CN101609684B - Post-processing filter for decoding voice signal - Google Patents

Post-processing filter for decoding voice signal Download PDF

Info

Publication number
CN101609684B
CN101609684B CN 200810039224 CN200810039224A CN101609684B CN 101609684 B CN101609684 B CN 101609684B CN 200810039224 CN200810039224 CN 200810039224 CN 200810039224 A CN200810039224 A CN 200810039224A CN 101609684 B CN101609684 B CN 101609684B
Authority
CN
Grant status
Grant
Patent type
Prior art keywords
filter
order
speech signal
low
frame
Prior art date
Application number
CN 200810039224
Other languages
Chinese (zh)
Other versions
CN101609684A (en )
Inventor
林福辉
黄鹤云
Original Assignee
展讯通信(上海)有限公司
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 provides a post-processing filter for a decoding voice signal, which is used for filtering at all frequency bands to strengthen the fundamental tone element of the decoding voice signal. The post-processing filter comprises a highpass filter and a lowpass filter, wherein the highpass filter is used for filtering the front part of each frame of a voice signal; the lowpass filter is used for filtering the rear part of each frame of a voice signal; and the front part and the rear part of each frame of the voice signal are divided under a condition that delay caused by the highpass filter is no greater than that caused by the lowpass filter. The filter provided by the invention has an excellent suppression effect of interharmonics noises and low delay like that of a lowpass filter.

Description

解码语音信号的后处理滤波器 Postprocessing filter decoded speech signal

技术领域 FIELD

[0001] 本发明涉及解码语音信号的后处理技术,尤其涉及一种解码语音信号的后处理滤波器。 [0001] The present invention relates to post-processing the decoded speech signal, and more particularly to post-processing filter of decoding a speech signal.

背景技术 Background technique

[0002] 通常语音编解码算法都是基于语音产生模型,即用线性预测技术来模拟声道模型和码激励来表征语音的声源信号。 [0002] Generally voice codec algorithms are based on a speech generation model, i.e. linear prediction technique to simulate a sound source signal to the vocal tract model and representing the speech excitation code. 常见的编解码方式为码激励线性预测编解码方式,即CELP方法。 Common codec for the code excited linear prediction codec, i.e. CELP method. 语音的声源基本有三种产生方式,一种是由元音音素产生周期性的声源信号,通常可以用谐波或者正弦波来模拟。 Speech sound source generating three basic ways, one vowel phoneme is generated by a periodic excitation signal, typically a sine wave or harmonic may be simulated. 一种是由清辅音音素产生的类似于白噪声的声源信号, 另一种则是介乎两者之间的声源信号。 One is similar to white noise sound source signals generated by the voiceless phonemes, the other is a sound source signal ground between. 根据语音的这种特性,CELP方法采用了自适应码本和固定码本来构成整个激励信号。 According to such a characteristic of speech, CELP method employs an adaptive codebook and a fixed codebook excitation constitute the entire signal. 其中自适应码本主要表征了语音激励里的谐波,而固定码本则表征了剩余分量。 Wherein the adaptive codebook excitation is mainly characterized in harmonic speech, while fixed codebook is characterized by the residual component.

[0003] 由于码率的有限性,解码语音信号不可避免地带有一些失真。 [0003] Because of the limited bit rate, the decoded speech signal is inevitably some distortion. 目前的几类语音编解码算法,包括CELP,都花费了较多比特数(bit)表征语音的周期性,因此即使在低码率下,语音的周期性也会得到比较完整的保持。 The current types of speech coding algorithm, including the CELP, more bits are spent periodic number (bit) representing the speech, so that even at a low bit rate voice also periodically obtained more complete retention. 然而,谐波分量以外的其他分量则保持得比较差。 However, other components other than a harmonic component of the retention than poor. 因此,在各个谐波分量之间的噪声会相对较强,抑制这部分噪声是有必要的。 Thus, the noise between the respective harmonic components will be relatively strong, the noise suppressing portion that is necessary. 现有的语音后处理方法有很多,包括对固定码本的低频加重,对自适应码本的加重,对整个激励的梳状滤波等等。 There are many conventional speech post processing method, comprises a low frequency emphasis of the fixed codebook, the adaptive codebook is increased, the excitation of the entire comb filtering and the like.

[0004] 现有的后处理滤波器设计方法主要包括以下几种典型的实现方式: [0004] The conventional post-processing filter design method mainly includes the following exemplary ways:

[0005] 一种方法是加重自适应码本激励信号在总激励信号的比重。 [0005] One method is to increase the proportion of the adaptive codebook excitation signal in the total excitation signal. 实现方法是将解码的自适应码本激励信号乘以一个权重,再加到总激励上并最后进行能量归一化处理。 Implementation is the decoded adaptive codebook excitation signal is multiplied by a weight, and added to the total excitation energy and finally normalized.

[0006] 另一种方法则是对固定码本激励进行滤波处理,它的原理是将固定码本激励的高频成分进行加重,再和自适应码本激励相加,产生总激励。 [0006] Another approach is the fixed codebook excitation filter processing is performed, it is the principle of the fixed codebook excitation for the high frequency components increase, the adaptive codebook excitation and then summed to produce the total excitation.

[0007] 还有一种经典方法则是对某一段频率的信号而不是所有信号进行梳状滤波。 [0007] Another method is the classical signal instead of frequencies, all signals will be a comb filter. 发明内容 SUMMARY

[0008] 本发明所要解决的问题是提供一种解码语音信号的后处理滤波器,用以在所有频带进行滤波,以增强解码语音信号的基音成分。 [0008] The present invention aims to solve is to provide a post-processed decoded speech signal filter for filtering in all frequency bands, to enhance the pitch component of the decoded speech signal.

[0009] 本发明提出一种解码语音信号的后处理滤波器,包括:高阶滤波器,用以对语音信号的每一帧的前一部分进行滤波;低阶滤波器,用以对语音信号的每一帧的后一部分进行滤波;其中该语音信号的每一帧的前一部分与后一部分的划分满足使该高阶滤波器所导致的延迟不大于该低阶滤波器所导致的延迟。 [0009] The present invention provides a post-processing filter of decoding a speech signal, comprising: a high-order filter to the front part of the speech signal for each frame is filtered; low order filter to the speech signal after filtering a portion of each frame; wherein the front portion and rear portion of the divided meet each frame of the speech signal so that the high-order filter is no greater than the delay due to delay caused by low-order filter.

[0010] 在上述的滤波器中,该语音信号的每一帧的前一部分与后一部分是依据该高阶滤波器与该低阶滤波器的阶数划分。 [0010] In the filter, the front portion and the rear portion of each frame of the speech signal is a high order filter according to the order of the low order filter division number.

[0011] 在上述的滤波器中,该高阶滤波器的阶数为1,1为正整数,该低阶滤波器的阶数为m,m为正整数且1 >m,其中对于具有N个点的语音信号帧,该前一部分包含前(N-1+m)个点,该后一部分包含后(1-m)个点。 [0011] In the filter, the order of the higher order filters 1,1 is a positive integer, the order of the low order filter is m, m is a positive integer and 1> m, wherein for having N speech signal frame points before (N-1 + m) points of the front portion of which comprises, after the portion containing the rear (1-m) points.

[0012] 在上述的滤波器中,该高阶滤波器与该低阶滤波器为有限冲击响应滤波器。 [0012] In the filter, the filter order filter in response to the low-order filter is a finite impulse.

[0013] 在上述的滤波器中,该高阶滤波器为二阶滤波器,该低阶滤波器为一阶滤波器。 [0013] In the filter, the high-order filter is a second order filter, the low-order filter is a first order filter.

[0014] 在上述的滤波器中,该高阶滤波器满足: [0014] In the filter, the high-order filter is satisfied:

[0015] h(n) = δ (η)+G δ (η+Τ) +G δ (η+2Τ); [0015] h (n) = δ (η) + G δ (η + Τ) + G δ (η + 2Τ);

[0016] 而该低阶滤波器满足: [0016] The low-order filter and satisfies:

[0017] g (η) = δ (η) +G δ (η+Τ); [0017] g (η) = δ (η) + G δ (η + Τ);

[0018] 其中T为所述语音信号的基音周期,G为所述语音信号的基音增益。 [0018] wherein T is the pitch period of the speech signal, G is the pitch gain of the voice signal.

[0019] 在上述的滤波器中,对于具有N个点的语音信号帧,该高阶滤波器是对其中的前(NTI)个点滤波,该低阶滤波器是对其中的后(Τ+1)个点滤波。 [0019] In the filter, for a speech signal frame having N points, the high-order filter is a pair of front (NTI) which points in the filter, the filter is a low order in which the post (Τ + 1) point filter.

[0020] 本发明不同于目前的后处理技术的是,它是对所有频带进行的滤波,以增强解码语音信号的基音成分。 [0020] The present invention differs from the current post-processing techniques is that it is all of the filter bands to enhance the decoded speech signal pitch component. 本发明所提出的滤波器结合了高阶滤波器和低阶滤波器,因而具有较好的对谐波间噪声的抑制效果,又具有相当于低阶滤波器的较低延迟。 Filter proposed by the present invention combines the high-order and low-order filters filter, which has good inter-harmonic noise suppression effect, but also has low-order filter corresponding to low delay.

具体实施方式 detailed description

[0021] 本发明提出一种对语音信号的后处理滤波器,用以在所有频带对语音信号进行基音增强滤波。 [0021] The present invention proposes a post-processing filter of the speech signal, for performing pitch enhancement filtering the speech signal in all bands.

[0022] 可以预见的是,滤波器的性能直接取决于其滤波器阶数,更高阶的滤波器往往会取得更好的滤波效果。 [0022] It is contemplated, which depends directly on the performance of the filter order of the filter, higher order filters tend to achieve better filtering effect. 然而,同时也会带来更大的延时或者需要耗费更多的缓存。 However, it will also lead to greater delays or takes more cache. 为此,本发明的实施例提出一种滤波器设计方法,它应用于语音的后处理。 To this end, embodiments of the present invention provides a filter design method, which is applied to the post-processed speech.

[0023] 假设有两组滤波器h和g,它们的阶数分别为1和m且1 > m,对于每一帧语音信号s (η),η = 0,1,...,NI,本发明采用如下的滤波器: [0023] Suppose there are two filters h and g, respectively, their order and 1 and m 1> m, for each frame of the speech signal s (η), η = 0,1, ..., NI, filter the present invention is as follows:

[0024] [0024]

Figure CN101609684BD00041

[0025] 其中高阶滤波器h(n)对每一帧语音信号的前一部分进行滤波,低阶滤波器g(n) 对每一帧语音信号的后一部分进行滤波。 [0025] wherein the higher order filter h (n) of the front portion of each frame of the speech signal is filtered, the low order filter g (n) after a portion of each frame of the speech signal is filtered. 在此,前一部分和后一部分的划分满足使滤波器h(n)所导致的延迟不大于滤波器g (η)所导致的延迟。 Here, the front portion and the rear portion divided satisfy the condition that the filter h (n) is not greater than the delay caused by the filter g (η) resulting in a delay. 举例来说,滤波器h (η)本身的延迟为Tl,滤波器g(n)本身的延迟为T2,由于前一部分处理点的提前,使滤波器h(n)对整个滤波所造成的延迟减少T3,而滤波器g(n)所造成的延迟仍为T2,因此只要满足T1-T3 ^ T2, 即可满足滤波器的延迟设计要求。 For example, the filter h (η) delay itself is Tl, the filter g (n) delay itself is T2, due to the advance of the front part of the processing point, so that the filter h (n) of the delay caused by filtering whole reduction T3, delay filter g (n) is still caused by T2, so long as it satisfies T1-T3 ^ T2, to meet the design requirements of the filter delay.

[0026] 一般地,根据滤波器h(n)和g(n)的阶数来进行处理点的划分。 [0026] Generally, in accordance with the filter h (n) and g (n) in order to divide the processing point. 例如,在前N-1+m 个点,用高阶滤波器(即1阶)进行更佳的处理,而在后半部分,即1-m个点,用低延迟的低阶滤波器(即m阶)来保证延迟足够小。 For example, the previous N-1 + m points, a better order filter processing (i.e., step 1), and in the latter half, i.e., 1-m points with low delay low-order filters ( i.e. m-th order) to ensure that the delay is sufficiently small.

[0027] 需要指出的是,在滤波器表达式(1)里,可以采用任意一种HR滤波器(h和g),在此并不限定。 [0027] It is noted that, in the filter of Expression (1) where any one of HR filters (h and g) may be employed, not limited to this. 并且滤波器h和g可以有任意的阶数1和m,只要满足1 > m。 And the filter g and h can have any order and 1 m, as long as 1> m.

[0028] 下面描述一种主要的实施例来具体说明上述的滤波器,所描述的实施例仅作为举例的目的,并不意图限制本发明的范围。 [0028] The following describes a main embodiments specifically described above to a filter, the described embodiment example purposes only, not intended to limit the scope of the invention.

[0029] 考虑到语音的周期性,通常的后处理滤波器都是基于以基音周期T和基音增益G 为主要参数来构造有限冲击响应滤波器(FIR)。 [0029] Taking into account the periodicity of the speech, it is usually based on the post-processing filter to the pitch period T and pitch gain G as the main parameter to construct a finite impulse response filter (FIR). 一阶的FIR为:[0030] g(n) = δ (η) +G δ (η+Τ) (2) A FIR order of: [0030] g (n) = δ (η) + G δ (η + Τ) (2)

[0031] 其中,δ (η)为冲击响应。 [0031] wherein, δ (η) for the impulse response. 其频率响应为: Its frequency response is:

[0032] [0032]

Figure CN101609684BD00051

[0033] 表达式O)的滤波器需要往前帧借用T个采样点,需要一定的缓存保存。 [0033] Expression O) of the filter needs to borrow forward frame T sampling points requires a certain cache holds.

[0034] 二阶的FIR 为: [0034] The second-order FIR as follows:

[0035] h(n) = δ (η) +G δ (η+Τ) +G δ (η+2Τ) (4) [0035] h (n) = δ (η) + G δ (η + Τ) + G δ (η + 2Τ) (4)

[0036] 其频率响应为: [0036] The frequency response is:

[0037] Η(ω) = l+2Gcos(coT) (5) [0037] Η (ω) = l + 2Gcos (coT) (5)

[0038] 表达式的滤波器需要往前帧借用T个采样点的同时,还需要往后借用T个采样点,它同时需要一定的缓存以及一定的延迟。 [0038] Expression of the filter needs to borrow frame forward while sampling points T, T needs to borrow later sampling points, it also requires a certain cache, and a certain delay. 然而同时也可以看到,它对谐波之间频率响 However, also can be seen, it has a frequency between the harmonic response

TT 371 TT 371

应…n...附近频带)的抑制效果(增益越小抑制效果越好)比一阶滤波器要好。 Should ... ... n-band near) inhibitory effect (the smaller the better the inhibition gain) is better than a first order filter.

因此,滤波器可以设计为: Thus, the filter may be designed to:

[0039] [0039]

Figure CN101609684BD00052

[0040] 可以看出,此滤波器l(n)具有较好的对谐波间噪声的抑制效果,且具有相当于一阶滤波器的较低延迟T。 [0040] As can be seen, this filter l (n) has good inter-harmonic noise suppression effect, and having a lower order filter corresponding to a delay T.

[0041] 在上述的实施例中,基音周期T可以采用任意方法得到。 [0041] In the above embodiment, the pitch period T may be obtained using any method. 在一个实施例中,可以从语音解码器得到自适应码本延迟参数作为基音周期。 In one embodiment, can be obtained as an adaptive codebook pitch delay parameter from the speech decoder.

[0042] 在上述的实施例中,基音增益G也可以采用任意方法得到。 [0042] In the above embodiment, the pitch gain G can be obtained using any method. 在在一个实施例中,可以从语音解码器得到自适应码本增益作为基音增益,或者增益的二分之一。 In one embodiment, can be obtained from the speech decoder adaptive codebook pitch gain as a gain, or the gain of one-half.

[0043] 综上所述,根据本发明的上述实施例,当同时选用低阶滤波和高阶滤波时,可以设计一个和低阶滤波同样延迟的滤波器,但其质量也好于低阶滤波器,接近高阶滤波器的效 [0043] As described above, according to the embodiments of the present invention, when simultaneously selected low-order and higher-order filtering filter, and a low-order filter can be designed to delay the same filter, but better than the low-order filtering mass , proximity effects order filter

Claims (7)

  1. 1. 一种解码语音信号的后处理滤波器,包括:高阶滤波器,用以对语音信号的每一帧的前一部分处理点进行滤波;低阶滤波器,用以对语音信号的每一帧的后一部分处理点进行滤波;其中该语音信号的每一帧的前一部分处理点与后一部分处理点的划分满足使该高阶滤波器所导致的延迟不大于该低阶滤波器所导致的延迟。 A post-processing filter decoded speech signal, comprising: a high-order filter to the first part of the processing of each frame of the speech signal point filtering; low order filter for each speech signal after part of the processing point of the frame is filtered; wherein a portion of the processing division points before and after the treatment portion of the point of each frame of the speech signal satisfy the condition that the high order filter is no greater than the delay due to low-order filter resulting delay.
  2. 2.如权利要求1所述的滤波器,其特征在于,该语音信号的每一帧的前一部分处理点与后一部分处理点是依据该高阶滤波器与该低阶滤波器的阶数划分。 2. The filter according to claim 1, characterized in that the front part of the processing part of the processing point and the point of each frame of the speech signal according to the high-order filter is the low-order dividing filter order .
  3. 3.如权利要求2所述的滤波器,其特征在于,该高阶滤波器的阶数为1,1为正整数,该低阶滤波器的阶数为m,m为正整数且1 > m,其中对于具有N个点的语音信号帧,该前一部分处理点包含前N-1+m个点,该后一部分处理点包含后1-m个点。 3. The filter according to claim 2, wherein the order of the higher order filters 1,1 is a positive integer, the order of the low order filter is m, m is a positive integer and 1> m, wherein for speech signal frame having N points, the front part of the processing site comprises N-1 + m before the point, the rear portion of the process comprising 1-m point after point.
  4. 4.如权利要求1所述的滤波器,其特征在于,该高阶滤波器与该低阶滤波器为有限冲击响应滤波器。 4. The filter according to claim 1, characterized in that the high-order filter and the low-order filter is a finite impulse response filter.
  5. 5.如权利要求4所述的滤波器,其特征在于,该高阶滤波器为二阶滤波器,该低阶滤波器为一阶滤波器。 5. The filter according to claim 4, wherein the high order filter is a second order filter, the low-order filter is a first order filter.
  6. 6.如权利要求5所述的滤波器,其特征在于,该高阶滤波器满足:h(n) = δ (η) +G δ (η+Τ) +G δ (η+2Τ);该低阶滤波器满足:g(n) = δ (η) +G δ (η+Τ);其中T为所述语音信号的基音周期,G为所述语音信号的基音增益,δ (η)为冲击响应。 6. The filter according to claim 5, wherein the higher order filter satisfies: h (n) = δ (η) + G δ (η + Τ) + G δ (η + 2Τ); the low order filter satisfies: g (n) = δ (η) + G δ (η + Τ); wherein T is the pitch period of the speech signal, G is the pitch gain of the speech signal, δ (η) of impulse response.
  7. 7.如权利要求6所述的滤波器,其特征在于,对于具有N个点的语音信号帧,该高阶滤波器是对其中的前NTI个点滤波,该低阶滤波器是对其中的后Τ+1个点滤波。 7. The filter according to claim 6, characterized in that for speech signal frame having N points, the high-order filter is a filter former NTI points therein, wherein the low order filter is the after filtering Τ + 1 points.
CN 200810039224 2008-06-19 2008-06-19 Post-processing filter for decoding voice signal CN101609684B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN 200810039224 CN101609684B (en) 2008-06-19 2008-06-19 Post-processing filter for decoding voice signal

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN 200810039224 CN101609684B (en) 2008-06-19 2008-06-19 Post-processing filter for decoding voice signal

Publications (2)

Publication Number Publication Date
CN101609684A true CN101609684A (en) 2009-12-23
CN101609684B true CN101609684B (en) 2012-06-06

Family

ID=41483408

Family Applications (1)

Application Number Title Priority Date Filing Date
CN 200810039224 CN101609684B (en) 2008-06-19 2008-06-19 Post-processing filter for decoding voice signal

Country Status (1)

Country Link
CN (1) CN101609684B (en)

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1424712A (en) 2002-12-19 2003-06-18 北京工业大学 Method for encoding 2.3kb/s harmonic wave excidted linear prediction speech
CN1460992A (en) 2003-07-01 2003-12-10 北京阜国数字技术有限公司 Low-time-delay adaptive multi-resolution filter group for perception voice coding/decoding
CN1710810A (en) 2005-06-30 2005-12-21 天津通广三星电子有限公司 Parallel overlap frequency-domain digital filter array
EP1798724A1 (en) 2004-11-05 2007-06-20 Matsushita Electric Industrial Co., Ltd. Encoder, decoder, encoding method, and decoding method
EP1906705A1 (en) 2005-07-15 2008-04-02 Matsushita Electric Industrial Co., Ltd. Signal processing device

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1424712A (en) 2002-12-19 2003-06-18 北京工业大学 Method for encoding 2.3kb/s harmonic wave excidted linear prediction speech
CN1460992A (en) 2003-07-01 2003-12-10 北京阜国数字技术有限公司 Low-time-delay adaptive multi-resolution filter group for perception voice coding/decoding
EP1798724A1 (en) 2004-11-05 2007-06-20 Matsushita Electric Industrial Co., Ltd. Encoder, decoder, encoding method, and decoding method
CN1710810A (en) 2005-06-30 2005-12-21 天津通广三星电子有限公司 Parallel overlap frequency-domain digital filter array
EP1906705A1 (en) 2005-07-15 2008-04-02 Matsushita Electric Industrial Co., Ltd. Signal processing device

Also Published As

Publication number Publication date Type
CN101609684A (en) 2009-12-23 application

Similar Documents

Publication Publication Date Title
Yegnanarayana et al. Enhancement of reverberant speech using LP residual signal
Yegnanarayana et al. An iterative algorithm for decomposition of speech signals into periodic and aperiodic components
US5864794A (en) Signal encoding and decoding system using auditory parameters and bark spectrum
US5924061A (en) Efficient decomposition in noise and periodic signal waveforms in waveform interpolation
US6047254A (en) System and method for determining a first formant analysis filter and prefiltering a speech signal for improved pitch estimation
US20020052736A1 (en) Harmonic-noise speech coding algorithm and coder using cepstrum analysis method
US20060130637A1 (en) Method for differentiated digital voice and music processing, noise filtering, creation of special effects and device for carrying out said method
US6003000A (en) Method and system for speech processing with greatly reduced harmonic and intermodulation distortion
JPH08123495A (en) Wide-band speech restoring device
US8239190B2 (en) Time-warping frames of wideband vocoder
JP2005534950A (en) The method for concealing efficient frame erasure in a speech codec which is based on linear prediction, and device
CN101140759A (en) Band-width spreading method and system for voice or audio signal
CN1989548A (en) Audio decoding device and compensation frame generation method
JP2003280696A (en) Apparatus and method for emphasizing voice
US20080056511A1 (en) Audio Signal Interpolation Method and Audio Signal Interpolation Apparatus
Ramamoorthy et al. Enhancement of ADPCM speech coding with backward-adaptive algorithms for postfiltering and noise feedback
JP2009128906A (en) Method and system for denoising mixed signal including sound signal and noise signal
JP2009058708A (en) Voice processing system, method and program
JP2002366195A (en) Method and device for encoding voice and parameter
JP2004272292A (en) Sound signal processing method
JP2001154699A (en) Hiding for frame erasure and its method
CN101617362A (en) Audio decoding device and audio decoding method
US20130030798A1 (en) Method and apparatus for audio coding and decoding
US20110125507A1 (en) Method and System for Frequency Domain Postfiltering of Encoded Audio Data in a Decoder
Hagen et al. Removal of sparse-excitation artifacts in CELP

Legal Events

Date Code Title Description
C06 Publication
C10 Entry into substantive examination
C14 Grant of patent or utility model
TR01
EE01