composed of "length" and "codeword (hexadecimal number)". By adjusting the number of bits for each step using “length” and “codeword” corresponding to “index”, it is possible to reduce the number of bits and generate an audio signal with good transmission efficiency.

For example, when considering the case where the number of encoded bits is insufficient, it is assumed that the maximum quantized value at a certain scale factor is, for example, "12". At this time, the maximum quantization value is set to "7" instead of changing the scale factor value by one, so that the frequency error to be encoded can be reduced although the quantization error increases. Further, when the maximum quantization coefficient becomes small, it is expected that the length of the code word assigned to each Huffman code also becomes short, and as a result, the effect of reducing the number of bits can be expected. By such a method, it is possible to allocate the reduced bits to the important parts for the sound quality, so that the relatively perceptual sound quality can be improved.

Further, as compared with the method of changing the scale factor value by one, the amount of calculation can be reduced since the scale factor value is often changed greatly. With the contents described above, the number of bits can be reduced without deteriorating the sound quality.

Next, embodiments of the present invention will be described with reference to the drawings. In the example, 20
An example in which 48 samples are converted into 1024 DCT coefficients will be described.

FIG. 2 is an example of a block diagram showing the configuration of the encoding apparatus according to the present invention.

The encoding apparatus shown in FIG. 2 includes an auditory model 11,
Filter bank 12, scale factor 13,
It is configured to have a quantizer 14, noiseless coding 15, a rate / distortion controller 16, a step control controller 17, and a bitstream multiplexer 18.

In FIG. 1, the auditory model 11 calculates a masking pattern of quantization noise of an input voice signal. That is, the auditory masking threshold of the audio signal is calculated.

More specifically, the input audio signal is FFT (Fast Fourier Transform) so as to coincide with the DCT analysis position in the filter bank described later.
The maximum amount of noise (threshold) that can be masked by the voice signal is calculated by using, and the Signal to Mask ratio and threshold value for each scale factor band are output. It also outputs the result of whether to select the long, start, stop, or short block type. For details of the above-mentioned auditory model, see IS
O / IEC 13817-7 ANNEX B2 Chapter ENC
ORDER 2.1 Psychoacoustic Mo
It is described in del.

The filter bank 12 transforms the input voice signal in the time domain into a signal in the frequency domain using FFT transformation or DCT transformation.

The conversion method will be described more specifically. In the basic processing of encoding in AAC, an encoder converts a time domain signal into a frequency domain signal. On the contrary, in the decoder, the signal in the frequency domain is converted into the signal in the time domain. MD this
CT (Modified Discrete Cosine Transform) and IM
DCT (Inverse Modified Discrete Cosine Transfor
m) running by. In addition, MDCT, IMDC
For T, TDAC (time domain aliasing distortion removal technique) is used to reduce block distortion.

For details, refer to ISO / IEC 1381.
7-7 ANNEX B2 Chapter ENCORDER 2.3 F
ilterbank and block switch
It is described in ing.

The scale factor 13 creates a scale factor value expressing a gain for converting a frequency coefficient.

Here, an example of the formula showing the scale factor value is shown in the formula (1).

[0044]

[Equation 1] Note that scf [sb] is the scale factor value of the sbth scale factor band, and this value is output to the quantizer 14. QUANT_STEP is a quantization step value input from the step controller 17. Also, common_scf is a correction term, and for example, 100 is set. P in equation (1)
ow_spectrum [sb] is the filter bank 1
Using the value mdct [sb] input from 2, the formula (2)
Calculate as.

[0045]

[Equation 2] Note that mdct [sb] in the equation (2) is an arbitrary mdct coefficient in the sbth scale factor band, and for example, the maximum mdct value is set.

The quantizer 14 converts the frequency coefficient into a quantized value. The details of the conversion means are ISO / IEC.
13817-7 ANNEX B2 Chapter ENCORDER
2.7 Quantization.
Specifically, it is expressed by a mathematical expression as shown in Expressions (3) and (4).

[0047]

[Equation 3] Here, x_quant [i] is the quantized value of the frequency coefficient having the i-th index (“index” in FIG. 1) and is output to the noiseless coding 15. Also, mdct_line [i] is a filter bank D
The coefficient is the CT-transformed i-th index and is input from the scale factor 13. Also, M
0.405 as a general value for AGIC_NUMBER
4 (fixed value) is set.

The noiseless coding 15 converts scale factor values and quantized values into codewords such as Huffman coding. For details, see ISO / IEC 13817-7A.
NNEX B2 Chapter ENCORDER 2.8 Noise
It is described in less Coding. In addition, a procedure of an example of a mathematical expression for converting four coefficients into a Huffman code word will be described with reference to a flowchart before reference. The Huffman code table used here is as shown in FIG.
The Huffman code table is not limited to this within the scope of the invention.

FIG. 3 is a diagram showing an example of the program flow of the process of converting to a Huffman code word.

In FIG. 3, iterative processing is performed for each scale factor band (S1). First, the initial value is set (S2), and from offset (sb) to top (sb)
The following processes from S3 to S8 are performed until
3). Further, since the process is performed for each of the four coefficients, the increment in S3 is +4. Further, offset (sb) represents the index (i) of the lower limit DCT coefficient of each scale factor band, and top (sb) represents the index of the upper limit DCT coefficient of the sb band. The DCT of sb, offset (sb) and top (sb) described above
An example of band division of coefficients is shown in FIG. As shown in FIG. 4, four DCT coefficients are assigned to one sb (a total of 1024 0 to 1023).

Next, the index for referring to FIG. 1 is calculated (S4). Here, x_quant [i] is the quantized value of the i-th coefficient, and is input from the quantizer 14. For the index value calculated in S4, the codeword (Huffman code) and tmp are extracted with reference to FIG. 1 (S5, S6). For example, in S4, index
Is "10", the codeword is set to "72" and the tmp is set to "7". In addition, the number of bits required to encode the quantized value is calculated (S
7). The output of S7 is output to the rate / distortion controller 16. By repeating this for all bands divided for each scale factor band (for each step), all the coefficients can be converted into Huffman code words (S8, S9).

Next, the rate / distortion controller 16 and
The operation of the step controller 17 will be described with reference to a flowchart of mathematical expressions.

FIG. 5 is a flow chart showing an example of the operation of the rate / distortion controller and the step controller.

In FIG. 5, the rate / distortion controller 16 operates based on the value set in the step controller 17 (S11). The value set in the step controller is preferably the maximum value of the quantized value. In the present invention, the set value is set to {16,12,7,4,2,1}, but the set value is not limited to this.

First, the coefficient (mdct_lin) having the i-th index that has been DCT-transformed by the filter bank
e (i) is compared with the size of the allowable quantization noise (allwed_dist (sb)) input from the auditory model (S13), and if the quantization noise is larger, 0 is set in mdct_line [i]. Set (S1
4). This is repeated until the index becomes maximum (S12, S15).

Next, rate / distortion control is performed. First,
The subscript j in the STEP_MAT table in S11 is set to 0 (initialization) (S16). Next, the number of bits that can be used in the scale factor band (average_bit)
s) is the number of bits (bit_count ())
Is determined from S17 to S23.
This is confirmed by repeating the above process (S17).
It should be noted that average_bits is set in advance, and bit_count () is the sum of the number of bits calculated by the noiseless coding 15, and the number of bits required for transmission such as the scale factor or the codebook number of the Huffman table is also included. Including.

First, the maximum quantized value is set in QUANT_STEP (S18). It should be noted that the setting value is set from a larger value in S11. Next, processing is performed for each scale factor band (S19). First, the scale factor value (calc_scale ()) is calculated (S
20). The calculated scale factor value is output to the scale factor 13. Further, in S21, the quantization value (calc_quant ()) is calculated, and the quantizer 1
4 is output.

This is performed for each scale factor band (S22), and the bit count (bit_count) is calculated.
until t ()) no longer exceeds average_bits,
Repeatedly (S23).

If the average_bits is exceeded, the required number of bits cannot be obtained, and there arises a problem that encoding cannot be performed or sound quality is distorted when encoded. In that case, the value of the quantized value in the step controller is set low and the process is performed again.

That is, in the processing from S17 to S23, S17
If the condition is not satisfied, the next value (for example, 16 is set and 12 is set if the condition is not satisfied) is set, and S is again set.
The processing from 17 to S23 is performed. In this way, the QUANT_STEP having the first condition under the processing is the optimum value, and by encoding using this value, the highest sound quality can be obtained with the required number of bits.

The gain or the optimum value of the quantized value can be obtained by the flowchart shown in FIG.

Next, the bit stream multiplexer 1
At 8, the input coded data and the control information such as the scale factor or the codebook number of the Huffman table are converted into a bitstream. As a result, it is possible to efficiently encode the audio signal. For details of the bitstream multiplexer 18,
It is described in ISO / IEC13817-7 Chapter 1, syntax. That is, the parameters output from each block of the encoding device are rearranged into the format detailed in syntax and output.

The operation content of each block of the encoding device is not limited to this, and the calculation may be performed by different mathematical expressions for the same block.

For example, the scale factor value may be derived based on the maximum frequency coefficient value in one or a plurality of scale factor band groups by changing the formula (1) into the following formula (5).

[0065]

[Equation 4] Note that max_pow [sbm, sbn] can be calculated by equation (6).

[0066]

[Equation 5] Also, max_mdct [sbm, sbn] is from m to n
The largest MDCT coefficient in the second scale factor band (m, n: integer).

Further, by changing the expression (1) to the following expression (7), the scale factor value of an arbitrary scale factor band is used as the maximum frequency coefficient value in each scale factor band, and the numerator is further set to 1 or It can also be derived as an arbitrary integer value.

[0068]

[Equation 6] Note that the value of k in equation (7) is 1 or n <QUANT_
It is a positive integer that satisfies STEP. Also, max_po
w_spectrum can be calculated by equation (8).

[0069]

[Equation 7] Thereby, it is possible to control the sound quality in encoding the audio signal.

Next, an operation example of the encoding device in the case where the encoding device according to the present invention is equipped with the MS controller will be described with reference to a block diagram.

FIG. 6 is an example of a block diagram showing a configuration of an encoding device including an MS controller according to the present invention.

The encoding apparatus shown in FIG. 6 includes an auditory model 11,
Scale factor 13, quantizer 14, noiseless coding 15, rate / distortion controller 16
, Step controller 17, bitstream multiplexer 18, MS controller 19, M
/ S stereo tool 20.

Here, the operation of each block will be described with respect to parts different from the description of each block mainly with reference to FIG.

In FIG. 6, the auditory model 11 calculates a masking pattern of quantization noise. The MS controller 19 calculates M / S entropy. That is, from the time-frequency converted coefficients, the L component, R component, M
The energy of the component (L + R component) and the S component (LR component) are calculated, and based on the calculation result of the masking pattern input from the auditory model 11 and output information such as the number of bits required for encoding, the MS mode, Auxiliary information is created for determination of LR mode switching or bit allocation between channels in MS mode.

Here, the determination contents of the MS mode and the LR mode by the energy calculation in the MS controller 19 will be described. First, in the following equations (9) to (12), the energy (eM (sb), eS (s) of the M component, S component, L component, and R component of each scale factor band coefficient is calculated.
b), eL (sb), eR (sb)) are shown.

[0076]

[Equation 8] Also, eM (sb), eS (sb), eL (sb), and eR
The energy ratio of each of the (sb) components is calculated by using equations (13) and (1
Equation (15) shows the determination formula for switching between the MS mode and the LR mode calculated by using 4).

[0077]

[Equation 9] K in Expression (15) is a positive constant, for example, k =
Set to 1.

The method of switching between the MS mode and the LS mode is not limited to this, and is described in Japanese Patent Application No. 2001-70926 “Stereo signal encoding method and encoding apparatus” filed by the present applicant. As you can see, we use auditory entropy to calculate M / S entropy,
Depending on the result, it is possible to switch between the LR mode and the MS mode.

Next, the scale factor values (scf_M [sb], scf_S of the M component and the S component in the scale factor 13 when the MS mode is applied).
The formulas for calculating [sb]) are shown in formulas (16) and (17).

[0080]

[Equation 10] QUANT_M and QUANT_S are M components,
It is the quantization step value of each S component and is input from the rate / distortion controller 16.

Next, the rate / distortion controller 16
The program operation of the step controller 17 will be described with reference to a flowchart.

FIG. 7 is a flow chart showing an example of the operation of the rate / distortion controller and the step control controller having the MS controller. In addition, M
The operations other than the S mode are the same as those in FIG.

In FIG. 7, the rate / distortion controller 16 operates based on the value set in the step controller 17 (S31). First, the coefficients (dM [i], dS [i]) of the M component and the S component having the i-th index that are DCT-transformed by the filter bank 12 respectively.
And the magnitudes of allowable M and S component quantization noises input from the auditory model 11 (MS control) (allwed_dist_M (sb), allwed_
dist_S (sb)) is compared with the M component and the S component, and when the allowable quantization noise is larger, 0 is set to dM [i] and dS [i] (S33 to S).
37). This is repeated until the index becomes maximum (S32, S37).

Next, rate / distortion control is performed. First,
The subscript j of the STEP_MAT table in S11 is set to 0 (initialization) (S38). Next, the number of bits that can be used in the scale factor band (average_bit)
s) is the number of bits (bit_count ())
It is judged from S39 to S45 whether processing can be performed without exceeding
This is confirmed by repeating the above process (S39).

It should be noted that the average_bits is set in advance, and the bit_count () is the sum of the number of bits calculated by the noiseless coding 15, which is necessary for transmission such as the scale factor or the codebook number of the Huffman table. It also includes the number of bits (control information).

First, QUANT_STEP_M and QUA
The maximum value of the quantized value set in S31 is set in NT_STEP_S (S40).

It should be noted that a large value of the set value of S31 is set to QUA.
Set to NT_STEP_M. At the same time, the next largest value is set in QUANT_STEP_S. Next, processing is performed for each scale factor band (S4
1). First, the scale factor value (calc_sca
le ()) is calculated (S42). The calculated scale factor value is output to the scale factor 13.
In S43, the quantized value (calc_quant ())
Is calculated and output to the quantizer 14.

This is performed for each scale factor band (S44), and the bit count (bit_count) is calculated.
It is repeated until t () does not exceed average_bits (S45).

If the average_bits is exceeded, the required number of bits cannot be obtained, and there arises a problem that encoding cannot be performed or the encoded sound quality is distorted when encoded. So in that case,
The value of the quantized value in the step controller is set low and the process is performed again.

That is, if the conditions are not satisfied in the processes of S39 to S45, the next value (for example, 16 is set and 12 is not satisfied) is set, and the processes of S39 to S45 are repeated. In this way, the QUANT_STEP with the first condition is the optimum value, and by encoding using this value, the highest sound quality can be obtained with the required number of bits.

The gain of the M component and the S component or the optimum value of the quantized value can be obtained by the flowchart shown in FIG.

The flow chart shown in FIG. 7 shows the case where the MS mode is applied to all scale factor bands, but the present invention is not limited to this.
For example, by determining the MS mode and the LS mode for each scale factor band, the mode can be switched for each scale factor band, and the optimum values of the gain and the quantized value can be efficiently obtained.

The MS stereo tool 20 creates a sum signal (L + R) and a difference signal (LR) of frequency band signals, and switches between the MS mode and the LR mode by a control signal from the MS controller 19. The coefficient of the band in the MS mode is mdct_li of each LR component.
Let ne [i] be dL (i), dR (i), dM (i), dS
(i) can be calculated by equations (18) and (19), respectively.

[0094]

[Equation 11] Note that the other blocks are the same as the block description in FIG. 2, and thus detailed description will be omitted.

As described above, when MS stereo is applied to an arbitrary scale factor band, it has the same index number based on the magnitude of the energy or maximum frequency coefficient of the M component and S component in the scale factor band, or the auditory entropy. Since the quantized values of the maximum frequency coefficients of the M component and S component of the scale factor band can be made different and the number of bits can be adjusted by the component, efficient audio signal encoding is possible. Becomes

In addition, MS for each scale factor band
By determining and switching between the mode and the LS mode, more efficient audio signal encoding can be performed.

The operation contents of each block in FIG. 2 and FIG. 6 are not limited to those described above, and the same block may be calculated by different mathematical expressions.

For example, in the scale factor 13, the calculation formula for obtaining the scale factor value, the formula (5) and the formula (7) may be divided and processed under predetermined conditions.
Here, as an example, the mathematical expression under the condition 1 is expressed by the formula (2
0) and the mathematical expression under the condition 2 are shown in Expression (21).

[0099]

[Equation 12]

[0100]

[Equation 13] Note that scf_I [sb] represents the sb-th scale factor value under condition 1, and scf_II [sb] is
The sb-th scale factor value in the case of condition 2 is shown.

The condition for switching is that the quantizer 1
In the calculation formulas (Formulas (3) and (4)) based on Formula 4, when x_quant [i] is first calculated as the condition 1 and the quantized values in all scale factor bands become 0, The control is performed so that the condition 2 is changed.

Next, the operation of the rate / distortion controller 16 and the step controller 17 of the coding apparatus including the conditions 1 and 2 will be described with reference to a flow chart of mathematical expressions.

8 and 9 are flow charts showing an example of the operations of the rate / distortion controller and the step controller having the switching condition. Here, in FIG. 8, the processing from S51 to S56 is the same as the processing from S11 to S16 shown in FIG. 5, so description thereof will be omitted here.

In rate / distortion control, the number of bits available in the scale factor band (averag)
e_bits) is a bit number count (bit_co
Condition 1 to judge whether processing can be performed without exceeding unt ())
This is confirmed by repeating the processing from S57 to S72 including the processing for switching to Condition 2 and (S57).

It should be noted that the average_bits is set in advance, and the bit_count () is the sum of the number of bits calculated by the noiseless coding 15, and is necessary for transmission such as the scale factor or the codebook number of the Huffman table. Including the number of bits.

First, the maximum quantized value is set in QUANT_STEP (S58). Next, processing is performed on all the divided scale factor bands (S5).
9). First, 1 is set to flag [sb] as an initial value (S60). Next, the scale factor value (scf [sb]) of the sb-th scale factor band is obtained by the mathematical expression (Equation (20)) of Condition 1 (S61), and the quantized value (calc_quant ()) is calculated. (S6
2).

Next, 0 is set in flag [sb] (S
63), the quantized value calculated up to S62 is offse
Whether t (sb) to top (sb) are all 0 is confirmed (S64 to S67). First, it is judged whether or not the quantized value is 0 (S65), and if there is a quantized value other than 0, fla is determined.
1 is set to g [sb] (S66).

Next, it is judged whether or not flag [sb] is 0 (S68), and if it is 0, the scale factor value of the sbth scale factor band is calculated using the equation (21) (S69). . If "NO" in S68 or after the process of S69 is completed, the quantized value is calculated (S7).
0). Do this for each scale factor band (S
71) and the bit count (bit_count ()) is a
It is repeated until the number of average_bits is exceeded (S72).

If the average_bits is exceeded, the required number of bits cannot be obtained, and there arises a problem that the encoding cannot be performed or the sound quality is distorted when encoded. In that case, the value of the quantized value in the step controller is set low and the process is performed again.

The gain or the optimum value of the quantized value can be obtained by the flow charts shown in FIGS. 8 and 9.

Further, as another block configuration example of the encoding device, the quantizer 14 which respectively performs the calculation by the formulas of the above formulas (1), (6), (20) and (21).
Each module having (scale factor 13, quantizer 14, noiseless coding 15, rate / distortion controller 16 and step controller 1
7) is provided in the encoding device, and the one having a small bit count is used for the encoding, whereby the encoding can be performed efficiently.

An example of each block configuration of the coding apparatus having the above-mentioned module will be described with reference to FIG.

The coding apparatus shown in FIG.
, MS controller 19 and filter bank 12
, M / S stereo tool 20, and module A21
, Module B22, module C23, evaluation
It is configured to have a selection unit 24 and a bitstream multiplexer 18. In addition, the module A21
FIG. 11 shows a configuration example of each block of the module B22 and the module C23.

Here, for example, in the module A21,
Switching according to the conditions 1 and 2 described above (equation (20),
The module B22 has the operation described with reference to FIGS. 8 and 9 for calculating the quantized value by performing the expression (21), and the module B22 uses the quantized value according to the expressions (7) and (8). In addition, the module C23 has a process of calculating a quantized value according to the formulas (5) and (6).

The evaluation / selection unit 24 uses the module A21,
Based on the outputs of module B22 and module C,
It has a function of evaluating and selecting which module control signal and data signal are output to the bitstream multiplexer 18. The evaluation is performed by comparing the number of used bits output from each of the modules A21, B22, and C23 with the data output from each module, and using the module having the smallest number of used bits and the minimum data amount To convert.

With the block configuration shown in FIG. 10, efficient encoding can be selected and encoded for a wide variety of audio signals.

As described above, according to the present invention, for example, when a coding bit is insufficient, a quantized value is selected based on the characteristics of a code word such as Huffman coding, and the coding is performed. The word length can be reduced and at the same time the number of bits can be reduced. According to the present invention, the reduced bits can be assigned to a portion that is important for sound quality, so that the relatively perceptual sound quality can be improved.

Further, since the scale factor value is often changed largely based on the maximum value of the quantized value, the calculation amount can be reduced. In addition,
The standard of such transform coding is syntax of the decoder.
Therefore, the bitstream encoded by the present invention can be decoded by the existing decoder.

Note that the flow of operation contents used in the present invention is an example, and is not limited to the above description within the scope of the invention.

[0120]

As described above, according to the present invention, it is possible to efficiently encode an audio signal and reduce the number of usable bits. In addition, even when the number of bits that can be used in audio encoding is small, it is possible to reduce deterioration in detected sound quality.

[Brief description of drawings]

FIG. 1 is a diagram showing an example of a Huffman codeword table.

FIG. 2 is an example of a block diagram showing a configuration of an encoding device according to the present invention.

FIG. 3 is an example of programmatically showing a flow of processing for converting into a Huffman code word.

FIG. 4 is a diagram showing an example of band division of DCT coefficients.

FIG. 5 is a flowchart showing an example of operations of a rate / distortion controller and a step controller.

FIG. 6 is an example of a block diagram showing a configuration of an encoding device including an MS controller according to the present invention.

FIG. 7 is a flowchart showing an example of operations of a rate / distortion controller and a step control controller having an MS controller.

FIG. 8 is a flowchart (1) showing an example of operations of a rate / distortion controller having a switching condition and a step controller.

FIG. 9 is a flowchart (2) showing an example of the operations of the rate / distortion controller having a switching condition and the step controller.

FIG. 10 is a diagram showing an example of each block configuration of an encoding device having a module.

FIG. 11 is a diagram showing a block configuration example of each module in the present invention.

[Explanation of symbols]

11 Hearing model 12 filter banks 13 Scale factor 14 Quantizer 15 Noiseless coding 16 rate / distortion controller 17 step controller 18 bitstream multiplexer 19 MS controller 20 M / S Stereo Tool 21 Module A 22 Module B 23 Module C 24 Evaluation / Selection Section

   ─────────────────────────────────────────────────── ─── Continued front page    F-term (reference) 5D045 DA20                 5J064 AA02 BA09 BA16 BB05 BC01                       BC08 BC09 BC11 BC16 BC23                       BD01

Claims (9)

[Claims] 1. An audio signal encoding method, comprising: a conversion procedure for converting a time domain signal into a frequency domain signal; and a division procedure for dividing the frequency coefficient group converted in the conversion step into a plurality of bands. A control procedure for controlling the gain or the quantized value at a value of a frequency coefficient represented by a product of a gain and a quantized value in a stepwise manner, and an encoding procedure for encoding the quantized value. A characteristic audio signal encoding method. 2. The gain is set for each scale factor band, and encoding is performed so that the code word of the maximum frequency coefficient in the scale factor band has the maximum quantized value. 1. The audio signal encoding method according to 1. 3. The gain is set for each scale factor band, and the quantization value of the code word of the maximum frequency coefficient in the scale factor band is set to 1 or an arbitrary integer. 2. The audio signal encoding method as described in 2. 4. When MS stereo is applied to an audio signal, the energy of M component and S component in the scale factor band, or the magnitude of the maximum frequency coefficient, or the auditory entropy is used to determine the M component in the scale factor band. 4. The audio signal encoding method according to claim 2, wherein different quantization values are used as the maximum frequency coefficients of the S component and the S component, respectively. 5. An audio signal coding apparatus, comprising: a transforming unit that transforms a time domain signal into a frequency domain signal; and a splitting unit that splits the frequency coefficient group transformed in the transforming step into a plurality of bands. A control unit that controls the gain or the quantized value at a value of a frequency coefficient represented by a product of a gain and a quantized value in a stepwise manner, and an encoding unit that encodes the quantized value. A characteristic audio signal encoding device. 6. The control means sets the gain for each scale factor band, and the encoding means sets the code word of the maximum frequency coefficient within the scale factor band to be the maximum quantized value. The audio signal encoding apparatus according to claim 5, wherein the audio signal encoding apparatus performs encoding. 7. The control means sets the gain for each scale factor band, and the encoding means sets the quantized value of the code word of the maximum frequency coefficient in the scale factor band to 1 or an arbitrary integer. 7. The audio signal encoding device according to claim 5, wherein: 8. When MS stereo is applied to an audio signal, the energy of M component and S component in the scale factor band, or the magnitude of the maximum frequency coefficient, or the auditory entropy is used to determine the M component in the scale factor band. 9. The audio signal coding apparatus according to claim 6, wherein different quantization values are used as the maximum frequency coefficients of the S component and the S component, respectively. 9. One of the plurality of encoding means is evaluated / selected based on a plurality of encoding means having different encoding methods and a required number of bit rates for each scale factor band. 9. The audio signal encoding apparatus according to claim 5, further comprising an evaluation / selection unit.