JP2000338998A - Audio signal encoding method and decoding method, device therefor, and program recording medium - Google Patents

Abstract

PROBLEM TO BE SOLVED: To encode an input voice containing a tone component much in high area with high efficiency by dividing the frequency area coefficient of an input signal sample to generate a coefficient small piece group, classifying the coefficient small pieces to a plurality of groups followed by flattening, and encoding them. SOLUTION: An input signal is converted to the coefficient of frequency area, and a series of coefficients are bundled every band of about 100 Hz width to form coefficient small pieces, and the respective coefficient small pieces are classified to a high level group and low level group according to the intensities. For example, when the magnitude of the coefficient of frequency area is fluctuated as in a modified discrete conversion (MDCT) coefficient, the adjacent coefficients are bundled to form coefficient small pieces, and the coefficient small pieces are classified to a small coefficient group G0 and a large coefficient group G1. After the group G1 with high intensity and the group G0 with low intensity are independently flattened, the two groups G0 and G1 are collected to one followed by vector quantization.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention provides an encoding method for converting an audio signal into a digital code with high efficiency, and a decoding method for restoring the digital code into an audio signal. And
It can be used for transmission, broadcasting, etc. using audio signal communication channels.

[0002]

2. Description of the Related Art As a conventional technique for encoding an audio signal with high efficiency, for example, there is a transform encoding method shown in FIG. In the transform coding, an audio signal input as a discrete signal sample sequence is subjected to time / frequency conversion for each input of a fixed number of samples by a time / frequency conversion unit 11, and is converted into a frequency domain coefficient before coding. The processing unit 2 performs pre-processing on the coefficients in the frequency domain, and then the quantization unit 3 performs quantization. Transform-domain Weighted Interleave Vector Quantization
The ave Vector Quantization (TWINVQ) method applies to this example.

The TWINVQ system uses weighted interleave vector quantization as a quantization method at the last stage in the quantization unit 3. In addition, since the vector quantization can be performed more efficiently as the distribution of the input coefficients is smaller, the coefficient is flattened in two stages in the preprocessing by the preprocessing unit 2 in FIG. I have. In the first-stage flattening, the overall rough coefficient variation is flattened by normalizing the coefficients in the frequency domain with the spectrum obtained by the linear prediction method. In the second stage of flattening, the frequency domain coefficient is normalized for each small band divided so as to have an equal width on the Bark scale, which is one of the frequency scales.
Flattening is performed more finely than a linear prediction spectrum.

[0004] The Bark scale has a characteristic that if the band is divided into equal widths, the width becomes acoustically substantially equal. A small band having a uniform width on the Bark scale is approximately equal in width to the auditory sense. However, on the linear scale, as shown in FIG. It is wider. Therefore, when the coefficient group in the frequency domain is divided so as to have the same width on the Bark scale, a smaller band located at a higher frequency includes more coefficients.

[0005] The reason why the second level of flattening is performed in Bark scale units is to effectively distribute a limited amount of information in consideration of hearing. The operation of flattening by normalization for each small band on the Bark scale is performed in the expectation that the coefficients in the small band are stationary, but the small band located at a high frequency contains many coefficients. Therefore, there are many cases where the inside of the small band is not stationary as shown in FIG. 2, for example. This leads to a reduction in the efficiency of vector quantization and, as a result, to a deterioration in sound quality of the decoded audio signal. Such a problem is particularly likely to occur when an audio signal containing many tone components in a high frequency range is input.

[0006] Regarding the TWINVQ system, see Iwagami et al., “Frequency Domain Weighted Interleave Vector Quantization (TwinVQ)”.
Tone Coding by IEICE, IEICE Transactions Vol.J80-A,
See pp.830-837 (1997) for details. In the coding of the form shown in FIG. 1, there is also a coding method using scalar quantization of adaptive bit allocation as a quantization method. In such an encoding method, the coefficients in the frequency domain are divided into small bands, and optimal bit allocation is performed for each small band. As a method of dividing the small band, there is a case where the band is divided into equal widths on the Bark scale for the purpose of auditory optimization, but in this case, as in the case of TWINVQ, when the stationary band is not stationary in a small band located at a high frequency. There is a problem that the quantization efficiency is reduced.

As an encoding method for solving such a problem, when an input signal is converted into a frequency domain signal and flattened (normalized) for each small band, the normalized bandwidth is changed according to the shape of the spectrum distribution. Japanese Patent Application Publication No. 7-336232 discloses an adaptive change. In this method, the bandwidth of normalization is reduced in a small band including a tone component,
In other areas, the bandwidth for normalization is widened to reduce the number of flattened sub-bands as a whole, thereby improving coding efficiency. However, according to this method, when tone components are sparse, a fine normalized width is also applied to a flat portion near the tone components, and the encoding efficiency may decrease. Also, since the normalization information must be coded and attached for each tone component, if a large number of dispersed tone components exist, the code amount of the normalization information increases accordingly.

[0008] Japanese Patent Application Publication No. 7-168593 discloses that a tone component and other components are separated and coded to improve coding efficiency. In this method, for each spectrum of the maximum value, the spectrum and its neighboring spectrum are normalized and encoded as a group of tone component signals, so that information on the position and size of each maximum value spectrum is encoded and attached. There is a need to. For this reason, when there are a large number of tone components, a large amount of position and size information is required, which may hinder improvement in coding efficiency.

Japanese Patent Application Publication No. 7-248145 discloses that pitch components in which tone components are arranged at equal intervals are separated and encoded. The position information of the pitch component is given by the fundamental frequency of the pitch and requires only a small amount of information.However, there is a problem that the tone component cannot be separated well from a sound having a non-integer overtone structure such as a metal sound. .

[0010]

SUMMARY OF THE INVENTION An object of the present invention is to provide an encoding method and a coding method capable of encoding an input sound containing a large number of tone components in a high frequency band with good efficiency. An object of the present invention is to provide a decoding method, an apparatus using the method, and a recording medium on which the method is recorded as a computer-executable program.

[0011]

According to the present invention, an audio signal encoding method for encoding an input audio signal sample includes the following steps: (a) input audio signal samples are input to a fixed number of input audio signal samples; Time / frequency conversion is performed for each sample to obtain frequency domain coefficients,
(b) The frequency domain coefficient is divided into one or more bundled coefficient pieces to generate a coefficient piece group, (c) calculating the strength of each coefficient piece of the coefficient piece group, (d) By classifying the coefficient piece group into at least one of at least two groups based on their intensities, at least two systems of coefficient piece rows are generated, the classification information is encoded and output as a classification information code, e) Encode at least the two series of coefficient pieces and output them as coefficient codes.

According to the present invention, a decoding method for inputting a digital code and outputting an audio signal sample includes:
The method includes the following steps: (a) decoding the input digital code to obtain a group of coefficient pieces of a plurality of systems; and (b) decoding the input digital code to obtain classification information of the coefficient pieces. In step (b), a coefficient domain group composed of a plurality of systems is combined and restored to a frequency-domain coefficient composed of a series of coefficient series.
The frequency-to-time conversion is performed on the frequency domain coefficient obtained in step (1) and the result is output as an audio signal.

Alternatively, the decoding method includes the following steps: (a) decoding the input digital code to obtain a group of coefficient pieces each having a plurality of coefficients in the frequency domain; Decode the code, obtain the classification information of the coefficient pieces, classify the coefficient piece group into a series of coefficient pieces based on the classification information, and (c) decode the input digital code,
While obtaining the normalized information of the coefficient piece group, each of the coefficient piece rows classified into a plurality of systems in the step (b) is denormalized by the normalized information, and (d) obtained in the step (b). Based on the classification information of the coefficient pieces, the coefficient piece rows of a plurality of systems denormalized in the above step (c) are sequentially arranged in an array of one system before classification, and restored to frequency domain coefficients, and (e) The frequency domain coefficient obtained in step (d) is frequency / time converted, and the result is output as an audio signal.

According to the present invention, an encoding device that inputs an audio signal sample and outputs a digital code
A time / frequency conversion unit for performing time / frequency conversion on the input audio signal for each of a fixed number of inputs to obtain a frequency domain coefficient;
A coefficient-segment generator for dividing the frequency-domain coefficients from the time / frequency converter into small-slice groups obtained by bundling adjacent coefficients; and an intensity calculator for calculating the strength of each of the small-segment groups from the coefficient-segment generator. The coefficient pieces are divided into at least two groups based on the relative magnitudes of the intensities of the piece groups calculated by the strength calculation section, and the small piece groups generated by the coefficient piece generation section based on the information on the grouping. Are classified into at least two systems, the coefficient segment classifying unit that encodes the classification information and is output as a digital code, and each of the coefficients classified into at least two systems by the coefficient unit classifying unit is encoded, and the result is digitally encoded. And a quantizer that outputs the code.

[0015] Alternatively, the encoding apparatus may include a time / frequency conversion unit that obtains a frequency domain coefficient by performing time / frequency conversion on the input audio signal for each predetermined number of inputs, and a frequency domain coefficient from the time / frequency conversion unit. A coefficient piece generation unit that divides adjacent coefficients into small piece groups, a strength calculation unit that calculates the strength of each of the small piece groups generated by the coefficient piece generation unit, and a small piece group calculated by the strength calculation unit The coefficient pieces are divided into at least two groups based on the relative magnitudes of the intensities of the pieces, and the pieces group generated by the coefficient piece generation unit is classified into at least two systems based on the information of the grouping, and the classification is performed. A coefficient segment classification unit that encodes information and outputs it as a digital code, and normalizes the intensity of each of the coefficient segment groups classified into at least two systems by the coefficient segment classification unit, and encodes the normalized information. Conclusion A digital code, and a coefficient group of at least two systems normalized by the flattening unit, and a coefficient of one system of an array before classifying the coefficient pieces using the grouping information. It includes a coefficient synthesizing unit that re-synthesizes into a small piece group, and a quantization unit that quantizes the coefficient small piece group re-synthesized by the coefficient synthesizing unit and outputs the result as a digital code.

According to the present invention, a decoding apparatus which inputs a digital code and outputs an audio signal sample,
An inverse quantization unit that decodes the input digital code and obtains a group of coefficient pieces from a plurality of systems; and decodes the input digital code to obtain classification information of the coefficient pieces and divides the group of coefficient pieces from the plurality of systems based on this information. A coefficient synthesizing unit for synthesizing and sequentially restoring the frequency domain coefficients of one system, a frequency / time converting unit for frequency / time converting the frequency domain coefficients restored by the coefficient synthesizing unit, and outputting the result as an audio signal; And

Alternatively, the decoding device decodes the input digital code to obtain a group of coefficient pieces, and decodes the input digital code to obtain classification information of the coefficient pieces, and further, based on this information, A coefficient segment classification unit for classifying the coefficient segment into a plurality of systems; a decoding unit for decoding an input digital code to obtain normalized information of the coefficient segment group; The inverse flattening unit that inverse normalizes each of the above, based on the classification information of the coefficient small pieces obtained by the coefficient small piece classifying unit, the coefficient small piece group of a plurality of systems denormalized by the inverse flattening unit before classification. It includes a coefficient synthesizing unit that sequentially arranges them in one system and restores them to frequency domain coefficients, and a frequency / time converting unit that frequency / time converts the frequency domain coefficients from the coefficient synthesizing unit and outputs the result as an audio signal.

[0018]

DESCRIPTION OF THE PREFERRED EMBODIMENTS In the present invention, after converting an input signal into frequency domain coefficients, a series of coefficients are bundled for each band having a width of about 100 Hz to form coefficient small pieces, and each coefficient small piece corresponds to its intensity. At least two groups, for example, a high level group and a low level group. For example, as shown in FIG. 3A, the modified discrete cosine transform (MDC
T) When the magnitude of a coefficient in the frequency domain such as a coefficient fluctuates, adjacent frequency domain coefficients (FIG. 3B) are bundled to form coefficient pieces as shown in FIG. 3C. the group G ₀ coefficient magnitudes is small as shown in FIG. 3D, classified into large group G _1. Higher for the group G ₁ and the lower group G ₀ intensity, subjecting each independently of the process. In the independent processing after grouping, in addition to a method of performing quantization processing separately, a method of performing two independent groups G ₀ and G ₁ and performing vector quantization by performing flattening independently. There is.

Since the shapes of the coefficient pieces belonging to each group after the classification are often caused by the same sound source, the variation in intensity within the group is small. For this reason,
In the independent processing after grouping, for example, if processing is performed for each small band at equal intervals on the Bark scale, quantization can be performed with high efficiency while allocating an optimal amount of information perceptually. The grouping is not limited to two,
It may be divided into two or more.

As described above, according to the present invention, coefficient pieces are classified into a plurality of groups, and each group is flattened before coding, and the classification information is coded. This classification information is based on the aforementioned Japanese Patent Application Publication 7-1685.
Because compression is easier than the location information required in 93,
The amount of information can be reduced, and coding can be performed efficiently. First Embodiment FIG. 4 shows a first embodiment of the present invention.

The processing units 11 to 18 in FIG.
Then, an audio signal x that is a discrete sample sequence is input, and an encoded bit sequence C is output. In the decoding unit 30, the processing units 31 to 36 in FIG. 4 receive the encoded bit sequence C and output an audio signal x that is a discrete sample sequence. Time / frequency converter 11 The input audio signal x is input to the time / frequency converter 11 as a sequence of discrete samples, and is subjected to time / frequency conversion for each of a fixed number N of input samples to be converted into N frequency domain coefficients. I do. As a method of the time / frequency transform, a discrete cosine transform (DCT) or a modified discrete cosine transform (MD
CT) can be used. When the modified discrete cosine transform is used as a transform method, the N and the 2 × N input audio samples immediately before each of the N inputs are transformed to obtain N frequency domain coefficients. obtain. A window function such as a Hamming window or a Hanning window may be applied immediately before performing the time / frequency conversion processing. In particular, when the modified discrete cosine transform is selected as the transform method, it is desirable to apply a window represented by the following equation W to the input sample x.

W (i) = 0.5 {1-cos [2π (0.5 + i) / N]}, i = 0, 1,..., N−1 (1) The above processing is performed by the modified discrete cosine transform. When the case is represented by a mathematical expression as an example, the following expression is obtained.

[0023]

(Equation 1) Here, i is an input sample number, k is a number representing a frequency, and x represents an input sample. Coefficient small piece generator 12 The coefficients in the frequency domain obtained by the time / frequency converter 11 are input to the coefficient small piece generator 12. The coefficient piece generation unit 12 collects the input coefficients in the frequency domain for every M pieces to form coefficient pieces. As a result, the coefficient piece E
Is configured as follows:

E (q, m) = X (q · M + m) (3) where q = 0, 1,..., Q−1, and m = 0, 1,. -1 Here, q is a number representing a coefficient piece, m is a number representing a coefficient in the coefficient piece, and Q is the number of coefficient pieces. The size M of the coefficient piece can take any integer value of 1 or more,
The effect is high if the frequency width is set to, for example, about 100 Hz. For example, if the sampling frequency of the input signal is 48
If it is kHz, set M to about 8. Also, here, the value of M has been described as being constant for all coefficient pieces, but the coefficient pieces need not necessarily be of equal size, and the value of M may be set individually.

The coefficient pieces generated by the coefficient piece generating section 12 are input to a coefficient piece classification determining section 13 and a coefficient piece classifying section 14. FIG. 5 shows a detailed structure of the coefficient piece classification determining unit 13. As shown in FIG.
The coefficient piece classification determining unit 13 inputs the coefficient pieces and outputs the classification information. That is, the input coefficient piece is input to the coefficient piece strength calculator 3-1 and the strength I is calculated for each coefficient piece as in the following equation.

[0026]

(Equation 2) The coefficient small piece intensity is divided into small bands by a band dividing unit 3-2. The divided coefficient small piece strength is represented as I _sb (i _sb , q _sb ). Here, i _sb indicates a small band number, and q _sb indicates a small piece number in the small band. The number of coefficient pieces included in one small band is an arbitrary number of 2 or more, and is given by Q _sb (i _sb ). The relationship between I (q) and _Isb is expressed by the following equation.

I _sb (i _sb , q _sb ) = I (q) (5) Further, the relationship between i _sb , q _sb and q is represented by the following equation.

[0028]

(Equation 3) The coefficient small piece intensity band-divided by the band dividing section 3-2 is calculated by a threshold value determining section 3-3, a small piece classification determining section 3-4, and a separability calculating section 3-
Passed to 5. In the threshold decision unit 3-3, the band division unit 3-
The maximum value and the minimum value of the coefficient piece intensity received from 2 are obtained for each small band, and based on this value, the threshold value T for the piece classification is determined as in the following equation.

T _sb (i _sb ) = αI _sb (i _sb , q _max ) + (1−α) I _sb (i _sb , q _min ) (7) where q _min gives the minimum value of I _sb Coefficient piece number,
q _max is the number of the coefficient piece that gives the maximum value of I _sb , the constant α
Is 1 ≧ α> 0. The value of α is set to about 0.4. The threshold value T _sb determined here is passed to the small piece classification determining unit 3-4. In the small piece classification determining section 3-4, the coefficient strength I _sb received from the band dividing section 3-2 is compared with the threshold value T _sb received from the threshold value determining section 3-3 to determine the classification of the coefficient small piece. .
The classification information G of the coefficient pieces is determined as follows for each of q = 0, 1,..., Q-1.

G (q) = 0 for I _sb (i _sb , q _sb ) ≦ T _sb (i _sb ) = 1 for I _sb (i _sb , q _sb )> T _sb (i _sb ) (8) The classification information G (q) of the determined coefficient piece is passed to the separation degree calculation unit 3-5 and the classification information output unit 3-7. In the separation degree calculation unit 3-5, the coefficient intensity I received from the band division unit 3-2
_sb and classification information of the coefficient pieces received from the small piece classification judging section 3-4 G a (q) for each subband based on I _sb G (q) = 0 and G (q) = 1
And the degree of separation is calculated from the intensity of each group. Before calculating the degree of separation, the intensities of the two groups are determined. The intensity I _G0 of the group where G (q) = 0 is obtained as follows.

[0031]

(Equation 4) The intensity _IG1 of the group where G (q) = 1 is obtained as follows.

[0032]

(Equation 5) Separation D _sb is determined from I _G0 and I _G1 as follows. D _sb (i _sb ) = I _G1 (i _sb ) / I _G0 (i _sb ) (11) The _resolution D _sb (i _sb ) for each small band i _sb determined here is
It is passed to the use / non-use decision section 3-6 for small-piece classification.

The small-segment use / non-use determining unit 3-6 determines use / non-use of the small-segment classification for each small band according to the degree of separation determined by the separability calculating unit 3-5. When the separation degree D _sb exceeds the threshold value Dt, the small piece classification use flag F _sb (i _sb ) is set to 1. Otherwise, use small piece classification flag F
Set _sb (i _sb ) to 0. The small piece classification use flag F determined here is passed to the classification information output unit 3-7. In the classification information output unit 3-7, the classification information G (q) of the coefficient small pieces received from the small piece classification determination unit 3-4 is converted into the small piece classification use flag F _sb received from the small piece classification use / non-use determination unit 3-6. _Redetermined for each small band based on (i _sb ). When the value of F _sb (i _sb ) is 0, all the values of the classification information G (q) of the coefficient pieces belonging to the i _sb- th small band are set to 0. When the value of F _sb (i _sb ) is 1, the classification information G (q) of the coefficient pieces belonging to the i _sb- th small band is left as it is. G using F _sb
Although redetermination of (q) is not always necessary, _{redetermining} by F _sb makes it possible to set G (q) of a coefficient piece having a small variation in coefficient magnitude within a small band to 0, and , The encoding of the classification information G (q) can be performed with higher efficiency.

The classification information G (q) of the coefficient pieces re-determined by the classification information output section 3-7 is output from the coefficient piece classification determining section 13 in FIG. It is passed to the classification information compression unit 15. Coefficient small piece classifying unit 14 The coefficient small piece classifying unit 14 receives the coefficient small piece generated by the coefficient small piece generating unit 12 and the classification information G (q) of the coefficient small piece determined by the coefficient small piece classification determining unit 13 and receives all the coefficient small pieces. The coefficient pieces are group E _g0 of G (q) = 0 and group E _{g1 of} G (q) = 1.
Classify into.

The coefficient small piece classifying unit 14 selects the group E_g0And E_g1
Size S₀, S₁Has a memory (not shown) for storing
Shall be Also, the role of the counter for the small piece number q is
(Not shown). FIG.
7 is a flowchart of the process of the coefficient piece classifying unit 14. Coefficient piece
In the classifying unit 14, first, S₀, S₁, Q to 0
Rear. Next, the small piece number q of the memory is converted into a coefficient piece E (q, m).
If q is smaller than the number Q of
Proceed to step S3, otherwise E_g0(S₀, m) to glue
E _g0 And E_g1(S₁, m) to group E_g1 As each
Size S₀, S₁And exit (step S
2).

In step S3, the classification information G of the coefficient pieces
It is determined whether the value of (q) is 1; if this is true, the process proceeds to step (step S6); if false, the process proceeds to step (step S4). In step S4, a small piece E (q, m) indicated by the memory counter q is added to the small piece group E _g0 as in the following equation. E _g0 (S ₀ , m) = E (q, m), m = 0, 1,..., M−1 In step S5, the memory group size S ₀ is incremented by one, and the process proceeds to step S8. Proceed to

In step S6, a small piece E (q, m) indicated by the memory counter q is added to the small piece group E _g1 as in the following equation. _{E g1 (S 1, m)} = E (q, m), m = 0,1, ..., the M-1 step S7, the increment of the group size S ₁ of the memory by one, step S8 the process Proceed to Step S8
Then, the memory counter of the small piece number q is incremented by 1, and the process proceeds to step S2.

The small piece groups E _g0 and E _g1 classified by the coefficient small piece classifying unit 14 and their sizes S
₀ and S ₁ are the first and second quantizers 16 and 17 respectively.
Passed to. -Coefficient small piece classification information compression section 15 The coefficient small piece classification information compression section 15, a series of coefficient small piece classification information G (q), q = 0,
1, ..., Q-1 are compressed, and the compressed coefficient piece classification information G (q) ^*
To the multiplexing unit 18.

Since the appearance probabilities of the values 0 and 1 of the coefficient piece classification information G (q) are usually biased, any reversible compression coding using this can be used. However, Huffman coding or arithmetic coding can be used. The use of entropy coding such as In addition, the effect can be obtained by using run-length coding. Further, as shown in FIG. 7, the coefficient piece classification information G (q), q = 0, 1,..., Q-1 is divided into several blocks, and the coefficient piece classification information having a value of 1 in the block. If there is no G (q) is set to F _G flag, represented by 1 bit to 0,
Behalf of that block only the flag F _G, after setting the flag F _G to 1 when the value is one of the factors piece classification information G (q), add a flag F _G = 0 at the beginning of the block And
The number of bits may be reduced as a whole by expressing the coefficient piece classification information G (q) in the block with one bit each. Further, the coefficient piece classification information in which the number of bits is reduced may be applied to, for example, the above-described Huffman coding or arithmetic coding. First Quantizer 16 The first quantizer 16 encodes the coefficients constituting the small piece group E _g0 classified by the small coefficient classifier 14.

Before encoding, the small piece group E _g0 is converted into a single coefficient sequence as shown in the following equation. C ₀ (s · M + m) = E _g0 (s, m), s = 0, 1,..., S ₀ , m = 0, 1,. The coefficient forming the coefficient sequence C ₀ may be divided into a number of small blocks, and a method of scalar quantization by adaptively assigning the number of quantization bits to each small block may be used (method A). _Dividing the coefficient that constitutes ₀ into several small blocks, determining the optimal quantization width for each small block, performing scalar quantization, and using a method that performs entropy coding such as Huffman coding or arithmetic coding Alternatively, the coefficient sequence C ₀ may be vector-quantized collectively (method C), or the coefficient sequence C ₀ may be interleaved vector quantized (method D).

The information quantized here is obtained by the method A or C
Alternatively, when D is used, the quantization index In _E0 is converted into a bit string by converting the quantization index In _E0 into a bit string with a minimum necessary number of bits, and then passed to the multiplexing unit 18. In the case of using the method B, the bit string is passed to the multiplexing unit 18 as it is. Further, the size S ₀ of the small piece group E _g0 input from the coefficient small piece classifying unit 14 is also converted to a bit string by being converted into a binary number with a predetermined number of bits, and then passed to the multiplexing unit 18. Second Quantizer 17 The second quantizer 17 encodes the coefficients forming the small piece group E _g1 classified by the small coefficient classifier 14. The encoding is performed in the same procedure as that of the first quantization unit, but the encoding method does not necessarily need to be the same as that of the first quantization unit.

Before encoding, the small piece group E _g1 is converted into a single coefficient sequence as shown in the following equation. _{C 1 (s · M + m} ) = E g1 (s, m), s = 0, 1, ..., S 1, m = 0,
1, ..., the M-1 encoding method divides the coefficients constituting the coefficient sequence C ₁ into some small blocks, for scalar quantization adaptively allocated number of quantization bits to each small block A method may be used (method A), or a coefficient constituting the coefficient sequence C ₁ may be divided into several small blocks, an optimal quantization width is determined for each small block, scalar quantized, and then Huffman coding is performed. Alternatively, a method of performing entropy coding such as arithmetic coding may be used (method B), the coefficient sequence C ₁ may be vector-quantized at once (method C), or the coefficient sequence C ₁ may be interleaved vector quantization. (Method D).

The information encoded here is obtained by the method A or C
Alternatively, when D is used, the quantization index In _E1 is converted into a bit string by converting the quantization index In _E1 into a bit string with a minimum necessary number of bits, and then passed to the multiplexing unit 8. In the case of using the method B, the bit string is passed to the multiplexing unit 18 as it is. The size S ₁ of the small piece group E _g1 input from the coefficient small piece classifying section 4 is converted into a bit string by being converted into a binary number with a predetermined number of bits, and then passed to the multiplexing section 18.

Whichever method is used, the encoding method used in the second quantization unit 17 does not need to be the same as the encoding method used in the first quantization unit 16. Rather, the coding amount is reduced by focusing on the difference in the characteristics of the coefficient piece groups E _g0 and E _g1 given to the first and second quantization units 16 and 17 and using different coding methods suitable for each. Or distortion due to code errors can be reduced. Multiplexing unit 18 The multiplexing unit 18 includes a coefficient piece classification information compression unit 15 and a first
Input information G (q) from the quantization unit 16 and the second quantization unit 17
^* , In _E0 , and In _E1 are all output as output bit strings.
The bit string output from the multiplexing unit 18 becomes the output of the encoding unit 10 and is passed to the demultiplexing unit 31 of the decoding unit 30.

Hereinafter, the decoding unit 30 will be described. Demultiplexing unit 31 The demultiplexing unit 31 receives the bit string output from the encoding unit 10 and performs the reverse of the procedure performed by the multiplexing unit 18 on the bit string In _E0 and the bit string In _E0 to the first dequantization unit 32. 2 inverse quantization unit 33
Bit string an In _E1 and coefficient pieces classification information restoring unit decomposes the bit stream G (q) ^* to 34, each first inverse quantization unit 32 to,
Second inverse quantization unit 33 and coefficient piece classification information restoring unit 34
Send to First dequantization unit 32 The first dequantization unit 32 restores the bit string output from the demultiplexing unit 31 and outputs the coefficient piece group E _g0 and its size S ₀ . The size S ₀ is restored by converting a bit string representing a certain size, which is binarized with a predetermined number of bits, into an integer.

The bit string representing the small piece group E _g0 is the first bit string.
After the coefficient sequence C ₀ ^q is restored by inverse processing of the quantization method A, B, C or D used in the quantization unit 16, the small piece group E _g0 ^q is restored as in the following equation. E _g0 ^q (s, m) = C ₀ ^q (s · M + m), s = 0,1, ..., S ₁ -1, m = 0,
1, ..., M-1 Note that the superscripts ^q such as C ₀ ^q and E _g0 ^q are
Since the first quantizing unit 16 receives the quantization processing, a quantization error occurs, the C ₀ ^q and E _g0 ^q restored in the inverse quantization unit 32 are quantized with _respect to the original C ₀ and E _g0. It shows that the conversion error is included. The same applies to the superscript suffix q of other symbols. Second dequantizer 33 The second dequantizer 33 restores the bit string output from the demultiplexer 31 and outputs the coefficient piece group E _g1 and its size S ₁ . The size S ₁ is restored by converting a bit string representing a certain size, which is binarized by a predetermined number of bits, into an integer.

The bit string representing the small piece group E _g1 is represented by the second
After the coefficient sequence C ₁ ^q is restored by inverse processing of the quantization method A, B, C or D used in the quantization unit 17, the small piece group E _g1 ^q is restored as in the following equation. E _g1 ^q (s, m) = C ₁ ^q (s · M + m), s = 0,1, ..., S ₁ -1, m = 0,
1, ..., M-1 Coefficient piece classification information restoring section 34 In the coefficient piece classification information restoring section 34, the bit string output from the demultiplexing section 31 is reversibly used by the coefficient piece classification information compressing section 15. Restoration is performed by processing reverse to the compression encoding method, and coefficient piece classification information G (q), q = 0, 1,..., Q-1 is restored. Naturally, if different encoding methods are used in the first and second quantization units 16 and 17 in the encoding unit 10, the decoding used in the first and second inverse quantization units 32 and 33 of the corresponding decoding 30 will be described. The methods are also different. -Coefficient combining unit 35 The coefficient combining unit 35 combines the small piece groups output from the first inverse quantization unit 32 and the second inverse quantization unit 33 with the coefficient small piece classification information G output from the coefficient small piece classification information restoring unit 34. Based on (q), the system is reconstructed into one system and the coefficients in the frequency domain are output.

[0048] Figure 8, of the processing coefficient synthesis unit 35 is a processing procedure for obtaining a coefficient pieces group E ^q. Step S
In step 1, the initial values of S ₀ , S ₁ , and q are set to 0. In step S2, it is determined whether q <Q. If YES, the coefficient piece classification information G (q) is 1 in step S3. And N
If O, the coefficient piece E _g0 ^q (S ₀ , m) is changed to E ^q (q,
m) is defined to be the S ₀ and incremented by 1 in step S5, the flow returns to step S2 and incremented by 1 q in step S8. E _g1 ^q (S ₁ determined in step S6 if YES in step S3,
m) is defined as E ^q (q, m), S ₁ is incremented by ₁ in step S7, q is incremented by 1 in step S8, and the process returns to step S2. If it is determined in step S2 that q is not smaller than Q, the process ends, and the coefficient piece group E ^q (q, m), q = 0, 1,.
1, m = 0,1, ..., M-1 are obtained.

The coefficients piece group E ^q is a coefficient small piece generator 12 and the opposite manner, is reconstructed coefficients of frequency domain X ^q by the following equation. X ^q (qM + m) = E ^q (q, m), q = 0, 1, ..., Q-1; m = 0, 1,
..., M-1-Frequency / time conversion unit 36 The frequency / time conversion unit 36 performs frequency / time conversion on a series of coefficients X ^q (q · M + m) in the frequency domain output from the coefficient synthesis unit 35, and performs audio processing. Generate and output signal ^xq .

As a method of frequency / time conversion, an inverse discrete cosine transform (IDCT) or an inverse modified discrete cosine transform (IM
DCT) can be used. When the inverse transformed discrete cosine transform is used as the transform method, N input coefficients are transformed to obtain 2N time domain samples. After multiplying this sample by a window function W expressed by the following equation, an N sample obtained by adding the former N samples of the current frame and the latter N samples of the immediately preceding frame is output.

W (i) = 0.5 [1−cos [2π (0.5 + i) / N]｝, i =
0, 1,..., N−1 The above processing is expressed by the following equation when the inverse modified discrete cosine transform is used as an example.

[0052]

(Equation 6) x ^q (i) = Z ^t−1 (i + N) + Z (i), i = 0, 1,..., N−1 where x ^q (i) is an output audio sample signal. Second Embodiment FIG. 9 shows a second embodiment of the present invention. Processing unit 1 in FIG.
1, 12, 13, 14, 15, 19, and 20 are encoding units 1
0, an audio signal which is a discrete sample sequence is input, and an encoded bit sequence is output. Processing unit 31,
Reference numerals 34, 36 to 40 constitute the decoding unit 30, which inputs an encoded bit sequence and outputs an audio signal which is a discrete sample sequence.

Processing units corresponding to those of the first embodiment are given the same numbers. The processing units 11 to 15 in the encoding unit 10 according to the second embodiment perform the same processing as the corresponding units in the first embodiment, and a detailed description thereof will be omitted. FIG.
0 is referred to in the description of the flattening of the frequency domain coefficient in the embodiment described below. In row A, the frequency domain coefficient from the time / frequency conversion unit 11 is
Shows the state defined as the coefficient piece E (q, m), and rows B and C show the coefficient piece of G (q) = 1 and the coefficient of G (q) = 0 determined by the coefficient piece classification determining unit 13. Small pieces are shown separately, row D
And E represent the coefficient pieces whose classification has been determined by the coefficient piece classifying unit 14.
, The output is divided into two successive coefficient pieces, that is, two coefficient piece groups E _g0 and E _g1 . The processing of the coefficient pieces shown by these rows A to E is the first
This is the same as in the embodiment.

The coefficient piece groups E _g0 , E _g1 (rows E, D) and their sizes S ₀ , S ₁ from the coefficient piece classifying section 14 are input to the flattening / synthesizing section 20, and the coefficient piece class determining section 1
The coefficient piece classification information G (q) from 3 is also flattened / combined by the unit 20.
Is input to The flattening / synthesizing unit 20 sequentially determines coefficient pieces in each coefficient piece group based on their coefficient values for each original small band. L ₀ = L
₀₀ , L ₀₁ , L ₀₂ , L ₀₃ , L ₀₄ , L ₀₅ , L ₀₆ (row E) and L ₁ = L
₁₀ , L ₁₁ , L ₁₃ , and L ₁₅ (row D) to normalize by normalization, and these flattened coefficient pieces (rows G and F) of the two groups are used as coefficient piece classification information G (q). Based on this, it is incorporated into the original frequency position on the same frequency axis to form a series of flattened frequency domain system sequence e (q, m) (row H), which is provided to the vector quantization unit 19. Further, the coefficient piece flattening information L ₀ , L ₁ used for flattening is encoded and provided to the multiplexing unit 18 as L ₀ ^* , L ₁ ^* . Identical to each subband determines the representative value L ₀ and / or L ₁ in the coefficients pieces therein, the frequency than slightly apart (i.e. one or more of the small band accustomed #)
This is because there is a possibility that the cognate value may take a greatly different value, and when these are normalized collectively, the improvement in flatness is small. In vector quantization section 19 vector quantization section 19 sends the coefficients in the frequency domain supplied from the flattened and coefficient synthesizing unit 20 performs vector quantization, the coding index an In _e to the multiplexer 18. As a method of vector quantization, weighted interleaved vector quantization is desirable. The multiplexing unit 18 is a vector quantization unit 1
Compression classification information G of the coding index an In _e from 9 from the coefficient piece classification information compressing part 15 (q) ^*, flattening and combining unit 2
It is multiplexed with coefficient piece flattening information L ₀ ^* and L ₁ ^{* from} ₀ and transmitted to, for example, the decoding unit 30.

Hereinafter, the decoding unit 30 in the second embodiment will be described. The vector dequantizer 37 reproduces the vector quantization index In _e received from the demultiplexer 31 with reference to, for example, a codebook, and obtains a flattened frequency domain coefficient group e. ^q
(q, m) is obtained and sent to the coefficient piece generating unit 38. Coefficient small piece generation unit 38 The coefficient small piece generation unit 38 includes the coefficient small piece generation unit 1 of the first embodiment.
In the same manner as in FIG. 2 (FIG. 4), the flattened coefficient group e ^q (q, m) in the frequency domain is flattened into coefficient pieces e ^q (q), q = 0,
Divide into 1, ..., Q-1. Coefficient small piece classification section 39 The coefficient small piece classification section 39 includes a coefficient small piece classification information restoring section 34.
According to the coefficient piece classification information G (q) = 0 or 1 from FIG.
In the same manner as in the coefficient segment classification unit 14 in the above, the flattened coefficient segment e ^q (q) is converted into the flattened coefficient segment group e _g0
^q (size S ₀ ) and e _g1 ^q (size S ₁ ).

The inverse flattening / synthesizing unit 40 _{converts the} flattening coefficient small piece groups e _g0 ^q and e _g1 ^q into flattening information L _g = (L ₀ , L ₁ ), L ₀ = L ₀₀ , for each divided small area. L ₀₁ , L ₀₂ , L ₀₃ , L ₀₄ , L ₀₅ ,
L ₀₆ ; L ₁ = Inverse flattening by L ₁₀ , L ₁₁ , L ₁₃ , L ₁₅ , that is, E _g0 ^q
= e _g0 ^q L ₀ , E _g1 ^q = e _g1 ^q L ₁ is calculated, and coefficient pieces are sequentially sampled from E _g0 ^q or E _g1 ^q according to the classification information G (q) = 0 or 1,
By sequentially arranging them on the same frequency axis, coefficient pieces EA (q) of all bands are obtained. The frequency / time conversion unit 36 converts the whole band coefficient piece EA (q) into a time domain signal X and outputs it.

FIG. 11 shows an example of the configuration of the flattening / synthesizing unit 20 and the inverse flattening / synthesizing unit 40 in the second embodiment described with reference to FIG.
A and 11B. The coefficient piece group E _g0 and its size S ₀ output from the coefficient piece classification unit 14 are input to the first flattening unit 21, and the coefficient piece group E _g1 and its size S ₁ output from the coefficient piece classification unit 14. Is input to the second flattening unit 22. First flattening unit 21 The first flattening unit 21 flattens the coefficient small piece group E _g0 output from the coefficient small piece classification unit 14 using the coefficient small piece classification information G (q) as auxiliary information. Coefficient small piece group E
_{The g0} flattening process is a process of _obtaining a representative value for each of a plurality of coefficient small pieces (small bands), and normalizing the coefficients constituting all the coefficient small pieces of the small band using the representative values.

When the entire processing of the encoding unit 10 and the decoding unit 30 is executed by a computer program,
Handling all coefficient small pieces by specifying their positions on the linear frequency axis involves more processing steps that can be commonly used in encoding and decoding processing, and the configuration of the encoding and decoding program becomes easier. Here, an example is shown in which the coefficient pieces of the obtained coefficient piece group E _g0 are returned to the original frequency position, flattened, and then returned to a group of continuous coefficient pieces.
However, this method requires a larger amount of calculation and a larger memory capacity for processing than a method that does not restore a frequency position described later. The same applies to the second flattening unit 22.

FIG. 12 shows an example of the configuration of the first flattening section 21. In the frequency band restoration unit 21-1, the coefficient pieces E _g0 (s, m), s = 0,1, ..., S ₀ constituting the input coefficient piece group E _g0
Is developed into a coefficient piece group EA over the entire band based on the coefficient piece classification information G (q) (see row C in FIG. 10), and is passed to the small band dividing section 21-2. FIG. 13 shows a coefficient piece group E _g0 (s,
m), s = 0,1, ... , showing a recovery procedure by the frequency band restoring part 21-1 for S _0.

In step S1, the values of q and S are initialized to 0. In step S2, it is determined whether or not the coefficient piece classification information G (q) from the coefficient piece classification unit 13 is 0. If 0, the s-th coefficient piece E of the coefficient piece group E _{g0 in} step S3
_g0 (s, m) is arranged as the q-th coefficient piece EA (q) in the entire band (q = 0, 1,..., Q-1), and q and s are each advanced by one step. If G (q) = 0 is not satisfied in step S3, the coefficient 0 is set as the q-th coefficient piece EA (q) in the entire band in step S4.
(M) are inserted. In step S6, it is determined whether q is smaller than Q. If smaller than q, the process returns to step S2 and the same procedure S is performed.
2, S3, S4 and S5 are repeated. If q is not smaller than Q in step S6, the restoration of the coefficient piece group E _{g0 to} the entire band is completed.

The small band dividing section 21-2 divides the coefficient small piece group EA developed into all bands into small bands. The divided bandwidth of the small band may be constant in the entire band, or may be wider as the frequency is higher. The band-divided coefficient piece group is provided to a small-band representative value calculation unit 21-3 and a normalization unit 21-5. The small band representative value calculation unit 21-3 calculates a representative value for each divided small band. The representative value may be the maximum value of the absolute value of the coefficient in the small band or the square root of the average of the power of the coefficient in the small band that is larger than 0. The calculated representative value is provided to the small band representative value encoding unit 21-4.

The small band representative value coding section 21-4 codes the representative value of the small band. First, the representative value of the small band is scalar-quantized to obtain a quantization index L ₀ ^* . Those whose quantization index is 0 are not coded, and only those whose quantization index is larger than 0 are output to the multiplexing unit 18 as coefficient flattening information. As another method, the representative value group may be interleaved vector quantized. The representative value L ₀ of the subband quantized is transferred to the normalization unit 21-5.

In the normalizing section 21-5, the small coefficient group E _g0 divided into small bands provided from the small band dividing section 21-2 is quantized by the small band representative value encoding section 21-4. Is normalized by the representative value of the small band. The normalized coefficients piece group e _g0 i.e. is flattened is transferred to the coefficient pieces group restoring part 21-6. In the coefficient piece group restoring unit 21-6, the entire band coefficient piece group normalized by the reverse process of the frequency band restoring unit 21-1 is
The flattened coefficient segment group is restored and used as the output of the first flattening unit 21. Second flattening unit 22 The second flattening unit 22 is configured in the same manner as the first flattening unit 21, and by the same processing, converts the coefficient small piece group E _g1 given from the coefficient small piece classification unit 14 into a coefficient. Piece classification information G
Flatten using (q) as auxiliary information. The procedure is first
The same as the method in the flattening unit 21, but the frequency band restoring unit
In the part corresponding to 21-1 and coefficient piece group restoration unit 21-6,
The processing in the case where the value of the coefficient piece classification information G (q) is 0 and the processing in the case where the value is 1 are exchanged. In some small bands, the coefficient small piece group E _g1 does not exist, but in such a small band, the flattening by the second flattening unit 22 is not performed. This is applied to all processes by the second flattening unit 22 described below. -Coefficient combining unit 23 In the coefficient combining unit 23, the coefficient piece groups that have been flattened by the first flattening unit 21 and the second flattening unit 22, respectively, are used in the first embodiment.
And a flattened frequency domain coefficient is obtained by the same method as the coefficient synthesis unit 35 of FIG.

The inverse flattening / synthesizing section 40 shown in FIG.
Coefficient small piece group e received from the piece classification unit 39_g0 ^qAnd e_g1 ^qTo
Decoded coefficient piece flattening information L₀, L₁Inverse flat using
And the coefficients of these two inversely flattened groups
Piece E_g0 ^q, E_g1 ^qIs a series according to the coefficient piece classification information G (q).
Frequency domain coefficient E of^q(q, m) and output. -1st reverse flattening part 41 The 1st reverse flattening part 41 in FIG. 11B is shown in FIG.
FIG. 14 shows a configuration corresponding to the first flattening unit 21. Reverse
Flattening information L input from the multiplexing unit 31₀ ^*, L ₁ ^*use
And flattening coefficient small group e_g0 ^qIs inversely flattened. Immediately
That is, as shown in FIG.
Flattening coefficient subgroup e_g0 ^qMake up the flattener
A few pieces e_g0 ^q(s), s = 0,1, ..., S₀Is the coefficient piece classification information G
Based on (q), expand into coefficient sub-group EA (q) over the entire band
Is done. This coefficient piece group EA (q) is given to the sub-band splitter 41-2.
available.

The small band dividing section 41-2 divides the coefficient small piece group EA (q) expanded to the whole band into small bands. The bandwidth of the small band may be constant throughout the band, or may be wider as the frequency is higher. The band-divided coefficient small piece group is provided to the inverse normalization unit 41-5. Small band representative value decoding unit 41
-4, the input coefficient piece flattening information L ₀ ^* is decoded by the decoding method corresponding to the coding method in the small band representative value coding unit 21-4 (FIG. 12), and the small band representative value L ₀ is obtained.

In the inverse normalizing section 41-5, a flattening coefficient small piece group e _g0 ^q divided into small bands provided from the small band dividing section 41-2.
The inversely normalized representative value L ₀ of the small band decoded by the subband representative value decoding part 41-4. In the coefficient segment group restoring unit 41-6, the inversely normalized coefficient segment group is restored to a coefficient segment group by a process reverse to that of the frequency band restoring unit 41-1, and the output E _g0 ^{q of} the first inverse flattening unit 41 is restored. And Second de-flattening unit 42 The second de-flattening unit 42 in FIG. 11B is configured similarly to the first de-flattening unit 41 shown in FIG. 14, and the flattening information L input from the demultiplexing unit 31. using the subband representative value L ₁ obtained from ₁ ^* inverse flattening the flattened coefficient pieces group e _g1 ^q. The procedure is the same as that of the method in the first inverse flattening unit 41. However, in the part corresponding to the frequency band restoring unit 41-1 and the coefficient segment group restoring unit 41-6, the value of the coefficient segment classification information G (q) is The processing in the case of 0 and the processing in the case of 1 are interchanged. Although the small band there is the absence of flattened coefficient pieces group e _g1 ^q, in such a small bandwidth it does not execute the processing according to the second inverse flattening unit 26. This is applied to all processing by the second inverse flattening unit 26 described below.

The frequency / time converter 36 converts the frequency / time
The frequency domain coefficient X ^q = E ^q (q, m) from the inverse flattening / synthesizing section 40 is converted into a time domain signal x ^q and output in the same manner as the time converting section 36. The flattening section 21 (or 2) in FIG.
As an example of 2), FIG. 12 shows a case where the coefficient pieces are restored to the whole band, flattened by normalization, and returned to the coefficient piece group again. 6 shows a configuration example of a flattening unit 21 in the case of directly normalizing without restoring to a band. In this example, the coefficient segment group given together with the size S ₀ from the coefficient segment classification unit 14
E _g0 is the coefficient small piece classification determining unit 1 in the small band dividing unit 21-2.
3 (row E) based on the classification information G (q) from No. 3 to obtain the correspondence between those small bands and the classification information G (q). The small band representative value calculation unit 21-3 may use the maximum value of the absolute value of the coefficient value or the mean square of the coefficient values other than zero for each small band. The small band representative value is encoded by the small band representative value encoding unit 21-4, and the encoded representative value L ₀ ^* is given to the multiplexing unit 18 as coefficient piece flattening information, and the quantization obtained by decoding is obtained. given subband representative value L ₀ in the normalization unit 21-5 generates a flattened coefficient pieces group e _g0 by normalizing the coefficients pieces of subband. The second flattening section 22 can be similarly configured.

FIG. 16 shows a configuration example of the first inverse flattening unit 41 in the decoding unit 30 corresponding to the configuration in FIG. In this example, the flattening coefficient piece group e _g0 ^q from the coefficient piece classification unit 39 (FIG. 9) is divided into small bands associated with the coefficient piece classification information G (q) in the small band division unit 41-2. Are given to the inverse normalization section 41-5. On the other hand, the small band representative value decoding unit 41-4
Is the coding coefficient small piece flattening information L ₀ ^* from the demultiplexing unit 31 ^.
To obtain a subband representative value L ₀ by decoding, giving the inverse normalization unit 41-5. Inverse normalization unit 41-5 generates a coefficient pieces Group E _g0 ^q which is opposite flattened by inversely normalized by subband representative value L ₀ corresponding flattening coefficient pieces group e _g0 ^q for each subband ,Output.

FIGS. 17A and 17B show another configuration example of the flattening / combining unit 20 and the inverse flattening / combining unit 40 in FIG. In the flattening / combining unit 20 of the encoding unit 10, the first
The flattening information calculator 21A divides the given coefficient piece group E _g0 (row E in FIG. 10) into small areas, and represents the representative values L ₀₀ , L ₀₁ , L ₀₂ ,... Of the coefficient pieces in each small area. .. is calculated and given to the flattening information combining unit 23A as flattening information L ₀ (= L ₀₀ , L ₀₁ , L ₀₂ ,...), And the encoded L ₀ ^* is sent to the multiplexing unit 18. give. The small area is configured by combining input coefficient pieces belonging to the same small band when they are developed on the frequency axis. The small band is set in advance. Also, the representative value may be, for example, the maximum value of the absolute values of the coefficients in the small area, or the average value of the absolute values of the coefficients excluding 0. Similarly, the second flattening information calculation unit 22A also receives the given coefficient piece group E _g1
(Row D in FIG. 10) is divided into small areas of the same size as the first flattening information calculation unit 21A, and the representative value L ₀₁ ,
L _11, calculates a ..., with providing the flattening information to the flattening information combining part 23A, the multiplexer 1 L ₁ ^* obtained by encoding
Give 8

The flattening information combining unit 23A receives the flattening information L ₀₀ , L ₀₁ ,... From the first flattening information calculating unit 21A, and receives the flattening information L ₁₀ from the second flattening information calculating unit 22A. , L
_11, ... is given, q = 0, 1, ... first flattening information calculating unit 21 depending on whether the classification information G (q) is 1 or 0 for
A or the second flattening information calculator 22A
, And arranging them on the same frequency axis in order (that is, in the order of q = 0, 1,...), Thereby obtaining one series of flattening information over the entire band (FIG. 10 row I). Get.

On the other hand, the coefficient synthesizing unit 24 A
_g0 and E _g1 are given, and are obtained from E _g0 or E _g1 depending on whether G (q) is 0 or 1 by the same procedure as the combination of the flattening information by the flattening information combining unit 23A, and the same frequency is obtained. By sequentially arranging them on the axis, all bands (that is, q = 0,1, ..., Q-
A series of coefficient pieces E (q, m) is obtained over 1). Note that this series of coefficient pieces is the same as the coefficient piece series generated by the coefficient piece generation unit 12 (FIG. 9), and the coefficient synthesis unit 24A may be omitted by using this.

The flattening section 25 has a coefficient synthesizing section 24A (or
The coefficient piece E from the several piece generation unit 12) is
According to the flattened information sequence from the convenient unit 23A, the corresponding q
Flattening coefficient over the whole band by dividing by each
(Row H in FIG. 10). The obtained flattening coefficient is shown in FIG.
It is provided to the vector quantization unit 19. In the decryption unit 30
As shown in FIG. 17B, the inverse flattening / combining unit 40
Of the flattening unit 20 (FIG. 17A) in the converting unit 10
I do. That is, the first and second flattening information reproducing units 41A,
42A is the flattening information L from the demultiplexing unit 31A.₀ ^*, L ₁ ^* 
And the representative value L of the small area₀, L₁Flattening information combining unit 4
Give to 3A. The flattening information combining unit 43A performs coefficient piece classification.
Flattening information L according to information G (q)₀And L₁Over all bands
It is connected to the system and provided to the inverse flattening unit 45. Coefficient synthesis unit 4
4A is a small flattening coefficient from the coefficient piece classifying unit 39 (FIG. 9).
Piece group e_g0 ^q, e_g1 ^q Is given, and coefficient piece classification information G
e based on (q)_g0 ^qAnd e_g1 ^qOver the entire band, and 1
System flattening coefficient small piece e^q(q, m), and the inverse flattening unit 45
Given one system full band flattening coefficient e^q(q, m) 1 system
Frequency domain coefficients by inverse flattening using the entire band flattening information
E^q(q, m) is generated by the frequency / time conversion unit 36 (FIG. 9).
give.Third embodiment FIG. 18 shows a third embodiment of the present invention. Embodiment 3
In the configuration of the second embodiment shown in FIG.
The tanning unit 29 generates the time / frequency conversion unit 11 and the coefficient piece.
The decoding unit 30 includes the inverse flattening unit 4
9 is an inverse flattening / synthesizing unit 40 and a frequency / time converting unit 36
It is the structure provided between them. -Flattening unit 29 In the flattening unit 29, the time / frequency conversion unit 11 gives
The coefficient group in the frequency domain obtained is flattened, and the coefficient
Send to 2. As a method of flattening, for example, linear prediction spec
Normalization by vector is desirable. In that case, the linear prediction
The linear prediction coefficient LP used to generate the vector is encoded,
Auxiliary information LP^* To the multiplexing unit 18. Other processing
Is similar to the embodiment of FIG. Inverse flattening unit 49 In the inverse flattening unit 49, the linearity obtained from the demultiplexing unit 31 is obtained.
Prediction coefficient information LP^* Is a linear prediction coefficient L obtained by decoding
Generate a linear prediction spectrum from P, thereby inverse flat
Coefficient group E sent from the conversion / synthesis unit 40^qinverse flatten (q, m)
Frequency domain coefficient obtained by
It is output to the interval conversion unit 36. The operation of the other parts is shown in FIG.
This is the same as in the embodiment.

In the above description, if the number of samples is not required when quantizing the first and second coefficient piece groups E _g0 and E _g1 , the determination of the group sizes S ₀ and S ₁ can be omitted. In the above description, the coefficient pieces are classified into two groups, but may be classified into three or more groups.
The width of the coefficient small piece was set to about 100 Hz, but it can be set to an appropriate value at about 200 Hz or less, and it is conceivable that the width becomes narrower toward the lower frequency side. Further, the division of the coefficient pieces as described above does not necessarily have to be performed over the entire band, and it is within the scope of the present invention to perform the division only on a part of the bands.

In the third embodiment shown in FIG. 18, the first and second flattening units 21 and 22 of the flattening / combining unit 20 and the first and second inverse flattening units of the inverse flattening / combining unit 40 are used. 41,
Reference numeral 42 may be the same as the configuration of the flattening unit and the inverse flattening unit shown in FIGS. 12 and 14, respectively, or may be the same as the flattening unit and the inverse flattening unit shown in FIGS. Further, the flattening / combining unit 20 and the inverse flattening unit 40 in FIG. 18 may be replaced with the configurations shown in FIGS. 17A and 17B, respectively. The configuration in which the flattening unit 29 shown in FIG. 18 is provided between the time / frequency conversion unit 11 and the coefficient piece generation unit 12 can also be applied to the first embodiment shown in FIG.

FIG. 19 shows a configuration in which the encoding method and the decoding method according to the present invention are implemented by a computer. The computer 50 comprises a CPU 51, a RAM 52, a ROM 53, an input / output interface 54, which are mutually connected via a bus 50. A hard disk 55 is included. ROM
A basic program for operating the computer 50 is written in 53, and a program for executing the above-described encoding method and decoding method according to the present invention is stored in the hard disk 55 in advance. For example, when encoding
The PU 51 loads the encoding program from the hard disk 55 into the RAM 52, encodes the audio sample signal input from the interface 54 by processing according to the encoding program, and outputs the encoded audio sample signal from the interface 54. At the time of decoding, a decoding program is loaded from the hard disk 55 to the RAM 52, an input code is processed according to the decoding program, and an audio sample signal is output. A program for executing the encoding / decoding method according to the present invention may use a program recorded on an external disk device 57 connected to the internal bus 58 via the driving device 56. The recording medium on which the program for executing the encoding / decoding method according to the present invention is recorded may be any form of recording medium such as a magnetic recording medium, an IC memory, and a compact disk.

[0076]

As described above, according to the present invention, the frequency domain coefficients are successively bundled for each of a plurality of coefficients to form a plurality of coefficient pieces, and are classified into a plurality of groups according to the strength of each coefficient piece. Since coding is performed for each of the classified groups, the coefficient pieces in the same group have good flatness, and thus coding is performed efficiently. Using this invention,
Even a music signal in which a strong tone component such as a metallic sound is mixed in a high frequency band can be efficiently encoded.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a general form of a transform coding method.

FIG. 2 is a diagram illustrating an example of a tendency of an amplitude of a frequency domain coefficient.

FIG. 3 is a diagram showing a method of the present invention.

FIG. 4 is a block diagram showing a functional configuration of the first embodiment of the present invention.

FIG. 5 is a block diagram showing a detailed functional configuration of a coefficient piece classification determining unit 13 in the first, second, and third embodiments.

FIG. 6 is a diagram showing a processing flow of a coefficient piece classifying unit 14 in the first, second, and third embodiments of the present invention.

FIG. 7 is a schematic diagram showing the operation of a coefficient piece classification information compression unit 15 in the first, second, and third embodiments of the present invention.

FIG. 8 is a diagram showing a processing flow of a coefficient combining unit 35 in the first, second, and third embodiments of the present invention.

FIG. 9 is a block diagram showing a functional configuration of a second embodiment of the present invention.

FIG. 10 is a diagram for explaining flattening of frequency domain coefficients in the second and third embodiments.

11A is a block diagram illustrating a configuration example of a flattening / synthesizing unit 20 in FIG. 9; and FIG. 11B is a block diagram illustrating a configuration example of an inverse flattening / synthesizing unit 40 in FIG.

FIG. 12 is a block diagram showing a detailed functional configuration of a first flattening unit 21 in the second and third embodiments of the present invention.

FIG. 13 is a diagram showing a flow of processing of a frequency band restoring unit 21-1 in a flattening unit in the second and third embodiments of the present invention.

FIG. 14 is a block diagram showing a functional configuration example of a first inverse flattening unit 41 in FIG. 11B.

FIG. 15 is a block diagram showing another example of the functional configuration of the first flattening unit 21 in FIG. 11A.

FIG. 16 is a block diagram showing another example of the functional configuration of the first inverse flattening unit 41 in FIG. 11B.

17A is a block diagram illustrating another example of a functional configuration of the flattening / synthesizing unit 20 in FIG. 9, and FIG. 17B is a block diagram illustrating another example of a functional configuration of the inverse flattening / synthesizing unit 40 in FIG.

FIG. 18 is a block diagram showing a functional configuration of a third embodiment of the present invention.

FIG. 19 is a block diagram showing a configuration of a computer for executing encoding and decoding according to the present invention by a program.

[Procedure amendment]

[Submission date] March 23, 2000 (200.3.2.
3)

[Procedure amendment 1]

[Document name to be amended] Statement

[Correction target item name] Claim 41

[Correction method] Change

[Correction contents]

──────────────────────────────────────────────────続 き Continued on the front page (51) Int.Cl. ⁷ Identification FI FI Theme Court ゛ (Reference) G10L 9/18 M (72) Inventor Akio Kami 2-3-1 Otemachi, Chiyoda-ku, Tokyo Nippon Telegraph and Telephone Telephone Co., Ltd. (72) Inventor Kazuaki Chikira 2-3-1 Otemachi, Chiyoda-ku, Tokyo Nippon Telegraph and Telephone Corporation (72) Inventor Takeshi Mori 2-3-3, Otemachi, Chiyoda-ku, Tokyo No. 1 Nippon Telegraph and Telephone Corporation

Claims (46)

[Claims] 1. An audio signal encoding method for encoding an input audio signal sample, comprising the following steps: (a) time / frequency transforming an input audio signal sample for every fixed number of input samples; (B) divide the frequency domain coefficients into one or more bundled coefficient pieces to generate a coefficient piece group, and (c) generate a coefficient piece group for each of the coefficient piece groups. Calculating the strength, and (d) generating at least two series of coefficient pieces by classifying the coefficient piece group into at least two groups based on their strengths, and encoding the classification information. (E) Encode the at least two series of coefficient pieces and output as coefficient codes. 2. The audio signal encoding method according to claim 1, wherein said step (d) divides said coefficient piece group into a plurality of small bands for each of a plurality of coefficient pieces, and for each of said small bands, a coefficient small piece therein. Into one of the at least two groups based on their strength. 3. The audio signal encoding method according to claim 2, wherein said step (e) includes separately encoding said at least two series of coefficient pieces and outputting them as corresponding coefficient codes. . 4. The method of claim 2, wherein said step (e) comprises the following steps: (e-1) said at least two sequences of coefficient strips in said step (d). Are normalized separately, the normalized information is encoded, and the result is output as a normalized information code. (E-2) The coefficient pieces of the normalized coefficient row of at least two systems are Based on the classification information, re-synthesize into one group of coefficient small pieces of the array before classification, and (e-3) quantize the one-system coefficient small group re-synthesized in step (e-2), The result is output as the coefficient code. 5. The audio signal encoding method according to claim 3, wherein the number of the groups is two, and the step (d) is performed for each of the small bands in the distribution of the intensity of the coefficient piece of the small band. Is determined, and the threshold is compared with the intensity of each coefficient piece in the small band for classification. 6. The audio signal encoding method according to claim 5, wherein said step (d) calculates the sum of the intensities of the coefficient pieces belonging to the two groups in each of the small bands, and determines the ratio of the sums. Obtained as an index of intensity fluctuation in the small band, and when the above ratio is equal to or less than a predetermined value, all the coefficient small pieces of the small band are divided into two groups,
Reclassifying into the less intense group. 7. The audio signal encoding method according to claim 3, wherein said step (a) comprises: preliminarily normalizing the frequency domain coefficients with a spectral outline over the entire band of the input audio signal to thereby obtain a frequency domain coefficient. Flattening the coefficients, encoding the spectral outline information, and outputting the result as a spectral outline code. 8. The audio signal encoding method according to claim 4, wherein said step (e-1) comprises, for each small band of said at least two systems of coefficient small strings, a representative of the intensity of the coefficient small piece within said small band. A value is obtained, and all coefficient pieces of the small band are collectively normalized by a value corresponding to the representative value. 9. The audio signal encoding method according to claim 4, wherein said step (e-1) comprises separately restoring said at least two series of coefficient pieces into the entire band, and respectively representing a representative value for each small band. Is normalized by the representative value, and is output as a coefficient small piece sequence which is respectively flattened. 10. The audio signal encoding method according to claim 8, wherein said step (e-1) comprises calculating a representative value of coefficient piece intensity in each small band, quantizing the representative value, and quantizing the representative value. The small band is normalized by the representative value, and the quantization information is output as flattening information. 11. The audio signal encoding method according to claim 2, wherein said step (e) includes the following steps: (e-1) a step for each small band of said at least two systems of coefficient small strings; A value representative of the coefficient piece intensity in the band is obtained as flattening information. (E-2) The above-mentioned at least two systems of flattening information are combined into the whole band, and the at least two systems of coefficient piece rows are combined into the whole band. (E-3) normalizing the combined coefficient pieces with the combined flattening information to obtain a series of flattening coefficient pieces, and (e-4) the flattening coefficient pieces of the one system. The sequence is encoded and output as a coefficient code. 12. The audio signal encoding method according to claim 1, wherein the encoding of the classification information in the step (d) is performed by lossless compression. 13. The audio signal encoding method according to claim 1, wherein said step (e) comprises encoding at least one of said at least two series of coefficient pieces by adaptive bit allocation quantization. I do. 14. The audio signal encoding method according to claim 1, wherein the step (e) entropy-encodes after at least one of the at least two sequences of coefficient pieces is scalar-quantized. . 15. The audio signal encoding method according to claim 1, 3 or 11, wherein said step (e) encodes at least one of said at least two series of coefficient pieces by vector quantization. 16. The audio signal encoding method according to claim 1, wherein said step (e) comprises the step of encoding at least one of said at least two series of coefficient pieces in another system. And different encoding. 17. A decoding method for inputting a digital code and outputting an audio signal sample, comprising the steps of: (a) decoding an input digital code to obtain a plurality of sets of coefficient pieces; (b) The input digital code is decoded to obtain the classification information of the coefficient pieces, and based on this information, a group of coefficient pieces of a plurality of systems are synthesized to form a frequency domain coefficient composed of a group of coefficient pieces arranged in sequence. (C) The frequency domain coefficient obtained in step (b) is
Time conversion is performed, and the result is output as an audio signal. 18. A decoding method for inputting a digital code and outputting audio signal samples, comprising the steps of: (a) decoding an input digital code, wherein each coefficient piece has a plurality of coefficients in a frequency domain; And (b) decoding the input digital code to obtain classification information of the coefficient pieces, classifying the coefficient piece groups into a plurality of systems of coefficient piece columns based on the classification information, c) Decode the input digital code to obtain the normalized information of the coefficient sub-group, and denormalize each of the coefficient sub-sequences classified into a plurality of systems in the step (b) by the normalization information. ) Based on the coefficient piece classification information obtained in step (b) above, the coefficient piece rows of a plurality of systems denormalized in step (c) are sequentially arranged in an array of one system before classification, and the frequency domain (E) Step (d) above Obtained frequency domain coefficients of the frequency /
Time conversion is performed, and the result is output as an audio signal. 19. The decoding method according to claim 17, wherein said step (c) decodes the input digital code to obtain a spectral outline over the entire band, and inversely normalizes a frequency domain coefficient with the spectral outline. Including the step of transforming. 20. The decoding method according to claim 18, wherein in the step (d), the restored frequency domain coefficients are denormalized in a spectral outline and used as frequency domain coefficients. 21. The decoding method according to claim 18 or 19, wherein the step (c) restores the classified coefficient segment sequence to the original entire band according to the coefficient segment classification information, and Is denormalized by the normalization information. 22. The decoding method according to claim 17, wherein the decoding of the classification information in the step (b) is performed by decoding a lossless compression code. 23. The decoding method according to claim 17, wherein said step (a) comprises decoding an adaptive bit allocation quantization code for at least one series of coefficient pieces. 24. The decoding method according to claim 17 or 19, wherein the step (a) performs decoding of an entropy code for at least one series of coefficient pieces.
Obtain a scalar quantized coefficient. 25. The decoding method according to claim 17, wherein the step (a) decodes a vector quantization code for at least one system. 26. The decoding method according to claim 17 or 19, wherein the step (a) performs a decoding method different from the decoding method in another system for at least one series of coefficient pieces. 27. An encoding device for inputting audio signal samples and outputting a digital code, comprising: a time / frequency conversion unit for performing time / frequency conversion on an input audio signal for every predetermined number of inputs to obtain a frequency domain coefficient; And a coefficient-segment generator for dividing the frequency-domain coefficients from the time / frequency converter into small-slice groups in which adjacent coefficients are bundled; and an intensity calculation for calculating the strength of each of the small-segment groups from the coefficient-segment generator. The coefficient pieces are divided into at least two groups based on the relative magnitudes of the intensities of the small piece groups calculated by the strength calculation section, and the coefficient piece generation section generates the coefficient pieces based on the information of this grouping. Classify the group of small pieces into at least two systems, encode the classification information and output it as a digital code, and encode each of the coefficients classified by the coefficient small piece classifier into at least two systems. The audio signal encoding apparatus; and a quantization unit for outputting the result as a digital code. 28. An encoding apparatus for inputting audio signal samples and outputting a digital code, comprising: a time / frequency conversion unit for performing time / frequency conversion on an input audio signal for every predetermined number of inputs to obtain a frequency domain coefficient. And a coefficient segment generating unit that divides the frequency domain coefficients from the time / frequency converting unit into a small group of adjacent coefficients, and an intensity for calculating the strength of each of the small group generated by the coefficient small unit. A coefficient section is divided into at least two groups based on the relative magnitudes of the intensities of the small piece groups calculated by the calculation section and the intensity calculation section, and the coefficient section generation section generates the coefficient pieces based on the information on the grouping. A small coefficient group that classifies the divided small pieces into at least two systems, encodes the classification information and outputs it as a digital code, and a coefficient small piece group that is classified into at least two systems by the small coefficient classifier. A flattening unit that normalizes the respective intensities, encodes the normalization information, and outputs the result as a digital code, and a coefficient group of at least two systems normalized from the flattening unit are grouped into groups. A coefficient synthesizing unit that re-synthesizes into a group of coefficient small pieces of an array prior to classifying the coefficient small pieces using the information, and quantizes the coefficient small pieces re-synthesized by the coefficient synthesizing unit and digitally codes the result. An audio signal encoding device, comprising: a quantizing unit that outputs the audio signal as a signal. 29. The encoding apparatus according to claim 27, wherein the frequency domain coefficient from said time / frequency conversion unit is normalized by a spectral outline over the entire band of the input audio signal, thereby obtaining the frequency domain coefficient. A second flattening unit for flattening, encoding the spectral shape information, and outputting the result as a digital code is included. 30. The encoding apparatus according to claim 29, wherein the flattening unit is a unit that collectively normalizes the classified coefficient sub-groups having similar frequency bands to which they belonged. . 31. A decoding device for inputting a digital code and outputting an audio signal sample, comprising: an inverse quantization section for decoding the input digital code to obtain a group of coefficient pieces of a plurality of systems; Then, while obtaining the classification information of the coefficient small pieces, the coefficient combining section for combining the coefficient small piece groups of a plurality of systems based on this information and restoring them into one system of frequency domain coefficients arranged sequentially, An audio signal decoding device, comprising: a frequency / time conversion unit that performs frequency / time conversion of a frequency domain coefficient and outputs a result as an audio signal. 32. A decoding device for inputting a digital code and outputting an audio signal sample, comprising: an inverse quantization unit for decoding the input digital code to obtain a group of coefficient pieces; A coefficient segment classification unit for classifying the coefficient segment group into a plurality of systems based on the information of the segment, and decoding of the input digital code to obtain normalized information of the coefficient segment group. An inverse flattening unit that inversely normalizes each of the coefficient segment groups classified into multiple systems by the coefficient segment classifying unit, and an inverse flattening unit that inversely flattens the coefficient segment based on the coefficient segment classification information obtained by the coefficient segment classifying unit. A coefficient synthesis unit for sequentially arranging the coefficient small piece groups of the plurality of normalized systems into one system before classification, and restoring the frequency domain coefficients, and a frequency / time conversion of the frequency domain coefficients from the coefficient synthesis unit, and the result as audio Frequency output as No. /
An audio signal decoding device comprising: a time conversion unit. 33. A decoding apparatus according to claim 32, wherein an input digital code is decoded to obtain a spectrum outline over the entire band, and a frequency domain coefficient input to a frequency / time conversion unit based on the spectrum outline. And a second inverse flattening unit that inversely normalizes. 34. The decoding device according to claim 32, wherein the inverse flattening unit declassifies the group of classified coefficient pieces into ones having similar frequency bands to which the group belongs and collectively performs inverse normalization. It is a means to do. 35. An encoded program recording medium for inputting an audio signal sample and outputting a digital code, wherein the program includes the following steps: (a) input audio signal samples are output every fixed number of input samples; Time / frequency conversion to obtain frequency domain coefficients, (b) dividing the frequency domain coefficients into one or more bundled coefficient pieces to generate a coefficient piece group, and (c) generating a coefficient piece group. Calculate the strength of each coefficient piece, and (d) generate at least two systems of coefficient piece rows by classifying the coefficient piece group into at least one of two groups based on their strengths. The information is encoded and output as a classification information code. (E) The at least two series of coefficient pieces are encoded and output as a coefficient code. 36. The encoding program recording medium according to claim 35, wherein said step (d) divides said coefficient piece group into a plurality of small bands for each of a plurality of coefficient pieces, and for each of said small bands, a coefficient small piece therein. Are classified into any of the at least two groups based on their intensities. 37. The encoded program recording medium according to claim 36, wherein said step (e) includes a step of separately encoding each of said at least two series of coefficient pieces and outputting them as corresponding coefficient codes. . 38. The encoded program recording medium according to claim 36, wherein said step (e) includes the following steps: (e-1) In said step (d), said at least two systems of the coefficient small row In addition to normalizing the intensities separately, encoding the normalized information, outputting the result as a normalized information code, (e-2) the coefficient pieces of the normalized coefficient row of at least two systems are Based on the classification information, resynthesize into one group of coefficient small pieces of the array before classification, and (e-3) quantize the one set of coefficient small pieces recombined in step (e-2), and The result is output as the coefficient code. 39. The encoded program recording medium according to claim 37, wherein the number of the groups is two, and the step (d) is performed for each small band by the distribution of the intensity of the coefficient piece of the small band. , One threshold is determined, and the threshold is compared with the intensity of each coefficient piece in the small band to perform classification. 40. The encoding program recording medium according to claim 39, wherein in said step (d), the sum of the intensities of the coefficient pieces belonging to the two groups in each of the small bands is determined, and the ratio of the sum is calculated. A step of reclassifying all coefficient small pieces of the small band into the smaller one of the two groups if the ratio is equal to or less than a predetermined value. Including. 41. The encoded program recording medium according to claim 39, wherein the step (a) comprises: normalizing the frequency domain coefficients in advance by using a spectral outline over the entire band of the input audio signal. Flattening the coefficients, encoding the spectral shape information, and outputting the result as a spectral shape code. 42. A recording medium recording a decoding program for inputting a digital code and outputting an audio signal sample, said program including the following steps: (a) decoding an input digital code, (B) Decode the input digital code to obtain the classification information of the coefficient pieces, and based on this information, synthesize a group of coefficient pieces of multiple systems, (C) restore the frequency domain coefficients obtained in step (b)
Time conversion is performed, and the result is output as an audio signal. 43. A recording medium recording a decoding program for inputting a digital code, inputting a digital code, and outputting an audio signal sample, said program including the following steps: (a) input digital code (B) decoding the input digital code to obtain the classification information of the coefficient pieces, and to obtain the coefficient pieces based on the above classification information. The small group is classified into a plurality of series of coefficient small series, and (c) the input digital code is decoded to obtain the normalized information of the above-described small group of coefficients. Are denormalized by the above normalization information. (D) Based on the classification information of the coefficient pieces obtained in the step (b), the coefficient piece sequence of the plural systems denormalized in the step (c) based on the Before classification 1 Sequentially arranged in the sequence of integration, to restore the frequency domain coefficients, the frequency a frequency domain coefficients obtained in (e) the step (d) /
Time conversion is performed, and the result is output as an audio signal. 44. The decoding program recording medium according to claim 42, wherein in said step (c), the input digital code is decoded to obtain a spectral outline over the entire band, and a frequency domain coefficient is calculated using the spectral outline. Including denormalizing. 45. The decoding program recording medium according to claim 43, wherein in said step (d), said restored frequency domain coefficient is denormalized by a spectral outline and used as a frequency domain coefficient. 46. The decoding program recording medium according to claim 43, wherein the step (c) comprises restoring the classified coefficient piece sequence to the entire band according to the coefficient piece classification information. Is denormalized by the normalization information every time.