CN101436407B - Method for encoding and decoding audio - Google Patents

Method for encoding and decoding audio Download PDF

Info

Publication number
CN101436407B
CN101436407B CN 200810232760 CN200810232760A CN101436407B CN 101436407 B CN101436407 B CN 101436407B CN 200810232760 CN200810232760 CN 200810232760 CN 200810232760 A CN200810232760 A CN 200810232760A CN 101436407 B CN101436407 B CN 101436407B
Authority
CN
Grant status
Grant
Patent type
Prior art keywords
frequency
signal
error
residual
method
Prior art date
Application number
CN 200810232760
Other languages
Chinese (zh)
Other versions
CN101436407A (en )
Inventor
吴礼仲
徐雅俊
熊静
郭小川
马鸿飞
Original Assignee
西安电子科技大学
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Grant date

Links

Abstract

The invention discloses a method for coding and decoding an audio, which mainly solves the problems that the prior method for coding and decoding the audio has low compression ratio and poor reconstruction audio quality. The method comprises the following steps: adopting a time frequency translation and frequency domain filtering method or a time domain filtering and time frequency translation method to analyze an audio signal so as to obtain a frequency domain residual error signal; dividing the frequency domain residual error signal into a low frequency residual error signal and a high frequency residual error signal, and performing direct coding on the low frequency residual error signal and performing parameter coding on the high frequency residual error signal respectively; reconstructing the high frequency residual error signal through the decoded low frequency residual error signal and the decoded high frequency residual error; recombining the decoded low frequency residual error signal and the reconstructed high frequency residual error signal to obtain a reconstructed frequency domain residual error signal; and finally adopting a frequency domain inverse filtering and time frequency inverse transform method or a time frequency inverse transform and time domain inverse filtering method to obtain a reconstructed audio signal. The method eliminates the redundancy in the frequency domain residual error signal and improves the compression ratio of audio coding, the channel utilization rate and the audio transmission quality, and is applied to multimedia communication and consumer electronics.

Description

音频编解码方法 Audio codec method

技术领域 FIELD

[0001] 本发明涉及音频编解码技术领域,具体涉及一种基于高频残差重构的信号重构方法,用于多媒体通信和消费类电子领域。 [0001] The present invention relates to audio encoding and decoding technologies, and particularly relates to a signal reconstruction method based on the reconstructed high-frequency residual, for multimedia communications and consumer electronics.

背景技术 Background technique

[0002] 在多媒体通信领域,包括语音在内的音频、尤其是宽带音频已逐渐成为主要通信业务之一。 [0002] In the field of multimedia communications, including voice audio, especially wideband audio has become one of the main communication services. 但是音频信号频带较宽、编码数据量较大,这给音频信号的实时传输和有效存储带来很大的困难。 However, a wide band audio signal, the encoded data is large, which brings great difficulties for effective real-time transmission and storage of audio signals. 虽然MP3、AAC、EAAC和EAAC+等音频编码算法已经能够较好地音频信号进行压缩编码,满足了一定应用的要求,但还无法较好地胜任目前正在发展的移动多媒体通信和各种移动平台等业务。 Although MP3, AAC, EAAC and EAAC + and other audio coding algorithms have been better able to encode the audio signal is compressed to meet the requirements of certain applications, but not well qualified for mobile multimedia communications currently being developed and a variety of mobile platforms business. 所以有必要研究效率更高和质量更好的音频编码算法。 It is necessary to study higher efficiency and better quality audio coding algorithm. 近年来,在音频压缩编码的研究领域,音频信号高频成分的处理、压缩和重构成为相关研究的关键技术之一,如何利用低频频段的信号来重建高频信号是重要的研究内容。 In recent years, research in the field of compressed audio signal encoded audio signal processing of high frequency components, compression and reconstruction of one of the key technology-related research, how to use a low-frequency band to reconstruct the high-frequency signal is important research.

[0003] 现有技术利用频域低频进行频域高频重建的方法主要有两种,简单介绍如下: [0003] The prior art method of using frequency low-frequency reconstructed frequency domain There are two main, briefly as follows:

[0004] 现有技术1 [0004] The prior art 1

[0005] 把音频或语音的低频信号通过一个数字滤波组处理,得到一组低频子带信号;再把该组低频子带作为一个整块信号来进行高频信号的复制。 [0005] The audio or speech signal through a low-frequency digital filtering process group, to give a set of low frequency subband signals; then the set of low frequency subband signal block as a high-frequency signal to be reproduced. 整个高频频段信号的复制方法是将高频信号按照频率由低到高划分成若干频段,每段与上述整块低频信号的带宽大致相同;然后将整块低频子带组连续复制到高频频段的每一段。 Replication Method entire frequency band signal is a frequency signal according to the frequency from low to high into several frequency bands, each of the above-described monolith substantially the same as the bandwidth of the low frequency signal; frequency subband block is then copied to the set of successive frequency each piece band. 这样,整块低频子带组会被周期性地在高频频段使用若干次,直到整个需要恢复的高频频段都被复制完成为止。 Thus, the entire frequency subband is set in a high frequency band periodically used several times, until the entire high frequency band needs to be restored is copied is completed. 具体方式有两种:一是把整块低频子带组平移到对应的高频频段,二是把整块低频子带组先折叠,即颠倒了子带排列顺序,再把整块低频子带组平移到对应的高频频段;在复制过程中,这两种方式可能会交叉使用。 DETAILED two ways: First, the piece lower sub-band is translated to the high-frequency group corresponding to the band, two-piece lower sub-band is to be folded first group, i.e., reverse the order of the sub-band, then the low frequency sub-band block group shifted to a high frequency corresponding to the band; during replication, these two methods may be used cross. 这样,整块低频子带组会被周期性地使用,直到整个需要恢复的高频频段都被复制完成为止。 Thus, the entire set of frequency subband is used periodically, until the entire high frequency band needs to be restored is copied is completed.

[0006] 现有技术2 [0006] The prior art 2

[0007] 把低频信号通过低通滤波器组处理,得到一组低频子带信号。 [0007] The low-frequency signal through a low pass filter set to give a set of low frequency subband signals. 这里不再像现有技术1那样,把选取的低频子带组作为一个整体,整段地连续复制需要恢复的高频部分,而是利用低频子带组中的子带,分别对应地恢复一些离散分布的高频子带。 This is no longer as in the prior art 1, the low frequency subband to the selected group as a whole, need to copy the whole high frequency portion is continuously recovered, but the use of low frequency sub-band sub-band groups respectively corresponding to restore some discretely distributed frequency sub-band. 如果在高频部分有非常丰富的谐波分量,则其谐波分量的频率可能就是其相应基频的整数倍。 If there is very rich in the high frequency harmonic component portion, the frequency of the harmonic components which might be an integer multiple of the corresponding frequency group. 在此思想的指导下,现有技术2提出,如果高频部分某些子带的子带序号是2、3、4、5等自然数的整数倍, 即某些高频子带和低频子带之间存在倍数的对应关系,这些子带很可能存在丰富的谐波成分,需要重点恢复。 Under the guidance of this idea, the prior art 2 proposed, if some of the high frequency portion of the sub-subband number is an integral multiple of the natural numbers 2,3,4,5, etc., i.e. some high frequency subband and low frequency subband there is a corresponding relationship between the multiples of these sub-bands are likely to exist a wealth of harmonic components, we need to focus on recovery. 这样,用连续的低频子带组恢复离散分布的高频子带的过程就完成了。 Thus, the recovery process discretely distributed frequency sub-band with a continuous frequency subband set is complete. 最后,对于该方法遗漏的高频子带,还要选取波形与之相近的低频子带,对遗漏的高频子带进行恢复,从而完成所有高频子带的复制。 Finally, the method for the higher sub-band missing, but also to select a waveform with similar low frequency subband, the omission of high frequency subbands recovery, thereby completing the copying all frequency subbands.

[0008] 上述两种现有技术中无论按现有技术1把低频子带作为一个整块进行周期性地平移复制或折叠复制,还是按现有技术2进行倍频复制,都是机械性地恢复谐波,没有考虑音频语音信号的多样性和变化性,另外复制时候是按照子带序号依次提取复制,由于低频子带和高频子带的波形本来就不同,所以被复制的高频子带和原始的高频子带相比,将存在较大波形差别或峰值差异,因此重建的高频信号准确性不是太高,影响重构音频信号的质量。 [0008] Whether the above-described two prior art according to the prior art one-piece lower sub-band as a copy or translate periodically folded copy, or according to the prior art 2 multiplies replication, are mechanically recovery harmonic, without considering the diversity and variation of the audio speech signal, and copied in accordance with sub-band when the sequentially extracted copy number, since the lower sub-band and the high frequency sub-band waveforms have been different, it is copied to the high frequency sub and a high frequency band sub-band compared with the original, there will be a large difference in a waveform or peak value difference, and therefore the accuracy of the reconstructed high-frequency signal is not too high, influence the quality of the reconstructed audio signal.

发明内容 SUMMARY

[0009] 本发明的目的之一是提供一种基于频域滤波和高频残差重构的音频编解码方法, 目的之二是提供一种基于时域滤波和高频残差重构的音频编解码方法,以避免上述已有技术的不足,提高重建高频残差信号的准确性和音频编解码的压缩比,保证重构音频信号质量。 [0009] One object of the present invention is to provide an audio codec and frequency domain filtering high frequency reconstruction method based on a residual, bis object is to provide a time-domain based filtering and high-frequency reconstructed audio residuals encoding and decoding method, in order to avoid the above deficiencies of the prior art, to improve the compression ratio of the high-frequency reconstructed residual signal and an audio codec accuracy to ensure quality reconstructed audio signal.

[0010] 本发明的技术方案是这样实现的: [0010] aspect of the present invention is implemented as follows:

[0011] 技术方案一: [0011] Technical Solution a:

[0012] 基于频域滤波和高频残差重构的音频编解码方法,包括如下步骤: [0012] Audio encoding and decoding method and high frequency domain filtering based reconstruction residuals, comprising the steps of:

[0013] 1)在编码端,音频原始时域信号经过时频变换处理,得到原始频域信号; [0013] 1) at the encoder, the original time domain audio signal through frequency transform process, frequency domain to obtain the original signal;

[0014] 2)原始频域信号经过频域感知滤波处理,得到原始频域残差信号; [0014] 2) the raw frequency domain signal after sensing the frequency domain filtering process, to obtain the original frequency-domain residual signal;

[0015] 3)原始频域残差信号经过频域残差分析和编码处理,得到低频残差信号编码和高频残差参数编码,并输出到传输信道或存储介质。 [0015] 3) through the original frequency domain residual signal and encoding the frequency domain residual analysis process, to obtain a high-frequency and low-frequency residual signal coding parameters coded residual, and outputs it to a transmission channel or storage medium.

[0016] 4)在解码端,接收来自输出到传输信道或存储介质的低频残差信号编码和高频残差参数编码,并对其进行频域残差解码和重构处理,得到重构频域残差信号; [0016] 4) at the decoding end receiving the output from the low-frequency and high-frequency residual residual signal coding parameters coded transport channel or storage medium, and subjected to a frequency domain residual decoding and reconstruction processing to obtain frequency reconstruction domain residual signal;

[0017] 5)重构频域残差信号经过频域感知逆滤波处理,得到重构频域信号; [0017] 5) frequency-domain residual signal is reconstructed through inverse filtering perceptual frequency domain to obtain frequency-domain signal reconstruction;

[0018] 6)对重构频域信号进行时频反变换处理,得到音频重构时域信号。 [0018] 6) when the frequency inverse transform processing on the reconstructed frequency domain signal to obtain reconstructed time-domain audio signal.

[0019] 技术方案二: [0019] Technical Solution II:

[0020] 基于时域滤波和高频残差重构的音频编解码方法,包括如下步骤: [0020] The audio codec in time domain and frequency filtering based reconstruction residuals, comprising the steps of:

[0021] Tl)在编码端,音频原始时域信号经过时域感知滤波处理,得到原始时域残差信号; [0021] Tl) domain perceptual filtering process when time-domain signal through the encoder, the original audio, the original time domain to obtain a residual signal;

[0022] T2)原始时域残差信号经过时频变换处理,得到原始频域残差信号; [0022] T2) time-frequency conversion after the original time domain residual signal to obtain the original frequency-domain residual signal;

[0023] T3)原始频域残差信号经过频域残差分析和编码处理,得到低频残差信号编码和高频残差参数编码,并输出到传输信道或存储介质。 [0023] T3) after the original frequency domain residual signal and encoding the frequency domain residual analysis process to obtain the low frequency and the high residual residual signal coding parameter encoding, and outputs it to a transmission channel or storage medium.

[0024] T4)在解码端,接收来自输出到传输信道或存储介质的低频残差信号编码和高频残差参数编码,并对其进行频域残差解码和重构处理,得到重构频域残差信号; [0024] T4) at the decoding end receiving the output from the low-frequency and high-frequency residual residual signal coding parameters coded transport channel or storage medium, and subjected to a frequency domain residual decoding and reconstruction processing to obtain frequency reconstruction domain residual signal;

[0025] T5)对重构频域残差信号进行时频逆变换处理,得到重构时域残差信号; [0025] T5) for inverse transformation frequency reconstructed frequency domain residual signal, to give the reconstructed time-domain residual signal;

[0026] T6)将重构时域残差信号进行时域逆滤波处理,得到音频重构时域信号。 [0026] T6) the reconstructed time-domain time-domain residual signal inverse filtering process, to obtain reconstructed time-domain audio signal.

[0027] 上述两种方案的音频编解码方法,其中所述的频域残差分析和编码处理,包括如下步骤: [0027] The audio codec method of the above two schemes, a frequency domain residual analysis process, and wherein said encoding, comprising the steps of:

[0028] (Al)按照等带宽频带、或临界频带、或频程频带频的带划分方法,先对原始频域残差信号进行频带划分,然后根据音频编码器设定的编码速率选择一个频带划分端点,将原始频域残差信号分割成原始低频残差信号和原始高频残差信号两部分,使这两部分各具有若干个频带; [0028] (Al) and the like in accordance with the frequency bandwidth, or critical bands, or bands with frequency-range division method, the original first frequency-domain residual signal is band division, and select a frequency band audio encoder according to a coding rate set dividing endpoint and dividing the original signal into the original frequency-domain residual signal and the original residual low frequency residual signal of two parts, so that the two parts each having a plurality of frequency bands;

[0029] (A2)对原始低频残差信号进行编码,得到低频残差信号编码输出;再对低频残差信号编码在编码端进行本地解码,得到解码低频残差信号;[0030] (A3)根据解码低频残差信号与原始高频残差信号的相似性或相关性,对高频残差参数进行分析,即选择一个频带匹配策略对解码低频残差信号和原始高频残差信号进行频带匹配,并计算最佳匹配的高频残差参数,得到包括频带划分方法、最佳匹配频带频率位置、能量匹配因子、声道耦合参数和帧间扩展参数在内的原始高频残差参数; [0029] (A2) of the original low-frequency residual signal is encoded residual signal to obtain the low frequency encoded output; low again locally decoded residual signal encoded on the encoding side, the decoded low frequency to obtain a residual signal; [0030] (A3) of the decoded low frequency band residual signal and the original residual signal frequency similarity or correlation decoded low frequency residual signal and the original residual signal, the residual high-frequency analysis parameters, i.e. a selection policy based band matching matching, and calculates the best match frequency residual parameters obtained band division method, the residual original high frequency band parameters including the best matching position, matching the energy factor, and an inter-channel coupling parameters, including the extended parameters;

[0031] (A4)对原始高频残差参数进行编码,得到高频残差参数编码输出。 [0031] (A4) of the original high residual encoding parameters, the encoding parameters to obtain a residual high-frequency output.

[0032] 上述两种方案的音频编解码方法,其中所述的频域残差解码和重构处理,包括如下步骤: [0032] The audio codec method of the above two schemes, a frequency domain residual decoding and reconstruction processing therein, comprising the steps of:

[0033] (Cl)对接收到的低频残差信号编码进行解码,得到解码低频残差信号; Low residual signal coding [0033] (Cl) to decode the received, decoded lower band to obtain a residual signal;

[0034] (C2)对接收到的高频残差参数编码进行解码,得到解码高频残差参数; Encoding high-frequency residual parameters [0034] (C2) the received decoded to obtain decoded high frequency residual parameters;

[0035] (C3)利用解码低频残差信号、解码高频残差参数中的最佳匹配频带频率位置和能量匹配因子,重构高频残差信号频带&(如,1); [0035] (C3) using the decoded low frequency residual signal decoding best matching position and energy of the high frequency band parameter matches the residual factor, the reconstructed high frequency band residual signal & (e.g., 1);

[0036] (C4)利用解码高频残差参数中的频带划分方法、声道耦合参数和帧间扩展参数, 将得到的重构高频残差信号频带进行组合,得到重构高频残差信号; [0036] (C4) using the decoded high frequency band division method of residuals parameter, coupling parameters and inter-channel extended parameters, the high-frequency reconstructed residual signal of the frequency band obtained by combining a high-frequency reconstruction to obtain a residual signal;

[0037] (C5)利用解码高频残差参数中的频域残差高低频段分界频率,将解码低频残差信号与重构高频残差信号进行组合,得到重构频域残差信号。 [0037] (C5) using the frequency domain residual low frequency band decoded high frequency residuals boundary parameters, the decoded residual signal and the reconstructed high-frequency low-frequency residual signals are combined to obtain the reconstructed residual signal in frequency domain.

[0038] 上述两种方案的音频编解码方法,其中所述的频带匹配策略包括:高频残差信号频带在低频残差信号区域进行匹配、高频残差信号频带在低频残差信号区域及其扩展区域进行匹配、通过声道耦合进行高频残差信号匹配和通过帧间扩展进行高频残差信号匹配。 [0038] The audio codec scheme of the above two methods, wherein the band matching policy comprises: a high-frequency band residual signal in the residual signal to match the low frequency region, the high frequency band of the residual signal in a low frequency region and the residual signal expansion region which matches the residual signal by matching the high-frequency and high-frequency coupling channel by matching the residual signal inter extension.

[0039] 本发明由于充分考虑了频域残差信号频谱的白化特性和高频残差信号和低频残差信号之间的相关性和相似性,通过选择高频残差信号和低频残差信号的匹配策略,估算高频残差信号的重构参数,并用高频残差参数准确的重构高频残差信号,并以此为基础,实现音频信号的高效编码和解码,因而提高了音频信源编码的压缩比或编码效率,节省了传输音频信号所需传输带宽和节省存储音频信号所需存储容量,同时提高了音频压缩编码质量。 [0039] Since the present invention is fully considered and similarity correlation, by selecting a high frequency and low frequency residual signal between a residual signal characteristic whitened residual signal and the frequency spectrum of a frequency domain residual signal and the low frequency residual signal matching strategy, estimates a residual signal of a high frequency reconstruction parameters, and to accurately reconstruct the residual high-frequency high-frequency residual signal parameters, and on this basis, to achieve high efficiency encoding and decoding an audio signal, thereby enhancing the audio source coding compression ratio or the coding efficiency, the audio signal to save transmission bandwidth required storage capacity for storing and saving the desired audio signal, while improving the quality of the audio compression encoder.

附图说明 [0040] 图 1基于频域滤波的音频编解码方法流程图;[0041] 图 2基于时域滤波的音频编解码方法流程图;[0042] 图 3频域残差信号分析和编码方法流程图;[0043] 图 4频域残差信号解码和重构方法流程图;[0044] 图 5高低频频域残差信号及高频残差信号重构示意图;[0045] 图 6高频残差信号频带在低频残差信号区域进行匹配策略示意图;[0046] 图 7高频残差信号频带在低频残差信号区域及其扩展区域进行匹配策略示意图[0047] 图 8通过声道耦合进行高频残差信号匹配策略示意图;[0048] 图 9通过帧间扩展进行高频残差信号匹配策略示意图。 BRIEF DESCRIPTION [0040] FIG. 1 is a flowchart of audio coding and decoding based on frequency domain filtering; flowchart [0041] FIG. 2 based audio codecs temporal filtering method; [0042] FIG. 3 and a frequency domain residual signal coding analysis The method of flowchart; [0043] FIG. 4 the frequency domain residual signal decoded and reconstructed flowchart of a method; [0044] FIG. 5 low frequency domain residual signal and the high frequency reconstructed residual signal schematic; [0045] FIG 6 frequency band residual signal to match a schematic view of a low frequency region strategy residual signal; [0046] FIG. 7 matching policy schematic frequency band residual signal in a low frequency region and the extended region of the residual signal [0047] FIG. 8 by the coupling channel the residual high-frequency signal matching policy schematic; [0048] FIG. 9 by residual inter-frequency matching a schematic view of the policy extension signal.

具体实施方式 detailed description

[0049] 实施例一 [0049] Example a

[0050] 本实施例提供一种音频编解码的方法,该方法是基于频域滤波和高频残差重构的音频编码方法。 [0050] The present embodiment provides an audio codec method, which is an audio encoding method and high frequency domain filtering residual Reconstruction. 高频残差信号是频域残差信号的高频部分,高频残差信号的分析和重构是为了有效的压缩频域残差信号的数据量,提高音频信号编码和传输效率。 Residual signal frequency is the high frequency portion of the frequency domain residual signal, frequency analysis and reconstruction of the residual signal in order to effectively compress the data of a frequency domain residual signal, an audio signal coding and improve transmission efficiency. 高频残差信号的分析和重构是按照一定的规则将频域残差信号分成低频残差信号和高频残差信号两部分, 利用高频残差信号与低频残差信号的相关性或相似性,提取高频残差重构所需要的参数, 丢弃高频残差信号,然后用高频残差参数来重构高频残差信号,从而重构整个频域残差信号。 Analysis of the high-frequency reconstructed residual signal and according to certain rules residual signal into low frequency domain residual signal part and two high-frequency residual signal, the residual signal using the correlation and the low-frequency signal or the residual similarity, the residual high-frequency extraction reconstruction parameters required, discarding the residual high-frequency signal, and the residual high-frequency high-frequency reconstructed residual signal parameters, to reconstruct the entire frequency domain residual signal.

[0051] 参照图1,本实施例的音频编解码步骤如下: [0051] Referring to FIG 1, the audio codec step of the present embodiment is as follows:

[0052] 步骤101,在编码端,音频原始时域信号经过时频变换处理,得到原始频域信号。 [0052] Step 101, the encoding side, the original time domain audio signal through frequency transform processing to obtain the original frequency domain signal.

[0053] 音频原始时域信号是包括语音信号、音频信号或任何人耳可以听到的各种声音信号的混合声音;音频信号的频率范围主要在OHz到20kHz之间,音频信号的采样频率为96kHz、48kHz、44. IkHz、32kHz、22. 05kHz、16kHz、11. 025kHz 和8kH ;音频信号的编码通常是以音频帧为单位,常用音频帧的大小按照实际应用一般在50毫秒之内。 [0053] The audio signal comprising the original time domain mixed audio speech signal, an audio signal or any of various sound signals in the human ear can hear; a frequency range of the audio signal is primarily OHz between 20kHz, the sampling frequency of the audio signal is 96kHz, 48kHz, 44 IkHz, 32kHz, 22 05kHz, 16kHz, 11 025kHz and 8 kHz;... usually encoded audio signal in units of audio frames, audio frames according to the size of conventional practice is generally within 50 milliseconds. 时域变换采用但不限于修正离散余弦变换、修正重叠变换和快速傅里叶变换方法进行变换。 Using time domain but not limited to modified discrete cosine transform, modified lapped transform and fast Fourier transform method to transform.

[0054] 步骤102,原始频域信号经过频域感知滤波处理,得到原始频域残差信号。 [0054] Step 102, the original frequency domain signal after sensing the frequency domain filtering process, frequency domain to obtain the original residual signal.

[0055] 频域感知滤波器是反映人耳听觉特性的频域滤波器,它对来步骤101的频域信号进行频域滤波,得到在感知意义上白化了的频域残差信号,如果用HM(f)表示频域感知滤波器的传输函数,用M(f)表示由感知参数表征的感知曲线,则HM(f)可以表示为 [0055] The frequency domain filter is a perceptual characteristics of the human auditory reflect the frequency domain filter, its frequency domain signal to the step 101 is filtering in the frequency domain, resulting in a whitened sense perceptual frequency domain residual signal if HM (f) represents the frequency domain perceptual transfer function of the filter, the curve represented by a perceptual awareness parameter characterized by M (f), the HM (f) can be expressed as

1(/) = ^y,其中f表示频率,单位为Hz。 1 (/) = ^ y, where f represents the frequency, in Hz.

[0056] 步骤103,频域残差分析和编码。 [0056] Step 103, and encoding the frequency domain residual analysis.

[0057] 参见3所示,频域残差分析和编码的具体步骤包括: [0057] Referring to FIG 3, the specific steps and encoding the frequency domain residual analysis comprises:

[0058] 步骤301,频带划分和高低频残差信号分割,即按照等带宽频带、或临界频带、或频程频带等方法,先对原始频域残差信号进行频带划分,然后根据音频编码器的编码速率选择其中一个划分频带的端点将原始频域残差信号分割成原始低频残差信号和原始高频残差信号两部分,使这两部分各具有若干个频带。 [0058] Step 301, and high frequency band division dividing the residual signal, i.e., in equal frequency bandwidth, or critical bands, octave band or the like, to the original frequency-domain residual signal is band division, and an audio encoder the coding rate select one of the endpoints of the original band dividing the frequency-domain residual signal is divided into the original high-frequency low-frequency signal and the original residual residual signal in two parts, so that the two parts each having a plurality of frequency bands.

[0059] 如图5所示,完整的频域残差信号用R表示,为了有效地对频域残差信号进行分析、编码和重构,将频域残差信号按照等带宽频带、或临界频带、或频程频带等方法进行频带划分,把具有一定属性,如听觉特性的相邻频率分量划分到相同的频带中去。 [0059] shown in Figure 5, the complete frequency domain residual signal is represented by R, in order to efficiently frequency domain residual signal is analyzed, coded and reconstructed, frequency-domain residual signal in equal frequency bandwidth, or critical bands, octave band or band division method such as, to a certain attributes, such as the hearing characteristics of adjacent frequency components divided into the same frequency band to. 设fra、fu和fHi分别为第i个频带的中心频率、低端边界频率和高端边界频率,h为频带带宽,选择某个频带的低端边界频率L或高端边界频率fHi为分界频率fD,将频域残差信号在频域内分割成低频残差信号&和高频残差信号&两部分,在0到fD之间就具有连续队个低频残差信号频带,在fD与fz之间就具有连续〜个高频残差信号频带,高于频率fz的高频残差信号频带中的信号为零。 Set fra, fu and fHi are the i-th center frequency of the band, low limit frequency and high boundary frequency, h is the frequency bandwidth to select a frequency band of the low-end boundary L of the boundary frequency or high frequency fD fHi as a boundary, dividing the frequency domain residual signal in the frequency domain residual signal into a low frequency and high frequency residual signal & two & portions, to between 0 and a continuous band fD low frequency band residual signal, and fz to between fD - having a continuous high frequency band of the residual signal, the residual signal of the high frequency band higher than the frequency fz of zero. 设f表示频率,单位为Hz、fs表示采样频率,归一化频率用/ = //为表示; 如果用fB表示时域音频信号的带宽,则fo = fs/fs = 0.5。 Let f represents the frequency in Hz, fs represents the sampling frequency, the normalized frequency by / = @ is represented; if the bandwidth represented by a time domain audio signal fB, the fo = fs / fs = 0.5.

[0060] 步骤302,对原始低频残差信号进行编码,得到低频残差信号编码输出。 [0060] Step 302, the low frequency the original residual signal is encoded residual signal to obtain the low frequency encoded output. 原始低频残差信号的编码,采用各种有失真的编码方法,如线性或非线性标量量化编码、矢量量化编码,或者同时采用各种无失真的编码方法,如Huffman编码和算术编码; Low-frequency encoding the original residual signal, using a variety of lossy encoding method, such as a linear or non-linear scalar quantization and coding, vector quantization coding, or while using a variety of lossless coding methods, such as Huffman coding and arithmetic coding;

[0061] 步骤303,对低频残差信号编码在编码端进行本地解码,得到解码低频残差信号和低频残差信号的编码;[0062] 步骤304,根据解码低频残差信号和原始高频残差信号的相似性或相关性,对高频残差参数进行分析,即选择一个频带匹配策略对解码低频残差信号与原始高频残差信号进行频带匹配,并计算最佳匹配的高频残差参数,得到原始高频残差参数。 [0061] Step 303, the low frequency locally decoded residual signal encoded on the encoding side, to obtain a residual signal encoding and decoding residual signal and LF; [0062] Step 304, based on the decoded residual signal and the original low-frequency residual the similarity or correlation difference signals, high-frequency residual parameters analyzed, i.e., select a low frequency band matching policy decoded residual signal and the original residual signal frequency band matching, and calculates the best match for the high frequency residues difference parameter, to obtain the original high frequency residual parameters.

[0063] 如图5所示,频带匹配就是在低频残差信号内寻找与高频残差频带特性最接近的频带,目的是在解码端高质量地重构高频残差信号。 As shown in [0063] FIG. 5, the high frequency band matching is to find the closest residual band characteristics in the low frequency band residual signal is a high quality reconstructed object frequency residual signal at the decoder. 频带匹配采用不同的频带匹配策略,所谓频带匹配策略是指如何在低频残差信号&中选择一段频域残差信号,并用它来重构高频残差信号&中具有相同频带宽度的高频残差信号的方法。 Employ a different frequency band matching policy to match the so-called band matching strategy is how to select the frequency domain residual signal for a period in the low frequency & residual signal, and used to reconstruct the high frequency signal & residual frequency band having the same width the method of the residual signal. 在图5中,因为频率fz以上的残差信号已经归零,所以没有必要再对这部分残差进行重构。 In FIG. 5, since the above residual signal has zero frequency fz, so there is no need to reconstruct this residual portion.

[0064] 所述频带匹配策略包括但不限于以下几种: [0064] The band matching policy including but not limited to, the following:

[0065] (Pl)高频残差信号频带在低频残差信号区域进行匹配 [0065] (Pl) frequency band residual signal in the residual signal to match the low frequency region

[0066] 如图6所示,将处于fD与fz之间的〜个高频残差频带分别独立地在处于0到fD 之间的低频残差信号区域&进行匹配,寻找误差最小或相关性最大的最佳匹配频带。 [0066] As shown in FIG 6, the residual high-frequency band is ~ between fD and fz are independently a residual signal in a low frequency region is between 0 to fD & match, find the minimum error or correlation the biggest match of the best bands. 这时, 频带划分方法、最佳匹配频带频率位置和能量匹配因子等即为高频残差重构参数。 In this case, band division method, frequency band position of the best matching and energy matching parameter is the high-frequency reconstructed residual factor.

[0067] (P2)高频残差信号频带在低频残差信号区域及其扩展区域进行匹配 [0067] (P2) to match the high-frequency band residual signal in the low frequency region and the extended region a residual signal

[0068] 如图7所示,采用上述频带匹配策略(Pl),先将处于fD与fz之间的频率最低的高频残差信号频带在处于0到fD之间的低频残差信号区域&进行匹配,寻找误差最小或相关性最大的最佳匹配频带,并用此最佳匹配频带来重构频率最低的高频残差信号频带Rhi ;再将这个重构的频率最低的高频残差信号频带加入原来的低频残差信号&,形成一个新组成的低频残差信号Ru = RL+RH1 ;然后,以这个新组成的低频残差信号为基础,采用上述频带匹配策略(Pl),继续重构更高频率的高频残差信号频带;依此类推,直至完成全部fD 到&之间的所有高频残差信号频带的匹配和重构。 [0068] 7, the above-described band matching policy (Pl is), the first is the lowest frequency of the high band residual signal fz in between fD and a residual signal in the low frequency region to between 0 & fD matching error to find the minimum or maximum of the best correlation matching band, and with this best match the frequency band of the lowest frequency band Rhi reconstructed residual signal frequency; then the lowest frequency of high-frequency reconstructed residual signal the original low frequency band residual signal is added & forming a new residual signal composed of low frequency Ru = RL + RH1; then, the new residual signal composed of low basis using the bandwidth matching strategy (Pl is), continued heavy configuration of the high frequency band residual signal of higher frequency; and so on until completion of all fD to all matching and reconstruction of the high frequency band between the residual signal &. 这时,频带划分方法、最佳匹配频带频率位置、能量匹配因子和低频残差区域扩展参数即为高频残差重构参数。 In this case, band division method, frequency band position of the best match, and a low residual energy matching factor is the high-frequency region extension parameter residual reconstruction parameters.

[0069] (P3)通过声道耦合进行高频残差信号匹配 [0069] (P3) by matching the residual signal is a high frequency coupling channel

[0070] 如图8所示,设音频信源是具有C (C ^ 2)个主声道的多声道音频信源,比如5. 1 声道音频就有5个主声道和一个低重音效声道,这时存在C个声道的低频残差信号。 [0070] As shown, the audio source is provided in FIG. 8 is a C (C ^ 2) multi-channel audio source in the main channel, such as 5.1-channel audio there are five main channels and a low heavy sound channels, the low frequency channels of the residual signal C is present at this time. 因为各声道的音频信号具有较大相关性,所以各声道的频域残差信号也具有较大相关性,因此任何一个声道不仅能够利用其所在声道的低频残差信号进行高频残差信号重构,而且还能够利用其它声道的低频残差信号进行高频残差信号重构;这样可以得到更多的匹配选择, 提高高频残差信号重构质量。 Since each channel audio signal having a large correlation, so each channel frequency domain residual signal also has a large correlation and therefore only one channel can be any frequency in the use of their low residual signal channel Reconstruction of the residual signal, but also other channels using a low frequency signal is a residual high-frequency reconstructed residual signal; this can be obtained more match selection to improve the quality of the high-frequency reconstructed residual signal. 以第1声道为例给以说明,首先可以采用上述频带匹配策略(Pl)和(P2),利用所在声道的处在0到fD之间的低频残差信号&匹配和重构处于fD与fz 之间的高频残差信号。 To give an example of the first channel described above may be employed first band matching policy (Pl is) and (P2), using a channel where the residual signal at a low frequency between 0 and fD fD & matching and reconstruction in a residual signal between the high frequency fz. 除此之外,还可以利用其它声道的处在0到fD之间的低频残差信号Rl匹配和重构第1声道内处于fD与fz之间的高频残差信号。 In addition, the other channels may also be utilized in the low-frequency reconstructed residual signal Rl and the matching of the first channel between the residual signal at a high frequency and fD fD 0 to fz. 与第1声道一样,所有C个声道的频域残差信号多可以采用相同的方法进行处理。 As with the first channel, all the channels C frequency domain residual signal may be processed by the same multiple method.

[0071] (P4)通过帧间扩展进行高频残差信号匹配 [0071] (P4) the residual high-frequency signal match extended by inter

[0072] 上述频带匹配策略(Pl)、(P2)和(P3)所描述的高频残差信号的匹配策略是利用当前帧的低频残差信号重构当前帧的高频残差信号。 [0072] The band matching policy (Pl), (P2) and (P3) of the residual signal frequency matching strategy described are high-frequency reconstructed residual signal of the current frame using the residual signal of the current frame is low. 因为相邻帧音频信号之间通常存在很大的相关性,所以相邻帧的残差也具有较大的相关性。 Because often there is a strong correlation between adjacent frames of the audio signal, the residual neighboring frames have greater relevance. 因此,当前音频帧的高频残差信号, 不仅可以用当前帧的低频残差信号进行重构,也可以用当前帧之前若干帧的包括重构低频残差信号和重构高频残差信号的重构残差信号进行重构。 Thus, the high frequency current signal of a residual audio frames, only low frequency may be reconstituted with a residual signal of the current frame, it may be used a number of frames before the current frame includes a reconstructed residual signal and the low frequency reconstructed residual signal the reconstructed residual signal is reconstructed. 如图9所示,采用上述频带匹配 9, the above-described band matching

9策略(Pl)、(P2)和(P;?)的方法利用第j帧的低频残差信号对第j帧的高频残差信号进行重构,或采用上述频带匹配策略(Pl)、(P2)和(P;?)的方法利用第j-Ι帧的低频残差信号对第j帧的高频残差信号进行重构,还可以采用上述频带匹配策略(Pl)、(P2)和(P;?)的方法利用第j_l帧的重构高频残差信号对第j帧的高频残差信号进行重构。 Strategies 9 (Pl), (P2) and methods (P ;?) using a low frequency of a residual signal of the j-th frame of the residual signal of the high frequency reconstructed frame j, or use of the bandwidth matching strategy (Pl is), (P2) and methods (P ;?) of the j-th residual signal using a low frequency-high frequency iota frame j-th frame of the residual signal is reconstructed, the bandwidth can also be employed matching strategy (Pl is), (P2) and (P ;?) a method of using a high-frequency reconstructed frame j_l residual signal to the residual signal of the high-frequency reconstructed frame j. 此外,这些方法还可以扩展到第j_2帧、第j_3帧去。 Further, these methods may also be extended to j_2 first frame, the first frame to j_3.

[0073] 上述各种频带匹配策略都需要在低频残差信号和高频残差信号之间寻找最佳匹配频带,这里所说的最佳,可以用误差或失真最小来表达,也可以用相关性最大来表示。 [0073] The various band matching strategies need to find the best match between the low frequency band residual signal and the residual signal frequency, here the best, can be expressed with a minimum error or distortion can also be associated with greatest to represent. 具体计算最佳匹配的高频残差参数的步骤包括: Step specific calculation frequency residual parameters best match comprises:

[0074] (Bi)计算归一化低频残差频带信号屯O,/)和归一化高频残差频带信号:^…,/): 设进行匹配的高频残差信号和低频残差信号的频带宽度为L个频点 [0074] (Bi) calculated normalized residual low frequency band signal Tun O, /) and the residual normalized frequency band signals: ... ^, /): set to match the residual signal and the low frequency residual signal bandwidth frequency points is L

Figure CN101436407BD00101

为低频残差信号频带、&(q,1), 1 =0,1, 为高频残差信号频带,其中ρ和q分别表示&(P,1)和&(9,1)所在频带的起始位置,P e [0,fD]、qe [fD,fz],计算兔Q9,/)和 A low frequency band residual signal, & (q, 1), 1 = 0,1, a high-frequency band of the residual signal, wherein q and ρ denote & (P, 1) and & (9,1) band where starting position, P e [0, fD], qe [fD, fz], calculated rabbit Q9, /), and

Figure CN101436407BD00102

其中,屯0,/)表示&(?,1),1 = 0,1,…LI被其自身绝对值的最大值Rtaax (ρ)进行归一化得到的归一化低频残差信号、^Kp,/)表示&(P,1), 1 = 0,1,"化-1被其自身绝对值的最大值RHmax(q)进行归一化得到的归一化高频残差信号。 Wherein, Tun 0, /) represents & (?, 1), 1 = 0,1, ... LI is the maximum absolute value of its own Rtaax (ρ) are normalized residual signal normalized low frequency obtained, ^ Kp, /) represents & (P, 1), 1 = 0,1, "of -1 is the maximum absolute value of its own RHmax (q) normalized frequency normalized residual signal obtained.

[0075] (B2)计算匹配失真测度d (p,q)或匹配相关函数r (p,q),如果用失真最小来进行匹配分析,则匹配的失真测度表示为: [0075] (B2) is calculated to match the distortion measure d (p, q) or matching correlation function r (p, q), if the matching analysis to minimize distortion, the distortion measure of the matching is represented as:

[0076] [0076]

Figure CN101436407BD00103

[0077] 其中w(p,q)是频率影响因子,·η<|^/&,/) I)是残差幅度影响因子。 [0077] where w (p, q) is a frequency influence factor, · η <| ^ / &, /) I) is the amplitude of the residual impact factor. 如果w(p,q)= 1.0,<|云/^,/)丨)=1.0,那么式(14)可以简化为: If w (p, q) = 1.0, <| cloud / ^, /) Shu) = 1.0, then the formula (14) can be simplified to:

[0078] [0078]

Figure CN101436407BD00104

[0079] 如果用相关性最大进行匹配分析,归一化低频残差信号和归一化高频残差信号的相关函数表示为: [0079] If the analysis of the maximum correlation match, normalized residual signal and low-frequency residual signal normalized correlation function is expressed as:

[0080] [0080]

Figure CN101436407BD00105

[0081] 同样,w(p,q)表示频率影响因子,I)表示残差幅度影响因子。 [0081] Similarly, w (p, q) represents a frequency influence factor, I) represents the amplitude of the residual impact factor. 如果w(p, q) = 1·0,>ν(| ^(仏/)1) = 1.0,那么式(16)可以简化为: If w (p, q) = 1 · 0,> ν (| ^ (Fo /) 1) = 1.0, then the formula (16) can be simplified to:

[0082] [0082]

Figure CN101436407BD00106

[0083] (B3)确定最佳频带匹配位置Ptl和(¾,该最佳匹配就是失真测度d (p,q)的最小值或相关函数r(p,q)的最大值所对应的ρ和q的最佳值Ptl和(¾,它们确定了高频残差与低频残差所匹配的频带。 [0083] (B3) determining the best matching position Ptl band and (¾, the best match is the maximum value of the minimum distortion measure d (p, q) or correlation function r (p, q) and the corresponding ρ and q optimal value Ptl (¾, which determines the frequency band with low residual residual matched.

[0084] (B4)利用如下公式计算最佳能量匹配因子: [0084] (B4) is calculated using the formula optimal energy matching factor:

[0085] [0085]

Figure CN101436407BD00107

[0086] 这样,步骤304所得到的频带划分结果、最佳匹配频带频率位置、频域残差高低频段分界频率fD、能量匹配因子、声道耦合参数和帧间扩展参数,就是原始高频残差参数。 [0086] Thus, the band division result obtained in step 304, the position of the best match the frequency band, the frequency domain residual boundary frequency and low frequency fD, energy matching factor, coupling parameters and inter-channel enhancement parameter is the original high frequency residual difference parameter.

[0087] 步骤305,高频残差参数编码。 [0087] Step 305, a high frequency residual coding parameters.

[0088] 对步骤304产生的原始高频残差参数进行编码,得到高频残差信号编码输出。 [0088] The original high frequency residual parameters 304 generated by the step of encoding the residual signal to obtain high frequency encoded output. 原始高频残差参数的编码,采用各种有失真的编码方法,如线性或非线性标量量化编码、矢量量化编码,或者同时采用各种无失真的编码方法,如Huffman编码和算术编码。 It encodes the original high frequency residual parameters, various lossy encoding method, such as a linear or non-linear scalar quantization and coding, vector quantization coding, or while using a variety of lossless coding methods, such as Huffman coding and arithmetic coding.

[0089] 步骤104,频域残差解码和重构。 [0089] Step 104, the frequency domain residual decoded and reconstructed.

[0090] 在解码端,接收到的低频残差信号编码和高频残差参数编码经过频域残差解码和重构处理,得到重构频域残差信号,如图4所示。 [0090] On the decoding side, the received low-frequency and the high residual signal coding parameters coded residual through a frequency domain residual decoding and reconstruction processing to obtain the reconstructed residual signal in the frequency domain, as shown in FIG. 具体步骤包括: These steps include:

[0091] 步骤401,对接收到的低频残差信号编码进行解码,得到解码低频残差信号。 [0091] Step 401, the low frequency the received encoded residual signal is decoded to obtain a decoded lower band residual signal.

[0092] 步骤402,对接收到的高频残差参数编码进行解码,得到解码高频残差参数,该解码高频残差参数包括频带划分结果、频域残差高低频段分界频率fD、最佳匹配频带频率位置、能量匹配因子、声道耦合参数和帧间扩展参数。 [0092] Step 402, a high frequency residual parameters for decoding the received encoded, resulting in a decoded residual frequency parameters, the decoded high frequency residual parameters includes a band division result, the frequency domain residual boundary frequency and low frequency fD, most best match the frequency band position, energy matching factor, coupling parameters and inter-channel extension parameters.

[0093] 步骤403,高频残差信号重构。 [0093] Step 403, the high-frequency reconstructed residual signal.

[0094] 根据解码高频残差参数,利用解码低频残差信号复制和重构高频残差信号,具体步骤包括: [0094] The high-frequency residual parameters decoded, using the decoded low-frequency reconstructed residual signal and a residual signal replication, particularly comprising the step of:

[0095] (Dl)根据步骤401和402得到的解码低频残差信号、最佳匹配频带频率位置和能量匹配因子,用下式进行高频残差信号复制和能量匹配: [0095] (Dl) The matching factor decoded lower band residual signal obtained in step 401 and 402, and the position of the best match frequency band energy, high-frequency energy of the residual signal and a matching copy of the formula:

[0096] Rh (q0, 1) = GE(p0, q0) · Rl (p0, 1),1 = 0,1,-LI (6) [0096] Rh (q0, 1) = GE (p0, q0) · Rl (p0, 1), 1 = 0,1, -LI (6)

[0097] 得到重构高频残差信号频带4(¾,1); [0097] The residual signal to obtain the reconstructed high frequency band 4 (¾, 1);

[0098](拟)根据步骤401和402得到的频带划分结果、声道耦合参数和帧间扩展参数,将所有重构高频残差信号频带进行组合,得到重构高频残差信号。 [0098] (TBC) according to the frequency band division result obtained in step 401 and 402, coupling parameters and inter-channel extended parameters, all the reconstructed high frequency band residual signal are combined to obtain a high-frequency reconstructed residual signal.

[0099] 步骤404,根据解码得到的频域残差高低频段分界频率fD,将解码低频残差信号与重构高频残差信号进行组合,得到重构频域残差信号。 [0099] Step 404, the frequency domain residual The boundary frequency and low frequency fD is decoded, the decoded residual signal and the reconstructed high-frequency low-frequency residual signals are combined to obtain the reconstructed residual signal in frequency domain.

[0100] 步骤105,重构频域残差信号经过频域逆滤波处理,得到重构频域信号;如果用He(f)表示频域感知逆滤波器,则Hk(f)表示为Α(/) = ^Γ^ = Μ(/),其中f表示频率,单位为Hz。 [0100] Step 105, the reconstructed frequency domain residual signal through inverse filtering in the frequency domain, the frequency domain signal to obtain reconstructed; if He (f) represents the frequency domain inverse perceptual filter, then Hk (f) is represented by [alpha] ( /) = ^ Γ ^ = Μ (/), where f represents the frequency, in Hz.

[0101] 步骤106,对重构频域信号进行时频反变换处理,得到音频重构时域信号。 [0101] Step 106, when the frequency inverse transform processing on the reconstructed frequency domain signal to obtain an audio domain signal reconstruction. 与时频变换相对应,时域反变换采用反向修正离散余弦变换、反向修正重叠反变换或反向快速傅里叶变换方法。 With the corresponding time-frequency transform, inverse transform time domain using inverse modified discrete cosine transform, inverse transform or inverse modified overlapping inverse fast Fourier transform method.

[0102] 实施例二 [0102] Second Embodiment

[0103] 本实施例提供一种基于时域滤波和高频残差重构的音频编解码方法。 [0103] The audio codec method of the present embodiment provides a high frequency and time domain filtering based on the reconstructed residual embodiment.

[0104] 参见图2,该方法步骤如下: [0104] Referring to Figure 2, the steps are as follows:

[0105] 步骤201,在编码端,音频原始时域信号经过时域感知滤波处理,得到原始时域残差信号。 [0105] Step 201, the encoding side, the original time domain audio signal is time domain perceptual filtering process, to obtain the original time domain residual signal. 其中,时域感知滤波器是反映人耳听觉特性的时域滤波器,它对音频原始时域信号进行时域滤波,得到在感知意义上白化了的时域残差信号;时域感知滤波器的传输函数用Hm(Z)表示,时域感知滤波器采用但不限于线性预测滤波器。 Wherein the time domain filter is a perceptual characteristics of the human auditory reflected time domain filter, its original time domain audio signal in time domain filtering to obtain a perceptually whitened sense a time domain residual signal; temporal perceptual filter the transfer function is represented by Hm (Z), time-domain, but not limited to a perceptual filter using linear prediction filter.

[0106] 步骤202,原始时域残差信号经过时频变换处理,得到原始频域残差信号,其中,时域变换采用但不限于修正离散余弦变换、修正重叠变换和快速傅里叶变换方法进行变换。 [0106] Step 202, the original time domain residual signal through frequency transform process, frequency domain to obtain the original residual signal, wherein the time domain transform using, but not limited to modified discrete cosine transform, modified lapped transform and fast Fourier transform method transformed.

[0107] 步骤203,频域残差分析和编码。 [0107] Step 203, and encoding the frequency domain residual analysis.

[0108] 参见3所示,频域残差分析和编码的具体步骤包括: [0108] Referring to FIG 3, the specific steps and encoding the frequency domain residual analysis comprises:

[0109] 步骤301,频带划分和高低频残差信号分割,即按照等带宽频带、或临界频带、或频程频带等方法,先对原始频域残差信号进行频带划分,然后根据音频编码器的编码速率选择其中一个划分频带的端点将原始频域残差信号分割成原始低频残差信号和原始高频残差信号两部分,使这两部分各具有若干个频带。 [0109] Step 301, and high frequency band division dividing the residual signal, i.e., in equal frequency bandwidth, or critical bands, octave band or the like, to the original frequency-domain residual signal is band division, and an audio encoder the coding rate select one of the endpoints of the original band dividing the frequency-domain residual signal is divided into the original high-frequency low-frequency signal and the original residual residual signal in two parts, so that the two parts each having a plurality of frequency bands.

[0110] 如图5所示,完整的频域残差信号用R表示,为了有效地对频域残差信号进行分析、编码和重构,将频域残差信号按照等带宽频带、或临界频带、或频程频带等方法进行频带划分,把具有一定属性,如听觉特性的相邻频率分量划分到相同的频带中去。 [0110] shown in Figure 5, the complete frequency domain residual signal is represented by R, in order to efficiently frequency domain residual signal is analyzed, coded and reconstructed, frequency-domain residual signal in equal frequency bandwidth, or critical bands, octave band or band division method such as, to a certain attributes, such as the hearing characteristics of adjacent frequency components divided into the same frequency band to. 设fra、fu和fHi分别为第i个频带的中心频率、低端边界频率和高端边界频率,h为频带带宽,选择某个频带的低端边界频率L或高端边界频率fHi为分界频率fD,将频域残差信号在频域内分割成低频残差信号&和高频残差信号&两部分,在0到fD之间就具有连续队个低频残差信号频带,在fD与fz之间就具有连续〜个高频残差信号频带,高于频率fz的高频残差信号频带中的信号为零。 Set fra, fu and fHi are the i-th center frequency of the band, low limit frequency and high boundary frequency, h is the frequency bandwidth to select a frequency band of the low-end boundary L of the boundary frequency or high frequency fD fHi as a boundary, dividing the frequency domain residual signal in the frequency domain residual signal into a low frequency and high frequency residual signal & two & portions, to between 0 and a continuous band fD low frequency band residual signal, and fz to between fD - having a continuous high frequency band of the residual signal, the residual signal of the high frequency band higher than the frequency fz of zero. 设f表示频率,单位为泡、&表示采样频率,归一化频率用/ = //>表示; 如果用fB表示时域音频信号的带宽,则Jb 二JBI fs = 0.5。 Let f represents the frequency in the bubble, & represents a sampling frequency, the normalized frequency by / = //>; if one represents the bandwidth with time domain audio signal fB, the two Jb JBI fs = 0.5.

[0111] 步骤302,对原始低频残差信号进行编码,得到低频残差信号编码输出。 [0111] Step 302, the low frequency the original residual signal is encoded residual signal to obtain the low frequency encoded output. 原始低频残差信号的编码,采用各种有失真的编码方法,如线性或非线性标量量化编码、矢量量化编码,或者同时采用各种无失真的编码方法,如Huffman编码和算术编码; Low-frequency encoding the original residual signal, using a variety of lossy encoding method, such as a linear or non-linear scalar quantization and coding, vector quantization coding, or while using a variety of lossless coding methods, such as Huffman coding and arithmetic coding;

[0112] 步骤303,对低频残差信号编码在编码端进行本地解码,得到解码低频残差信号和低频残差信号的编码; [0112] Step 303, the low frequency locally decoded residual signal encoded on the encoding side, the low frequency decoded residual signal to obtain encoded residual signal and the low frequency;

[0113] 步骤304,根据解码低频残差信号和原始高频残差信号的相似性或相关性,对高频残差参数进行分析,即选择一个频带匹配策略对解码低频残差信号与原始高频残差信号进行频带匹配,并计算最佳匹配的高频残差参数,得到原始高频残差参数。 [0113] Step 304, the similarity or relevance of the decoded low frequency signal and the original residual residual signal, high-frequency residual parameters analyzed, i.e., select a low frequency band matching policy decoded residual signal and the original high residual signal frequency band matching, and calculates the best match frequency residual parameters to obtain the original high frequency residual parameters.

[0114] 如图5所示,频带匹配就是在低频残差信号内寻找与高频残差频带特性最接近的频带,目的是在解码端高质量地重构高频残差信号。 As shown in [0114] FIG. 5, the high frequency band matching is to find the closest residual band characteristic band residual signal in a low frequency, high frequency reconstructed object is high quality residual signal at the decoder. 频带匹配采用不同的频带匹配策略,所谓频带匹配策略是指如何在低频残差信号&中选择一段频域残差信号,并用它来重构高频残差信号&中具有相同频带宽度的高频残差信号的方法。 Employ a different frequency band matching policy to match the so-called band matching strategy is how to select the frequency domain residual signal for a period in the low frequency & residual signal, and used to reconstruct the high frequency signal & residual frequency band having the same width the method of the residual signal. 在图5中,因为频率fz以上的残差信号已经归零,所以没有必要再对这部分残差进行重构。 In FIG. 5, since the above residual signal has zero frequency fz, so there is no need to reconstruct this residual portion.

[0115] 所述频带匹配策略包括但不限于以下几种: [0115] The band matching policy including but not limited to, the following:

[0116] (Pl)高频残差信号频带在低频残差信号区域进行匹配 [0116] (Pl) frequency band residual signal in the residual signal to match the low frequency region

[0117] 如图6所示,将处于fD与fz之间的〜个高频残差频带分别独立地在处于0到fD 之间的低频残差信号区域&进行匹配,寻找误差最小或相关性最大的最佳匹配频带。 [0117] As shown in FIG 6, the residual high-frequency band is ~ between fD and fz are independently a residual signal in a low frequency region is between 0 to fD & match, find the minimum error or correlation the biggest match of the best bands. 这时, 频带划分方法、最佳匹配频带频率位置和能量匹配因子等即为高频残差重构参数。 In this case, band division method, frequency band position of the best matching and energy matching parameter is the high-frequency reconstructed residual factor.

[0118] (P2)高频残差信号频带在低频残差信号区域及其扩展区域进行匹配 [0118] (P2) to match the high-frequency band residual signal in the low frequency region and the extended region a residual signal

[0119] 如图7所示,采用上述频带匹配策略(Pl),先将处于之间的频率最低的高频残差信号频带在处于0到fD之间的低频残差信号区域&进行匹配,寻找误差最小或相关性最大的最佳匹配频带,并用此最佳匹配频带来重构频率最低的高频残差信号频带Rhi ;再将这个重构的频率最低的高频残差信号频带加入原来的低频残差信号&,形成一个新组成的低频残差信号Ru = RL+RH1 ;然后,以这个新组成的低频残差信号为基础,采用上述频带匹配策略(Pl),继续重构更高频率的高频残差信号频带;依此类推,直至完成全部fD 到&之间的所有高频残差信号频带的匹配和重构。 [0119] As shown in FIG. 7, the above-described band matching policy (Pl is), the lowest frequency of the high frequency band between the first residual signal in the residual signal in the low frequency region between 0 to fD & match, find the minimum error or maximum correlation best matching band, and with this best match the frequency band of the lowest frequency reconstructed frequency band residual signal Rhi; then the frequency of the lowest frequency reconstructed residual signal added to the original frequency band & residual signal of low frequency, low frequency component to form a new residual signal Ru = RL + RH1; then, the new residual signal composed of low basis using the bandwidth matching strategy (Pl is), the higher continue reconstruction residual signal of the high frequency band frequency; and so on until completion of all fD to all matching and reconstruction of the high frequency band between the residual signal &. 这时,频带划分方法、最佳匹配频带频率位置、能量匹配因子和低频残差区域扩展参数即为高频残差重构参数。 In this case, band division method, frequency band position of the best match, and a low residual energy matching factor is the high-frequency region extension parameter residual reconstruction parameters.

[0120] (P3)通过声道耦合进行高频残差信号匹配 [0120] (P3) by matching the residual signal is a high frequency coupling channel

[0121] 如图8所示,设音频信源是具有C(C ^ 2)个主声道的多声道音频信源,比如5. 1 声道音频就有5个主声道和一个低重音效声道,这时存在C个声道的低频残差信号。 [0121] As shown, the audio source is provided in FIG. 8 is a C (C ^ 2) multi-channel audio source in the main channel, such as 5.1-channel audio there are five main channels and a low heavy sound channels, the low frequency channels of the residual signal C is present at this time. 因为各声道的音频信号具有较大相关性,所以各声道的频域残差信号也具有较大相关性,因此任何一个声道不仅能够利用其所在声道的低频残差信号进行高频残差信号重构,而且还能够利用其它声道的低频残差信号进行高频残差信号重构;这样可以得到更多的匹配选择, 提高高频残差信号重构质量。 Since each channel audio signal having a large correlation, so each channel frequency domain residual signal also has a large correlation and therefore only one channel can be any frequency in the use of their low residual signal channel Reconstruction of the residual signal, but also other channels using a low frequency signal is a residual high-frequency reconstructed residual signal; this can be obtained more match selection to improve the quality of the high-frequency reconstructed residual signal. 以第1声道为例给以说明,首先可以采用上述频带匹配策略(Pl)和(P2),利用所在声道的处在0到fD之间的低频残差信号&匹配和重构处于fD与fz 之间的高频残差信号。 To give an example of the first channel described above may be employed first band matching policy (Pl is) and (P2), using a channel where the residual signal at a low frequency between 0 and fD fD & matching and reconstruction in a residual signal between the high frequency fz. 除此之外,还可以利用其它声道的处在0到fD之间的低频残差信号Rl匹配和重构第1声道内处于fD与fz之间的高频残差信号。 In addition, the other channels may also be utilized in the low-frequency reconstructed residual signal Rl and the matching of the first channel between the residual signal at a high frequency and fD fD 0 to fz. 与第1声道一样,所有C个声道的频域残差信号多可以采用相同的方法进行处理。 As with the first channel, all the channels C frequency domain residual signal may be processed by the same multiple method.

[0122] (P4)通过帧间扩展进行高频残差信号匹配 [0122] (P4) the residual high-frequency signal match extended by inter

[0123] 上述频带匹配策略(Pl)、(P2)和(P3)所描述的高频残差信号的匹配策略是利用当前帧的低频残差信号重构当前帧的高频残差信号。 [0123] The band matching policy (Pl), (P2) and (P3) of the residual signal frequency matching strategy described are high-frequency reconstructed residual signal of the current frame using the residual signal of the current frame is low. 因为相邻帧音频信号之间通常存在很大的相关性,所以相邻帧的残差也具有较大的相关性。 Because often there is a strong correlation between adjacent frames of the audio signal, the residual neighboring frames have greater relevance. 因此,当前音频帧的高频残差信号, 不仅可以用当前帧的低频残差信号进行重构,也可以用当前帧之前若干帧的包括重构低频残差信号和重构高频残差信号的重构残差信号进行重构。 Thus, the high frequency current signal of a residual audio frames, only low frequency may be reconstituted with a residual signal of the current frame, it may be used a number of frames before the current frame includes a reconstructed residual signal and the low frequency reconstructed residual signal the reconstructed residual signal is reconstructed. 如图9所示,采用上述频带匹配策略(Pl)、(P2)和(P;?)的方法利用第j帧的低频残差信号对第j帧的高频残差信号进行重构,或采用上述频带匹配策略(Pl)、(P》和(P;?)的方法利用第j_l帧的低频残差信号对第j帧的高频残差信号进行重构,还可以采用上述频带匹配策略(Pl)、(P2)和(P;?)的方法利用第j_l帧的重构高频残差信号对第j帧的高频残差信号进行重构。此外,这些方法还可以扩展到第j_2帧、第j_3帧去。 9, the above-described band matching policy (Pl), (P2) and methods (P ;?) using a low frequency of a residual signal of the j-th frame of the residual signal of the high frequency reconstructed frame j, or With the above band matching policy (Pl), (P "method and (P ;?) using the residual signal of low frequency to high frequency j_l frame j-th frame of the residual signal is reconstructed, the bandwidth can also be employed matching strategy (Pl), (P2) and (P ;?) a method of using a high-frequency reconstructed frame j_l residual signal to the residual signal of the high-frequency reconstructed frame j. further, these methods can be extended to the first j_2 frame, the frame j_3 go.

[0124] 上述各种频带匹配策略都需要在低频残差信号和高频残差信号之间寻找最佳匹配频带,这里所说的最佳,可以用误差或失真最小来表达,也可以用相关性最大来表示。 [0124] The various band matching strategies need to find the best match between the low frequency band residual signal and the residual signal frequency, here the best, can be expressed with a minimum error or distortion can also be associated with greatest to represent. 具体计算最佳匹配的高频残差参数的步骤包括: Step specific calculation frequency residual parameters best match comprises:

[0125] (Bi)计算归一化低频残差频带信号兔和归一化高频残差频带信号:^…,/): 设进行匹配的高频残差信号和低频残差信号的频带宽度为L个频点、&(p,l),l = 0,1,… LI为低频残差信号频带、&(q,1), 1 =0,1, 为高频残差信号频带,其中ρ和q分别表示&(P,1)和&(9,1)所在频带的起始位置,P e [0,fD]、qe [fD,fz],计算屯(ρ,/)和^O,/),其中,^O,/)表示& (P,1),1 = 0,1,…LI被其自身绝对值的最大值Rtaax (ρ)进行归一化得到的归一化低频残差信号、:^O?,/)表示&(P,1), 1 = 0,1,"化-1被其自身绝对值的最大值RHmax(q)进行归一化得到的归一化高频残差信号。 Bandwidth provided to match the residual signal and the low frequency residual signal: [0125] (Bi) calculated normalized residual low frequency band signal and rabbit a normalized frequency band signal of the residual: ... ^, /) frequency points is L, & (p, l), l = 0,1, ... LI residual signal is a low frequency band, & (q, 1), 1 = 0,1, the residual signal into a high frequency band, wherein and q are [rho] & (P, 1) and & (9,1) where the starting position of the band, P e [0, fD], qe [fD, fz], calculated Tun (ρ, /) ^ O, and , /), where, ^ O, /) represents & (P, 1), 1 = 0,1, ... LI is the maximum absolute value of its own Rtaax (ρ) normalizing a normalized low residue obtained difference signal,:? ^ O, /) represents & (P, 1), 1 = 0,1, "of -1 is the maximum absolute value of its own RHmax (q) normalized normalized high obtained frequency residual signal.

[0126] (B2)计算匹配失真测度d (p,q)或匹配相关函数r (p,q),如果用失真最小来进行匹配分析,则匹配的失真测度表示为:[0127] [0126] (B2) is calculated to match the distortion measure d (p, q) or matching correlation function r (p, q), if the matching analysis to minimize distortion, the distortion measure of the matching is represented as: [0127]

Figure CN101436407BD00141

[0128] 其中w(p,q)是频率影响因子,:^(彳,/)|)是残差幅度影响因子。 [0128] where w (p, q) is a frequency influence factor: ^ (left foot, /) |) is the amplitude of the residual impact factor. 如果w(p,q)= 1.0,w(|^(g,/)1) = 1.0,那么式(14)可以简化为: [0129] If w (p, q) = 1.0, w (| ^ (g, /) 1) = 1.0, the formula may then (14) simplifies to: [0129]

Figure CN101436407BD00142

[0130] 如果用相关性最大进行匹配分析,归一化低频残差信号兔…,/)和归一化高频残差信号;^Ο?,/)的相关函数表示为: ? [0130] If the analysis of the maximum correlation match, normalized residual signal low rabbit ..., /) and the residual signal normalized frequency; ^ o, /) is the correlation function is expressed as:

[0131] [0131]

Figure CN101436407BD00143

[0132] 同样,w(p,q)表示频率影响因子,W(|^f(^,/)丨)表示残差幅度影响因子。 [0132] Similarly, w (p, q) represents a frequency influence factor, W (| ^ f (^, /) Shu) represents the magnitude of the residual impact factor. 如果w(p, q) = 1.0,w(| = 1.0,那么式(16)可以简化为: [0133] If w (p, q) = 1.0, w (| = 1.0, then the formula (16) can be simplified as: [0133]

Figure CN101436407BD00144

[0134] (B3)确定最佳频带匹配位置Ptl和(¾,该最佳匹配就是失真测度d (p,q)的最小值或相关函数r(p,q)的最大值所对应的ρ和q的最佳值Ptl和(¾,它们确定了高频残差与低频残差所匹配的频带。 [0134] (B3) determining the best matching position Ptl band and (¾, the best match is the maximum value of the minimum distortion measure d (p, q) or correlation function r (p, q) and the corresponding ρ and q optimal value Ptl (¾, which determines the frequency band with low residual residual matched.

[0135] (B4)利用如下公式计算最佳能量匹配因子: [0135] (B4) is calculated using the formula optimal energy matching factor:

[0136] [0136]

Figure CN101436407BD00145

[0137] 这样,步骤304所得到的频带划分结果、最佳匹配频带频率位置、频域残差高低频段分界频率fD、能量匹配因子、声道耦合参数和帧间扩展参数,就是原始高频残差参数。 [0137] Thus, the band division result obtained in step 304, the position of the best match the frequency band, the frequency domain residual boundary frequency and low frequency fD, energy matching factor, coupling parameters and inter-channel enhancement parameter is the original high frequency residual difference parameter.

[0138] 步骤305,高频残差参数编码。 [0138] Step 305, a high frequency residual coding parameters.

[0139] 对步骤304产生的原始高频残差参数进行编码,得到高频残差信号编码输出。 [0139] original high residual parameters 304 generated by the step of encoding the residual signal to obtain high frequency encoded output. 原始高频残差参数的编码,采用各种有失真的编码方法,如线性或非线性标量量化编码、矢量量化编码,或者同时采用各种无失真的编码方法,如Huffman编码和算术编码。 It encodes the original high frequency residual parameters, various lossy encoding method, such as a linear or non-linear scalar quantization and coding, vector quantization coding, or while using a variety of lossless coding methods, such as Huffman coding and arithmetic coding.

[0140] 步骤204,频域残差解码和重构。 [0140] Step 204, the frequency domain residual decoded and reconstructed.

[0141] 在解码端,接收到的低频残差信号编码和高频残差参数编码经过频域残差解码和重构处理,得到重构频域残差信号,如图4所示。 [0141] On the decoding side, the received low-frequency and the high residual signal coding parameters coded residual through a frequency domain residual decoding and reconstruction processing to obtain the reconstructed residual signal in the frequency domain, as shown in FIG. 具体步骤包括: These steps include:

[0142] 步骤401,对接收到的低频残差信号编码进行解码,得到解码低频残差信号。 [0142] Step 401, the low frequency the received encoded residual signal is decoded to obtain a decoded lower band residual signal.

[0143] 步骤402,对接收到的高频残差参数编码进行解码,得到解码高频残差参数,该解码高频残差参数包括频带划分结果、频域残差高低频段分界频率fD、最佳匹配频带频率位置、能量匹配因子、声道耦合参数和帧间扩展参数。 [0143] Step 402, a high frequency residual parameters for decoding the received encoded, resulting in a decoded residual frequency parameters, the decoded high frequency residual parameters includes a band division result, the frequency domain residual boundary frequency and low frequency fD, most best match the frequency band position, energy matching factor, coupling parameters and inter-channel extension parameters.

[0144] 步骤403,高频残差信号重构。 [0144] Step 403, the high-frequency reconstructed residual signal.

[0145] 根据解码高频残差参数,利用解码低频残差信号复制和重构高频残差信号,具体步骤包括: [0145] The high-frequency residual parameters decoded, using the decoded low-frequency reconstructed residual signal and a residual signal replication, particularly comprising the step of:

[0146] (Dl)根据步骤401和402得到的解码低频残差信号、最佳匹配频带频率位置和能量匹配因子,用下式进行高频残差信号复制和能量匹配:[0147] Rh (q0, 1) = Ge (p0, q0) · Rl (p0, 1) , 1 = 0, 1,…L_l(6) [0146] (Dl) The matching factor decoded lower band residual signal obtained in step 401 and 402, and the position of the best match frequency band energy, high-frequency matching the residual signal replication formula and energy: [0147] Rh (q0 , 1) = Ge (p0, q0) · Rl (p0, 1), 1 = 0, 1, ... L_l (6)

[0148] 得到重构高频残差信号频带&(qQ,1); [0148] to obtain the reconstructed high frequency band residual signal & (qQ, 1);

[0149](拟)根据步骤401和402得到的频带划分结果、声道耦合参数和帧间扩展参数,将所有重构高频残差信号频带进行组合,得到重构高频残差信号。 [0149] (TBC) according to the frequency band division result obtained in step 401 and 402, coupling parameters and inter-channel extended parameters, all the reconstructed high frequency band residual signal are combined to obtain a high-frequency reconstructed residual signal.

[0150] 步骤404,根据解码得到的频域残差高低频段分界频率fD,将解码低频残差信号与重构高频残差信号进行组合,得到重构频域残差信号。 [0150] Step 404, the frequency domain residual The boundary frequency and low frequency fD is decoded, the decoded residual signal and the reconstructed high-frequency low-frequency residual signals are combined to obtain the reconstructed residual signal in frequency domain.

[0151] 步骤205,重构频域残差信号经过时频反变换处理,得到重构时域残差信号;与时频变换相对应,时域反变换采用反向修正离散余弦变换、反向修正重叠反变换或反向快速傅里叶变换方法。 [0151] Step 205, the reconstructed frequency domain residual signal passes through a frequency inverse transform process, to obtain reconstructed time-domain residual signal; and the corresponding time-frequency transform, inverse transform time domain using inverse modified discrete cosine transform, inverse overlap correction or inverse fast Fourier transform inverse transform method.

[0152] 步骤206,对重构时域残差信号进行时域感知逆滤波处理,得到音频重构时域信号输出;如果用Hk(Z)表示时域感知逆滤波器,则Hk(f)可以表示为^(2) = ^^。 [0152] Step 206, when the sensing field of the reconstructed residual signal inverse temporal filtering process, to obtain reconstructed domain audio output signals; temporal perceptual inverse filter, then Hk (f) if Hk (Z) represents ^ can be expressed as (2) = ^^.

[0153] 本发明上述实施例提供的音频编码方法和解码方法,能够对包括语音信号在内的音频信号进行高效高质量的压缩编码,提高音频传效率。 Invention, the above-described audio encoding method and decoding method provided in the embodiment, it is possible for an audio signal including a speech signal is performed, including [0153] the present high efficiency compression coding to improve audio transfer efficiency.

[0154] 以上实施例只是用于帮助理解本发明的方法及其核心思想;同时,对于本领域的一般技术人员,依据本发明的思想,在具体实施方式及应用范围上均会有改变之处,综上所述,本说明书内容不应理解为对本发明的限制。 [0154] the above embodiments are merely used to help understanding the method and core ideas of the present invention; Meanwhile, those of ordinary skill in the art, according to the ideas of the present invention, there are changes in the embodiments and application scope of the , Therefore, the specification shall not be construed as limiting the present invention.

Claims (10)

  1. 1. 一种基于频域滤波和高频残差重构的音频编解码方法,包括如下步骤:1)在编码端,音频原始时域信号经过时频变换处理,得到原始频域信号;2)原始频域信号经过频域感知滤波处理,得到原始频域残差信号;3)原始频域残差信号经过频域残差分析和编码处理,得到低频残差信号编码和高频残差参数编码,并输出到传输信道或存储介质。 An audio codec based on frequency domain filtering and high-frequency reconstructed residuals, comprising the following steps: 1) at the encoder, the original time domain audio signal through frequency transform process, frequency domain to obtain the original signal; 2) raw frequency domain signal after sensing the frequency domain filtering process, to obtain the original frequency-domain residual signal; 3) through the original frequency domain residual signal and encoding the frequency domain residual analysis process, to obtain a high-frequency and low-frequency residual signal coding parameters coded residual , and outputs it to a transmission channel or storage medium. 4)在解码端,接收来自输出到传输信道或存储介质的低频残差信号编码和高频残差参数编码,并对其进行频域残差解码和重构处理,得到重构频域残差信号;5)重构频域残差信号经过频域感知逆滤波处理,得到重构频域信号;6)对重构频域信号进行时频反变换处理,得到音频重构时域信号。 4) on the decoding side, the received frequency and the low frequency residual signal coding parameter encoding residual from the output to a transmission channel or storage medium, and subjected to a frequency domain residual decoding and reconstruction processing, the frequency domain to obtain the reconstructed residual signal; 5) through the reconstructed residual signal in the frequency domain a frequency domain sensing inverse filtering, the reconstructed signal to obtain a frequency domain; 6) when the frequency inverse transform processing on the reconstructed frequency domain signal to obtain reconstructed time-domain audio signal.
  2. 2.根据权利要求1所述的音频信号编解码方法,其中步骤幻所述的频域残差分析和编码处理,包括如下步骤:(Al)按照等带宽频带、或临界频带、或频程频带的频带划分方法,先对原始频域残差信号进行频带划分,然后根据音频编码器设定的编码速率选择一个频带划分端点,将原始频域残差信号分割成原始低频残差信号和原始高频残差信号两部分,使这两部分各具有若干个频带;(A2)对原始低频残差信号进行编码,得到低频残差信号编码输出;再对低频残差信号编码在编码端进行本地解码,得到解码低频残差信号;(A3)根据解码低频残差信号与原始高频残差信号的相似性或相关性,对高频残差参数进行分析,即选择一个频带匹配策略对解码低频残差信号和原始高频残差信号进行频带匹配,并计算最佳匹配的高频残差参数,得到包括频带划分结果、频域残 The audio signal encoding and decoding method according to claim 1, wherein the frequency domain analysis of the residual and the coding process of magic, comprising the steps of: (Al) and the like in accordance with the frequency bandwidth, or critical bands, octave band or the band division method, the original first frequency-domain residual signal is band division, and select a band division coding rate according to the endpoint audio encoder settings, the original residual signal into the frequency domain of the original signal and the original residual low frequency high residual signal frequency in two parts, so that the two parts each having a plurality of frequency bands; (A2) of the original low-frequency residual signal is encoded residual signal to obtain the low frequency encoded output; low again locally decoded residual signal encoded on the encoding side to obtain a decoded lower band residual signal; (A3) according to similarity or correlation between the decoded residual signal and the original low-frequency residual signal, the residual high-frequency analysis parameters, i.e. a selection of the decoded low-band matching policy residues difference signal and the original residual signal frequency band matching, and calculates the best match frequency residual parameters to obtain results of divisions including band, frequency-domain residues 高低频段分界频率、最佳匹配频带频率位置、能量匹配因子、声道耦合参数和帧间扩展参数在内的原始高频残差参数;(A4)对原始高频残差参数进行编码,得到高频残差参数编码输出。 Boundary frequency and low frequency, the frequency band of the best matching position, the energy matching factor, coupling parameters and inter-channel extension parameters including parameters original high residuals; (A4) of the original high residual coding parameters, obtain high encoded output frequency residual parameters.
  3. 3.根据权利要求2所述的音频信号编解码方法,其中步骤(Α;?)所述的频带匹配策略包括:高频残差信号频带在低频残差信号区域进行匹配、高频残差信号频带在低频残差信号区域及其扩展区域进行匹配、通过声道耦合进行高频残差信号匹配和通过帧间扩展进行高频残差信号匹配。 3. The method of encoding and decoding an audio signal according to claim 2, wherein the step (Α ;?) of the band matching policy comprises: a high-frequency band residual signal to match, a high-frequency region of the residual signal in the residual signal low matching in the low frequency band residual signal region and the expansion region, high-frequency coupling channel by matching the residual signal and a high frequency signal by matching the residual inter extension.
  4. 4.根据权利要求3所述的音频信号编解码方法,其中所述的高频残差信号频带在低频残差信号区域进行匹配,是用各个高频残差信号频带在低频残差信号区域进行匹配,寻找误差最小或相关性最大的最佳匹配频带。 The audio signal encoding and decoding method according to claim 3, wherein said high frequency band of the residual signal to match the low frequency region in the residual signal, the residual signal is in the low frequency region of the respective frequency band residual signal match, find the minimum error or maximum correlation best matching band.
  5. 5.根据权利要求3所述的音频信号编解码方法,其中所述的高频残差信号频带在低频残差信号区域及其扩展区域进行匹配,是先将频率最低的高频残差信号频带在低频残差信号区域进行匹配,寻找误差最小或相关性最大的最佳匹配频带,并用此最佳匹配频带重构频率最低的高频残差信号频带;再将这个重构的频率最低的高频残差信号频带加入原来的低频残差信号,形成一个新组成的低频残差信号;然后以这个新组成的低频残差信号为基础,对更高频率的高频残差信号频带进行匹配,直至完成所有高频残差信号频带的匹配。 The audio signal encoding and decoding method according to claim 3, wherein said high frequency band of the residual signal to match the residual signal in a low frequency region and the expansion region, the lowest frequency of the first frequency band residual signal in the low frequency region matching the residual signal, to find the minimum error or maximum correlation best matching band, and with this best match the frequency of the lowest frequency band reconstruction band residual signal; then the lowest frequency reconstructed high frequency band residual signal is added to the original low-frequency residual signal, the residual signal to form a new low frequency component; then the new residual signal composed of low frequency based on the frequency band of the residual signal of higher frequency matching, until the completion of all the matching frequency band residual signal.
  6. 6.根据权利要求3所述的音频信号编解码方法,其中所述的通过声道耦合进行高频残差信号匹配,是指多声道音频信号情况下,任意一个声道的高频残差信号频带采用高频残差信号频带在低频残差信号区域进行匹配或高频残差信号频带在低频残差信号区域及其扩展区域进行匹配,利用所在声道的低频残差信号对所在声道的高频残差信号进行匹配, 并采用高频残差信号频带在低频残差信号区域进行匹配或高频残差信号频带在低频残差信号区域及其扩展区域进行匹配,利用其它声道的低频残差信号对所在声道的高频残差信号进行匹配。 6. The method of encoding and decoding an audio signal according to claim 3, wherein said coupling channel through the high frequency matching the residual signal, refers to a case where multi-channel audio signals, a channel of any residual HF a high frequency signal band residual signal frequency band for matching or matching residual signal in the low frequency band residual signal region and the extended region in the low frequency region of the residual signal, where the use of low-frequency channels where channel residual signal residual signal frequency matches the residual signal and high frequency or high frequency band residual signal to match the residual signal in a low frequency region in the low frequency band residual signal to match the region and the expansion region, with other channels low residual signal high frequency channel where the residual signal to match.
  7. 7.根据权利要求3所述的音频信号编解码方法,其中所述的通过帧间扩展进行高频残差信号匹配,是指当前帧的高频残差信号频带采用高频残差信号频带在低频残差信号区域进行匹配、高频残差信号频带在低频残差信号区域及其扩展区域进行匹配或通过声道耦合进行高频残差信号匹配,利用当前帧的低频残差信号对当前帧的高频残差信号进行匹配, 并采用高频残差信号频带在低频残差信号区域进行匹配、高频残差信号频带在低频残差信号区域及其扩展区域进行匹配或通过声道耦合进行高频残差信号匹配,利用前一帧或前若干帧的重构频域残差信号,包括重构低频残差信号和重构高频残差信号,对当前帧的高频残差信号进行匹配。 The audio signal encoding and decoding method according to claim 3, wherein said high frequency signal by matching the residual inter extension, refers frequency band residual signal of the current frame high-frequency band in the residual signal residual signal to match the low frequency region, the residual high-frequency signal in the low frequency band residual signal to match its extension region or a high-frequency region by matching the residual signal coupling channel, using a low frequency signal of a current frame of a residual current frame residual signal frequency matching, and a high frequency band residual signal in the residual signal to match the low area, high frequency band residual signal matching in the low frequency region and a residual signal by channel extension region or coupling residual signal to match the high-frequency reconstructed using the previous frame or several frames before a frequency domain residual signal including the reconstructed residual signal and the low frequency reconstructed residual signal, the residual high-frequency signal of a current frame are match.
  8. 8.根据权利要求2所述的音频信号编解码方法,其中步骤(Α;?)所述的计算最佳匹配的高频残差参数,包括如下步骤:(Bi)在ρ e [0,fD]、qe [fD, fz]范围,计算归一化低频残差频带信号兔(ρ,/)和归一化高频残差频带信号^/(P,/),其中,P和q分别表示屯(/?,/)和:^Q?,/)所在频带的起始位置,fz 为残差信号不为零所对应的最高频率,fD为低频残差信号和高频残差信号分界频率,1为频带内频点指针。 8. The audio signal encoding and decoding method according to claim 2, wherein said high-frequency residual parameters calculated in step (Α ;?) best match, comprising the steps of: (Bi) at ρ e [0, fD ], qe [fD, fz] range, compute a normalized residual low frequency band signal rabbit (ρ, /) and the residual normalized frequency band signal ^ / (P, /), where, P and q are Tun (/, /?), and: ^ Q, /) where the band starting position, FZ is not zero residual signal corresponding to the highest frequency, low frequency fD residual signal and the high frequency boundaries residual signal frequency? , a pointer within a frequency band. (B2)在pe [0,fD]、qe [fD,fz]范围,计算匹配失真测度d(p,q)或匹配相关函数r(p,q);(B3)在ρ e [0,fD]、qe [fD, fz]范围,将失真测度d(p,q)的最小值或相关函数r (p, q)的最大值所对应的P和q的取值Ptl和%为最佳频带匹配位置; (B4)计算最佳能量匹配因子(^(Ptl, Q0)。 (B2) in the pe [0, fD], qe [fD, fz] range, calculates a matching distortion measure d (p, q) or matching correlation function r (p, q); (B3) at ρ e [0, fD ], qe [fD, fz] range, the minimum maximum distortion measure d (p, q) or correlation function r (p, q) corresponding to the values ​​of P and q% of the optimal frequency band and Ptl matching position; (B4) calculates the optimum energy matching factor (^ (Ptl, Q0).
  9. 9.根据权利要求1所述的音频信号编解码方法,其中步骤4)所述的频域残差解码和重构处理,包括如下步骤:(Cl)对接收到的低频残差信号编码进行解码,得到解码低频残差信号; (C2)对接收到的高频残差参数编码进行解码,得到解码高频残差参数; (C3)利用解码低频残差信号、解码高频残差参数中的最佳匹配频带频率位置和能量匹配因子,重构高频残差信号频带4(¾,1);(C4)利用解码高频残差参数中的频带划分方法、声道耦合参数和帧间扩展参数,将得到的重构高频残差信号频带进行组合,得到重构高频残差信号;(C5)利用解码高频残差参数中的频域残差高低频段分界频率,将解码低频残差信号与重构高频残差信号进行组合,得到重构频域残差信号。 Decoding (Cl) low received residual signal coding: 9. The audio signal encoding and decoding method according to claim 1, wherein the frequency domain residual decoding and reconstruction processing according to step 4), comprising the steps of to obtain a decoded lower band residual signal; a high residual coding parameters (C2) the received decoded to obtain decoded high frequency residual parameters; (a C3) residual signal using the decoded low frequency, high frequency residual parameters decoded best matching position and energy band frequency matching factor, the reconstructed high frequency band residual signal 4 (¾, 1); band division method (C4) using the decoded residual frequency parameter, coupling parameters and inter-channel extension parameters, the high-frequency reconstructed residual signal of the frequency band obtained by combining a high-frequency reconstruction to obtain a residual signal; (C5) using the frequency domain residual low frequency band decoded high frequency residuals boundary parameters, the decoded low-residue frequency difference signal reconstructed residual signal are combined to obtain the reconstructed residual signal in frequency domain.
  10. 10. 一种基于时域滤波和高频残差重构的音频编解码方法,包括如下步骤:Tl)在编码端,音频原始时域信号经过时域感知滤波处理,得到原始时域残差信号; T2)原始时域残差信号经过时频变换处理,得到原始频域残差信号; T3)原始频域残差信号经过频域残差分析和编码处理,得到低频残差信号编码和高频残差参数编码,并输出到传输信道或存储介质。 An audio codec in time domain filtering and high-frequency reconstruction based on the residuals, comprising the steps of: Tl) at the encoder, the original time domain audio signal is time domain perceptual filtering process, to obtain the original time domain residual signal ; T2) original time-domain residual signal after the frequency transform processing to obtain the original frequency-domain residual signal; T3) after the original frequency domain residual signal and encoding the frequency domain residual analysis process, to obtain a high-frequency and low-frequency residual signal coding residual coding parameter, and outputs it to a transmission channel or storage medium. T4)在解码端,接收来自输出到传输信道或存储介质的低频残差信号编码和高频残差参数编码,并对其进行频域残差解码和重构处理,得到重构频域残差信号; T5)对重构频域残差信号进行时频逆变换处理,得到重构时域残差信号; T6)将重构时域残差信号进行时域逆滤波处理,得到音频重构时域信号。 T4) decoding side receives the output from the low-frequency and high-frequency residual residual signal coding parameters coded transport channel or storage medium, and subjected to a frequency domain residual decoding and reconstruction processing, the frequency domain to obtain the reconstructed residual when T6) the reconstructed time-domain time-domain residual signal inverse filtering process, to obtain reconstructed audio; signal; T5) inverse transform processing on the reconstructed frequency-domain residual signal is frequency reconstruction to obtain a time domain residual signal domain signal.
CN 200810232760 2008-12-22 2008-12-22 Method for encoding and decoding audio CN101436407B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN 200810232760 CN101436407B (en) 2008-12-22 2008-12-22 Method for encoding and decoding audio

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN 200810232760 CN101436407B (en) 2008-12-22 2008-12-22 Method for encoding and decoding audio

Publications (2)

Publication Number Publication Date
CN101436407A true CN101436407A (en) 2009-05-20
CN101436407B true CN101436407B (en) 2011-08-24

Family

ID=40710814

Family Applications (1)

Application Number Title Priority Date Filing Date
CN 200810232760 CN101436407B (en) 2008-12-22 2008-12-22 Method for encoding and decoding audio

Country Status (1)

Country Link
CN (1) CN101436407B (en)

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106409299A (en) 2012-03-29 2017-02-15 华为技术有限公司 Signal coding and decoding method and equipment
CN105976824A (en) 2012-12-06 2016-09-28 华为技术有限公司 Signal decoding method and device

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1535459A (en) 2001-07-26 2004-10-06 日本电气株式会社 Speech bandwidth extension and speech bandwidth extension method
WO2006103488A1 (en) 2005-03-30 2006-10-05 Nokia Corporation Source coding and/or decoding
CN1849648A (en) 2003-09-16 2006-10-18 松下电器产业株式会社 Coding apparatus and decoding apparatus
CN101048814A (en) 2004-11-05 2007-10-03 松下电器产业株式会社 Encoder, decoder, encoding method, and decoding method
CN101276587A (en) 2007-03-27 2008-10-01 北京天籁传音数字技术有限公司 Audio encoding apparatus and method thereof, audio decoding device and method thereof

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1535459A (en) 2001-07-26 2004-10-06 日本电气株式会社 Speech bandwidth extension and speech bandwidth extension method
CN1849648A (en) 2003-09-16 2006-10-18 松下电器产业株式会社 Coding apparatus and decoding apparatus
CN101048814A (en) 2004-11-05 2007-10-03 松下电器产业株式会社 Encoder, decoder, encoding method, and decoding method
WO2006103488A1 (en) 2005-03-30 2006-10-05 Nokia Corporation Source coding and/or decoding
CN101276587A (en) 2007-03-27 2008-10-01 北京天籁传音数字技术有限公司 Audio encoding apparatus and method thereof, audio decoding device and method thereof

Also Published As

Publication number Publication date Type
CN101436407A (en) 2009-05-20 application

Similar Documents

Publication Publication Date Title
US20070219785A1 (en) Speech post-processing using MDCT coefficients
US20110173006A1 (en) Audio Signal Synthesizer and Audio Signal Encoder
JP2010020251A (en) Speech coder and method, speech decoder and method, speech band spreading apparatus and method
WO2004010415A1 (en) Audio decoding device, decoding method, and program
JP2003044097A (en) Method for encoding speech signal and music signal
US20080120118A1 (en) Method and apparatus for encoding and decoding high frequency signal
CN1677490A (en) Intensified audio-frequency coding-decoding device and method
CN101083076A (en) Method and apparatus to encode and/or decode signal using bandwidth extension technology
CN101025918A (en) Voice/music dual-mode coding-decoding seamless switching method
CN1787078A (en) Stereo based on quantized singal threshold and method and system for multi sound channel coding and decoding
CN1498396A (en) Audio coding and decoding equipment and method thereof
CN1677493A (en) Intensified audio-frequency coding-decoding device and method
CN101751926A (en) Signal coding and decoding method and device, and coding and decoding system
JP2004206129A (en) Improved method and device for audio encoding and/or decoding using time-frequency correlation
US20100292994A1 (en) method and an apparatus for processing an audio signal
CN101120615A (en) Near-transparent or transparent multi-channel encoder/decoder scheme
CN102208188A (en) Audio signal encoding-decoding method and device
CN102194457A (en) Audio encoding and decoding method, system and noise level estimation method
JP2010538317A (en) Noise filler methods and apparatus
CN1677491A (en) Intensified audio-frequency coding-decoding device and method
US20100211400A1 (en) Method and an apparatus for processing a signal
JP2010538318A (en) Adaptation of the transition frequency between the noise filler and bandwidth extension
CN101849258A (en) Technique for encoding/decoding of codebook indices for quantized MDCT spectrum in scalable speech and audio codecs
CN101276587A (en) Audio encoding apparatus and method thereof, audio decoding device and method thereof
CN101568959A (en) Method, medium, and apparatus with bandwidth extension encoding and/or decoding

Legal Events

Date Code Title Description
C06 Publication
C10 Entry into substantive examination
C14 Grant of patent or utility model
EXPY Termination of patent right or utility model