CN106328158A - Voice signal processing device and voice signal processing method - Google Patents

Abstract

The present invention provides a voice signal processing device and a voice signal processing method. The method comprises the steps: determining a last sampling point, in phase matching with a sampling point corresponding to a phase reference sampling point serial number, in an m-th original frequency reduction signal frame according to a phase reference sampling point serial number, corresponding to a central sampling point in an m-th updated frequency reduction signal frame, in the m-th original frequency reduction signal frame; and taking the continuous p sampling points from the last sampling point as the sampling points of an (m+1) updated frequency reduction signal frame. The method effectively improves the signal distortion caused by the mismatching phase during signal frame overlapping.

Description

Speech signal processing device and audio signal processing method

Technical field

The present invention relates to a kind of signal processing apparatus, particularly relate to a kind of speech signal processing device and voice Signal processing method.

Background technology

For Hearing Impaired, it often cannot clearly receive the voice signal of higher-frequency, example Such as consonant signal, but the signal for low frequency can clearly be heard.General prior art is by inciting somebody to action The voice signal of high frequency carries out frequency reducing, and signal frame is carried out overlap, to solve this problem.But, by In prior art when carrying out the overlap action of signal frame, can't consider whether its phase place mates, therefore Will appear from the addition of a part of signal, the situation of a part of signal subtraction at overlapping, in turn result in signal Distortion.

Summary of the invention

The present invention provides a kind of speech signal processing device and audio signal processing method, can effectively improve During signal frame overlap, phase place does not mate caused distorted signals situation.

The speech signal processing device of the present invention includes processing unit, and it is in order to frequency reducing sampled voice signal, To produce the down-scaled signals of the original down-scaled signals frame including a sequence, and produce according to original down-scaled signals frame Raw corresponding renewal down-scaled signals frame, the most each original down-scaled signals frame includes p sample point.Process single Unit also updates the middle sample point of down-scaled signals frame according in m-th original down-scaled signals frame with m-th Corresponding phase reference sample point numbering determines to take with phase reference in m-th original down-scaled signals frame Last sample point of the sample point phase matched that sampling point numbering is corresponding, will from phase reference sample point Continuous p the sample point that last sampling of the sample point phase matched that numbering is corresponding is lighted is as the M+1 the sample point updating down-scaled signals frame, and the renewal down-scaled signals frame that aliasing is adjacent, to produce friendship Folded voice signal, wherein p be positive integer, m be the positive integer more than 1.

In one embodiment of this invention, above-mentioned adjacent two renewal down-scaled signals frames are respectively provided with 50% Overlap section.

In one embodiment of this invention, above-mentioned processing unit is also according in m-th original down-scaled signals frame The sampling value of sample point adds up the first count value and the second count value, wherein when down-scaled signals is by positive half cycle When transferring negative half period to, processing unit makes zero the first count value, when down-scaled signals is transferred to negative half period by negative half period Processing unit makes zero the second count value.Processing unit by original for m-th down-scaled signals frame with phase reference The first count value corresponding to sample point that sample point numbering is corresponding or the second count value as reference value, and Benchmark value determines sampling corresponding with phase reference sample point numbering in m-th original down-scaled signals frame Last sample point of some phase matched.

In one embodiment of this invention, during above-mentioned processing unit also judges m-th original down-scaled signals frame Whether first count value sample point corresponding to corresponding with phase reference sample point numbering be less than or equal to m The second counting corresponding to sample point corresponding with phase reference sample point numbering in individual original down-scaled signals frame Value.If corresponding to sample point corresponding with phase reference sample point numbering in m-th original down-scaled signals frame The first count value corresponding with phase reference sample point numbering less than or equal in m-th original down-scaled signals frame The second count value corresponding to sample point, will in original for m-th down-scaled signals frame with phase reference sample The first count value corresponding to sample point corresponding to some numbering is as reference value, and by original for m-th frequency reducing The sample point conduct of last sampling in the sample point that in signal frame, the first count value is corresponding equal to during reference value In m-th original down-scaled signals frame with phase reference sample point sample point phase matched corresponding to numbering Later sample point, if numbering corresponding taking with phase reference sample point in m-th original down-scaled signals frame The first count value corresponding to sampling point does not takes with phase reference less than or equal in m-th original down-scaled signals frame The second count value corresponding to sample point that sampling point numbering is corresponding, by original for m-th down-scaled signals frame with The second count value corresponding to sample point corresponding to phase reference sample point numbering is as reference value, and by the The sample point that in m original down-scaled signals frame, the second count value is corresponding equal to during reference value finally samples Sample point as in m-th original down-scaled signals frame with phase reference sample point sample point corresponding to numbering Last sample point of phase matched.

In one embodiment of this invention, down-scaled signals is also multiplied by a Hamming window by above-mentioned processing unit.

The audio signal processing method of the present invention comprises the following steps.Frequency reducing sampled voice signal, to produce Including the down-scaled signals of the original down-scaled signals frame of a sequence, the most each original down-scaled signals frame includes p Sample point, wherein p is positive integer.Fall is updated with m-th according in m-th original down-scaled signals frame Frequently the numbering of the phase reference sample point corresponding to middle sample point of signal frame determines in the original frequency reducing of m-th Last sample point of sample point phase matched corresponding with phase reference sample point numbering in signal frame, Wherein m is the positive integer more than 1.By from the sample point phase place corresponding with phase reference sample point numbering Continuous p the sample point that last sampling joined is lighted updates taking of down-scaled signals frame as the m+1 Sampling point.The renewal down-scaled signals frame that aliasing is adjacent, to produce overlapping voice signal.

In one embodiment of this invention, above-mentioned adjacent two renewal down-scaled signals frames are respectively provided with 50% Overlap section.

In one embodiment of this invention, above-mentioned foundation in m-th original down-scaled signals frame with m-th The numbering of the phase reference sample point corresponding to middle sample point updating down-scaled signals frame determines former in m-th Last of sample point phase matched corresponding with phase reference sample point numbering in beginning down-scaled signals frame takes The step of sampling point comprises the following steps.Tire out according to the sampling value of sample point in m-th original down-scaled signals frame Count the first count value and the second count value, wherein make zero when down-scaled signals is transferred to negative half period by positive half cycle First count value, make zero the second count value when down-scaled signals is transferred to negative half period by negative half period.By m-th The first count value corresponding to sample point corresponding with phase reference sample point numbering in original down-scaled signals frame Or second count value as reference value.With phase place in benchmark value decision m-th original down-scaled signals frame Last sample point of the phase matched of the sample point that benchmark sample point numbering is corresponding.

In one embodiment of this invention, above-mentioned will original for m-th down-scaled signals frame take with phase reference Sampling point numbers corresponding the first count value corresponding to sample point or the second count value step as reference value Comprise the following steps.Judge in m-th original down-scaled signals frame corresponding with phase reference sample point numbering Whether the first count value corresponding to sample point is less than or equal in m-th original down-scaled signals frame and phase place base The second count value corresponding to sample point that quasi-sample point numbering is corresponding.If m-th original down-scaled signals frame In be less than or equal to m-th with the first count value corresponding to phase reference sample point sample point corresponding to numbering The second counting corresponding to sample point corresponding with phase reference sample point numbering in original down-scaled signals frame Value, corresponding to sample point corresponding with phase reference sample point numbering in original for m-th down-scaled signals frame The first count value as reference value.If m-th original down-scaled signals frame is compiled with phase reference sample point The first count value number corresponding to corresponding sample point less than or equal in m-th original down-scaled signals frame with The second count value corresponding to sample point that phase reference sample point numbering is corresponding, by original for m-th frequency reducing The second count value corresponding to sample point corresponding with phase reference sample point numbering in signal frame is as benchmark Value.

In one embodiment of this invention, if in m-th original down-scaled signals frame with phase reference sample point The first count value corresponding to sample point corresponding to numbering less than or equal in m-th original down-scaled signals frame with The second count value corresponding to sample point that phase reference sample point numbering is corresponding, above-mentioned Speech processing Method includes, by the first count value in original for m-th down-scaled signals frame equal to corresponding taking during reference value In sampling point, the sample point of last sampling is compiled with phase reference sample point as in m-th original down-scaled signals frame Last sample point of number corresponding sample point phase matched.

In one embodiment of this invention, if in m-th original down-scaled signals frame with phase reference sample point First count value corresponding to sample point of numbering correspondence is less than or equal in m-th original down-scaled signals frame Second count value sample point corresponding to corresponding with phase reference sample point numbering, at above-mentioned voice signal Reason method includes, the second count value in original for m-th down-scaled signals frame is corresponding equal to during reference value In sample point the sample point of last sampling as in m-th original down-scaled signals frame with phase reference sample point Last sample point of the sample point phase matched that numbering is corresponding.

In one embodiment of this invention, above-mentioned audio signal processing method includes, is multiplied by by down-scaled signals Hamming window.

Based on above-mentioned, embodiments of the invention according in m-th original down-scaled signals frame with m-th more Phase reference sample point corresponding to the middle sample point numbering of new down-scaled signals frame determines original in m-th Last sampling of sample point phase matched corresponding with phase reference sample point numbering in down-scaled signals frame Point, will light from last sampling of the sample point phase matched corresponding with phase reference sample point numbering Continuous p sample point as the m+1 update down-scaled signals frame sample point, effectively to improve letter During number frame overlap, phase place does not mate caused distorted signals situation.

For the features described above of the present invention and advantage can be become apparent, special embodiment below, and coordinate Accompanying drawing is described in detail below.

Accompanying drawing explanation

Fig. 1 illustrates the schematic diagram of the speech signal processing device of one embodiment of the invention；

Fig. 2 illustrates the signal processing schematic diagram of the sampled voice signal of one embodiment of the invention；

Fig. 3 illustrates the schematic diagram of the original down-scaled signals frame WL2 of one embodiment of the invention；

Fig. 4 illustrates the schematic flow sheet of the audio signal processing method of one embodiment of the invention；

Fig. 5 illustrates the schematic flow sheet of the audio signal processing method of another embodiment of the present invention.

Description of reference numerals:

102: processing unit；

104: sampling unit；

S1: primary speech signal；

S2: sampled voice signal；

W1～W4: sampled signal frame；

WL1～WL4: original down-scaled signals frame；

WL1 '～WL4 ': update down-scaled signals frame；

WH1～WH4: be multiplied by the renewal down-scaled signals frame after Hamming window；

SL, SL ', SH: down-scaled signals；

First count value；

N: sample point is numbered；

SO: overlapping voice signal；

The step of S402～S410, S502～S512: audio signal processing method.

Detailed description of the invention

Fig. 1 illustrates the schematic diagram of the speech signal processing device of one embodiment of the invention, refer to Fig. 1. Speech signal processing device includes processing unit 102 and sampling unit 104, and processing unit 102 couples Sampling unit 104, wherein processing unit 102 can such as be implemented with CPU, and samples list Unit 104 then can such as implement with logic circuit, is so not limited.Sampling unit 104 sampling is former Beginning voice signal S1, to produce sampled voice signal S2.Processing unit 102 can frequency reducing sampled voice signal S2 is to produce the down-scaled signals of the down-scaled signals frame including a sequence, and sampled voice as Figure 2 illustrates is believed Shown in the signal processing schematic diagram of number S2, sampled voice signal S2 can include the sampled signal frame of a sequence, For simplifying explanation, Fig. 2 embodiment only illustrates 4 sampled signal frame W1～W4, the most not as Limit.Down-scaled signals SL includes multiple original down-scaled signals frame WL1～WL4, owing to down-scaled signals SL is Obtained by frequency reducing sampled voice signal S2, the length of the most original down-scaled signals frame is more than sampled voice signal The length of the sampled signal frame of S2.

Processing unit 102 by adjusting the sample point that original down-scaled signals frame is comprised, and can be updated Down-scaled signals frame (the renewal down-scaled signals frame WL1 '～WL4 ' of such as Fig. 2), makes each update down-scaled signals The middle sample point of frame updates the phase matched of the initial sample point of down-scaled signals frame, Jin Ergai with the next one During kind signal frame overlap, phase place does not mate caused distorted signals situation.

Furthermore, it is understood that each original down-scaled signals frame can include p sample point, wherein p is positive integer, Processing unit 102 can will update in down-scaled signals frame with m-th in m-th original down-scaled signals frame Between sample point numbering corresponding to sample point number as phase reference sample point, and take according to phase reference Sampling point numbering determines to number corresponding taking with this phase reference sample point in m-th original down-scaled signals frame Last sample point of sampling point phase matched, and last sampling continuous p that lights since then is taken Sampling point is as the m+1 sample point updating down-scaled signals frame, so that m-th updates down-scaled signals frame Middle sample point and the phase matched of the m+1 initial sample point updating down-scaled signals frame, wherein m is Positive integer more than 1.Consequently, it is possible to m-th updates down-scaled signals frame and the m+1 renewal frequency reducing letter When number frame carries out the signal frame aliasing of 50% (that is m-th updates down-scaled signals frame and the m+1 renewal Down-scaled signals frame has the overlap section with 50% respectively), the unmatched situation of phase place just can subtract significantly Few, and improve the situation of distorted signals.

Specifically, processing unit 102 can be according to the sampling of sample point in m-th original down-scaled signals frame Accumulative first count value of value and the second count value, wherein transferred to negative half period as down-scaled signals SL by positive half cycle Time make zero the first count value, make zero the second count value when down-scaled signals is transferred to negative half period by negative half period.Tool For body, the bulk billing system of above-mentioned count value can represent according to following formula (1)～(4):

${\mathrm{PN}}_m\left(n\right)=\{\begin{array}{cc}1,& s_m\left(n\right)\&GreaterEqual;0\\ 0,& s_m\left(n\right)<0\end{array}---\left(1\right)$

${\mathrm{PN}}_m^D\left(n\right)={\mathrm{PN}}_m\left(n\right)-{\mathrm{PN}}_m(n-1)---\left(2\right)$

${\mathrm{Cot}}_m^{+}\left(n\right)=\left\{\begin{array}{cc}0,& {\mathrm{PN}}_m^D\left(n\right)=1\\ {\mathrm{Cot}}_m^{+}(n-1)+1,& else\end{array}\right.---\left(3\right)$

${\mathrm{Cot}}_m^{-}\left(n\right)=\{\begin{array}{cc}0,& {\mathrm{PN}}_m^D\left(n\right)=-1\\ {\mathrm{Cot}}_m^{-}(n-1)+1,& else\end{array}---\left(4\right)$

Wherein m is the positive integer more than 1, n=0,1,2 ..., 2N-2, N are the positive integer more than 1, s_m(n) For the sampling value of the sample point of numbering n, PN in m-th original down-scaled signals frame_mN () is by sampling value s_mN () transfers the value represented with " 1 ", " 0 ", wherein PN to_m(-1)=PN_m(0)。Original for m-th The first count value corresponding to the sample point of numbering n in down-scaled signals frame, andOriginal for m-th The second count value corresponding to the sample point of numbering n in down-scaled signals frame, whereinAndFrom formula (1), (2)For the corresponding down-scaled signals stored count when positive half cycle Value, andFor the corresponding down-scaled signals stored count value when negative half period.As shown in formula (1)～(4), In the present embodiment, by sampling value s_mN () is more than or equal to 0 and s_mSampling value when () is less than 0 n sets respectively It is 1,0, in counting the first count valueTime handleThe first count value corresponding during equal to 1 Zero, additionally and counting the second count valueTime handleThe second meter corresponding during equal to-1 Numerical value makes zero.

Processing unit 102 can will update down-scaled signals frame with in m-th in original for m-th down-scaled signals frame First count value or second corresponding to sample point of the phase reference sample point numbering correspondence obtained by Count value is as reference value, and determines in m-th original down-scaled signals frame according to this reference value and and phase place Last sample point of the sample point phase matched that benchmark sample point numbering is corresponding.Such as processing unit 102 Can determine whether corresponding to sample point corresponding with phase reference sample point numbering in m-th original down-scaled signals frame The first count value whether less than or equal in m-th original down-scaled signals frame and phase reference sample point numbering Corresponding the second count value corresponding to sample point, it can represent with following formula (5):

${\mathrm{Cot}}_m^{+}(n_{{\mathrm{Cot}}_{m-1}}-N+1)\&le;{\mathrm{Cot}}_m^{-}(n_{{\mathrm{Cot}}_{m-1}}-N+1)---\left(5\right)$

WhereinBe in m-1 original down-scaled signals frame with in m-1 original down-scaled signals frame Between the numbering corresponding to last sample point of phase matched of sample point, it is equal to the original fall of m-th Frequently in signal frame, corresponding m-th updates the numbering of the sample point of down-scaled signals last sample point of frame.Example As, in FIG. 2, it is assumed that each original down-scaled signals frame WL1～WL4 includes 0,1,2 respectively ..., 200 Deng 201 sample points, with the centre of original down-scaled signals frame WL1 in original down-scaled signals frame WL1 Corresponding to last sample point of the phase matched of sample point numbered 188, and believe in original frequency reducing In number frame WL2, the numbering of the corresponding sample point updating down-scaled signals frame last sample point of WL2 ' is also 188。For updating in down-scaled signals frame with m-th in m-th original down-scaled signals frame Between corresponding to sample point phase reference sample point numbering, the most in fig. 2, at original down-scaled signals frame In WL2 with update down-scaled signals frame WL2 ' middle sample point corresponding to sample point sample point numbered 88 (that is phase reference sample point numbered 88, N is 101).Former for m-th The first count value corresponding to sample point corresponding with phase reference sample point numbering in beginning down-scaled signals frame,For corresponding with phase reference sample point numbering in m-th original down-scaled signals frame The second count value corresponding to sample point.

If corresponding to sample point corresponding with phase reference sample point numbering in m-th original down-scaled signals frame The first count value corresponding with phase reference sample point numbering less than or equal in m-th original down-scaled signals frame The second count value corresponding to sample point, processing unit 102 by original for m-th down-scaled signals frame with First count value first count value corresponding to sample point of phase reference sample point numbering correspondence is as benchmark Value, and by sample point corresponding equal to during reference value for the first count value in original for m-th down-scaled signals frame In the sample point of last sampling as last sample point, it can represent with following formula (6), (7):

$n_{{\mathrm{Cot}}_m}^{+}\left(n\right)=\{\begin{array}{cc}n,& {\mathrm{Cot}}_m^{+}\left(n\right)={\mathrm{Cot}}_m^{+}(n_{{\mathrm{Cot}}_{m-1}}-N+1)\\ 0,& else\end{array}---\left(6\right)$

$n_{{\mathrm{Cot}}_m}=Max\left\{n_{{\mathrm{Cot}}_m}^{+}\left(n\right)\right\}---\left(7\right)$

From formula (6), (7), when corresponding to the sample point of numbering n in m-th original down-scaled signals frame First count value is equal to sampling corresponding with phase reference sample point numbering in m-th original down-scaled signals frame During the first count value corresponding to Dian,Equal to numbering n corresponding to sample point, otherwiseDeng In 0.AndIt is then allIn maximum, its represent at the original down-scaled signals of m-th The volume of last sample point of the phase matched of sample point corresponding with phase reference sample point numbering in frame Number, this sample point in order to update the initial sample point of down-scaled signals frame as the m+1.

If on the contrary, sampling corresponding with phase reference sample point numbering in m-th original down-scaled signals frame The first count value corresponding to Dian does not samples with phase reference less than or equal in m-th original down-scaled signals frame The second count value (that is formula (5) is false) corresponding to sample point that some numbering is corresponding, then processing unit 102 Will in original for m-th down-scaled signals frame with the corresponding to phase reference sample point sample point corresponding to numbering Two count values are as reference value, and the second count value in original for m-th down-scaled signals frame is equal to reference value Time corresponding sample point in the sample point of last sampling as last sample point, it can be with following formula Son (8), (9) represent:

$n_{{\mathrm{Cot}}_m}^{-}\left(n\right)=\{\begin{array}{cc}n,& {\mathrm{Cot}}_m^{-}\left(n\right)={\mathrm{Cot}}_m^{-}(n_{{\mathrm{Cot}}_{m-1}}-N+1)\\ 0,& else\end{array}---\left(8\right)$

$n_{{\mathrm{Cot}}_m}=Max\left\{n_{{\mathrm{Cot}}_m}^{-}\left(n\right)\right\}---\left(9\right)$

From formula (8), (9), whenFor the sampling of numbering n in m-th original down-scaled signals frame The second count value corresponding to Dian is numbered with phase reference sample point equal in m-th original down-scaled signals frame During corresponding the second count value corresponding to sample point,Equal to numbering n corresponding to sample point, OtherwiseEqual to 0.AndIt is then allIn maximum, its represent at m Last with the phase matched of phase reference sample point sample point corresponding to numbering in individual original down-scaled signals frame The numbering of one sample point, this sample point in order to update the initial sample of down-scaled signals frame as the m+1 Point.

As an example it is assumed that each original down-scaled signals frame WL1～WL4 in Fig. 2 includes 201 respectively Individual sample point, that is each original down-scaled signals frame WL1～WL4 include 0,1,2 respectively ..., 200 Deng 201 sample points.With the centre updating down-scaled signals frame WL2 ' in original down-scaled signals frame WL2 The phase reference sample point numbering corresponding to sample point the first count value corresponding to (it is 88) It is less than or equal in original down-scaled signals frame WL2 and the middle sample point institute updating down-scaled signals frame WL2 ' The second count value that corresponding phase reference sample point numbering sample point is correspondingAnd in original frequency reducing The middle sample point of signal frame WL2 (that is the sampling of numbered 88 in original down-scaled signals frame WL2 Point) corresponding to the first count valueIt is 18.

For finding out the initial sample point updating down-scaled signals frame WL3 ', processing unit 102 is countable original First count value in down-scaled signals frame WL2The numbering of sample point corresponding during equal to 18 (due to The first count value corresponding to the sample point of numbered 88 in original down-scaled signals frame WL2Little In the second corresponding count valueTherefore with the first count valueAs reference value).Such as figure Shown in the schematic diagram of 3 depicted original down-scaled signals frame WL2, in the fig. 3 embodiment, original fall Frequently the first count value in signal frame WL2The numbering of sample point corresponding during equal to 18 (that is not Equal to 0Value) include numbering 20,40,63,79,100,125,142,163,192 Deng sample point, wherein the sample point of numbering 192 is the first count value in original down-scaled signals frame WL2The sample point of last sampling in the sample point corresponding equal to time reference value (its value is 18), thereforeEqual to 192, processing unit 102 as update down-scaled signals frame WL3 ' initial sample point, And continuous 201 sample points that the sampling of numbering 192 in original down-scaled signals frame WL2 lighted as Update the sample point of down-scaled signals frame WL3 '.Include former as in figure 2 it is shown, update down-scaled signals frame WL3 ' Beginning down-scaled signals frame WL2 compiles in the sample point of numbering 183～200 and original down-scaled signals frame WL3 Numbers 192 and numbering 192 before sample point, numbering 92 (its in the most original down-scaled signals frame WL2 For updating middle sample point corresponding taking in original down-scaled signals frame WL3 of down-scaled signals frame WL3 ' Sampling point is numbered) can number as phase reference sample point, it is in order to update down-scaled signals frame as finding The foundation of the initial sample point of WL4 '.Similarly, also the initial sample point of down-scaled signals frame WL4 ' is updated Can obtain in a similar fashion, therefore not repeat them here.

Significantly, since original down-scaled signals frame WL1 is first original down-scaled signals frame, because of This sample point updating down-scaled signals frame WL1 ' is the sampling included by original down-scaled signals frame WL1 In point, and original down-scaled signals frame WL1 with update down-scaled signals frame WL1 ' middle sample point corresponding to Phase reference sample point numbered 100.In the present embodiment, with former in original down-scaled signals frame WL1 The volume corresponding to last sample point of the phase matched of the middle sample point of beginning down-scaled signals frame WL1 Number it is 188 (being so not limited), the wherein side of asking for of last sample point (sample point of numbering 188) Formula is similar to the above embodiments, and those skilled in the art should push away to obtain its embodiment party according to foregoing Formula, does not therefore repeat them here.

Adjust the sample point that each original down-scaled signals frame is comprised, and obtain the renewal frequency reducing of correspondence After signal frame, processing unit 102 just can carry out the aliasing of 50% to the adjacent down-scaled signals frame that updates, with Produce overlapping voice signal, owing to now each updates the middle sample point of down-scaled signals frame with the next one more The phase matched of the initial sample point of new down-scaled signals frame, therefore during signal frame overlap, phase place is not mated and is led The distorted signals situation caused will be enhanced significantly.Additionally, in part embodiment, also can obtain respectively After the renewal down-scaled signals frame that individual original down-scaled signals frame is corresponding, down-scaled signals is multiplied by Hamming window (Hamming Window), to increase the seriality updating down-scaled signals frame left and right end.As in figure 2 it is shown, After the down-scaled signals SL ' including renewal down-scaled signals frame WL1 '～WL4 ' is multiplied by Hamming window, available Including the down-scaled signals SH of renewal down-scaled signals frame WH1～WH4, down-scaled signals frame will be updated the most again WH1～WH4 carries out aliasing, i.e. can obtain overlapping voice signal SO.

Fig. 4 illustrates the schematic flow sheet of the audio signal processing method of one embodiment of the invention, refer to figure 4.From above-described embodiment, the audio signal processing method of speech signal processing device can include following step Suddenly.First, primary speech signal is sampled, to produce sampled voice signal (step S402).Then, fall Frequently sampled voice signal, to produce the down-scaled signals (step of the original down-scaled signals frame including a sequence S404), the most each original down-scaled signals frame includes p sample point, and wherein p is positive integer.Then, depend on According in m-th original down-scaled signals frame with m-th update down-scaled signals frame middle sample point corresponding to Phase reference sample point numbering determine in m-th original down-scaled signals frame with phase reference sample point compile Last sample point (step S406) of number corresponding sample point phase matched, wherein m is more than 1 Positive integer.Afterwards, certainly last of corresponding sample point phase matched will be numbered with phase reference sample point Continuous p the sample point that individual sampling is lighted is as the m+1 sample point (step updating down-scaled signals frame S408).Finally, then the renewal down-scaled signals frame that aliasing is adjacent, to produce overlapping voice signal (step S410), the most adjacent two renewal down-scaled signals frames can such as be respectively provided with the overlap section of 50%.

Fig. 5 illustrates the schematic flow sheet of the audio signal processing method of another embodiment of the present invention, refer to Fig. 5.Specifically, step S406 of Fig. 4 embodiment can include step S502～S506 in the present embodiment, That is first add up the first count value and the according to the sampling value of sample point in m-th original down-scaled signals frame Two count values, wherein make zero the first count value when down-scaled signals is transferred to negative half period by positive half cycle, works as frequency reducing Signal makes zero when being transferred to negative half period by negative half period the second count value (step S502), then that m-th is original With the first count value corresponding to phase reference sample point sample point corresponding to numbering or the in down-scaled signals frame Two count values are as reference value (step S504), and benchmark value determines m-th original frequency reducing letter the most again Last sample point of the phase matched of sample point corresponding with phase reference sample point numbering in number frame (step S506).For further, step S504 can include, first judges that the original frequency reducing of m-th is believed Whether the first count value corresponding to sample point corresponding with phase reference sample point numbering in number frame is less than With the corresponding to phase reference sample point sample point corresponding to numbering in m-th original down-scaled signals frame Two count values (step S508).If it is right with phase reference sample point numbering in m-th original down-scaled signals frame The first count value corresponding to the sample point answered is less than or equal in m-th original down-scaled signals frame and phase place base The second count value corresponding to sample point that quasi-sample point numbering is corresponding, by original for m-th down-scaled signals frame In (walk as reference value with the first count value corresponding to phase reference sample point sample point corresponding to numbering Rapid S510), in this case, can be by the first counting in original for m-th down-scaled signals frame in step S506 In the sample point that value is corresponding equal to during reference value, the sample point of last sampling is believed as the original frequency reducing of m-th Last sample point of sample point phase matched corresponding with phase reference sample point numbering in number frame.Phase Instead, if sample point institute corresponding with phase reference sample point numbering in m-th original down-scaled signals frame is right The first count value answered is not numbered with phase reference sample point less than or equal in m-th original down-scaled signals frame Corresponding the second count value corresponding to sample point, by original for m-th down-scaled signals frame with phase reference Second count value corresponding to sample point of sample point numbering correspondence is as reference value (step S512), at this Under situation, can be by the second count value in original for m-th down-scaled signals frame equal to reference value in step S506 Time corresponding sample point in last sampling sample point as in m-th original down-scaled signals frame with phase place Last sample point of the sample point phase matched that benchmark sample point numbering is corresponding.

In sum, embodiments of the invention according in m-th original down-scaled signals frame with m-th more Phase reference sample point corresponding to the middle sample point numbering of new down-scaled signals frame determines original in m-th Last sampling of sample point phase matched corresponding with phase reference sample point numbering in down-scaled signals frame Point, will light from last sampling of the sample point phase matched corresponding with phase reference sample point numbering Continuous p sample point as the m+1 update down-scaled signals frame sample point, effectively to improve letter When number frame overlaps, phase place does not mate caused distorted signals situation.

Last it is noted that various embodiments above is only in order to illustrate technical scheme, rather than right It limits；Although the present invention being described in detail with reference to foregoing embodiments, this area common Skilled artisans appreciate that the technical scheme described in foregoing embodiments still can be modified by it, Or the most some or all of technical characteristic is carried out equivalent；And these amendments or replacement, and The essence not making appropriate technical solution departs from the scope of various embodiments of the present invention technical scheme.

Claims (12)

1. a speech signal processing device, it is characterised in that including:

Processing unit, frequency reducing sampled voice signal, the original down-scaled signals frame of a sequence is included with generation Down-scaled signals, and produce corresponding renewal down-scaled signals frame according to those original down-scaled signals frames, respectively Described original down-scaled signals frame includes p sample point, and described processing unit is according in the original frequency reducing of m-th The phase reference sample point corresponding to middle sample point updating down-scaled signals frame in signal frame with m-th is compiled Number determine in described m-th original down-scaled signals frame with described phase reference sample point number corresponding taking Last sample point of sampling point phase matched, will number corresponding taking from described phase reference sample point Continuous p the sample point that last sampling described of sampling point phase matched is lighted is as the m+1 renewal The sample point of down-scaled signals frame, the renewal down-scaled signals frame that aliasing is adjacent, to produce overlapping voice signal, Wherein said phase reference sample point numbered in described m-th original down-scaled signals frame with described m The numbering of the sample point corresponding to middle sample point of individual renewal down-scaled signals frame, p is that positive integer, m are Positive integer more than 1.

Speech signal processing device the most according to claim 1, it is characterised in that described adjacent two Individual renewal down-scaled signals frame is respectively provided with the overlap section of 50%.

Speech signal processing device the most according to claim 2, it is characterised in that described process list Unit also according to the sampling value of sample point in described m-th original down-scaled signals frame add up the first count value and Second count value, wherein make zero described first counting when described down-scaled signals is transferred to negative half period by positive half cycle Value, make zero described second count value when described down-scaled signals is transferred to negative half period by negative half period, by described the The institute corresponding to sample point corresponding with described phase reference sample point numbering in m original down-scaled signals frame State the first count value or described second count value as reference value, and determine described the according to described reference value Sample point phase matched corresponding with described phase reference sample point numbering in m original down-scaled signals frame Last sample point described.

Speech signal processing device the most according to claim 3, it is characterised in that described process list Unit also judges to number corresponding taking with described phase reference sample point in described m-th original down-scaled signals frame Described first count value corresponding to sampling point whether less than or equal in described m-th original down-scaled signals frame with Described second count value corresponding to sample point that described phase reference sample point numbering is corresponding, if described the The institute corresponding to sample point corresponding with described phase reference sample point numbering in m original down-scaled signals frame State the first count value to be less than or equal in described m-th original down-scaled signals frame and described phase reference sample point Described second count value corresponding to sample point that numbering is corresponding, by original for described m-th down-scaled signals frame In with described first count value corresponding to described phase reference sample point sample point corresponding to numbering as institute State reference value, and by the first count value described in original for described m-th down-scaled signals frame equal to described benchmark In sample point corresponding during value, the sample point of last sampling is as in described m-th original down-scaled signals frame Last sample point described in the sample point phase matched corresponding with described phase reference sample point numbering, If sample point institute corresponding with described phase reference sample point numbering in described m-th original down-scaled signals frame Corresponding described first count value is not less than or equal in described m-th original down-scaled signals frame and described phase place Described second count value corresponding to sample point that benchmark sample point numbering is corresponding, by original for described m-th Described second meter corresponding to sample point corresponding with described phase reference sample point numbering in down-scaled signals frame Numerical value is as described reference value, and by the second count value described in original for described m-th down-scaled signals frame etc. In sample point corresponding when described reference value, the sample point of last sampling is as the original fall of described m-th Frequently in signal frame with described in described phase reference sample point sample point phase matched corresponding to numbering last Individual sample point.

Speech signal processing device the most according to claim 1, it is characterised in that described process list Described down-scaled signals is also multiplied by Hamming window by unit.

6. an audio signal processing method, it is characterised in that including:

Frequency reducing sampled voice signal, to produce the down-scaled signals of the original down-scaled signals frame including a sequence, The most each described original down-scaled signals frame includes p sample point, and wherein p is positive integer；

The middle sample point of down-scaled signals frame is updated with m-th according in m-th original down-scaled signals frame Corresponding phase reference sample point numbering determine in described m-th original down-scaled signals frame with described phase Last sample point of the sample point phase matched that position benchmark sample point numbering is corresponding, wherein said phase place Benchmark sample point is numbered updates frequency reducing letter in described m-th original down-scaled signals frame with described m-th The numbering of the sample point corresponding to middle sample point of number frame, m is the positive integer more than 1；And

Will from described in described phase reference sample point sample point phase matched corresponding to numbering last Continuous p the sample point that sampling is lighted is as the m+1 sample point updating down-scaled signals frame；And

The renewal down-scaled signals frame that aliasing is adjacent, to produce overlapping voice signal.

Audio signal processing method the most according to claim 6, it is characterised in that described adjacent two Individual renewal down-scaled signals frame is respectively provided with the overlap section of 50%.

Audio signal processing method the most according to claim 7, it is characterised in that according at m Individual original down-scaled signals frame updates the phase place base corresponding to middle sample point of down-scaled signals frame with m-th Quasi-sample point numbering determines to compile with described phase reference sample point in described m-th original down-scaled signals frame The step of last sample point of number corresponding sample point phase matched includes:

According to the sampling value of sample point in described m-th original down-scaled signals frame add up the first count value and Second count value, wherein make zero described first counting when described down-scaled signals is transferred to negative half period by positive half cycle Value, make zero described second count value when described down-scaled signals is transferred to negative half period by negative half period；

By sampling corresponding with described phase reference sample point numbering in original for described m-th down-scaled signals frame Described first count value corresponding to Dian or described second count value are as reference value；And

Determine described m-th original down-scaled signals frame samples with described phase reference according to described reference value Last sample point described of the phase matched of the sample point that some numbering is corresponding.

Audio signal processing method the most according to claim 8, it is characterised in that by described m Described in corresponding to sample point corresponding with described phase reference sample point numbering in individual original down-scaled signals frame First count value or described second count value include as the step of described reference value:

Judge described m-th original down-scaled signals frame numbers corresponding taking with described phase reference sample point Described first count value corresponding to sampling point whether less than or equal in described m-th original down-scaled signals frame with Described second count value corresponding to sample point that described phase reference sample point numbering is corresponding；

If sampling corresponding with described phase reference sample point numbering in described m-th original down-scaled signals frame Described first count value corresponding to Dian is less than or equal in described m-th original down-scaled signals frame and described phase Described second count value corresponding to sample point that position benchmark sample point numbering is corresponding, by former for described m-th Corresponding with described phase reference sample point numbering in beginning down-scaled signals frame corresponding to sample point described first Count value is as described reference value；And

If sampling corresponding with described phase reference sample point numbering in described m-th original down-scaled signals frame Described first count value corresponding to Dian is not less than or equal in described m-th original down-scaled signals frame with described Described second count value corresponding to sample point that phase reference sample point numbering is corresponding, by described m-th With described the corresponding to described phase reference sample point sample point corresponding to numbering in original down-scaled signals frame Two count values are as described reference value.

Audio signal processing method the most according to claim 9, it is characterised in that if described The institute corresponding to sample point corresponding with described phase reference sample point numbering in m original down-scaled signals frame State the first count value to be less than or equal in described m-th original down-scaled signals frame and described phase reference sample point Described second count value corresponding to sample point that numbering is corresponding, described audio signal processing method includes:

By right equal to institute during described reference value for the first count value described in original for described m-th down-scaled signals frame In the sample point answered the sample point of last sampling as in described m-th original down-scaled signals frame with described phase Last sample point described of the sample point phase matched that position benchmark sample point numbering is corresponding.

11. audio signal processing methods according to claim 9, it is characterised in that if described The institute corresponding to sample point corresponding with described phase reference sample point numbering in m original down-scaled signals frame State the first count value not sample with described phase reference less than or equal in described m-th original down-scaled signals frame Described second count value corresponding to sample point that some numbering is corresponding, described audio signal processing method includes:

By right equal to institute during described reference value for the second count value described in original for described m-th down-scaled signals frame In the sample point answered the sample point of last sampling as in described m-th original down-scaled signals frame with described phase Last sample point described of the sample point phase matched that position benchmark sample point numbering is corresponding.

12. audio signal processing methods according to claim 6, it is characterised in that including:

Described down-scaled signals is multiplied by Hamming window.