JP4963955B2 - Signal processing method, signal processing apparatus, and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a signal processing method in which an aliasing noise due to no bit allocation can be reduced. <P>SOLUTION: A quantized bit number detector 11 detects the number of quantized bits of a coefficient of the frequency band of a reversely quantized sound signal at the time of quantization. When the detected number of quantized bits of the coefficient is zero, a butterfly operation unit 31 performs butterfly operation as to the coefficient. The butterfly operation unit 31 generates an object frequency component to be used for band composition for canceling the aliasing noise through the butterfly operation as to a coefficient belonging to one frequency band when the number of quantized bits of the coefficient detected by the quantized bit number detector 11 is zero and the quantized bit number detector 11 detects the number of quantized bit of a coefficient at a loop-back position in another frequency band adjacent to the one frequency band that the coefficient belongs to being not zero. <P>COPYRIGHT: (C)2008,JPO&amp;INPIT

Description

  The present invention relates to a signal processing method for processing an acoustic signal obtained by dequantizing an encoded acoustic signal, a signal processing device, and a program for causing the signal processing device to function as a computer.

  As a technique for encoding an acoustic signal, MP3 (MPEG 1 Audio Layer 3), AAC (Advanced Audio Coding), ATRAC (Adaptive TRansform Acoustic Coding), WMA (Windows (registered trademark) Media Audio), or AC-3 (Audio Code Number). 3) etc. are known. For example, in the MP3 system, in order to compress with high efficiency, the acoustic signal is divided into a plurality of frequency bands and is blocked in units of variable length. The blocked digital data is converted into a spectrum signal by MDCT (Modified Discrete Cosine Transform) processing, and each spectrum signal is encoded with the number of bits assigned using the psychoacoustic characteristics (for example, Patent Documents 1 to 3).

  The acoustic signal encoded in this way is decoded by a decoding device. FIG. 18 is a block diagram showing a hardware configuration of a conventional decoding device. In the figure, reference numeral 100 denotes a conventional decoding device, which includes an unpacking unit 101, an inverse quantization unit 102, a frequency time conversion unit 103, a frequency band synthesis unit 104, and an acoustic signal output unit 105. The encoded acoustic signal is input to the unpacking unit 101, and the quantization coefficient, scale factor, scale factor multiplexer, global gain, and sub-block gain are unpacked from the frame information of the acoustic signal. Then, the inverse quantization unit 102 performs inverse quantization on the IMDCT coefficient using the quantization coefficient, the scale factor, the scale factor multiplexer, the global gain, and the sub-block gain.

  The IMDCT coefficient (Inverse Modified Discrete Cosine Transform) inversely quantized by the inverse quantization unit 102 is subjected to IMDCT processing by the frequency time conversion unit 103 for each frequency band, and converted to time-axis data. Further, the inversely converted frequency band is subjected to band synthesis by an IPFB (Inverse Polyphase Filter Bank), which is a band synthesis filter, in the frequency band synthesis unit 104 and then output to the acoustic signal output unit 105 (for example, Patent Literature 1). 3).

  Further, in order to compensate for the lack of power feeling associated with compression, a technique for supplementing a spectrum for power adjustment with a spectrum at the time of decoding is disclosed (for example, see Patent Document 4). In the technique described in Patent Literature 4, power adjustment information to be supplemented is generated in the power adjustment information determination unit in the encoding device based on the characteristics of the input audio signal during encoding. Next, the power adjustment information is encoded together with the encoded audio signal. Then, the power adjustment information encoded in the power adjustment information decoding unit in the decoding apparatus is decoded, and further, the power adjustment information is generated in the power correction spectrum generation / synthesis unit to supplement the decoded audio signal.

In the MP3 system, the data obtained by MDCT is subjected to removal of aliasing distortion components caused by PFB (Polyphase Filter Bank) in the frequency band by a butterfly circuit. That is, butterfly computation is performed on the 32 adjacent PFB bands by canceling out aliasing distortion (aliasing noise) components using 8 samples from the samples close to the band boundary as input (see, for example, Non-Patent Documents 1 to 3). On the other hand, in the ATRAC system, frequency band division by QMF (Quadrature Mirror Filter) is performed before frequency conversion such as MDCT, and frequency band synthesis is performed by IQMF (Inverse Quadrature Mirror Filter) after time conversion such as IMDCT. The filter is configured so that the aliasing noise generated by the QMF can be almost canceled (removed) by the IQMF.
JP 2002-351500 A JP-A-2005-195983 JP 2005-26940 A JP 2003-323198 A Supervised by Hiroshi Fujiwara, Hiroshi Yasuda, edited by Multimedia Communications Research Group, "Point Illustrated Standard Broadband + Mobile MPEG Textbook", ASCII, February 2003, first edition, p265 Jay Princen and A. Brandley, i-Triple Transactions on Acoustics, Speech, and Signal Processing, Volume, AS-32, No. 2 April 1984 Page 353-361 “Quadrature Mirror Filter Bank Design Based on Time Domain” ( "J. Prince and A. Brandley, IEEE Transactions on Acoustics, spech, and signal processing, vol. ASSP-32, No. 2 April. 1984 pp. 353-361" Quadratur Mr. Hiroshi Ochi “Digital Signal Processing Learned by Simulation Vol.9”, July 2001, CQ Publishing, p152-p154

  However, as described in Non-Patent Documents 2 and 3, these band division and band synthesis filters are designed on the assumption that some processing is not performed on the band-divided band signals, and some processing is performed. In such a case, the condition of complete reconstruction (see equation (2)) is not satisfied, and the aliasing distortion does not become zero. For example, in the case of QMF, the input signal for band division and band synthesis is x (n), the output signal is x ′ (n), and the transfer function X ′ (z) of x ′ (n) is expressed by Equation (1). The

Here, H ₀ is a division filter having a low-pass characteristic, H ₁ is a division filter having a high-pass characteristic, F ₀ is a synthesis filter having a low-pass characteristic, and F ₁ is a high-pass characteristic. Is a synthesis filter. The above-described equation (1) is completely reconstructed, that is, the conditions for returning X ′ (z) to X (z) (completely reconstructed) are equations (2) and (3).

Here, z ⁻¹ represents a delay of 1 sample, and if the expression (2) is satisfied, the distortion in the frequency amplitude characteristic and the frequency phase characteristic is eliminated, and the term of the aliasing component X (−z) becomes 0. That is, if Expression (3) is satisfied, the aliasing component is not included in the output x ′ (z). The filter coefficients of ATRAC QMF and IQMF are designed as shown in Equation (4) and satisfy Equation (5).

Here, h ₀ (n) is the impulse response of H ₀ , h ₁ (n) is the impulse response of H ₁ , f ₀ (n) is the impulse response of F ₀ , and f ₁ (n) is the impulse response of F _1. To express.

  Substituting this into equation (1) yields equation (6). In the characteristics of only the band division and band synthesis filters, the X (-z) term is 0, and the aliasing component is cancelled. become.

  As a factor that the aliasing distortion does not become zero, in the case of the ATRAC method, an aliasing distortion component is added after the band division filter, but quantization noise is generated by the irreversible encoding process, and especially the number of quantization bits is zero. Since the components that should originally be canceled by the band synthesis filter are completely removed, there is a problem that aliasing distortion occurs during decoding.

  On the other hand, in the case of the MP3 method, the aliasing distortion component in the frequency band coefficient is removed by the aliasing distortion reduction butterfly circuit in the frequency domain. Therefore, when the number of quantization bits is zero, the frequency component of the original sound is removed by quantization, but the aliasing distortion component that should be canceled is already removed, so that the problem as in the ATRAC method does not occur. However, even in the MP3 system, when the number of quantization bits is zero, there is a problem that the aliasing component for cancellation does not occur in the symmetrical band where the number of quantization bits is zero in the aliasing distortion reduction butterfly circuit. Patent Documents 1 to 4 and Non-Patent Documents 1 to 3 do not describe means for solving the problems.

  The present invention has been made in view of such circumstances, and the object thereof is not to perform any processing on the band-divided and band-synthesized band-band signals such as ATRAC and the like, such as ATRAC. In the method designed on the assumption that the number of quantization bits is detected, and when the number of quantization bits of the detected coefficient is zero, by generating a folding component in the frequency band for the coefficient by the arithmetic unit, An object of the present invention is to provide a signal processing method, a signal processing device, and a program for causing the signal processing device to function as a computer that can reduce the occurrence of aliasing distortion due to zero bit allocation.

  Another object of the present invention is a method in which the aliasing distortion component in the frequency band coefficient is removed by the aliasing distortion reduction butterfly circuit in the frequency domain, such as the MP3 method (hereinafter referred to as the MP3 method). A plurality of coefficients are selected from the coefficients in the frequency band of the dequantized acoustic signal, and the coefficients are not selected and the number of quantization bits is zero by interpolation using the selected plurality of coefficients. A signal processing method, a signal processing device, and a function for causing the signal processing device to function as a computer capable of generating a target frequency component used for aliasing reduction butterfly computation with higher accuracy by calculating an interpolation coefficient for the coefficient To provide a program.

The signal processing method according to the present invention is a signal processing method for processing an acoustic signal obtained by inversely quantizing a band-divided encoded acoustic signal, wherein quantization is performed when a frequency band coefficient of the inversely quantized acoustic signal is quantized. A detection step for detecting the number of bits, and a quantization bit number of a coefficient in one frequency band detected by the detection step is zero , and other adjacent to the one frequency band detected by the detection step And a calculation step of performing a butterfly operation on the coefficient of the one frequency band and the coefficient of the other frequency band when the number of quantization bits of the coefficient at the folding position in the frequency band is non-zero. .

The signal processing apparatus according to the present invention is a signal processing apparatus that processes an acoustic signal obtained by dequantizing a band-divided encoded acoustic signal, and performs quantization at the time of quantization of a frequency band coefficient of the dequantized acoustic signal. A detection unit that detects the number of bits, and a quantization bit number of a coefficient in one frequency band detected by the detection unit is zero , and another adjacent to the one frequency band detected by the detection unit And a calculation unit that performs a butterfly operation on the coefficient of the one frequency band and the coefficient of the other frequency band when the number of quantization bits of the coefficient at the folding position in the frequency band is non-zero. .

  In the signal processing device according to the present invention, the detection unit includes a quantization bit number at the time of quantization of a coefficient existing within a predetermined frequency from the boundary of the band division among the coefficients of the frequency band of the dequantized acoustic signal. It is comprised so that it may detect.

  In the signal processing device according to the present invention, the predetermined frequency is a frequency between a frequency of the band division boundary and a frequency of a predetermined intensity or more based on a frequency response of a band division filter that performs band division. And

The program according to the present invention is a program for processing an acoustic signal obtained by dequantizing a band-divided encoded acoustic signal by a computer. A detection step for detecting the number of quantization bits, a quantization bit number of a coefficient in one frequency band detected by the detection step is zero , and the other adjacent to the one frequency band detected by the detection step And a step of performing a butterfly operation on the coefficient of the one frequency band and the coefficient of the other frequency band when the number of quantization bits of the coefficient at the folding position in the frequency band is non-zero. And

  The signal processing method according to the present invention is a signal processing method for processing an acoustic signal obtained by inversely quantizing a band-divided encoded acoustic signal, wherein a plurality of coefficients are selected from the frequency band coefficients of the inversely quantized acoustic signal. A selection step for selecting, a detection step for detecting a quantization bit number at the time of quantization of the coefficient of the frequency band of the dequantized acoustic signal, and an interpolation method using a plurality of coefficients selected in the selection step, And a calculation step of calculating an interpolation coefficient for a coefficient that is not selected in the selection step and whose quantization bit number detected in the detection step is zero.

  The signal processing device according to the present invention is a signal processing device that processes an acoustic signal obtained by inversely quantizing a band-divided encoded acoustic signal, and calculates a plurality of coefficients from the frequency band coefficients of the inversely quantized acoustic signal. A selection unit that selects, a detection unit that detects the number of quantization bits when quantizing the coefficient of the frequency band of the dequantized acoustic signal, and an interpolation method using a plurality of coefficients selected by the selection unit, And a calculation unit that calculates an interpolation coefficient for a coefficient that is not selected by the selection unit and whose quantization bit number detected by the detection unit is zero.

  The program according to the present invention is a program for processing an acoustic signal obtained by dequantizing a band-divided encoded acoustic signal by means of a computer, wherein a plurality of coefficients are selected from the frequency band coefficients of the dequantized acoustic signal. A selection step for selecting, a detection step for detecting the number of quantization bits at the time of quantization of the coefficient in the frequency band of the dequantized acoustic signal, and an interpolation method using a plurality of coefficients selected in the selection step, And a calculation step of calculating an interpolation coefficient for a coefficient that is not selected by the selection step and whose quantization bit number detected by the detection step is zero.

The signal processing method according to the present invention is a signal processing method for processing an acoustic signal obtained by dequantizing a band-divided encoded acoustic signal, and whether or not the encoding method of the encoded acoustic signal is a method for performing a butterfly operation. A method determining step for determining whether or not, a detecting step for detecting a quantization bit number at the time of quantization of a coefficient of a frequency band of an inversely quantized acoustic signal, and a coding method performing a butterfly operation in the method determining step It is determined that it is not a method, and the number of quantization bits of the coefficient in one frequency band detected by the detection step is zero , and another frequency adjacent to the one frequency band detected by the detection step If the number of quantization bits of the coefficients in the folded position within the band of non-zero, butterfly for the coefficients and coefficient of said other frequency band of the one frequency band Characterized in that it comprises a calculation step of performing calculation.

The signal processing apparatus according to the present invention is a signal processing apparatus that processes an acoustic signal obtained by dequantizing a band-divided encoded acoustic signal, and whether or not the encoding scheme of the encoded acoustic signal is a scheme that performs butterfly computation. A method determining unit for determining whether or not, a detecting unit for detecting a quantization bit number at the time of quantization of a coefficient of a frequency band of an inversely quantized acoustic signal, and a coding method performing a butterfly operation in the method determining unit It is determined that it is not a method, and the number of quantization bits of the coefficient in one frequency band detected by the detection unit is zero , and another frequency adjacent to the one frequency band detected by the detection unit If the number of quantization bits of the coefficients in the folded position within the band of non-zero, this and a calculation unit for performing butterfly computation for the coefficients and coefficient of said other frequency band of the one frequency band The features.

  The signal processing apparatus according to the present invention has determined that a selection unit that selects a plurality of coefficients from the coefficients in the frequency band of the dequantized acoustic signal, and that the encoding method is a method for performing a butterfly operation in the method determination unit. In this case, an interpolation coefficient for a coefficient that is not selected by the selection unit and whose number of quantization bits detected by the detection unit is zero is calculated by an interpolation method using a plurality of coefficients selected by the selection unit. And a calculating unit.

The program according to the present invention is a program for processing by a computer an acoustic signal obtained by dequantizing a band-divided encoded acoustic signal, and whether the encoding method of the encoded acoustic signal is a method for performing butterfly computation on the computer. A method determining step for determining whether or not, a detecting step for detecting a quantization bit number at the time of quantization of a frequency band coefficient of an inversely quantized acoustic signal, and the encoding method performing butterfly operation in the method determining step. The number of quantization bits of the coefficient in one frequency band detected by the detection step is zero , and another adjacent to the one frequency band detected by the detection step is determined. If the number of quantization bits of the coefficients in the folded position within the frequency band of the non-zero coefficients of the one frequency band and the other circumferential The coefficient number band, characterized in that to perform the calculation step of performing a butterfly operation.

  In the present invention, the detection unit detects the number of quantization bits at the time of quantization of the frequency band coefficient of the inversely quantized acoustic signal. Then, when the number of quantized bits of the coefficient detected by the detection unit is zero, butterfly calculation is performed by the calculation unit that reduces the aliasing distortion of the coefficient, so that the band division filter aliasing generated due to zero bit allocation Distortion can be reduced.

  In the present invention, the detection unit pays attention to the coefficients existing in the predetermined frequency from the boundary of the band division among the coefficients in the frequency band of the dequantized acoustic signal, and the quantis at the time of quantizing these coefficients Detect the number of bits. The predetermined frequency is, for example, a frequency between the frequency at the boundary of the band division and a frequency of a predetermined intensity or higher based on the frequency response of the band division filter that performs band division. As described above, since the coefficient for performing the butterfly calculation is limited to the coefficient within the predetermined frequency, the power consumption can be reduced.

  In the present invention, the arithmetic unit has zero coefficient quantization bits detected by the detection unit, and the coefficient at the folding position in another frequency band adjacent to the one frequency band to which the coefficient belongs. When the number of quantization bits is detected as non-zero by the detection unit, a butterfly operation is performed using the detected coefficient and the coefficient at the folding position. On the other hand, when the number of quantization bits of the coefficient is zero and the number of quantization bits of the coefficient in the symmetric band from the boundary of the band division is also zero, the aliasing distortion does not occur and the calculation by the calculation unit is not executed.

  In the present invention, the selection unit selects a plurality of coefficients from coefficients in the frequency band of the inversely quantized acoustic signal. The detection unit detects the number of quantization bits at the time of quantizing the frequency band coefficient of the inversely quantized acoustic signal. Since the calculation unit calculates the interpolation coefficient for the coefficient that is not selected by the interpolation method using the plurality of coefficients selected by the selection unit and the number of quantization bits detected by the detection unit is zero, It is possible to generate an aliasing distortion component for cancellation in an optimal amount. Further, the energy lost during quantization is interpolated by the interpolation process, and it becomes possible to compensate for the lack of energy associated with the quantization error.

  In the present invention, the method determination unit removes the aliasing distortion component in the frequency band coefficient by the aliasing distortion reduction butterfly circuit in the frequency domain, for example, the encoding method of the encoded acoustic signal is the MP3 method. It is determined whether or not it is a method. Then, it is determined that the encoding method is not a method for performing butterfly computation, and when the number of quantized bits of the coefficient detected by the detection unit is zero, an object used for band synthesis for canceling aliasing distortion by butterfly computation for the coefficient Since the frequency component is generated, it is possible to reduce the aliasing distortion of the band division filter caused by the bit distribution being zero.

  In the present invention, when it is determined in the method determination unit that the encoding method is a method for performing a butterfly operation, a coefficient that is not selected is detected by an interpolation method using a plurality of coefficients selected by the selection unit. Since the interpolation coefficient for the coefficient whose quantization bit number detected by the unit is zero is calculated, the aliasing distortion component for cancellation can be generated in an optimum amount.

  In the present invention, the detection unit detects the number of quantization bits at the time of quantization of the frequency band coefficient of the inversely quantized acoustic signal. When the number of quantized bits of the detected coefficient is zero, the calculation unit performs a butterfly operation on the coefficient, so that it is possible to reduce the aliasing distortion of the band division filter caused by the bit distribution being zero. .

  In the present invention, the detection unit pays attention to the coefficients existing in the predetermined frequency from the boundary of the band division among the coefficients in the frequency band of the dequantized acoustic signal, and the quantis at the time of quantizing these coefficients Detect the number of bits. As described above, since the coefficient for performing the butterfly calculation is limited to a coefficient within a predetermined frequency, distortion can be effectively reduced in a band having a large distortion component, and power consumption can be reduced. It becomes.

  In the present invention, when the number of quantized bits of the coefficient is zero and the number of quantized bits of the coefficient in the symmetric band is also zero, no aliasing distortion occurs, so that the calculation by the calculation unit is not executed. As a result, generation of the target frequency component used for unnecessary aliasing distortion reduction butterfly computation is avoided, and the computation speed can be improved.

  In the present invention, the selection unit selects a plurality of coefficients from coefficients in the frequency band of the inversely quantized acoustic signal. The detection unit detects the number of quantization bits at the time of quantizing the frequency band coefficient of the inversely quantized acoustic signal. Since the calculation unit calculates the interpolation coefficient for the coefficient that is not selected by the interpolation method using the plurality of coefficients selected by the selection unit and the number of quantization bits detected by the detection unit is zero, It becomes possible to generate the target frequency component used for canceling aliasing reduction butterfly computation in an optimal amount. Further, the energy lost during quantization is interpolated by the interpolation process, and it becomes possible to compensate for the lack of energy associated with the quantization error.

  In the present invention, the method determination unit determines whether or not the encoding method of the encoded acoustic signal is a method for performing a butterfly operation. Then, it is determined that the encoding method is not a method for performing a butterfly operation, and when the number of quantized bits of the coefficient detected by the detection unit is zero, the butterfly operation is performed for the coefficient. It is possible to reduce the aliasing distortion of the generated band division filter. In addition, since the coding method is determined in advance and the coefficient of the dequantized frequency band is used, energy can be interpolated independently on the decoding side, and energy can be appropriately interpolated for various standards. Become.

  In the present invention, when it is determined in the method determination unit that the encoding method is a method for performing a butterfly operation, a coefficient that is not selected is detected by an interpolation method using a plurality of coefficients selected by the selection unit. Since the interpolation coefficient for the coefficient whose quantization bit number detected by the unit is zero is calculated, the target frequency component used for the aliasing distortion reduction butterfly calculation for cancellation can be generated in an optimum amount. In addition, the coding method is judged in advance and the coefficient of the frequency band that has been dequantized is used, so that energy can be interpolated independently on the decoding side, and energy can be appropriately interpolated for various standards. Thus, the present invention has excellent effects.

Embodiment 1
Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing a hardware configuration of a decoding device as a signal processing device. In the figure, reference numeral 20 denotes a decoding device that decodes an encoded acoustic signal, and includes an acoustic signal input unit 21, an unpacking unit 22, an inverse quantization unit 23, an interpolation processing unit 1, a butterfly operation unit 240, and a frequency time conversion unit. 24, a frequency band synthesis unit 25 and an acoustic signal output unit 26. In the present embodiment, an example in which MP3 is applied as a compression encoding method will be described first. The frequency band synthesizer 25 is an IPFB.

  The encoded acoustic signal read from the recording medium or the encoded acoustic signal received by the digital tuner is input to the acoustic signal input unit 21, and the input encoded acoustic signal is input to the unpacking unit (demultiplexer) 22. Is output. The unpacking unit 22 unpacks the quantization coefficient, the number of quantization bits, the scale factor, the scale factor multiplexer, the global gain, and the sub-block gain from the frame information of the acoustic signal. Using the unpacked quantization coefficient, the number of quantization bits, the scale factor, the scale factor multiplexer, the global gain, and the sub-block gain, the inverse quantization unit 23 performs inverse quantization on the IMDCT coefficient. From the inverse quantization unit 23, the IMDCT coefficient represented by the following equation (7) is output for each frequency band according to the block length (long block or short block).

  In Expression (7), the variable m is an IMDCT coefficient index, MK (m) is a quantization coefficient (Huffman decoded value), sgn (MK (m)) is a sign of the quantization coefficient, scalefac_multiplier is 1 or 0.5, gr Is the index of the granule, wnd is the index of the window shape, sfb is the index of the scale factor band, preflag [gr] is the pre-emphasis flag, 0 or 1, and pretab [sfb] is obtained by a predetermined pre-emphasis table Represents a value. Note that the scale factor in ATRAC (for example, expressed by 6 bits each and can be specified in units of about 2 dB) is the same as the value related to the scale factor in MP3, and the value related to the scale factor in MP3 is expressed by Equation (7). As shown in FIG. 4, the scale factor, the scale factor multiplexer, the global gain, and the sub-block gain (locations after the multiplier of 2 in Equation (7)), the pre-emphasis presence / absence flag, and the value obtained from the pre-emphasis table are used. . Hereinafter, the scale factor in ATRAC and the value related to the scale factor in MP3 are collectively described as a scale factor. Here, the scale factor means an exponent part represented by a mantissa part and an exponent part in order to express a spectrum of each predetermined frequency band. For example, in MP3, the spectrum of each predetermined frequency band is normalized so that the maximum value is 1.0, and the exponent part is encoded as a scale factor, a global gain, and a sub-block gain. Yes. The exponents of the scale factor, global gain, and sub-block gain are collectively referred to as values relating to the scale factor.

  In the present embodiment, as shown in the figure, 32 frequency band block (0) to block (31) include IMDCT coefficients I (0), I (1),..., I (m),. ) Is output. When the sampling frequency is 44.1 kHz, the frequency of block (0) is 0 Hz to 689.0625 Hz, block (1) is 689.0625 Hz to 1378.125 Hz, and block (31) is 21360.9375 Hz to 22050 Hz. In the following, a block of an arbitrary frequency band is assumed to be block (k). Here, k is an integer and satisfies 0 ≦ k ≦ 31. The IMDCT coefficients I (0) to I (575) of each frequency band are input to the interpolation processing unit 1.

  The IMDCT coefficient of each frequency band is composed of a plurality of coefficients (spectrums) according to the block length. The long block consists of 18 coefficients, and the short block consists of 6 coefficients. In the present embodiment, description will be made assuming that the block length is a long block.

  FIG. 2 is a graph showing changes in the IMDCT coefficient with respect to frequency. The horizontal axis represents frequency, and the vertical axis represents the coefficient. In the case of a long block, an IMDCT coefficient (hereinafter, represented by a coefficient I (m)) has 18 coefficients I (18 × k) to I (18 × k + 17) in one frequency band. In the graph of FIG. 2, changes in coefficients I (18 × k), I (18 × k + 1),..., I (18 × k + 17) are associated with frequencies 18 × k, 18 × k + 1,. It is shown. This coefficient takes a positive, negative or zero value.

In FIG. 1, the coefficient I (m) is input to the interpolation processing unit 1, and the coefficient I ′ (m) after the interpolation processing is output from the interpolation processing unit 1. The butterfly computation unit 240 performs butterfly computation in each band on the coefficient I ′ (m) of 8 samples from the boundary of the band division, using a butterfly computation expression of Expression (33) described later. Here, c _i input to the expression (33) is −0.6 when i = 0, −0.535 when i = 1, −0.33 when i = 2, −0.185 when i = 3, , I = 4, -0.095, i = 5, -0.041, i = 6, -0.0142, i = 7, -0.0037. Note that i is an interval between the input In1 or the input In2 and the division boundary, and these coefficients and mathematical expressions are stored in a memory (not shown) in the butterfly calculation unit 240. The frequency time conversion unit 24 receives the coefficient I ′ (m) after the interpolation process in which the butterfly operation is performed on some coefficients. The frequency time conversion unit 24 performs IMDCT processing on the coefficient I ′ (m) and converts it into a time-axis acoustic signal. Further, the inversely converted acoustic signal is band-synthesized by the frequency band synthesizing unit 25 by the IPFB which is a band synthesizing filter, and then output to the acoustic signal output unit 26.

FIG. 3 is a block diagram showing a hardware configuration of the interpolation processing unit 1. The interpolation processing unit 1 includes a quantization bit number detection unit 11, an interpolation determination unit 12, a selection unit 13, and a calculation unit 14. The quantization bit number detection unit 11 detects the quantization bit number at the time of quantizing the frequency band coefficient based on the input frame side information. Specifically, the number of quantized bits of the coefficient I (m) is detected by referring to table_select [ch] [gr] [region] of the frame side information in the bit stream unpacked by the unpacking unit 22. be able to.
The table_select [ch] [gr] [region] is a select signal indicating a Huffman-encoded Huffman table. By decoding the indicated Huffman table, a Huffman decoded value, that is, a coefficient I (m ) The number of quantization bits can be detected by obtaining the maximum number of Huffman decoded values of one region in the above region, but exists in the Huffman table indicated by table_select [ch] [gr] [region] Since the maximum number is determined in advance, the word length is detected as the number of quantization bits.

  The quantization bit number detection unit 11 outputs the detected quantization bit number to the interpolation determination unit 12 and the calculation unit 14. The interpolation determination unit 12 determines whether there is a coefficient I (m) that is equal to or smaller than a predetermined number of quantization bits of the coefficient I (m) in the frequency band. For example, the interpolation determination unit 12 may determine whether there is a coefficient I (m) in which the number of quantization bits of the coefficient I (m) in the frequency band is 4 or less. When the interpolation determination unit 12 determines that the coefficient I (m) within the input frequency band has a coefficient I (m) equal to or smaller than the predetermined number of quantization bits, the interpolation determination unit 12 calculates the coefficient I (m) of the frequency band. The data is output to the selection unit 13 for interpolation processing. On the other hand, when the interpolation determining unit 12 determines that there is no coefficient I (m) equal to or less than the predetermined number of quantization bits in the input coefficient I (m), the coefficient I (m) of the frequency band. Without being subjected to interpolation processing, the corrected coefficient I ′ (m) is output to the frequency time conversion unit 24 without going through the butterfly calculation unit 240.

  The selection unit 13 selects a plurality of coefficients from the coefficients in the frequency band. For example, at least the coefficients at both ends of the coefficient in the frequency band, that is, the coefficient corresponding to the lowest band and the coefficient corresponding to the highest band are selected. In the example of FIG. 2, I (18 × k) and I (18 × k + 17) are selected. Further, in addition to these, the selection unit 13 may select the coefficient having the maximum and minimum spectrum from the coefficients in the frequency band. In the example of FIG. 2, the minimum spectrum I (18 × k + 3) and the maximum spectrum and the coefficient I (18 × k + 17) at the highest frequency are selected. The selection unit 13 outputs the input coefficient I (m) and information related to the selected plurality of coefficients to the calculation unit 14.

  The calculation unit 14 uses the coefficient selected by the selection unit 13 and calculates an interpolation coefficient of a coefficient that has not been selected using an interpolation method. In this case, the calculation unit 14 calculates an interpolation coefficient for a coefficient having a quantization bit number of zero based on the quantization bit number of the coefficient output from the quantization bit number detection unit 11. As this interpolation method, for example, a Lagrange interpolation method or a spline interpolation method is used. Hereinafter, an example using the spline interpolation method will be described.

N + 1 points (x ₀ , y ₀ ), (x ₁ , Y ₁ ),..., (X _N , y _N ). However, x ₀ <x ₁ <.. x _N. A spline interpolation that smoothly connects these points will be described. A curve obtained by cubic spline interpolation is assumed to be y = S (x). Assume that S (x) is defined piecewise in each section [x _j , y _j ]. S (x) = S _j (x) in each section x _j ≦ x ≦ x _{j + 1} . Further, S _j (x) is given by a third order polynomial expressed by equation (8).

The coefficients a _j , b _j , c _j , _dj are determined from the conditions described below. That is, the curve y = S (x) is continuous and the point (x _j , y _j ) (J = 0, 1,..., N) are all satisfied (condition 1). Further, at the section boundary x = x _j (j = 1, 2,..., N−1), the first and second order differential coefficients of y = S (x) are continuous (condition 2). From condition 1, equation (9) is derived.

  Also, from condition 2, equation (10) is derived.

The coefficients a _j , b _j , c _j , _dj are determined by using these equations (9) and (10). First, the second-order differential coefficient of S (x) at x = x _j (j = 1, 2,..., N−1) is set as in Expression (11).

  Since the definition of the cubic spline is Expression (8), the second-order differential coefficient is expressed by Expression (12).

As a result, b _j = u _j / 2. Further, the second-order differential coefficient can be expressed by Expression (13).

  Equation (14) is derived from Equation (13).

  Moreover, the condition of Formula (15) is automatically satisfy | filled from the above formula.

Since d _j = y _{j is} also clear, Expression (16) is derived using Condition 1.

  Further, equation (17) is finally obtained from equation (16).

  Here, the last condition shown by Formula (18) is used.

  Expression (18) can be expressed from a cubic polynomial as Expression (19).

By substituting a _j , b _j , and c _j into equation (19), equation (20) is derived.

  When these are arranged in order, the simultaneous equations shown in Expression (21) are formed.

However, it is assumed that h _j and v _j satisfy the condition shown in the following formula (22). Note that h _j and v _j are known constants because they can be calculated from only x _j and y _j given first.

There are N + 1 unknown variables u _j in total, but the number of simultaneous linear equations described above is N-1. Therefore, u _j cannot be uniquely determined from the simultaneous linear equations. Therefore, one boundary condition is added to each point (x ₀ , y ₀ ), (x _N , y _N ) at both ends of the curve. There are several possible boundary conditions, but here, the condition that the change rate of the slope of the curve is 0 at both ends is adopted. Since the second-order derivative is 0, Expression (23) is derived.

  Expression (24) is derived from Expression (23).

Since u ₀ = u _N = 0, simultaneous linear equations relating to u _{1 to} u _N-1 represented by the equation (25) are obtained.

Next, a spline interpolation algorithm will be described. First, if N + 1 points (x _j , y _j ) (j = 0, 1,..., N) are given, and the cubic spline satisfies equations (26) and (27) piecewise, To do.

When the boundary condition at both ends of the curve is represented by equation (28), u ₀ = u _N = 0.

By calculating h _j (j = 0, 1,..., N) and u _j (j = 0, 1,..., N) and solving the simultaneous linear equations, u _{1 to} u _N-1 Is obtained. Finally, the coefficients a _j , b _j , c _j , _dj are obtained, and the curve S (x) is determined. Calculation unit 14 based on the selected by the selection unit 13 coefficients, the coefficient a _j of curve _{S j (x), b j} , c j, obtaining the d _j. Then, a coefficient which is not selected, the number of quantization bits for the coefficients below a predetermined value, the interpolation factor to calculate the S _j (x), not the interpolation coefficient S _j after interpolation (x) and interpolation coefficients As a coefficient I ′ (m) and output to the frequency time conversion unit 24.

  FIG. 4 is a flowchart showing the procedure of the interpolation process. In the following description, for ease of explanation, the block length of the coefficient in the frequency band is assumed to be a long block. First, the quantization bit number detection unit 11 detects the number of quantization bits (step S41). The detected number of quantization bits is output to the interpolation determination unit 12 and the calculation unit 14, respectively. The interpolation determining unit 12 determines whether there is a coefficient having a quantization bit number equal to or less than a predetermined value among the coefficients in the frequency band (step S42). If the interpolation determining unit 12 determines that there is no coefficient quantization bit number equal to or smaller than the predetermined value (NO in step S42), the series of processing ends. In this case, the interpolation determination unit 12 outputs the frequency band coefficient to the frequency time conversion unit 24 without passing through the butterfly calculation unit 240.

  On the other hand, when the interpolation determination unit 12 determines that there is a coefficient whose quantization bit number is equal to or smaller than a predetermined value (YES in step S42), the interpolation determination unit 12 outputs the coefficient in the frequency band to the selection unit 13. To do. The selection unit 13 selects the coefficients at both ends in the frequency band, that is, the low-frequency side coefficient and the high-frequency side coefficient, as the nodes of the spline interpolation (step S43). Further, the selection unit 13 selects the coefficient of the maximum spectrum and the coefficient of the minimum spectrum of the coefficients in the frequency band to be the nodes of the spline interpolation (Step S44). In addition, since the coefficient of the maximum spectrum and the coefficient of the minimum spectrum may be coefficients at both ends in the frequency band, the number of nodes is 2 to 4.

Calculation unit 14, based on the coefficient selected in step S43 and S44 shown in equation (8), the coefficient a _j of the cubic spline function, b _j, c _j, calculates a d _j (step S45). The calculation unit 14 determines whether or not the number of quantization bits of the coefficient not selected in steps S43 and S44 is equal to or less than a predetermined value (0) (step S46). When the calculation unit 14 determines that the number of quantization bits of the unselected coefficient is equal to or less than the predetermined value (0) (YES in step S46), the calculated coefficients a _j , b _j , c _j , _dj and the equation (8) ) To calculate the interpolation coefficient (step S47). On the other hand, when it is determined that the number of quantized bits of the unselected coefficient is not equal to or smaller than the predetermined value (0) (NO in step S46), the interpolation process is not performed and the process of step S47 is skipped.

The calculation unit 14 determines whether or not the processing in step S46 has been completed for all coefficients that are not selected in steps S43 and S44 (step S48). If the calculation unit 14 determines that the process has not ended (NO in step S48), the calculation unit 14 proceeds to step S46 in order to obtain the number of complementary relationships of other unselected coefficients. On the other hand, when the calculation unit 14 determines that the processing for all unselected coefficients has been completed (YES in step S48), the series of processing ends. As a result of performing the above processing for all frequency bands and performing spline interpolation on a coefficient whose quantization bit number is 0, an optimum spectrum is obtained as an interpolation coefficient.
The resolution of the quantized coefficient can be increased, and reproduction can be performed without being unsatisfactory or uncomfortable. It should be noted that the above-described method of selecting a coefficient serving as a node and the value of the number of quantization bits are merely examples and are not limited thereto.

Embodiment 2
The second embodiment relates to a mode for correcting the interpolation coefficient. FIG. 5 is a block diagram illustrating a hardware configuration of the interpolation processing unit 1 according to the second embodiment. In addition to the configuration of the first embodiment, an effective range determination unit 15 and a correction unit 16 are added. The scale factor of each frequency band is extracted from the frame side information of the bit stream output from the inverse quantization unit 23, and the extracted scale factor is input to the effective range determination unit 15. The quantization bit number of the coefficient detected by the quantization bit number detection unit 11 and the interpolation coefficient calculated by the calculation unit 14 are input to the effective range determination unit 15. Further, the coefficient that is not subjected to the interpolation process output from the interpolation determination unit 12 is also input to the effective range determination unit 15.

  FIG. 6 is a graph for explaining the effective range. In the graph of FIG. 6, the horizontal axis indicates the frequency, and the vertical axis indicates the spectrum size. A circle indicates a coefficient I (m) that has not been interpolated by spline interpolation. Here, for illustration purposes, the number of coefficients is 1 ≦ m ≦ 4, the scale factor is SF, and the number of quantization bits is 2. Further, the x mark represents the MDCT coefficient (M (m)) of the original sound. M (m) of the original sound is quantized to a circle in the direction of the arrow by 2-bit quantization. For example, M (1) is quantized to I (1) = 0SF because it is located at about 0.3SF and is 0.5SF or less. Since M (2) is larger than 0.5SF, it is quantized to I (2) = SF.

  Here, as shown in the figure, when I (1) = 0, the original sound M (1) theoretically exists from −0.5 SF to +0.5 SF because the number of quantization bits is two. When I (2) = SF, the original sound M (2) theoretically exists within the range of the upper limit SF and the lower limit 0.5SF. The theoretical range where the original sound exists determined by the scale factor and the number of quantization bits with respect to the coefficient I (m) is defined as an effective range. Here, when the effective range of the coefficient I (m) is P (m), the number of quantization bits is W, and the scale factor is SF, the effective range P (m) is defined by, for example, the following equation (29).

  However, when I (m) = SF, the effective range P (m) is defined by Expression (30).

  Further, when I (m) = − SF, the effective range P (m) is defined by the equation (31).

  The definition of these effective ranges is merely an example, and the effective range P (m) may be defined using the absolute value of the coefficient I (m), and may be determined based on the scale factor and the number of quantization bits for the coefficient. It is not limited to this.

  In FIG. 6, the triangle marks indicate the interpolation coefficient S (m) calculated by the calculation unit 14 in FIG. Focusing on S (1), S (3), and S (4), it can be understood that an interpolation coefficient closer to the original sound is calculated, and that the interpolation coefficient exists within an effective range indicated by an arrow. However, with respect to the interpolation coefficient S (2), an error caused by the Runge phenomenon of the interpolation method or the like occurs due to the selection of an inappropriate node coefficient, which deviates from the theoretically possible effective range. I understand that. The correcting unit 16 in FIG. 5 corrects this error based on the effective range P (m) and the interpolation coefficient S (m) output from the effective range determining unit 15.

  If the correction unit 16 determines that the interpolation coefficient is within the effective range, the correction unit 16 outputs the interpolation coefficient to the frequency time conversion unit 24 without passing through the butterfly calculation unit 240 without correction. On the other hand, when the correction unit 16 determines that the interpolation coefficient does not exist within the effective range, the correction unit 16 corrects the interpolation coefficient to belong within the effective range. This correction processing is performed as follows, for example. For example, when the interpolation coefficient exceeds the upper limit value of the effective range defined by the equations (29) to (31), the upper limit value is corrected to become the interpolation coefficient. Further, when the interpolation coefficient is lower than the lower limit value defined by the equations (29) to (31), the lower limit value is corrected to be the interpolation coefficient.

  In addition, an interpolation coefficient may be multiplied by a predetermined gain g. The gain g is a ratio between the upper limit (or lower limit) of the effective range P (m) and the interpolation coefficient S (m). Then, this gain g is multiplied by other interpolation coefficients (for example, adjacent S (m−2), S (m−1), S (m + 1), S (m + 2)), and the other interpolation coefficients have their respective effective ranges. It is determined whether it belongs to (P (m−2), P (m−1), P (m + 1), P (m + 2)). If it is determined that the correction unit 16 belongs, the gain g is multiplied by the interpolation coefficient S (m), and this value is output to the frequency time conversion unit 24 without going through the butterfly calculation unit 240.

  On the other hand, when the correction unit 16 determines that the other interpolation coefficients do not belong to each effective range, the value of the gain g is changed by a predetermined value (for example, 1.5 g, 1.4 g, 1.3 g,... 0). And this is repeated until the other interpolation coefficients belong to their respective effective ranges. Even when the above processing is performed, if other interpolation coefficients do not belong to the respective effective ranges, the upper limit value (or lower limit value) is set as the interpolation coefficient only for the interpolation coefficient S (m) as described above. Correct so that As a result, even if an interpolation error occurs due to some reason, the interpolation coefficient can be corrected within the theoretical range of quantization that the original sound can take, and the signal processing at the time of decoding can be stabilized. The correction process described above is merely an example, and other forms may be used as long as the correction is performed so that the interpolation coefficient is within the effective range.

  FIG. 7 is a flowchart showing the procedure of the correction process. The effective range determination unit 15 receives a scale factor, the number of quantization bits, an interpolation coefficient, and a coefficient (step S81). The effective range determination unit 15 determines an effective range for the coefficient based on the input scale factor and the number of quantization bits and the equations (29) to (31) (step S82). The effective range determination unit 15 outputs the effective range and the interpolation coefficient for the determined coefficient to the correction unit 16 (step S83).

  The correction unit 16 compares the interpolation coefficient with the effective range, and determines whether the interpolation coefficient is within the effective range (step S84). If the correction unit 16 determines that the interpolation coefficient is within the effective range (YES in step S84), the correction unit 16 outputs the interpolation coefficient to the frequency time conversion unit 24 without correction (step S87). On the other hand, when it is determined that the interpolation coefficient does not exist within the effective range (NO in step S84), the correction unit 16 corrects the interpolation coefficient to belong to the effective range (step S85). When the interpolation coefficient exceeds the upper limit value of the effective range defined by Expressions (29) to (31), the correction unit 16 corrects the upper limit value to be the interpolation coefficient. Further, when the interpolation coefficient is lower than the lower limit value defined by the equations (29) to (31), the lower limit value is corrected to be the interpolation coefficient. Then, the correction unit 16 outputs the corrected interpolation coefficient to the frequency time conversion unit 24 (step S86).

  FIG. 8 is a flowchart showing a procedure of gain g calculation processing. In the processing in step S85, the gain g may be calculated as described above, and the gain g may be multiplied by an interpolation coefficient to be corrected. The correction unit 16 calculates the ratio (g) between the upper limit (or lower limit) of the effective range output from the effective range determination unit 15 and the interpolation coefficient (step S91), and sets this as the gain g. The correction unit 16 substitutes the calculated g for the initial gain value g ′. The correction unit 16 multiplies the gain g 'by another interpolation coefficient (step S92). This may be performed, for example, for an interpolation coefficient in a frequency band that has the same number of quantization bits as the target interpolation coefficient S (m).

  The correction unit 16 determines whether or not another interpolation coefficient multiplied by the gain g ′ is within the effective range related to the other interpolation coefficient (step S <b> 93). When the correction unit 16 determines that all the other interpolation coefficients multiplied by the gain g ′ are within the effective range related to each interpolation coefficient (YES in step S93), the correction unit 16 converts the gain g ′ to the interpolation coefficient S (m). Is multiplied (step S94), and the process is terminated. On the other hand, when the correction unit 16 determines that at least one of the other interpolation coefficients multiplied by the gain g ′ is not within the effective range related to the other interpolation coefficient (NO in step S93), the correction unit 16 determines the gain g ′. The following processing is performed in order to make changes.

  The correcting unit 16 substitutes n + 1 for the variable n (step S95). Note that the initial value of n is 0. The correcting unit 16 subtracts (n / 10) g from 1.5 times the gain g (initial value gain g ') to calculate a new gain g' (step S96). That is, a process of changing in steps of 10% within a range of ± 50% from the gain g is performed. Also, as the number of quantization bits increases to 2 and 3, (n / 10) g is subtracted from 1.5 times the gain g, and (n / 10) g is subtracted from 1.25 times. The resolution may be increased by narrowing the range. The correcting unit 16 determines whether or not the variable n is 10 (step S97). When the correction unit 16 determines that the variable n is not 10 (NO in Step S97), the correction unit 16 proceeds to Step S92 and multiplies another interpolation coefficient by a new gain g '. In this way, the process of incrementing the variable n and changing the gain g stepwise is repeated.

  When the correction unit 16 determines that n is 10 (YES in step S97), that is, when the gain g is 1.5 g or more and 0.5 g or less, it is determined that correction by the gain g is difficult, and an interpolation coefficient is set. Correction is made to the upper limit (or lower limit) of the effective range (step S98). In the present embodiment, the process of multiplying g by 1.5 is performed in step S96. However, this is merely an example, and an appropriate value may be multiplied.

  The second embodiment is configured as described above, and the other configurations and operations are the same as those of the first embodiment. Therefore, corresponding parts are denoted by the same reference numerals, and detailed description thereof is omitted.

Embodiment 3
FIG. 9 is a block diagram showing the configuration of the signal processing device 20 according to the third embodiment. Each process of the signal processing device 20 according to the third embodiment may be realized by software executed on a personal computer. Hereinafter, the signal processing device 20 will be described as being a personal computer 20. The personal computer 20 is a publicly known computer, and a CPU (Central Processing Unit) 61 via a bus 67, a RAM (Random Access Memory) 62, a storage unit 65 such as a hard disk, an input unit 63, an output unit 64 such as a speaker, the Internet. The communication unit 66 is connectable to a communication network such as the above.

  A computer program for operating the personal computer 20 can also be provided by a portable recording medium 1A such as a CD-ROM, MO, or DVD-ROM as in the third embodiment. Further, the computer program can be downloaded from a server computer (not shown) via the communication unit 66. The contents will be described below.

  A portable recording medium 1A (CD-ROM, MO, DVD-ROM or the like) on which a computer program for causing a reader / writer (not shown) of the personal computer 20 shown in FIG. 9 to select a coefficient and calculate an interpolation coefficient is recorded. Insert this program into the control program of the storage unit 65. Alternatively, such a program may be downloaded from an external server computer (not shown) via the communication unit 66 and installed in the storage unit 65. Such a program is loaded into the RAM 62 and executed. Thereby, it functions as the signal processing apparatus 20 of the present invention as described above.

  The third embodiment is configured as described above, and the other configurations and operations are the same as those of the first and second embodiments. Therefore, corresponding parts are denoted by the same reference numerals, and detailed description thereof is omitted. To do.

Embodiment 4
FIG. 10 is a block diagram showing a hardware configuration of the decoding apparatus according to the fourth embodiment. In the figure, reference numeral 20A denotes a decoding device that decodes an encoded acoustic signal, and is applied to an AAC or ATRAC system other than the MP3 system. Below, the example which applied decoding apparatus 20A to the ATRAC system is demonstrated. The decoding device 20A includes an acoustic signal input unit 21, an unpacking unit 22, an inverse quantization unit 23, a quantization bit number detection unit 11, a butterfly calculation unit 31, which is a calculation unit, a frequency time conversion unit 24, and a frequency band synthesis unit 25. And an acoustic signal output unit 26. The frequency band synthesizer 25 is IQMF.

  The encoded acoustic signal read from the recording medium or the encoded acoustic signal received by the digital tuner is input to the acoustic signal input unit 21, and the input encoded acoustic signal is input to the unpacking unit (demultiplexer) 22. Is output. The unpacking unit 22 unpacks the quantization coefficient, the number of quantization bits, and the scale factor from the frame information of the acoustic signal. The unpacked quantization coefficient, the number of quantization bits, and the scale factor are inversely quantized by the inverse quantization unit 23. From the inverse quantization unit 23, the IMDCT coefficient represented by Expression (32) is output to the quantization bit number detection unit 11 for each frequency band.

  Here, the variable m in Equation 32 is the index of the IMDCT coefficient, i is the index of the inverse quantization frequency band, MK (m) is the quantization coefficient, WL (i) is the number of quantization bits, and SF (i) is the scale. Represents a factor.

  FIG. 11 is a graph for explaining the coefficients in the symmetric band. In the graph of FIG. 11, the horizontal axis represents frequency, the vertical axis represents spectrum, and a 512-sample PCM is divided into two by QMF, which is a band division filter, and a low frequency band block (L) and a high frequency band block. The IMDCT coefficient in (H) is displayed. FIG. 12 is a graph showing the frequency response of QMF. The horizontal axis represents each frequency. When the sampling frequency is 44.1 kHz, π / 2 is 22.05 Hz. The vertical axis represents intensity and the unit is decibel. As shown in FIG. 12, the response from the center of the frequency (band division boundary) to the low frequency side (H0 (z)) and the high frequency side (H1 (z)) are designed to have symmetrical characteristics. Therefore, the frequency where the stopband attenuation between H0 (z) and H1 (z) is about −70 dB is referred to as a low-frequency boundary and a high-frequency boundary. The division boundary between block (L) and block (H) is between I (255) and I (256). The low frequency boundary is described as being between I (191) and I (192), and the high frequency boundary is described as being between I (319) and I (320). The quantization bit number detection unit 11 determines whether or not the input coefficient of one frequency band exists between the division boundary and the low band boundary of one frequency band, and the number of quantization bits is 0. It is determined whether or not. If the coefficient is not included between the division boundary and the low frequency boundary of one frequency band, the aliasing distortion is very small, and the coefficient is output to the frequency time conversion unit 24 without going through the butterfly operation unit 31. To do. When the quantization bit number detection unit 11 determines that the coefficient quantization bit number is 0, the quantization bit number detection unit 11 determines the quantization bit number of the high-frequency coefficient folded around the band division boundary. That is, the quantization bit number detection unit 11 does not detect all the coefficients in the band, but only the coefficients within a predetermined frequency are detected and the butterfly calculation unit 31 performs the calculation. This predetermined frequency is, for example, as described above, a frequency (I) that is equal to or higher than a predetermined intensity (−70 dB) based on the band division boundary (between I (255) and I (256)) and the frequency response of FIG. (191) and I (192)). The range of coefficients detected by the quantization bit number detection unit 11 may be set as appropriate according to the characteristics of the band division filter, sound quality, power consumption, and the like. The range of coefficients to be detected is not illustrated in the quantization bit number detection unit 11 beforehand. Just store it in memory. In the present embodiment, the intensity based on the frequency response is a frequency of −70 dB or more. However, the present invention is not limited to this. For example, in the frequency response on the high frequency side, the intensity decreases as the frequency decreases from the band division boundary. However, the frequency may be higher than the intensity (-74 dB in FIG. 12) that takes the extreme value first. In this case, the coefficient detected by the quantization bit number detection unit 11 includes a frequency corresponding to the division boundary of the band and a frequency corresponding to an intensity that first takes an extreme value through the division boundary in the high frequency response. Coefficients that fall between them.

  When the quantization bit number detection unit 11 determines that the quantization bit number of the higher frequency coefficient folded around the band division boundary is 0, the folding component is also 0, and no aliasing distortion occurs. The bit number detection unit 11 outputs the coefficient of one frequency band to the frequency time conversion unit 24 without passing through the butterfly calculation unit 31.

  If either one of the coefficients or the number of quantization bits of the high frequency side coefficient folded around the band division boundary is not 0, the butterfly operation unit calculates the coefficient and the high frequency side coefficient folded around the band division boundary. To 31.

  The butterfly computation unit 31 performs a butterfly computation using the high-frequency side coefficient and the butterfly coefficient folded around the coefficient and the band division boundary, and outputs the coefficient after the butterfly computation to the frequency time conversion unit 24. The butterfly calculation is performed in order to reduce the aliasing distortion of the band-divided boundary band. The butterfly coefficient will be described later. The frequency time conversion unit 24 performs an IMDCT process and converts it into a time axis acoustic signal. Further, the inversely converted acoustic signal is subjected to band synthesis by the frequency band synthesis unit 25 using IQMF, which is a band synthesis filter, and then output to the acoustic signal output unit 26.

  Centering on I (255) and I (256) shown in FIG. 11, the low frequency side 64 IMDCT coefficients and the high frequency side 64 IMDCT coefficients up to about −70 dB of the frequency response of the QMF shown in FIG. On the other hand, the butterfly operation for canceling the aliasing distortion component is performed on I (192) to I (255) up to the low frequency boundary and I (256) to I (319) up to the high frequency boundary. Here, attention is focused on the coefficients I (193) and I (317) of one frequency band block (L). It is assumed that the number of quantization bits of I (193) and I (317) is zero.

  For I (193), a comparison is made with I (318) at the symmetrical position of the dividing boundary. The quantization bit number detection unit 11 detects the number of quantization bits of the coefficient I (318) and determines whether or not it is zero. Here, since the number of quantization bits of the coefficient I (318) is not 0, a butterfly operation using I (193) and I (318) is performed.

  FIG. 13 is an explanatory diagram showing the procedure of the butterfly calculation. The butterfly operation unit 31 obtains outputs Out1 and Out2 by substituting the inputs In1 and In2 into the equation (33). In the above example, the coefficients I (193) and I (318) are input, and the coefficients I ′ (193) and I ′ (318) are output from the butterfly operation unit 31 to the frequency time conversion unit 24.

Here, the coefficient c _i in the butterfly calculation is a coefficient determined according to the interval between the input In1 or the input In2 and the division boundary, and is obtained from the frequency response characteristic of the QMF in FIG. FIG. 14 is an explanatory diagram showing a record layout of the butterfly calculation coefficient storage unit 310. Butterfly computation coefficient storage unit 310 (not shown) is stored in the internal butterfly operation unit 31, the coefficient c _i in association with the distance i between the input In1 or input In2 and division boundary are stored. In the case of the coefficients I (193) and I (318), the interval with the divided region i = 62, and c _i = −0.00054932 is selected. The coefficients cs _i and ca _i are calculated from the coefficient c _{i according} to the equation (33). The calculated cs _i and ca _i may be stored in the butterfly calculation unit 31 in advance. On the other hand, in the case of the coefficient I (194), a comparison is made with I (317) at the symmetrical position of the division boundary. Here, assuming that the quantization bit number of I (317) is 0, the quantization bit number detection unit 11 passes the coefficients I (194) and I (317) as they are in the frequency time without passing through the butterfly operation unit 31. The data is output to the conversion unit 24.

  FIG. 15 is a flowchart showing the procedure of the butterfly calculation process. First, the quantization bit number detection unit 11 performs initial selection of coefficients (step S161). The quantization bit number detection unit 11 determines whether or not the selected coefficient exists between the division boundary and the low band boundary of one frequency band (step S162).

If the quantization bit number detection unit 11 determines that the selected coefficient exists between the division boundary and the low band boundary of one frequency band (YES in step S162), the quantization bit number of the selected coefficient is It is determined whether it is 0 (step S163). When the quantization bit number detection unit 11 determines that the number of quantization bits of the selected coefficient is 0 (YES in step S163), the quantization bit number of the coefficient in the high frequency band folded around the division boundary Whether or not is 0 is determined (step S164). When the quantization bit number detection unit 11 determines that the quantization bit number of the coefficient in the high frequency band turned around the division boundary is not 0 (NO in step S164), the selected coefficient and the high frequency band The side frequency band coefficient is output to the butterfly computing unit 31. The butterfly calculation unit 31 performs the butterfly calculation using the coefficient of one frequency band and the coefficient of the frequency band on the high frequency side (step S166). Specifically, the butterfly calculation unit 31 reads the expression 33, and substitutes the coefficient C _i read from the butterfly calculation coefficient storage unit 310, the coefficient of one frequency band, and the coefficient of the high frequency band into the read expression. The coefficient after the calculation and the coefficient on the high frequency side are obtained. The frequency band coefficient after the butterfly calculation and the high frequency band coefficient are output to the frequency time conversion unit 24.

  If the quantization bit number detection unit 11 determines that the number of quantization bits of the selected coefficient is not 0 (NO in step S163), the quantization of the coefficient in the high frequency band that is turned back around the division boundary It is determined whether or not the number of bits is 0 (step S165). When the quantization bit number detection unit 11 determines that the quantization bit number of the high frequency band coefficient folded around the division boundary is 0 (YES in step S165), the selected coefficient and the high frequency side Are output to the butterfly computing unit 31 and the processing described in step S166 is performed.

  In the case of NO in step S162, in the case of YES in step S164, and in the case of NO in step S165, the process of step S166, that is, the butterfly calculation process, is skipped. In this case, the selected coefficient passes through the butterfly calculation unit 31. Without being output to the frequency time conversion unit 24. Thereafter, the butterfly calculation unit 31 determines whether or not the processing has been completed for all the coefficients of block (L) (step S167). When it is determined that the processing has not been completed for all the coefficients (NO in step S167), the butterfly calculation unit 31 changes the processes in steps S162 to S166 to coefficients that have not been completed, and repeats the processes in and after step S162. On the other hand, when the butterfly calculation unit 31 determines that the processing has been completed for all the coefficients (YES in step S167), the series of processing ends.

  The fourth embodiment is configured as described above, and the other configurations and operations are the same as those of the first to third embodiments. Therefore, the corresponding parts are denoted by the same reference numerals and detailed description thereof is omitted. To do.

Embodiment 5
FIG. 16 is a block diagram showing a hardware configuration of decoding apparatus 20 and decoding apparatus 20A as signal processing apparatuses according to the fifth embodiment. FIG. 16 is a combination of the decoding device 20 according to the first embodiment and the decoding device 20A according to the fourth embodiment, to which a method determining unit 32 is added. The method determination unit 32 determines the encoding method of the input acoustic signal and outputs the input acoustic signal according to the encoding method to either the decoding device 20 or the decoding device 20A. When determining that the input encoding method is a method for performing a butterfly operation to reduce aliasing distortion, the method determination unit 32 outputs an acoustic signal to the decoding device 20 including the interpolation processing unit 1. On the other hand, when it is determined that the input coding method is not a method for performing butterfly computation to reduce aliasing distortion, an acoustic signal is output to the decoding device 20A including the butterfly computation unit 31. For example, when the method determination unit 32 determines that the encoding method is the MP3 method by referring to a header or the like in the input sound signal, the method determination unit 32 outputs the input sound signal to the decoding device 20. Thereafter, the interpolation processing described in the first embodiment is performed on the input acoustic signal. On the other hand, when determining that the encoding method of the input sound signal is the AAC method or the ATRAC method, the method determining unit 32 outputs the input sound signal to the decoding device 20A. Thereafter, the butterfly calculation process described in the fourth embodiment is performed on the input acoustic signal.

  The fifth embodiment has the above-described configuration, and the other configurations and operations are the same as those of the first to fourth embodiments. Therefore, the corresponding parts are denoted by the same reference numerals and detailed description thereof is omitted. To do.

Embodiment 6
The processing according to the fifth embodiment may be realized as software processing using the personal computer shown in FIG. FIG. 17 is a block diagram showing the configuration of the signal processing device 20 according to the sixth embodiment. A computer program for operating the personal computer 20 as a signal processing device can be provided by a portable recording medium 1A such as a CD-ROM, MO, or DVD-ROM as in the sixth embodiment. . Further, the computer program can be downloaded from a server computer (not shown) via the communication unit 66. The contents will be described below.

  A portable recording medium 1A (CD-ROM, MO or recording medium) on which a computer program for causing a reader / writer (not shown) of the personal computer 20 shown in FIG. 17 to determine a method, detect the number of quantization bits, and perform a butterfly operation is recorded. DVD-ROM or the like) is inserted, and this program is installed in the control program of the storage unit 65. Alternatively, such a program may be downloaded from an external server computer (not shown) via the communication unit 66 and installed in the storage unit 65. Such a program is loaded into the RAM 62 and executed. Thereby, it functions as the signal processing apparatus 20 of the present invention as described above.

  The sixth embodiment has the above-described configuration, and the other configurations and operations are the same as those of the first to fifth embodiments. Therefore, the corresponding parts are denoted by the same reference numerals and detailed description thereof is omitted. To do.

It is a block diagram which shows the hardware constitutions of the decoding apparatus which is a signal processing apparatus. It is a graph which shows the change of IMDCT coefficient with respect to frequency. It is a block diagram which shows the hardware constitutions of an interpolation process part. It is a flowchart which shows the procedure of an interpolation process. 6 is a block diagram illustrating a hardware configuration of an interpolation processing unit according to Embodiment 2. FIG. It is a graph for demonstrating an effective range. It is a flowchart which shows the procedure of a correction process. It is a flowchart which shows the procedure of the calculation process of a gain. FIG. 10 is a block diagram illustrating a configuration of a signal processing device according to a third embodiment. FIG. 10 is a block diagram illustrating a hardware configuration of a decoding device according to a fourth embodiment. It is a graph for demonstrating the coefficient in a symmetrical band. It is a graph which shows the frequency response of QMF. It is explanatory drawing which shows the procedure of a butterfly calculation. It is explanatory drawing which shows the record layout of a butterfly calculation coefficient memory | storage part. It is a flowchart which shows the procedure of a butterfly calculation process. FIG. 10 is a block diagram showing a decoding device as a signal processing device according to Embodiment 5 and a hardware configuration of the decoding device. FIG. 10 is a block diagram illustrating a configuration of a signal processing device according to a sixth embodiment. It is a block diagram which shows the hardware constitutions of the conventional decoding apparatus.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Interpolation processing part 11 Quantization bit number detection part 12 Interpolation determination part 13 Selection part 14 Calculation part 15 Effective range determination part 16 Correction | amendment part 20, 20A Decoding apparatus (signal processing apparatus)
DESCRIPTION OF SYMBOLS 21 Acoustic signal input part 22 Unpacking part 23 Dequantization part 24 Frequency time conversion part 25 Frequency band synthetic | combination part 26 Acoustic signal output part 31 Butterfly calculation part 32 Method judgment part 1A Portable recording medium

Claims (9)

In a signal processing method for processing an acoustic signal obtained by dequantizing a band-divided encoded acoustic signal,
A detection step for detecting the number of quantization bits at the time of quantizing the frequency band coefficient of the dequantized acoustic signal;
A coefficient in which the number of quantization bits of a coefficient in one frequency band detected by the detection step is zero and a coefficient at a folding position in another frequency band adjacent to the one frequency band detected by the detection step And a calculation step of performing a butterfly operation on the coefficient of the one frequency band and the coefficient of the other frequency band when the number of quantization bits is non-zero . In a signal processing apparatus that processes an acoustic signal obtained by dequantizing a band-divided encoded acoustic signal,
A detection unit for detecting the number of quantization bits at the time of quantizing the frequency band coefficient of the dequantized acoustic signal;
A coefficient in which the number of quantization bits of a coefficient in one frequency band detected by the detection unit is zero , and a coefficient at a folding position in another frequency band adjacent to the one frequency band detected by the detection unit A signal processing apparatus comprising: an arithmetic unit that performs a butterfly operation on the coefficient of the one frequency band and the coefficient of the other frequency band when the number of quantization bits is non-zero . The detector is
The coefficient is configured to detect the number of quantization bits at the time of quantization of a coefficient existing within a predetermined frequency from the boundary of the band division among the coefficients of the frequency band of the inversely quantized acoustic signal. Item 3. The signal processing device according to Item 2. The predetermined frequency is
The signal processing apparatus according to claim 3, wherein the signal processing apparatus has a frequency between a frequency at a boundary of the band division and a frequency having a predetermined intensity or more based on a frequency response of a band division filter that performs band division. In a program for processing by a computer an acoustic signal obtained by dequantizing a band-divided encoded acoustic signal,
On the computer,
A detection step for detecting the number of quantization bits at the time of quantizing the frequency band coefficient of the dequantized acoustic signal;
A coefficient in which the number of quantization bits of a coefficient in one frequency band detected by the detection step is zero and a coefficient at a folding position in another frequency band adjacent to the one frequency band detected by the detection step When the number of quantization bits is non-zero , a program for executing a butterfly operation on the coefficient of the one frequency band and the coefficient of the other frequency band . In a signal processing method for processing an acoustic signal obtained by dequantizing a band-divided encoded acoustic signal,
A method determination step for determining whether or not the encoding method of the encoded acoustic signal is a method for performing a butterfly operation;
A detection step for detecting the number of quantization bits at the time of quantizing the frequency band coefficient of the dequantized acoustic signal;
It is determined that the encoding method is not a method for performing a butterfly operation in the method determining step , and the number of quantized bits of the coefficient in one frequency band detected by the detecting step is zero , and the detecting step When the number of quantization bits of the coefficient at the folding position in the other frequency band adjacent to the one frequency band detected by the non-zero is non-zero , the butterfly for the coefficient of the one frequency band and the coefficient of the other frequency band A signal processing method comprising: an operation step for performing an operation. In a signal processing apparatus that processes an acoustic signal obtained by dequantizing a band-divided encoded acoustic signal,
A method determination unit for determining whether or not the encoding method of the encoded acoustic signal is a method for performing a butterfly operation;
A detection unit for detecting the number of quantization bits at the time of quantizing the frequency band coefficient of the dequantized acoustic signal;
The method determining unit determines that the encoding method is not a method for performing a butterfly operation , and the number of quantized bits of a coefficient in one frequency band detected by the detecting unit is zero , and the detecting unit When the number of quantization bits of the coefficient at the folding position in the other frequency band adjacent to the one frequency band detected by the non-zero is non-zero , the butterfly for the coefficient of the one frequency band and the coefficient of the other frequency band A signal processing apparatus comprising: an arithmetic unit that performs an arithmetic operation. A selection unit that selects a plurality of coefficients from coefficients in the frequency band of the dequantized acoustic signal;
When it is determined that the encoding method is a method for performing a butterfly operation in the method determination unit, a coefficient that is not selected by the selection unit and is not selected by the selection unit by an interpolation method using a plurality of coefficients selected by the selection unit. The signal processing apparatus according to claim 7 , further comprising: a calculating unit that calculates an interpolation coefficient for a coefficient having a detected number of quantization bits of zero. In a program for processing by a computer an acoustic signal obtained by dequantizing a band-divided encoded acoustic signal,
On the computer,
A method determination step for determining whether or not the encoding method of the encoded acoustic signal is a method for performing a butterfly operation;
A detection step for detecting the number of quantization bits at the time of quantizing the frequency band coefficient of the dequantized acoustic signal;
It is determined that the encoding method is not a method for performing a butterfly operation in the method determining step , and the number of quantized bits of the coefficient in one frequency band detected by the detecting step is zero , and the detecting step When the number of quantization bits of the coefficient at the folding position in the other frequency band adjacent to the one frequency band detected by the non-zero is non-zero , the butterfly for the coefficient of the one frequency band and the coefficient of the other frequency band A program that executes a calculation step that performs a calculation.

Images