CN100527850C

CN100527850C - Method and apparatus for synthetic subband filtering in an audio signal decoder

Info

Publication number: CN100527850C
Application number: CNB2006100088592A
Authority: CN
Inventors: 洪智玮; 张志舷; 蔡宪明
Original assignee: Quanta Computer Inc
Current assignee: Quanta Computer Inc
Priority date: 2006-02-22
Filing date: 2006-02-22
Publication date: 2009-08-12
Anticipated expiration: 2026-02-22
Also published as: CN101026763A

Abstract

The invention provides a device and a method for synthesis subband filtering in an audio signal decoder. The apparatus according to the invention is implemented for M groups of signals. Each group of signals respectively comprises N sub-band sampling signals. The apparatus includes a processor for processing an ith signal of the M signals, where i is an integer ranging from 0 to (M-1). The processor includes a discrete cosine transform module and a generation module. The discrete cosine transform module is used for converting the N sub-band sampling signals in the ith group of signals into N transformed vectors. If i is an odd number, the conversion module multiplies a (2j-1) th subband sample signal of the N subband sample signals by minus one, wherein j is an integer ranging from 1 to (N/2). The generation module generates N pulse code modulation signals based on the N converted vectors.

Description

Method and apparatus for synthetic subband filtering in an audio signal decoder

技术领域 technical field

本发明关于一种用于音频信号解码器中的合成次频带(subband)滤波的方法及装置。并且特别地，根据本发明的合成次频带滤波方法及装置应用于音频信号解码器中。The present invention relates to a method and device for synthetic subband filtering in an audio signal decoder. And especially, the synthesis sub-band filtering method and device according to the present invention are applied in audio signal decoders.

背景技术 Background technique

由国际标准组织所制定的MPEG(Motion Pictures Experts Group)音频信号标准提供了一种将音频信号编/解码的演算法，可大幅降低传输音频信号的频宽需求，并可提供低失真的信号质量。目前的MPEG音频信号标准包含Layer I，Layer II以及Layer III三级不同的处理方法。The MPEG (Motion Pictures Experts Group) audio signal standard formulated by the International Standards Organization provides an algorithm for encoding/decoding audio signals, which can greatly reduce the bandwidth requirements for transmitting audio signals and provide low-distortion signal quality . The current MPEG audio signal standard includes three different processing methods: Layer I, Layer II, and Layer III.

请参阅图1。图1表示MPEG-1 Layer III(MP3)标准中将一音频帧(audioframe)解码的流程图。步骤S11为解开(unpack)该音频帧的首标(header)，并且将该音频帧中的附属信息(side information)解码。步骤S12为将以霍夫曼编码(Huffman encoding)方式压缩的音频信号信息解码，并对解码后的频谱信号进行逆量化(re-quantization)处理。步骤S13为根据附属资料中的音频信号模式对频谱信号进行立体声处理(joint stereo processing)。步骤S14为对频谱信号进行频迭消除(alias reduction)。步骤S15为对频谱信号进行反向改良式离散余弦转换(inverse modified discrete cosinetransform，IMDCT)。步骤S16为将IMDCT后的取样信号频率反转(frequencyinversion)。步骤S17为以次频带合成滤波(subband synthesis filtering)方法合成出脉冲代码调制(pulse code modulation，PCM)信号。步骤S18为输出PCM信号，完成将该音频帧解码的操作。See Figure 1. Fig. 1 shows a flowchart of decoding an audio frame (audioframe) in the MPEG-1 Layer III (MP3) standard. Step S11 is to unpack (unpack) the header of the audio frame, and decode the side information in the audio frame. Step S12 is to decode the audio signal information compressed by Huffman encoding, and perform inverse quantization (re-quantization) processing on the decoded spectral signal. Step S13 is to perform joint stereo processing on the spectrum signal according to the audio signal mode in the attached materials. Step S14 is performing alias reduction on the spectrum signal. Step S15 is performing an inverse modified discrete cosine transform (IMDCT) on the spectrum signal. Step S16 is frequency inversion of the sampling signal after IMDCT. Step S17 is to synthesize a pulse code modulation (PCM) signal by means of subband synthesis filtering. Step S18 is to output the PCM signal, and complete the operation of decoding the audio frame.

于MP3标准中，经过步骤S15的IMDCT后的取样信号为576个，其中包含有32个次频带的取样信号，每个次频带又各自包含18个取样信号。步骤S16的频率反转为将第奇数个次频带中的第奇数个取样信号乘上负号。In the MP3 standard, there are 576 sampled signals after the IMDCT in step S15, including 32 sub-band sampled signals, and each sub-band includes 18 sampled signals. The frequency inversion in step S16 is to multiply the odd-numbered sampling signal in the odd-numbered sub-band by a negative sign.

请参阅图2。图2为表示在先前技术中进行频率反转的流程图。该流程由步骤S20开始。步骤S21为读取该32个次频带中的一次频带。步骤S22为判断于步骤S21中读取的次频带是否为该32个次频带中的第奇数个次频带，如果为是，则执行步骤S23，如果为否，则执行步骤S27。步骤S23为读取该次频带中18个信号中的一取样信号。步骤S24为判断于步骤S23中读取的取样信号是否为该18个取样信号中的第奇数个取样信号，如果为是，则执行步骤S25，如果为否，则执行步骤S26。步骤S25为将步骤S23中读取的取样信号乘上负号。步骤S26为判断是否该次频带中的18个取样信号皆已完成，如果为是，则执行步骤S27，如果为否，则执行步骤S23。步骤S27为判断是否该32个次频带皆已完成，如果为是，则执行步骤S28，如果为否，则执行步骤S21。步骤S28为结束该流程。See Figure 2. FIG. 2 is a flowchart showing frequency inversion in the prior art. The process starts with step S20. Step S21 is to read the primary frequency band in the 32 sub-frequency bands. Step S22 is to judge whether the sub-band read in step S21 is the odd-numbered sub-band among the 32 sub-bands, if yes, execute step S23, if no, execute step S27. Step S23 is to read a sampled signal among the 18 signals in the sub-band. Step S24 is to judge whether the sampled signal read in step S23 is the odd-numbered sampled signal among the 18 sampled signals, if yes, then execute step S25, if no, then execute step S26. Step S25 is to multiply the sampled signal read in step S23 by a minus sign. Step S26 is to judge whether all the 18 sampled signals in the sub-band have been completed, if yes, execute step S27, and if no, execute step S23. Step S27 is to judge whether the 32 sub-bands have been completed, if yes, execute step S28, if no, execute step S21. Step S28 is to end the process.

如图2所示，由于先前技术中的频率反转流程包含多次对存储器的读取、存入以及多个判断步骤，频率反转的流程耗费相当多的运算时间，降低了将MP3音频帧解码的效率。As shown in Figure 2, since the frequency inversion process in the prior art includes multiple times of reading, storing and multiple judgment steps to the memory, the process of frequency inversion consumes a considerable amount of computing time, reducing the MP3 audio frame decoding efficiency.

合成次频带滤波为MP3解码过程中的最后一个步骤。公知的离散余弦转换(discrete cosine transform，DCT)技术已提出了利用32点DCT将该32个次频带取样信号转换成32个转换后的向量的方法。该方法合成次频带滤波的步骤为依序将该18组经过IMDCT及频率反转的次频带取样信号转换为18组PCM信号。Synthetic subband filtering is the final step in the MP3 decoding process. The well-known discrete cosine transform (DCT) technique has proposed a method of converting the 32 sub-band sampled signals into 32 transformed vectors using a 32-point DCT. The step of synthesizing sub-band filtering in the method is to sequentially convert the 18 sub-band sampling signals after IMDCT and frequency inversion into 18 groups of PCM signals.

请参阅图3。图3为表示先前技术中一合成次频带滤波方法的流程图。该18组经过IMDCT及频率反转的次频带取样信号中的每一组信号皆包含32个次频带取样信号。步骤S31为将该组正在被处理中的32个次频带取样信号输入合成次频带滤波的方法或装置。步骤S32为将该32个次频带取样信号转换成32个转换后的向量(vector)。步骤S33为根据该32个转换后的向量计算并合成出32个PCM信号。See Figure 3. FIG. 3 is a flow chart showing a synthetic subband filtering method in the prior art. Each of the 18 sets of IMDCT-inverted sub-band samples includes 32 sub-band samples. Step S31 is a method or device for inputting the group of 32 sub-band sampling signals being processed into a synthetic sub-band filter. Step S32 is converting the 32 sub-band sampled signals into 32 converted vectors. Step S33 is to calculate and synthesize 32 PCM signals according to the 32 transformed vectors.

因此，本发明提供一种次频带合成滤波的方法及装置。根据本发明的方法及装置将先前技术中的频率反转方法结合到次频带合成滤波方法中。藉此，可大幅提升将MP3音频帧解码的速度。此外，本发明亦结合频率反转方法与此以DCT产生转换后的向量的方式，藉以提升将MP3音频帧解码的效率。Therefore, the present invention provides a method and device for sub-band synthesis filtering. The method and device according to the present invention combine the frequency inversion method in the prior art into the sub-band synthesis filtering method. Thereby, the speed of decoding MP3 audio frames can be greatly improved. In addition, the present invention also combines the frequency inversion method with DCT to generate converted vectors, so as to improve the efficiency of decoding MP3 audio frames.

发明内容 Contents of the invention

本发明的主要目的在于提供一种用于音频信号解码器中的合成次频带滤波的方法及装置。根据本发明的方法及装置将先前技术中的频率反转方法结合到次频带合成滤波方法中，藉此大幅提升将MP3音频帧解码的速度。The main object of the present invention is to provide a method and apparatus for synthetic subband filtering in an audio signal decoder. According to the method and device of the present invention, the frequency inversion method in the prior art is combined into the sub-band synthesis filter method, thereby greatly increasing the speed of decoding MP3 audio frames.

根据本发明的一较佳具体实施例的合成次频带滤波装置针对M组信号执行，该M组信号中的每一组信号分别包含N个次频带取样信号。该合成次频带滤波装置包含一处理器。该处理器用以处理该M组信号中的第i组信号，其中i为一范围在0至(M-1)之间的整数。该处理器包含一离散余弦转换模件与一产生模件。该离散余弦转换模件利用离散余弦转换，将该第i组信号中的N个次频带取样信号转换为N个转换后的向量。若i为奇数，于产生该N个转换后的向量的过程中，该离散余弦转换模件将该N个次频带取样信号中的第(2j-1)个次频带取样信号乘以负一，其中j为一范围在1至(N/2)之间的整数。该产生模件则是根据该N个转换后的向量，产生N个PCM信号。According to a preferred embodiment of the present invention, the device for synthesizing sub-band filtering is performed on M groups of signals, and each group of signals in the M groups of signals respectively includes N sub-band sampling signals. The synthetic subband filtering device includes a processor. The processor is used for processing the i-th group of signals among the M groups of signals, wherein i is an integer ranging from 0 to (M-1). The processor includes a discrete cosine transform module and a generation module. The discrete cosine transform module transforms the N sub-band sampling signals in the i-th group of signals into N transformed vectors by using discrete cosine transform. If i is an odd number, during the process of generating the N converted vectors, the discrete cosine transform module multiplies the (2j-1)th sub-band sample signal among the N sub-band sample signals by minus one, Where j is an integer ranging from 1 to (N/2). The generating module generates N PCM signals according to the N converted vectors.

举例而言，若该M组信号为MP3标准中经过IMDCT后的次频带取样信号，则M等于18，N等于32。For example, if the M groups of signals are sub-band sampling signals after IMDCT in the MP3 standard, M is equal to 18, and N is equal to 32.

关于本发明的优点与精神可以藉由以下的发明详述及附图得到进一步的了解。The advantages and spirit of the present invention can be further understood through the following detailed description of the invention and the accompanying drawings.

附图说明 Description of drawings

图1为表示先前技术在MPEG-1 Layer III(MP3)标准中将一音频帧(audioframe)解码的流程图。FIG. 1 is a flowchart showing prior art decoding an audio frame in the MPEG-1 Layer III (MP3) standard.

图2为表示在先前技术中进行频率反转的流程图。FIG. 2 is a flowchart showing frequency inversion in the prior art.

图3为表示先前技术中一合成次频带滤波方法的流程图。FIG. 3 is a flow chart showing a synthetic subband filtering method in the prior art.

图4为根据本发明的一较佳具体实施例的合成次频带滤波装置的方块图。FIG. 4 is a block diagram of a synthetic subband filtering device according to a preferred embodiment of the present invention.

图5为表示本发明结合频率反转与合成次频带滤波方法的概念。FIG. 5 shows the concept of the present invention combining frequency inversion and synthetic subband filtering method.

图6为根据本发明的一较佳具体实施例的合成次频带滤波方法的流程图。FIG. 6 is a flowchart of a synthetic subband filtering method according to a preferred embodiment of the present invention.

主要元件符号说明Description of main component symbols

S11-S18：流程步骤 S20-S28：流程步骤S11-S18: Process steps S20-S28: Process steps

S31-S33：流程步骤 40：合成次频带滤波装置S31-S33: Process steps 40: Composite sub-band filtering device

41：处理器 411：DCT模件41: Processor 411: DCT module

412：产生模件 411A：第一DCT单元412: Generate a module 411A: The first DCT unit

411A1：第一判断模件 411A2：第二判断模件411A1: first judgment module 411A2: second judgment module

411A3：计算模件 411B：第二DCT单元411A3: Calculation module 411B: Second DCT unit

S4A：N个次频带取样信号 S4B：N个中间信号S4A: N sub-band sampling signals S4B: N intermediate signals

S4C：N个转换后的向量 S4D：N个PCM信号S4C: N converted vectors S4D: N PCM signals

S601-S609：流程步骤S601-S609: Process steps

具体实施方式 Detailed ways

请参阅图4。图4为根据本发明的一较佳具体实施例的合成次频带滤波装置40的方块图。合成次频带滤波装置40针对M组信号执行，该M组信号中的每一组信号分别包含N个次频带取样信号S4A。举例而言，若该M组信号为MPEG-1 Layer III标准中经过IMDCT后的次频带取样信号，则M等于18，N等于32。See Figure 4. FIG. 4 is a block diagram of a synthetic subband filtering device 40 according to a preferred embodiment of the present invention. The synthesized sub-band filtering device 40 is performed on M groups of signals, each of which includes N sub-band sampling signals S4A. For example, if the M groups of signals are sub-band sampling signals after IMDCT in the MPEG-1 Layer III standard, then M is equal to 18 and N is equal to 32.

合成次频带滤波装置40包含一处理器41。处理器41用以处理该M组信号中的第i组信号，其中i为一范围在0至(M-1)之间的整数。处理器41包含一DCT模件411与一产生模件412。DCT模件411利用DCT，将该第i组信号中的N个次频带取样信号S4A转换为N个转换后的向量S4C。若i为奇数，于产生该N个转换后的向量S4C的过程中，DCT模件411将该N个次频带取样信号S4A中的第奇数个次频带取样信号乘以负一，亦即将第(2j-1)个次频带取样信号乘以负一，其中j为一范围在1至(N/2)之间的整数。产生模件412根据该N个转换后的向量S4C，产生N个PCM信号S4D。The synthesized subband filtering device 40 includes a processor 41 . The processor 41 is used for processing the i-th group of signals in the M groups of signals, wherein i is an integer ranging from 0 to (M−1). The processor 41 includes a DCT module 411 and a generation module 412 . The DCT module 411 uses DCT to convert the N sub-band sampled signals S4A in the i-th group of signals into N converted vectors S4C. If i is an odd number, in the process of generating the N converted vectors S4C, the DCT module 411 multiplies the odd-numbered sub-band sampling signal among the N sub-band sampling signals S4A by minus one, that is, the ( 2j-1) sub-band sampled signals are multiplied by minus one, where j is an integer ranging from 1 to (N/2). The generating module 412 generates N PCM signals S4D according to the N transformed vectors S4C.

以MPEG-1 Layer III标准的音频信号信号为例，由于直接以32点DCT将该18组信号中的32个次频带取样信号分别转换为32个PCM信号太复杂，因此可利用分解递回的运算方式，将32点DCT分解为8个4点DCT，以简化计算的复杂度。如本领域技术人员所知，32点DCT的方程式可表示如下：Taking the audio signal signal of MPEG-1 Layer III standard as an example, it is too complicated to convert the 32 sub-band sampling signals in the 18 groups of signals into 32 PCM signals directly by 32-point DCT, so the decomposed and recursive method can be used In the operation mode, the 32-point DCT is decomposed into eight 4-point DCTs to simplify the calculation complexity. As known to those skilled in the art, the equation for a 32-point DCT can be expressed as follows:

$X (n) = Σ_{k = 0}^{31} x (k) C_{64}^{(2 k + 1) n} for n = 0,1, . . ., 31,$ ......(式一) $x (no) = Σ_{k = 0}^{31} x (k) C_{64}^{(2 k + 1) no} for n = 0,1, . . ., 31,$ ...... (Formula 1)

其中的x(k)与X(n)分别为该32点DCT的输入信号与输出信号，该32个次频带取样信号的第k个次频带取样信号为x(k-1)，k为范围在1至32间的整数，并且 $C_{64}^{(2 k + 1) n} = \cos (\frac{nπ (2 k + 1)}{64}) .$ Among them, x(k) and X(n) are the input signal and output signal of the 32-point DCT respectively, and the kth sub-band sampling signal of the 32 sub-band sampling signals is x(k-1), and k is the range an integer between 1 and 32, and $C_{64}^{(2 k + 1) no} = \cos (\frac{nπ (2 k + 1)}{64}) .$

式一中的32点DCT的X(n)可被分解为两个16点DCT的F1(n)与F2(n)：The X(n) of the 32-point DCT in formula 1 can be decomposed into two 16-point DCTs F1(n) and F2(n):

X(2n)＝F1(n) for n＝0，1，...，15X(2n)=F1(n) for n=0, 1,..., 15

，......(式二), ... (Formula 2)

X(2n+1)＝F2(n)+F2(n+1) for n＝0，1，...，15X(2n+1)＝F2(n)+F2(n+1) for n＝0,1,...,15

其中，in,

$F f 11 ((n no)) = = {Σ Σ}_{k k = = 00}^{1515} f f 11 ((k k)) {C C}_{3232}^{((22 k k + + 11)) n no} for n for n = = 0,1 0,1,, . . . . . .,, 1515,,$

......(式三)... (Formula 3)

$F f 22 ((n no)) = = {Σ Σ}_{k k = = 00}^{1515} f f 22 ((k k)) {C C}_{3232}^{((22 k k + + 11)) n no} for n for n = = 0,1 0,1,, . . . . . .,, 1515$

其中，in,

f1(k)＝x(k)+x(31-k) for k＝0，1，...，15f1(k)=x(k)+x(31-k) for k=0, 1,..., 15

......(式四)...(Formula 4)

$f f 22 ((k k)) = = ((\frac{11}{22} {C C}_{6464}^{((22 k k + + 11))})) [[x x ((k k)) - - x x ((3131 - - k k))]] for k for k = = 0,1 0,1,, . . . . . .,, 1515$

并且 $C_{32}^{(2 k + 1) n} = \cos (\frac{nπ (2 k + 1)}{32}) .$ and $C_{32}^{(2 k + 1) no} = \cos (\frac{nπ (2 k + 1)}{32}) .$

式三中的16点DCT的F1(n)可再被分解为两个8点DCT的G1(n)与G2(n)：The F1(n) of the 16-point DCT in Equation 3 can be decomposed into two G1(n) and G2(n) of the 8-point DCT:

F1(2n)＝G1(n) for n＝0，1，...，7F1(2n)＝G1(n) for n＝0,1,...,7

，......(式五), ... (Form 5)

F1(2n+1)＝G2(n)+G2(n+1) for n＝0，1，...，7F1(2n+1)＝G2(n)+G2(n+1) for n＝0,1,...,7

其中，in,

$G G 11 ((n no)) = = {Σ Σ}_{k k = = 00}^{77} g g 11 ((k k)) {C C}_{1616}^{((22 k k + + 11)) n no} for n for n = = 0,1 0,1,, . . . . . .,, 77$

，......(式六), ... (Form 6)

$G G 22 ((n no)) = = {Σ Σ}_{k k = = 00}^{77} g g 22 ((k k)) {C C}_{1616}^{((22 k k + + 11)) n no} for n for n = = 0,1 0,1,, . . . . . .,, 77$

其中，in,

g1(k)＝f1(k)+f1(15-k) for k＝0，1，...，7g1(k)=f1(k)+f1(15-k) for k=0, 1,..., 7

...(式七)...(Equation 7)

$g g 22 ((k k)) = = ((\frac{11}{22} {C C}_{3232}^{((22 k k + + 11))})) [[f f 11 ((k k)) - - f f 11 ((1515 - - k k))]] for k for k = = 0,1 0,1,, . . . . . .,, 77$

并且 $C_{16}^{(2 k + 1) n} = \cos (\frac{nπ (2 k + 1)}{16}) .$ and $C_{16}^{(2 k + 1) no} = \cos (\frac{nπ (2 k + 1)}{16}) .$

式三中的16点DCT的F2(n)也可再被分解为两个8点DCT的G3(n)与G4(n)：The F2(n) of the 16-point DCT in Equation 3 can also be decomposed into two G3(n) and G4(n) of the 8-point DCT:

F2(2n)＝G3(n) for n＝0，1，...，7F2(2n)＝G3(n) for n＝0,1,...,7

，......(式八), ...... (Eight)

F2(2n+1)＝G4(n)+G4(n+1) for n＝0，1，...，7F2(2n+1)＝G4(n)+G4(n+1) for n＝0,1,...,7

其中，in,

$G G 33 ((n no)) = = {Σ Σ}_{k k = = 00}^{77} g g 33 ((k k)) {C C}_{1616}^{((22 k k + + 11)) n no} for n for n = = 0,1 0,1,, . . . . . .,, 77$

，......(式九), ... (Formation 9)

$G G 44 ((n no)) = = {Σ Σ}_{k k = = 00}^{77} g g 44 ((k k)) {C C}_{1616}^{((22 k k + + 11)) n no} for n for n = = 0,1 0,1,, . . . . . .,, 77$

其中，in,

g3(k)＝f2(k)+f2(15-k) for k＝0，1，...，7g3(k)=f2(k)+f2(15-k) for k=0, 1,..., 7

。(式十) (formula ten)

$g g 44 ((k k)) = = ((\frac{11}{22} {C C}_{3232}^{((22 k k + + 11))})) [[f f 22 ((k k)) - - f f 22 ((1515 - - k k))]] for k for k = = 0,1 0,1,, . . . . . .,, 77$

同理，式五与式八中的四个8点DCT皆可分别被进一步分解为两个4点DCT，也就是说，式一中的32点DCT最后可被分解八个4点DCT。Similarly, the four 8-point DCTs in Equation 5 and Equation 8 can be further decomposed into two 4-point DCTs, that is, the 32-point DCT in Equation 1 can be finally decomposed into eight 4-point DCTs.

根据式四、式七和式十，四个8点DCT的结果与该32点DCT的输入信号x的关系可被归纳为：According to Equation 4, Equation 7 and Equation 10, the relationship between the results of the four 8-point DCTs and the input signal x of the 32-point DCT can be summarized as:

g1(p)＝x(p)+x(31-p)+x(15-p)+x(16+p)g1(p)=x(p)+x(31-p)+x(15-p)+x(16+p)

$g g 22 ((p p)) = = ((\frac{11}{22} {C C}_{3232}^{((22 p p + + 11))})) [[x x ((p p)) + + x x ((3131 - - p p)) - - x x ((1515 - - p p)) - - x x ((1616 + + p p))]]$

$g g 33 ((p p)) = = ((\frac{11}{22} {C C}_{6464}^{((22 p p + + 11))})) [[x x ((p p)) - - x x ((3131 - - p p))]] + + ((\frac{11}{22} {C C}_{6464}^{((3131 - - 22 p p))})) [[x x ((1515 - - p p)) - - x x ((1616 + + p p))]] . .$

$g g 44 ((p p)) = = ((\frac{11}{22} {C C}_{3232}^{((22 p p + + 11))})) {{((\frac{11}{22} {C C}_{6464}^{((22 p p + + 11))})) [[x x ((p p)) - - x x ((3131 - - p p))]] - - ((\frac{11}{22} {C C}_{6464}^{((3131 - - 22 p p))})) [[x x ((1515 - - p p)) - - x x ((1616 + + p p))]]}}$

......(式十一)...... (Equation 11)

其中p为一范围在0至7之间的整数。Where p is an integer ranging from 0 to 7.

本发明将32点DCT分为三个阶段进行运算，并且将频率反转结合到第一阶段的运算中。如图4所示，DCT模件411包含一第一DCT单元411A与一第二DCT单元411B。The present invention divides 32-point DCT into three stages for operation, and combines frequency inversion into the operation of the first stage. As shown in FIG. 4 , the DCT module 411 includes a first DCT unit 411A and a second DCT unit 411B.

本发明在进行第一阶段的运算时，首先将32点DCT分解为两个16点DCT，然后再将每一个16点DCT进一步分解为2个8点DCT，以得到每个8点DCT的中间结果。如图4所示，于第一DCT单元411A中，该32个次频带取样信号被分为八个群组，每一个群组分别包含4个次频带取样信号：x(p)、x(31-p)、x(15-p)以及x(16+p)，其中p为一范围在0至7之间的整数。第一DCT单元411A包含一第一判断模件411A1、一第二判断模件411A2以及一计算模件411A3，并且第一DCT单元411A依序或同时处理该八个群组的次频带取样信号。When the present invention performs the calculation of the first stage, the 32-point DCT is first decomposed into two 16-point DCTs, and then each 16-point DCT is further decomposed into two 8-point DCTs to obtain the middle of each 8-point DCT result. As shown in FIG. 4, in the first DCT unit 411A, the 32 sub-band sampling signals are divided into eight groups, and each group includes 4 sub-band sampling signals: x(p), x(31 -p), x(15-p) and x(16+p), wherein p is an integer ranging from 0 to 7. The first DCT unit 411A includes a first judging module 411A1 , a second judging module 411A2 and a calculating module 411A3 , and the first DCT unit 411A processes the eight groups of sub-band sampling signals sequentially or simultaneously.

第一判断模件411A1用以判断该组正在处理中的次频带取样信号是否为该18组次频带取样信号中的第奇数组次频带取样信号，亦即判断i是否为一奇数。The first judging module 411A1 is used for judging whether the group of sub-band sampling signals being processed is an odd-th sub-band sampling signal among the 18 groups of sub-band sampling signals, that is, judging whether i is an odd number.

第二判断模件411A2由第一判断模件411A1所控制。如果第一判断模件411A1的判断结果为否，则第二判断模件411A2不操作。如果第一判断模件411A1的判断结果为是，第二判断模件411A2接着判断正在处理中的x(p)、x(31-p)、x(15-p)以及x(16+p)是否对应于一奇数p。The second judging module 411A2 is controlled by the first judging module 411A1. If the judgment result of the first judging module 411A1 is negative, the second judging module 411A2 does not operate. If the judgment result of the first judging module 411A1 is yes, the second judging module 411A2 then judges x(p), x(31-p), x(15-p) and x(16+p) being processed Does it correspond to an odd p.

计算模件411A3由第一判断模件411A1与第二判断模件411A2所控制。如果第一判断模件411A1的判断结果为否，表示该组正在处理中的次频带取样信号为第偶数组次频带取样信号。根据MP3规范中的频率反转的原则，该组次频带取样信号不需要被乘以负一，等同于未结合频率反转方法时的情况。因此，计算模件411A3根据式十一计算出四个8点DCT的中间结果(g1(p)、g2(p)、g3(p)以及g4(p))。如果第一判断模件411A1的判断结果为是，表示该组正在处理中的次频带取样信号为该18组次频带取样信号中的第奇数组次频带取样信号。根据MP3规范中的频率反转的原则，该组次频带取样信号中的第奇数个次频带取样信号必须被乘以负一。因此，第二判断模件411A2进一步判断p是否为一奇数。如果p为偶数，则x(31-p)与x(15-p)为该组次频带取样信号中的第奇数个次频带取样信号。举例来说，如果p等于2，x(31-p)与x(15-p)即为x(29)与x(13)，29与13皆为奇数。相对而言，如果p为奇数，则x(p)与x(16+p)为该组次频带取样信号中的第奇数个次频带取样信号。举例来说，如果p等于1，x(p)与x(16+p)即为x(1)与x(17)，1与17皆为奇数。因此，如果第二判断模件411A2为否，计算模件411A3于计算8点DCT的中间结果时，将x(31-p)与x(15-p)分别乘以负一。如果第二判断模件411A2为是，计算模件411A3于计算8点DCT的中间结果时，将x(p)与x(16+p)分别乘以负一。The calculating module 411A3 is controlled by the first judging module 411A1 and the second judging module 411A2. If the judgment result of the first judging module 411A1 is negative, it means that the group of sub-band sampling signals being processed is an even-th sub-band sampling signal. According to the principle of frequency inversion in the MP3 specification, the group of sub-band sampled signals does not need to be multiplied by minus one, which is equivalent to the situation when no frequency inversion method is combined. Therefore, the calculation module 411A3 calculates four intermediate results (g1(p), g2(p), g3(p) and g4(p)) of the 8-point DCT according to Equation 11. If the judgment result of the first judging module 411A1 is yes, it means that the group of sub-band sampling signals being processed is an odd group of sub-band sampling signals among the 18 groups of sub-band sampling signals. According to the principle of frequency inversion in the MP3 specification, the odd-numbered sub-band sample signal in the set of sub-band sample signals must be multiplied by minus one. Therefore, the second judging module 411A2 further judges whether p is an odd number. If p is an even number, then x(31-p) and x(15-p) are the odd-numbered sub-band sampled signals in the group of sub-band sampled signals. For example, if p is equal to 2, x(31-p) and x(15-p) are x(29) and x(13), and both 29 and 13 are odd numbers. Relatively speaking, if p is an odd number, then x(p) and x(16+p) are the odd-numbered sub-band sampled signals in the set of sub-band sampled signals. For example, if p is equal to 1, x(p) and x(16+p) are x(1) and x(17), and both 1 and 17 are odd numbers. Therefore, if the second judging module 411A2 is negative, the calculation module 411A3 multiplies x(31-p) and x(15-p) by minus one when calculating the intermediate result of the 8-point DCT. If the second judgment module 411A2 is yes, the calculation module 411A3 multiplies x(p) and x(16+p) by minus one when calculating the intermediate result of the 8-point DCT.

综上所述，如果第二判断模件411A2为否，计算模件411A3根据下列关系式计算8点DCT的中间结果：In summary, if the second judgment module 411A2 is no, the calculation module 411A3 calculates the intermediate result of 8-point DCT according to the following relational expression:

g1(p)＝x(p)-x(31-p)-x(15-p)+x(16+p)g1(p)=x(p)-x(31-p)-x(15-p)+x(16+p)

$g g 22 ((p p)) = = ((\frac{11}{22} {C C}_{3232}^{((22 p p + + 11))})) [[x x ((p p)) - - x x ((3131 - - p p)) + + x x ((1515 - - p p)) - - x x ((1616 + + p p))]]$

$g g 33 ((p p)) = = ((\frac{11}{22} {C C}_{6464}^{((22 p p + + 11))})) [[x x ((p p)) + + x x ((3131 - - p p))]] - - ((\frac{11}{22} {C C}_{6464}^{((3131 - - 22 p p))})) [[x x ((1515 - - p p)) + + x x ((1616 + + p p))]] . .$

$g g 44 ((p p)) = = ((\frac{11}{22} {C C}_{3232}^{((22 p p + + 11))})) {{((\frac{11}{22} {C C}_{6464}^{((22 p p + + 11))})) [[x x ((p p)) + + x x ((3131 - - p p))]] + + ((\frac{11}{22} {C C}_{6464}^{((3131 - - 22 p p))})) [[x x ((1515 - - p p)) + + x x ((1616 + + p p))]]}}$

......(式十二)...... (Formula 12)

如果第二判断模件411A2为是，计算模件411A3根据下列关系式计算8点DCT的中间结果：If the second judgment module 411A2 is yes, the calculation module 411A3 calculates the intermediate result of 8 points of DCT according to the following relational expression:

g1(p)＝-x(p)+x(31-p)+x(15-p)-x(16+p)g1(p)=-x(p)+x(31-p)+x(15-p)-x(16+p)

$g g 22 ((p p)) = = ((\frac{11}{22} {C C}_{3232}^{((22 p p + + 11))})) [[- - x x ((p p)) + + x x ((3131 - - p p)) - - x x ((1515 - - p p)) + + x x ((1616 + + p p))]]$

$g g 33 ((p p)) = = ((\frac{11}{22} {C C}_{6464}^{((22 p p + + 11))})) [[- - x x ((p p)) - - x x ((3131 - - p p))]] + + ((\frac{11}{22} {C C}_{6464}^{((3131 - - 22 p p))})) [[x x ((1515 - - p p)) + + x x ((1616 + + p p))]] . .$

$g g 44 ((p p)) = = ((\frac{11}{22} {C C}_{3232}^{((22 p p + + 11))})) {{((\frac{11}{22} {C C}_{6464}^{((22 p p + + 11))})) [[- - x x ((p p)) - - x x ((3131 - - p p))]] - - ((\frac{11}{22} {C C}_{6464}^{((3131 - - 22 p p))})) [[x x ((1515 - - p p)) + + x x ((1616 + + p p))]]}}$

......(式十三)...... (formula thirteen)

如式十二与式十三所示，根据本发明的合成次频带滤波装置40将频率反转结合到计算模件411A3计算8点DCT的中间结果的方法中。藉此，本发明可简化MP3规范中的解码方法并大幅提升将MP3音频帧解码的速度。As shown in Equation 12 and Equation 13, the synthesized subband filtering device 40 according to the present invention combines frequency inversion into the calculation module 411A3 to calculate the intermediate result of the 8-point DCT. Therefore, the present invention can simplify the decoding method in the MP3 specification and greatly increase the speed of decoding the MP3 audio frame.

以一实际应用时的情形为例，假设第一DCT单元411A正在处理第1组次频带取样信号中对应于p等于0的x(0)、x(31)、x(15)以及x(16)。由于第1组次频带取样信号为第奇数组次频带取样信号，第一判断模件411A1的判断结果为是，第二判断模件411A2接着进一步判断x(0)、x(31)、x(15)以及x(16)对应的p是否为偶数。由于p等于0，为一偶数，计算模件411A3根据式十二计算对应于x(0)、x(31)、x(15)以及x(16)的中间结果(g1(0)、g2(0)、g3(0)以及g4(0))。将p等于0代入式十二可得到下列关系式：Taking an actual application situation as an example, assume that the first DCT unit 411A is processing x(0), x(31), x(15) and x(16 ). Since the first group of sub-frequency band sampling signals is an odd group of sub-frequency band sampling signals, the judgment result of the first judging module 411A1 is yes, and the second judging module 411A2 then further judges x(0), x(31), x( 15) and whether p corresponding to x(16) is an even number. Because p is equal to 0 and is an even number, the calculation module 411A3 calculates the intermediate results corresponding to x(0), x(31), x(15) and x(16) (g1(0), g2( 0), g3(0), and g4(0)). Substituting p equal to 0 into formula 12 can get the following relational formula:

g1(0)＝x(0)-x(31)-x(15)+x(16)g1(0)=x(0)-x(31)-x(15)+x(16)

$g g 22 ((00)) = = ((\frac{11}{22} {C C}_{3232}^{11})) [[x x ((00)) - - x x ((3131)) + + x x ((1515)) - - x x ((1616))]]$

。......(式十四) ...... (Formula 14)

$g g 33 ((00)) = = ((\frac{11}{22} {C C}_{6464}^{11})) [[x x ((00)) + + x x ((3131))]] - - ((\frac{11}{22} {C C}_{6464}^{3131})) [[x x ((1515)) + + x x ((1616))]]$

$g g 44 ((00)) = = ((\frac{11}{22} {C C}_{3232}^{11})) {{((\frac{11}{22} {C C}_{6464}^{11})) [[x x ((00)) + + x x ((3131))]] + + ((\frac{11}{22} {C C}_{6464}^{3131})) [[x x ((1515)) + + x x ((1616))]]}}$

由于x(31)与x(15)为第奇数个次频带取样信号，相较于未结合频率反转的式十一，x(31)与x(15)于式十四中分别被乘以负一。Since x(31) and x(15) are odd-numbered sub-band sampling signals, compared to Equation 11 without frequency inversion, x(31) and x(15) in Equation 14 are respectively multiplied by minus one.

计算模件411A3于依序处理该八个群组的次频带取样信号时，每次计算出四个中间结果(g1(p)、g2(p)、g3(p)以及g4(p))。因此，将该八个群组接着处理完成后，计算模件411A3总共产生32个中间结果(g1(0)-g1(7)、g2(0)-g2(7)、g3(0)-g3(7)以及g4(0)-g4(7))。The calculating module 411A3 calculates four intermediate results (g1(p), g2(p), g3(p) and g4(p)) each time when sequentially processing the eight groups of sub-band sampled signals. Therefore, after the subsequent processing of the eight groups is completed, the calculation module 411A3 produces a total of 32 intermediate results (g1(0)-g1(7), g2(0)-g2(7), g3(0)-g3 (7) and g4(0)-g4(7)).

而本发明在进行第二阶段的运算时，则是将每个8点DCT进一步分解为2个4点DCT，并依据由32点DCT经频率反转及分解得出的8个4点DCT，计算得出32个转换后的向量。如图4所示，第二DCT单元411B将式五与式八中的四个8点DCT分解成八个4点DCT。根据计算模件411A3计算出的32个中间结果，第二DCT单元411B可计算出该八个4点DCT的结果，再进一步利用该八个4点DCT的结果产生该32个转换后的向量。However, when the present invention performs the second-stage calculation, each 8-point DCT is further decomposed into two 4-point DCTs, and based on the 8 4-point DCTs obtained by frequency inversion and decomposition of the 32-point DCT, The calculation yields 32 transformed vectors. As shown in FIG. 4 , the second DCT unit 411B decomposes the four 8-point DCTs in Equation 5 and Equation 8 into eight 4-point DCTs. According to the 32 intermediate results calculated by the calculation module 411A3 , the second DCT unit 411B can calculate the eight 4-point DCT results, and further use the eight 4-point DCT results to generate the 32 converted vectors.

本发明在进行第三阶段的运算时，将32点DCT经频率反转及分解得出的32个转换后向量处理合成32个PCM信号。如图4所示，处理器41中的产生模件412根据由第二DCT单元411B产生的32个转换后的向量，产生32个PCM信号。The present invention processes and synthesizes 32 PCM signals into 32 converted vectors obtained by frequency inversion and decomposition of 32-point DCT when performing the third-stage operation. As shown in FIG. 4 , the generating module 412 in the processor 41 generates 32 PCM signals according to the 32 transformed vectors generated by the second DCT unit 411B.

请参阅图5。图5为表示本发明结合频率反转与合成次频带滤波方法的概念。图5中的第一阶段表示第一DCT单元411A所执行的操作、第二阶段表示第二DCT单元411B所执行的操作、而第三阶段则表示产生模件412所执行的操作。线条交叉代表输入信号相互之间的相加或相减操作，图5中的CYX则代表上述方程式中的

。如图5所示，本发明在第一阶段将该32个次频带取样信号转换为八组中间结果，每组中间结果各自包含四个8点DCT的中间结果。公知技术的频率反转方法即结合到该第一阶段中。在第二阶段中，第二DCT单元411B根据该第一DCT单元411A计算出的总共32个中间结果，利用DCT产生该32个转换后的向量。在第三阶段中，产生模件412根据该32个转换后的向量，产生32个脉冲代码调制(pulse code modulation，PCM)信号。See Figure 5. FIG. 5 shows the concept of the present invention combining frequency inversion and synthetic subband filtering method. The first stage in FIG. 5 represents the operations performed by the first DCT unit 411A, the second stage represents the operations performed by the second DCT unit 411B, and the third stage represents the operations performed by the generation module 412 . The intersection of the lines represents the addition or subtraction of the input signals to each other, and the CYX in Figure 5 represents the

. As shown in FIG. 5 , the present invention converts the 32 sub-band sampling signals into eight groups of intermediate results in the first stage, and each group of intermediate results includes four 8-point DCT intermediate results. Known art frequency inversion methods are incorporated into this first stage. In the second stage, the second DCT unit 411B uses DCT to generate the 32 transformed vectors according to a total of 32 intermediate results calculated by the first DCT unit 411A. In the third stage, the generating module 412 generates 32 pulse code modulation (PCM) signals according to the 32 transformed vectors.

请参阅图6。图6为表示根据本发明的一较佳具体实施例的合成次频带滤波方法的流程图。该合成次频带滤波方法依序处理该18组频带取样信号。在处理每一组频带取样信号时，该方法首先执行步骤S601，设定一整数p为0。接着执行步骤S602，选取出该组次频带取样信号中的x(p)、x(31-p)、x(15-p)以及x(16+p)。步骤S603判断该组正在处理中的次频带取样信号是否为该18组次频带取样信号中的第奇数组次频带取样信号。如果步骤S603的判断结果为否，该合成次频带滤波方法执行步骤S605A，根据式十一计算g1(p)、g2(p)、g3(p)以及g4(p)。如果步骤S603的判断结果为是，该合成次频带滤波方法执行步骤S604，判断p是否为一奇数。如果步骤S604的判断结果为否，该合成次频带滤波方法执行步骤S605B，根据式十二计算g1(p)、g2(p)、g3(p)以及g4(p)。如果步骤S604的判断结果为是，该合成次频带滤波方法则执行步骤S605C，根据式十三计算g1(p)、g2(p)、g3(p)以及g4(p)。在步骤S605A、步骤S605B或步骤S605C之后，该合成次频带滤波方法执行步骤S606，判断p是否小于7，亦即判断是否该组次频带取样信号中的32个次频带取样信号都已被处理完成。如果步骤S606的判断结果为是，该合成次频带滤波方法执行步骤S607，设定p＝p+1，并重复执行步骤S602至步骤S606。如果步骤S606的判断结果为否，表示该组次频带取样信号中的32个次频带取样信号都已被处理完成，该合成次频带滤波方法即执行步骤S608与步骤S609。步骤S608根据g1(0)-g1(7)、g2(0)-g2(7)、g3(0)-g3(7)以及g4(0)-g4(7)计算出八个4点DCT的结果，再进一步利用该八个4点DCT的结果产生32个转换后的向量。步骤S609根据步骤S608中产生的32个转换后的向量，产生32个PCM信号。See Figure 6. FIG. 6 is a flowchart showing a synthetic subband filtering method according to a preferred embodiment of the present invention. The synthetic sub-band filtering method sequentially processes the 18 groups of frequency-band sampled signals. When processing each group of frequency band sample signals, the method first executes step S601, setting an integer p to 0. Next, step S602 is executed to select x(p), x(31-p), x(15-p) and x(16+p) in the group of sub-band sampled signals. Step S603 determines whether the group of sub-band sample signals being processed is an odd-th group of sub-band sample signals among the 18 groups of sub-band sample signals. If the judgment result of step S603 is negative, the combined sub-band filtering method executes step S605A to calculate g1(p), g2(p), g3(p) and g4(p) according to Equation 11. If the determination result of step S603 is yes, the synthetic sub-band filtering method executes step S604 to determine whether p is an odd number. If the judgment result of step S604 is negative, the combined sub-band filtering method executes step S605B to calculate g1(p), g2(p), g3(p) and g4(p) according to Equation 12. If the judgment result of step S604 is yes, the combined sub-band filtering method executes step S605C to calculate g1(p), g2(p), g3(p) and g4(p) according to Equation 13. After step S605A, step S605B or step S605C, the combined sub-band filtering method executes step S606 to determine whether p is less than 7, that is, to determine whether all 32 sub-band sampling signals in the group of sub-band sampling signals have been processed . If the judgment result of step S606 is yes, the synthetic subband filtering method executes step S607, sets p=p+1, and repeatedly executes steps S602 to S606. If the judgment result of step S606 is negative, it means that all 32 sub-band sampled signals in the set of sub-band sampled signals have been processed, and the combined sub-band filtering method executes steps S608 and S609. Step S608 calculates eight 4-point DCTs according to g1(0)-g1(7), g2(0)-g2(7), g3(0)-g3(7) and g4(0)-g4(7) As a result, the results of the eight 4-point DCTs are further utilized to generate 32 transformed vectors. Step S609 generates 32 PCM signals according to the 32 transformed vectors generated in step S608.

本发明的主要精神在于将32点DCT分为三个阶段进行运算：第一阶段将32点DCT递回分解以得到4个8点DCT的中间结果，并且将频率反转结合到第一阶段的运算中；第二阶段是将4个8点DCT进一步分解得出的8个4点DCT，并计算出32个转换后的向量；第三阶段则是将32个转换后向量处理合成32个PCM信号。由于部分32点DCT计算可以共用相同的系数，因此可以只读取一次后共用，藉以减少读取存储器的次数、加快运算速度。再者，结合频率反转与32点DCT的第一阶段运算于同一环路中，相较于先前技术使用双回圈个别计算频率反转与次频带合成，可节省方法代码大小，并可节省先前技术将频率反转结果存入存储器的次数，以及节省自存储器读取该结果进行次频带合成的次数，因而加快运算速度。于实际应用中，根据本发明的具体实施例未必需要将32点DCT分解为8个4点DCT，亦可能为将频率反转与1个32点DCT、2个16点DCT或是4个8点DCT结合。The main spirit of the present invention is to divide the 32-point DCT into three stages for calculation: the first stage recursively decomposes the 32-point DCT to obtain four intermediate results of the 8-point DCT, and combines the frequency inversion into the first stage In operation; the second stage is to further decompose the four 8-point DCTs to obtain 8 4-point DCTs, and calculate 32 converted vectors; the third stage is to process the 32 converted vectors into 32 PCMs Signal. Since some 32-point DCT calculations can share the same coefficients, they can be read only once and then shared, thereby reducing the number of memory reads and speeding up the calculation. Furthermore, combining frequency inversion and the first-stage operation of 32-point DCT in the same loop, compared with the prior art using double loops to separately calculate frequency inversion and sub-band synthesis, can save method code size and save The prior art saves the number of times the frequency inversion result is stored in the memory and saves the number of times the result is read from the memory for sub-band synthesis, thereby speeding up the operation speed. In practical application, according to the specific embodiment of the present invention, it is not necessarily necessary to decompose the 32-point DCT into eight 4-point DCTs, and it is also possible to combine the frequency inversion with one 32-point DCT, two 16-point DCTs or four 8-point DCTs. Click DCT to combine.

藉由以上较佳具体实施例的详述，希望能更加清楚描述本发明的特征与精神，而并非以上述所公开的较佳具体实施例来对本发明的范畴加以限制。相反地，其目的是希望能涵盖各种改变及具等效性的安排于本发明所欲申请的专利范围的范畴内。Through the above detailed description of the preferred embodiments, it is hoped that the features and spirit of the present invention can be described more clearly, rather than limiting the scope of the present invention by the preferred embodiments disclosed above. On the contrary, the intention is to cover various changes and equivalent arrangements within the scope of the claimed invention.

Claims

1. A synthetic sub-band filtering device used in an audio signal decoder, the device is implemented for M groups of signals, each group of signals in the M groups of signals includes N sub-band sampling signals, and M and N are both For natural numbers, the device contains:

A processor, the processor is used to process the i-th group of signals in the M groups of signals, where i is an integer ranging from 0 to M-1, the processor includes:

A discrete cosine transform DCT module, the DCT module uses DCT to convert the N sub-band sampling signals in the i-th group of signals into N converted vectors, if i is an odd number, the N converted vectors are generated During the vector process, the DCT module multiplies the 2j-1th sub-band sample signal among the N sub-band sample signals by minus one, where j is an integer ranging from 1 to N/2; and

A generating module, which generates N pulse code modulation signals according to the N transformed vectors.

2. The synthetic sub-band filtering device used in an audio signal decoder as claimed in claim 1, wherein the M group of signals conforms to the MPEG-1 Layer III standard, and M is equal to 18, and N is equal to 32.

3. The synthetic sub-band filtering device used in the audio signal decoder as claimed in claim 2, wherein the kth sub-band sampling signal of the 32 sub-band sampling signals in the i-th group signal is x(k- 1), k is an integer ranging from 1 to 32, and the DCT module includes:

A first DCT unit for processing x(p), x(31-p), x(15-p) and x(16+p), where p is a range from 0 to 7 an integer between , and the first DCT unit further comprises:

A first judging module, the first judging module is used to judge whether i is an odd number; and

A calculation module, the calculation module is controlled by the first judgment module, if the judgment result of the first judgment module is no, the calculation module calculates four intermediate results g corresponding to p according to the following equation ( p), g2(p), g3(p), and g4(p):

g1(p)=x(p)+x(31-p)+x(15-p)+x(16+p)

g g 22 ((p p)) = = ((\frac{11}{22} {C C}_{3232}^{((22 p p + + 11))})) [[x x ((p p)) + + x x ((3131 - - p p)) - - x x ((1515 - - p p)) - - x x ((1616 + + p p))]]

g g 33 ((p p)) = = ((\frac{11}{22} {C C}_{6464}^{((22 p p + + 11))})) [[x x ((p p)) - - x x ((3131 - - p p))]] + + ((\frac{11}{22} {C C}_{6464}^{((3131 - - 22 p p))})) [[x x ((1515 - - p p)) - - x x ((1616 + + p p))]],,

g g 44 ((p p)) = = ((\frac{11}{22} {C C}_{3232}^{((22 p p + + 11))})) {{((\frac{11}{22} {C C}_{6464}^{((22 p p + + 11))})) [[x x ((p p)) - - x x ((3131 - - p p))]] - - ((\frac{11}{22} {C C}_{6464}^{((3131 - - 22 p p))})) [[x x ((1515 - - p p)) - - x x ((1616 + + p p))]]}}

in

equal

equal

equal

\cos (\frac{2 p + 1}{32} π);

as well as

A second DCT unit, the second DCT unit is an intermediate result calculated according to the calculation module, and uses DCT to generate 32 transformed vectors.

4. The synthetic sub-band filtering device used in an audio signal decoder as claimed in claim 2, wherein the kth sub-band sampling signal of the 32 sub-band sampling signals in the i-th group signal is x(k- 1), k is an integer ranging from 1 to 32, and the DCT module includes:

A first judging module, the first judging module is used to judge whether i is an odd number;

A second judging module, the second judging module is controlled by the first judging module, if the judgment result of the first judging module is yes, the second judging module judges whether p is an odd number; and

A calculation module, the calculation module is controlled by the second judgment module, if the judgment result of the second judgment module is no, the calculation module calculates four intermediate results g corresponding to p according to the following equation ( p), g2(p), g3(p), and g4(p):

g1(p)=x(p)-x(31-p)-x(15-p)+x(16+p)

g g 22 ((p p)) = = ((\frac{11}{22} {C C}_{3232}^{((22 p p + + 11))})) [[x x ((p p)) - - x x ((3131 - - p p)) + + x x ((1515 - - p p)) - - x x ((1616 + + p p))]]

g g 33 ((p p)) = = ((\frac{11}{22} {C C}_{6464}^{((22 p p + + 11))})) [[x x ((p p)) + + x x ((3131 - - p p))]] - - ((\frac{11}{22} {C C}_{6464}^{((3131 - - 22 p p))})) [[x x ((1515 - - p p)) + + x x ((1616 + + p p))]],,

g g 44 ((p p)) = = ((\frac{11}{22} {C C}_{3232}^{((22 p p + + 11))})) {{((\frac{11}{22} {C C}_{6464}^{((22 p p + + 11))})) [[x x ((p p)) + + x x ((3131 - - p p))]] + + ((\frac{11}{22} {C C}_{6464}^{((3131 - - 22 p p))})) [[x x ((1515 - - p p)) + + x x ((1616 + + p p))]]}}

in

equal

equal

equal

\cos (\frac{2 p + 1}{32} π);

as well as

5. The synthetic sub-band filtering device used in the audio signal decoder as claimed in claim 2, wherein the kth sub-band sampling signal of the 32 sub-band sampling signals in the i-th group signal is x(k- 1), k is an integer ranging from 1 to 32, and the DCT module includes:

A second judging module, the second judging module is controlled by the first judging module, if the judging result of the first judging module is yes, then the second judging module judges whether p is an odd number; as well as

A calculation module, the calculation module is controlled by the second judgment module, if the judgment result of the second judgment module is yes, the calculation module is to calculate four intermediate results g1 corresponding to p according to the following equation (p), g2(p), g3(p), and g4(p):

g1(p)=-x(p)+x(31-p)+x(15-p)-x(16+p)

g g 22 ((p p)) = = ((\frac{11}{22} {C C}_{3232}^{((22 p p + + 11))})) [[- - x x ((p p)) + + x x ((3131 - - p p)) - - x x ((1515 - - p p)) + + x x ((1616 + + p p))]]

g g 33 ((p p)) = = ((\frac{11}{22} {C C}_{6464}^{((22 p p + + 11))})) [[- - x x ((p p)) - - x x ((3131 - - p p))]] + + ((\frac{11}{22} {C C}_{6464}^{((3131 - - 22 p p))})) [[x x ((1515 - - p p)) + + x x ((1616 + + p p))]],,

g g 44 ((p p)) = = ((\frac{11}{22} {C C}_{3232}^{((22 p p + + 11))})) {{((\frac{11}{22} {C C}_{6464}^{((22 p p + + 11))})) [[- - x x ((p p)) - - x x ((3131 - - p p))]] - - ((\frac{11}{22} {C C}_{6464}^{((3131 - - 22 p p))})) [[x x ((1515 - - p p)) + + x x ((1616 + + p p))]]}}

in

equal

equal

equal

\cos (\frac{2 p + 1}{32} π);

as well as

A second DCT unit, the second DCT unit is an intermediate result calculated according to the calculation module, and uses DCT to generate 32 converted vectors.

6. A method for synthetic subband filtering in an audio signal decoder, the method being performed for M groups of signals, each group of signals in the M groups of signals respectively comprising N subband sampling signals, both M and N is a natural number, the method includes the following steps:

Processing the i-th group of signals among the M groups of signals, wherein i is an integer ranging from 0 to M-1, and performing the following steps when processing the i-th group of signals:

(a) According to the N sub-band sampling signals and discrete cosine transform in the i-th group of signals, N converted vectors are generated. If i is an odd number, in the process of generating the N converted vectors, the N The 2j-1th sub-band sample signal among the sub-band samples is multiplied by minus one, where j is an integer ranging from 1 to N/2; and

(b) Generate N pulse code modulation signals according to the N converted vectors.

7. The synthetic sub-band filtering method used in the audio signal decoder as claimed in claim 6, wherein the M group of signals conforms to the MPEG-1 LayerIII standard, and M is equal to 18, and N is equal to 32.

8. The synthetic sub-band filtering method used in the audio signal decoder as claimed in claim 7, wherein the kth sub-band sampling signal of the 32 sub-band sampling signals in the i-th group signal is x(k- 1), k is an integer ranging from 1 to 32, and the following steps are performed when generating the 32 converted vectors in step (a):

(a1) process x(p), x(31-p), x(15-p) and x(16+p), where p is an integer ranging from 0 to 7, and is processed in x(p ), x(31-p), x(15-p) and x(16+p) perform the following sub-steps:

(a1-1) judge whether i is an odd number, if not, perform step (a1-2); and

(a1-2) Calculate four intermediate results g1(p), g2(p), g3(p) and g4(p) corresponding to p according to the following equations:

g1(p)=x(p)+x(31-p)+x(15-p)+x(16+p)

g g 22 ((p p)) = = ((\frac{11}{22} {C C}_{3232}^{((22 p p + + 11))})) [[x x ((p p)) + + x x ((3131 - - p p)) - - x x ((1515 - - p p)) - - x x ((1616 + + p p))]]

g g 33 ((p p)) = = ((\frac{11}{22} {C C}_{6464}^{((22 p p + + 11))})) [[x x ((p p)) - - x x ((3131 - - p p))]] + + ((\frac{11}{22} {C C}_{6464}^{((3131 - - 22 p p))})) [[x x ((1515 - - p p)) - - x x ((1616 + + p p))]],,

g g 44 ((p p)) = = ((\frac{11}{22} {C C}_{3232}^{((22 p p + + 11))})) {{((\frac{11}{22} {C C}_{6464}^{((22 p p + + 11))})) [[x x ((p p)) - - x x ((3131 - - p p))]] - - ((\frac{11}{22} {C C}_{6464}^{((3131 - - 22 p p))})) [[x x ((1515 - - p p)) - - x x ((1616 + + p p))]]}}

in

equal

equal

equal

\cos (\frac{2 p + 1}{32} π);

as well as

(a2) Using DCT to generate 32 converted vectors according to the intermediate results calculated in step (a1).

9. The synthetic sub-band filtering method used in the audio signal decoder as claimed in claim 7, wherein the kth sub-band sampling signal of the 32 sub-band sampling signals in the i-th group signal is x(k- 1), k is an integer ranging from 1 to 32, and the following steps are performed when generating the 32 converted vectors in step (a):

(a1-1) judge whether i is an odd number, if yes, execute step (a1-2); and

(a1-2) judge whether p is an odd number, if not, execute step (a1-3); and

(a1-3) Calculate four intermediate results g1(p), g2(p), g3(p) and g4(p) corresponding to p according to the following equations:

g1(p)=x(p)-x(31-p)-x(15-p)+x(16+p)

g g 22 ((p p)) = = ((\frac{11}{22} {C C}_{3232}^{((22 p p + + 11))})) [[x x ((p p)) - - x x ((3131 - - p p)) + + x x ((1515 - - p p)) - - x x ((1616 + + p p))]]

g g 33 ((p p)) = = ((\frac{11}{22} {C C}_{6464}^{((22 p p + + 11))})) [[x x ((p p)) + + x x ((3131 - - p p))]] - - ((\frac{11}{22} {C C}_{6464}^{((3131 - - 22 p p))})) [[x x ((1515 - - p p)) + + x x ((1616 + + p p))]],,

g g 44 ((p p)) = = ((\frac{11}{22} {C C}_{3232}^{((22 p p + + 11))})) {{((\frac{11}{22} {C C}_{6464}^{((22 p p + + 11))})) [[x x ((p p)) + + x x ((3131 - - p p))]] + + ((\frac{11}{22} {C C}_{6464}^{((3131 - - 22 p p))})) [[x x ((1515 - - p p)) + + x x ((1616 + + p p))]]}}

in

equal

equal

equal

\cos (\frac{2 p + 1}{32} π);

as well as

10. The synthetic sub-band filtering method used in an audio signal decoder as claimed in claim 7, wherein the kth sub-band sampling signal of the 32 sub-band sampling signals in the i-th group signal is x(k- 1), k is an integer ranging from 1 to 32, and the following steps are performed when generating the 32 converted vectors in step (a):

(a1-1) judge whether i is an odd number, if yes, execute step (a1-2); and

(a1-2) judge whether p is an odd number, if yes, perform step (a1-3); and

g1(p)=-x(p)+x(31-p)+x(15-p)-x(16+p)