CN102930871B - Linear predication analysis method, device and system - Google Patents

Abstract

The invention discloses a linear prediction analysis method, device and system and relates to the field of communication. According to the linear prediction analysis method, device and system, the predicting performance of a linear prediction code can be improved and the complexity of analysis operation is low. The solution is as follows: the linear prediction analysis method comprises the following steps: acquiring signal characteristic information of at least one sample point of an input signal; carrying out comparative analysis on the signal characteristic information to obtain an analysis result; selecting a window function according to the analysis result for self-adaptively windowing the input signal to obtain a windowed signal; and processing the windowed signal to obtain a linear prediction coding coefficient for linear prediction. The linear prediction analysis method, device and system disclosed by the invention are used for linear predication coding.

Description

A kind of Linear prediction analysis method, Apparatus and system

Technical field

The present invention relates to the communications field, relate in particular to a kind of Linear prediction analysis method, Apparatus and system.

Background technology

For saving the bandwidth of voice and audio signal transmission and storage, corresponding voice and audio decoding techniques are widely used, mainly be divided at present lossy coding and lossless coding, it is in full accord that the reconstruction signal of lossy coding and original signal can not keep, but can reduce to the full extent according to sound source feature and people's perception feature the redundant information of signal.Lossless coding must guarantee that reconstruction signal and original signal are in full accord, can so that last decoding quality without any damage, lossy coding compressibility is higher in general, but reconstructed speech quality does not guarantee, lossless coding can guarantee voice quality, but compressibility is lower, about 50% left and right.

No matter in lossy coding or lossless coding, linear predictive coding (LPC, Linear Prediction Coding) model is widely used in voice coding field, and in lossy coding, Qualcomm Code Excited Linear Prediction (QCELP) model is the success of its typical case's application.Ultimate principle is: first utilize linear prediction in short-term to remove the nearly sampling point redundance of voice signal, with long-term prediction, remove again the sampling point redundance far away of voice signal, finally to the parameter producing in forecasting process and through two-stage, predict that the residual signals obtaining carries out coding transmission.

Most damages with the linear prediction analysis of lossless audio encoding and decoding and generally comprises windowing, asks auto-correlation and three modules of Levinson Algorithm for Solving, by linear prediction, obtain residual signals, then with entropy coding, residual signals is encoded to realize audio compression.

State in realization in the process of linear predictive coding, inventor finds at least to exist in prior art following problem:

During windowing, adopt fixed window function, can make linear prediction performance not reach optimum;

Or, input signal is carried out respectively to twice linear prediction analysis, once to signal, add short window, another time lengthens window to signal, can, because input signal has been carried out to twice linear prediction analysis, make the complexity of linear prediction analysis larger.

Summary of the invention

Embodiments of the invention provide a kind of Linear prediction analysis method, Apparatus and system, can improve linear prediction performance, reduce analytic operation complexity.

A Linear prediction analysis method, comprising:

The amplitude of first sampling point and the amplitude of last sampling point of obtaining input signal, described input signal comprises N sampling point, and N is positive integer;

Amplitude to the amplitude of described first sampling point and last sampling point is analyzed, and according to analysis result, input signal is carried out to self-adaptation windowing, obtains signal after windowing;

Signal after windowing is processed, obtained linear forecast coding coefficient for linear prediction.

The Linear prediction analysis method that the embodiment of the present invention provides, by obtaining the amplitude of first sampling point He last sampling point of input signal, and according to this sampling point amplitude, input signal is carried out to self-adaptation windowing, therefore, can be in the situation that the less encoder complexity of increase have improved the estimated performance of linear predictive coding.

A Linear prediction analysis method, comprising:

A Linear prediction analysis method, is characterized in that, comprising:

Obtain input signal coded system, described input signal is signal G.711;

Input signal is changed, obtained PCM signal;

Input signal coded system is analyzed, and according to analysis result, PCM signal is carried out to self-adaptation windowing, obtain signal after windowing;

Signal after described windowing is processed, obtained linear forecast coding coefficient for linear prediction.

The Linear prediction analysis method that the embodiment of the present invention provides, by input signal coded system is analyzed, according to this Signal coding mode, input signal is carried out to self-adaptation windowing, therefore, can be in the situation that the less encoder complexity of increase have improved the estimated performance of linear predictive coding.

Accompanying drawing explanation

In order to be illustrated more clearly in the embodiment of the present invention or technical scheme of the prior art, to the accompanying drawing of required use in embodiment or description of the Prior Art be briefly described below, apparently, accompanying drawing in the following describes is only some embodiments of the present invention, for those of ordinary skills, do not paying under the prerequisite of creative work, can also obtain according to these accompanying drawings other accompanying drawing.

The FB(flow block) of the Linear prediction analysis method that Fig. 1 provides for the embodiment of the present invention;

The FB(flow block) of the Linear prediction analysis method that Fig. 2 provides for the embodiment of the present invention one;

The FB(flow block) of the Linear prediction analysis method that Fig. 3 provides for the embodiment of the present invention two;

The FB(flow block) of the Linear prediction analysis method that Fig. 4 provides for the embodiment of the present invention three;

The FB(flow block) of the Linear prediction analysis method that Fig. 5 provides for the embodiment of the present invention four;

The FB(flow block) of the Linear prediction analysis method that Fig. 6 provides for the embodiment of the present invention five;

The FB(flow block) of the Linear prediction analysis method that Fig. 7 provides for the embodiment of the present invention six;

The FB(flow block) of the Linear prediction analysis method that Fig. 8 provides for the embodiment of the present invention seven;

The structured flowchart of the linear prediction analysis device that Fig. 9 provides for the embodiment of the present invention;

The structured flowchart of the linear prediction analysis device that Figure 10 provides for another embodiment of the present invention;

The structure block diagram of the linear predictive coding system that Figure 11 provides for the embodiment of the present invention;

The structure block diagram of the linear predictive coding system that Figure 12 provides for another embodiment of the present invention.

Embodiment

Below in conjunction with the accompanying drawing in the embodiment of the present invention, the technical scheme in the embodiment of the present invention is clearly and completely described, obviously, described embodiment is only the present invention's part embodiment, rather than whole embodiment.Embodiment based in the present invention, those of ordinary skills, not making the every other embodiment obtaining under creative work prerequisite, belong to the scope of protection of the invention.

Embodiments of the invention provide a kind of Linear prediction analysis method, Apparatus and system, can improve linear prediction performance, reduce analytic operation complexity.

Below in conjunction with accompanying drawing, the embodiment of the present invention is described in detail.

The Linear prediction analysis method that the embodiment of the present invention provides, as shown in Figure 1, the step of the method comprises:

S101, obtain the characteristics of signals information of at least one sampling point of input signal;

S102, characteristics of signals information is compared to analysis, obtain analysis result;

S103, according to analysis result, select window function to carry out self-adaptation windowing to input signal, obtain signal after windowing;

S104, signal after windowing is processed, obtained linear forecast coding coefficient for linear prediction.

The Linear prediction analysis method that the embodiment of the present invention provides, by input signal is analyzed, obtains result, and distribute the required window function of windowing according to analysis result self-adaptation, therefore, can be in the situation that the less encoder complexity of increase have improved the estimated performance of linear predictive coding.

Wherein, characteristics of signals information comprises any one or any number of in amplitude, energy, zero-crossing rate, signal type, frame length, coded system.

Below by specific embodiment, be illustrated.

Embodiment mono-:

The Linear prediction analysis method that the embodiment of the present invention one provides, as shown in Figure 2, the method step comprises:

S201, obtain the amplitude of first sampling point of input signal | x[0] | and the amplitude of last sampling point | x[N-1] |, wherein, x[i], i=0,1 ..., N-1 is input signal, the sampling point number that N is input signal (as 40,80,160,240,320 etc.); Input signal here refers to inputs the signal that carries out lpc analysis, it may be a frame signal, also may be a frame signal add history buffer a segment signal (as the L of a history buffer sampling point, L can adopt according to different codecs different positive integers, as 40,80,160,240,320 etc.);

S202, to sampling point amplitude | x[0] | and | x[N-1] | analyze, and according to analysis result, input signal carried out to self-adaptation windowing:

As when input sample number is 40:

If the amplitude of first sampling point of input signal | x[0] | be less than certain and preset threshold value thr (as thr=128), 4 of the foremosts point of window function is set to:

w(n)＝0.23+0.77·cos(2·π·(31-8·n)/127)，n＝0，1，2，3

Otherwise 4 of the foremosts point to window function is set to:

w(n)＝0.26+0.74·cos(2·π·(31-8·n)/127)，n＝0，1，2，3

The the 5th to the 36th point of window function is all made as to 1, that is:

w(n)＝1，n＝4，...，35

If the amplitude of last sampling point of input signal | x[39] | be less than certain and preset threshold value thr (as thr=128), 4 points backmost of window function are set to:

w(n)＝0.23+0.77·cos(2·π·(8·n-281)/127)，n＝36，37，38，39

Otherwise 4 points backmost to window function are set to:

w(n)＝0.26+0.74·cos(2·π·(8·n-281)/127)，n＝36，37，38，39

Then the window function w (n) after arranging by above-mentioned self-adaptation, n=1,2 ..., 38,39 couples of signal x (n), n=1,2 ..., 38,39 carry out windowing,

xd[n]＝x[n]·w[n]，n＝0，1，...，38，39

Obtain the signal xd[n after self-adaptation windowing]], n=0,1 ..., 38,39

And for example when input sample number is 80:

If the amplitude of first sampling point of input signal | x[0] | be less than certain and preset threshold value thr (as thr=128), 8 of the foremosts point of window function is set to:

w(n)＝0.26+0.74·cos(2·π·(31-4·n)/127)，n＝0，1，2，...，7

Otherwise 8 of the foremosts point to window function is set to:

w(n)＝0.16+0.84·cos(2·π·(31-4·n)/127)，n＝0，1，2，...，7

The the 9th to the 72nd point of window function is all made as to 1, that is:

w(n)＝1，n＝8，...，71

If the amplitude of last sampling point of input signal | x[79] | be less than certain and preset threshold value thr (as thr=128), 8 points backmost of window function are set to:

w(n)＝0.26+0.74·cos(2·π·(4·n-285)/127)，n＝72，73，74，...，79

Otherwise 8 points backmost to window function are set to:

w(n)＝0.16+0.84·cos(2·π·(4·n-285)/127)，n＝72，73，74，...，79

Then the window function w (n) after arranging by above-mentioned self-adaptation, n=0,1 ..., 78,79 couples of signal x (n), n=0,1 ..., 78,79 carry out windowing,

xd[n]＝x[n]·w[n]，n＝0，1，...，78，79

Obtain the signal xd[n after self-adaptation windowing]], n=0,1 ..., 78,79

Window function w[n] adjustment strategy can by great many of experiments, select according to different audio encoding devices, be applicable to respectively different signals.Threshold value thr selectes by great many of experiments, as thr=128 or thr=157 etc.;

S203, signal after windowing is processed, obtained linear forecast coding coefficient, for linear prediction.

The Linear prediction analysis method that the embodiment of the present invention one provides, by obtaining the amplitude of first sampling point He last sampling point of input signal, and according to this sampling point amplitude, input signal is carried out to self-adaptation windowing, therefore, can be in the situation that the less encoder complexity of increase have improved the estimated performance of linear predictive coding.

Embodiment bis-:

The Linear prediction analysis method that the embodiment of the present invention two provides, as shown in Figure 3, the method step comprises:

S301, obtain the amplitude of first sampling point of input signal | x[0] |, wherein, x[i], i=0,1 ..., N-1 is input signal, the sampling point number that N is input signal; Input signal here refers to inputs the signal that carries out lpc analysis, it may be a frame signal, also may be the frame signal segment signal that adds history buffer (as the L of a history buffer sampling point, L can adopt according to different codecs different positive integers, as 40,80 etc.);

S302, to sampling point amplitude | x[0] | analyze, and according to analysis result, input signal carried out to self-adaptation windowing:

If the amplitude of first sampling point of input signal | x[0] | certain presets threshold value thr to be greater than (or being more than or equal to), with the first window function, input signal is carried out to windowing, even xd[i]=x[i] w1[i], i=0,1 ..., N-1, xd[i wherein] be the signal after windowing, w1[i] be the first window function;

Otherwise, with the second window function, input signal is carried out to windowing, even xd[i] and=x[i] w2[i], i=0,1 ..., N-1, wherein w2[i] be the second window function;

Window function w1[i] and w2[i] can by great many of experiments, select according to different audio encoding devices, be applicable to respectively different signals, for example w1[i] be sinusoidal windows, w2[i] be Hamming window; Or w1[i] be hamming window, w2[i] be sinusoidal windows.Threshold value thr selectes by great many of experiments, as thr=128 or thr=157;

In a concrete realization, thr=128, when frame length N=80,

$w1\left[i\right]=\left\{\begin{array}{cc}0.16+0.84\·\cos (2\·\mathrm{\π}\·(31-4\·i)/127),& i=\mathrm{0,1},...,7\\ 1,& i=\mathrm{8,9},...,71\\ 0.16+0.84\·\cos (2\·\mathrm{\π}(4\·i-285)/127),& i=\mathrm{72,73},...,79\end{array}\right.$

$w2\left[i\right]=\left\{\begin{array}{cc}0.26+0.74\·\cos (2\·\mathrm{\π}(31-4\·i)/127),& i=\mathrm{0,1},...,7\\ 1,& i=\mathrm{8,9},...,71\\ 0.26+0.74\·\cos (2\·\mathrm{\π}\·(4\·i-285)/127),& i=\mathrm{72,73},...,79\end{array}\right.$

When frame length N=40,

$w1\left[i\right]=\left\{\begin{array}{cc}0.26+0.74\·\cos (2\·\mathrm{\π}(31-8\·i)/127),& i=\mathrm{0,1},...,3\\ 1,& i=\mathrm{4,5},...,35\\ 0.26+0.74\·\cos (2\·\mathrm{\π}\·(8\·i-281)/127),& i=\mathrm{36,37},...,39\end{array}\right.$

$w1\left[i\right]=\left\{\begin{array}{cc}0.23+0.77\·\cos (2\·\mathrm{\π}\·(31-8\·i)/127),& i=\mathrm{0,1},...,3\\ 1,& i=\mathrm{4,5},...,35\\ 0.23+0.77\·\cos (2\·\mathrm{\π}\·(8\·i-281)/127),& i=\mathrm{36,37},...,39\end{array}\right.$

S303, signal after windowing is processed, obtained linear forecast coding coefficient, for linear prediction.

The Linear prediction analysis method that the embodiment of the present invention two provides, by obtaining the amplitude of first sampling point of input signal, and according to this sampling point amplitude, input signal is carried out to self-adaptation windowing, therefore, can be in the situation that the less encoder complexity of increase have improved the estimated performance of linear predictive coding.

Embodiment tri-:

The Linear prediction analysis method that the embodiment of the present invention three provides, as shown in Figure 4, the method step comprises:

S401, obtain input signal before the amplitude mean value of (or rear) M sampling point wherein, x[i], i=0,1 ..., N-1 is input signal, the sampling point number that N is input signal; Input signal here refers to inputs the signal that carries out lpc analysis, it may be a frame signal, also may be the frame signal segment signal that adds history buffer (as the L of a history buffer sampling point, L can adopt according to different codecs different positive integers, as 40,80 etc.);

S402, the amplitude mean value to front (or rear) M sampling point analyze, and according to analysis result, input signal carried out to self-adaptation windowing:

The amplitude mean value of front if (or rear) M sampling point certain presets threshold value thr to be greater than (or being more than or equal to), with the first window function, input signal is carried out to windowing, even xd[i]=x[i] w1[i], i=0,1 ..., N-1, wherein xd[i] be the signal after windowing, w1[i] be the first window function;

Otherwise, with the second window function, input signal is carried out to windowing, even xd[i] and=x[i] w2[i], i=0,1 ..., N-1, wherein w2[i] be the second window function;

Window function w1[i] and w2[i] can by great many of experiments, select according to different audio encoding devices, be applicable to respectively different signals, for example w1[i] be sinusoidal windows, w2[i] be Hamming window; Or w1[i] be hamming window, w2[i] be sinusoidal windows.Threshold value thr selectes by great many of experiments, as thr=127 or thr=152;

In a concrete realization, thr=128, when frame length N=80,

$w1\left[i\right]=\left\{\begin{array}{cc}0.15+0.85\·\cos (2\·\mathrm{\π}\·(31-4\·i)/127),& i=\mathrm{0,1},...,7\\ 1,& i=\mathrm{8,9},...,71\\ 0.15+0.85\·\cos (2\·\mathrm{\π}\·(4\·i-285)/127),& i=\mathrm{72,73},...,79\end{array}\right.$

$w2\left[i\right]=\left\{\begin{array}{cc}0.27+0.73\·\cos (2\·\mathrm{\π}\·(31-4\·i)/127),& i=\mathrm{0,1},...,7\\ 1,& i=\mathrm{8,9},...,71\\ 0.27+0.73\·\cos (2\·\mathrm{\π}\·(4\·i-285)/127),& i=\mathrm{72,73},...,79\end{array}\right.$

When frame length N=40,

$w1\left[i\right]=\left\{\begin{array}{cc}0.26+0.74\·\cos (2\·\mathrm{\π}\·(31-8\·i)/127),& i=\mathrm{0,1},...,3\\ 1,& i=\mathrm{4,5},...,35\\ 0.26+0.74\·\cos (2\·\mathrm{\π}\·(8\·i-281)/127),& i=36.37,...,39\end{array}\right.$

$w1\left[i\right]=\left\{\begin{array}{cc}0.23+0.77\·\cos (2\·\mathrm{\π}\·(31-8\·i)/127),& i=\mathrm{0,1},...,3\\ 1,& i=\mathrm{4,5},...,35\\ 0.23+0.77\·\cos (2\·\mathrm{\π}\·(8\·i-281)/127),& i=\mathrm{36,37},...,39\end{array}\right.$

S403, signal after windowing is processed, obtained linear forecast coding coefficient, for linear prediction.

The Linear prediction analysis method that the embodiment of the present invention three provides, by obtaining the amplitude mean value of (or rear) M sampling point before input signal, and according to this mean value, input signal is carried out to self-adaptation windowing, therefore, can be in the situation that the less encoder complexity of increase have improved the estimated performance of linear predictive coding.

Embodiment tetra-:

The Linear prediction analysis method that the embodiment of the present invention four provides, as shown in Figure 5, the method step comprises:

S501, obtain input signal before the average energy of (or rear) M sampling point wherein, x[i], i=0,1 ..., N-1 is input signal, the sampling point number that N is input signal.Input signal here refers to inputs the signal that carries out lpc analysis, it may be a frame signal, also may be the frame signal segment signal that adds history buffer (as the L of a history buffer sampling point, L can adopt according to different codecs different positive integers, as 40,80 etc.);

S502, the average energy to front (or rear) M sampling point analyze, and according to analysis result, input signal carried out to self-adaptation windowing:

The amplitude mean value of front if (or rear) M sampling point certain presets threshold value thr to be greater than (or being more than or equal to), with the first window function, input signal is carried out to windowing, even xd[i]=x[i] w1[i], i=0,1 ..., N-1, wherein xd[i] be the signal after windowing, w1[i] be the first window function.

Otherwise, with the second window function, input signal is carried out to windowing, even xd[i] and=x[i] w2[i], i=0,1 ..., N-1, wherein w2[i] be the second window function.

Window function w1[i] and w2[i] can by great many of experiments, select according to different audio encoding devices, be applicable to respectively different signals, for example w1[i] be sinusoidal windows, w2[i] be Hamming window; Or w1[i] be hamming window, w2[i] be sinusoidal windows.Threshold value thr selectes by great many of experiments, as thr=1024 or thr=2573.

In a concrete realization, thr=1280, when frame length N=80,

$w1\left[i\right]=\left\{\begin{array}{cc}0.18+0.82\·\cos (2\·\mathrm{\π}\·(31-4\·i)/127),& i=\mathrm{0,1},...,7\\ 1,& i=\mathrm{8,9},...,71\\ 0.18+0.82\·\cos (2\·\mathrm{\π}\·(4\·i-285)/127),& i=\mathrm{72,73},...,79\end{array}\right.$

$w2\left[i\right]=\left\{\begin{array}{cc}0.25+0.75\·\cos (2\·\mathrm{\π}\·(31-4\·i)/127),& i=\mathrm{0,1},...,7\\ 1,& i=\mathrm{8,9},...,71\\ 0.25+0.75\·\cos (2\·\mathrm{\π}\·(4\·i-285)/127),& i=\mathrm{72,73},...,79\end{array}\right.$

When frame length N=40,

$w1\left[i\right]=\left\{\begin{array}{cc}0.25+0.75\·\cos (2\·\mathrm{\π}\·(31-8\·i)/127),& i=\mathrm{0,1},...,3\\ 1,& i=\mathrm{4,5},...,35\\ 0.25+0.75\·\cos (2\·\mathrm{\π}\·(8\·i-281)/127),& i=\mathrm{36,37},...,39\end{array}\right.$

$w1\left[i\right]=\left\{\begin{array}{cc}0.24+0.76\·\cos (2\·\mathrm{\π}\·(31-8\·i)/127),& i=\mathrm{0,1},...,3\\ 1,& i=\mathrm{4,5},...,35\\ 0.24+0.76\·\cos (2\·\mathrm{\π}\·(8\·i-281)/127),& i=\mathrm{36,37},...,39\end{array}\right.$

S503, signal after windowing is processed, obtained linear forecast coding coefficient, for linear prediction.

The Linear prediction analysis method that the embodiment of the present invention four provides, by obtaining the average energy of (or rear) M sampling point before input signal, and according to this average energy, input signal is carried out to self-adaptation windowing, therefore, can be in the situation that the less encoder complexity of increase have improved the estimated performance of linear predictive coding.

Embodiment five:

The Linear prediction analysis method that the embodiment of the present invention five provides, as shown in Figure 6, the method step comprises:

S601, obtain the zero-crossing rate of input signal wherein, x[i], i=0,1 ..., N-1 is input signal, the sampling point number that N is input signal, for operating with (AND).Input signal here refers to inputs the signal that carries out lpc analysis, it may be a frame signal, also may be the frame signal segment signal that adds history buffer (as the L of a history buffer sampling point, L can adopt according to different codecs different positive integers, as 40,80 etc.);

S602, zero-crossing rate zc is analyzed, and according to analysis result, input signal is carried out to self-adaptation windowing:

If zero-crossing rate zc is greater than (or being more than or equal to), certain presets threshold value thr, with the first window function, input signal is carried out to windowing, even xd[i]=x[i] w1[i], i=0,1 ..., N-1, wherein xd[i] be the signal after windowing, w1[i] be the first window function;

Otherwise, with the second window function, input signal is carried out to windowing, even xd[i] and=x[i] w2[i], i=0,1 ..., N-1, wherein w2[i] be the second window function;

Window function w1[i] and w2[i] can by great many of experiments, select according to different audio encoding devices, be applicable to respectively different signals, for example w1[i] be sinusoidal windows, w2[i] be Hamming window; Or w1[i] be hamming window, w2[i] be sinusoidal windows.Threshold value thr selectes by great many of experiments, as thr=15 or thr=23;

In a concrete realization, thr=18, when frame length N=80,

$w\left[i\right]=\left\{\begin{array}{cc}0.16+0.84\·\cos (2\·\mathrm{\π}\·(31-4\·i)/127),& i=\mathrm{0,1},...,7\\ 1,& i=\mathrm{8,9},...,71\\ 0.16+0.84\·\cos (2\·\mathrm{\π}(4\·i-285)/127),& i=\mathrm{72,73},...,79\end{array}\right.$

$w2\left[i\right]=\left\{\begin{array}{cc}0.26+0.74\·\cos (2\·\mathrm{\π}(31-4\·i)/127),& i=\mathrm{0,1},...,7\\ 1,& i=\mathrm{8,9},...,71\\ 0.26+0.74\·\cos (2\·\mathrm{\π}\·(4\·i-285)/127),& i=\mathrm{72,73},...,79\end{array}\right.$

When frame length N=40,

$w1\left[i\right]=\left\{\begin{array}{cc}0.27+0.73\·\cos (2\·\mathrm{\π}\·(31-8\·i)/127),& i=\mathrm{0,1},...,3\\ 1,& i=\mathrm{4,5},...,35\\ 0.27+0.73\·\cos (2\·\mathrm{\π}\·(8\·i-281)/127),& i=\mathrm{36,37},...,39\end{array}\right.$

$w1\left[i\right]=\left\{\begin{array}{cc}0.25+0.75\·\cos (2\·\mathrm{\π}\·(31-8\·i)/127),& i=\mathrm{0,1},...,3\\ 1,& i=\mathrm{4,5},...,35\\ 0.25+0.75\·\cos (2\·\mathrm{\π}\·(8\·i-281)/127),& i=\mathrm{36,37},...,39\end{array}\right.$

S603, signal after windowing is processed, obtained linear forecast coding coefficient, for linear prediction.

The Linear prediction analysis method that the embodiment of the present invention five provides, by obtaining the zero-crossing rate of input signal, and according to this zero-crossing rate, input signal is carried out to self-adaptation windowing, therefore, can be in the situation that the less encoder complexity of increase have improved the estimated performance of linear predictive coding.

Embodiment six:

The Linear prediction analysis method that the embodiment of the present invention six provides, as shown in Figure 7, the method step comprises:

S701, obtain the zero-crossing rate of input signal average energy with front (or rear) M sampling point wherein, x[i], i=0,1 ..., N-1 is input signal, the sampling point number that N is input signal, for operating with (AND).Input signal here refers to inputs the signal that carries out lpc analysis, it may be a frame signal, also may be the frame signal segment signal that adds history buffer (as the L of a history buffer sampling point, L can adopt according to different codecs different positive integers, as 40,80 etc.);

S702, the average energy to zero-crossing rate zc and front (or rear) M sampling point analyze, and according to analysis result, input signal carried out to self-adaptation windowing:

If zero-crossing rate zc be greater than (or being more than or equal to) certain preset threshold value thr1 or be less than or equal to certain predefined threshold value thr2, with the first window function, input signal carried out to windowing, even xd[i]=x[i] w1[i], i=0,1 ..., N-1, wherein xd[i] be the signal after windowing, w1[i] be the first window function.

Otherwise, with the second window function, input signal is carried out to windowing, even xd[i] and=x[i] w2[i], i=0,1 ..., N-1, wherein w2[i] be the second window function.

Window function w1[i] and w2[i] can by great many of experiments, select according to different audio encoding devices, be applicable to respectively different signals, for example w1[i] be sinusoidal windows, w2[i] be Hamming window; Or w1[i] be hamming window, w2[i] be sinusoidal windows.Threshold value thr1 and thr2 select by great many of experiments, as thr1=15, thr2=1023 or thr1=23, thr2=1012.

In a concrete realization, thr1=17, thr2=1012, when frame length N=80,

$w1\left[i\right]=\left\{\begin{array}{cc}0.16+0.84\·\cos (2\·\mathrm{\π}\·(31-4\·i)/127),& i=\mathrm{0,1},...,7\\ 1,& i=\mathrm{8,9},...,71\\ 0.16+0.84\·\cos (2\·\mathrm{\π}(4\·i-285)/127),& i=\mathrm{72,73},...,79\end{array}\right.$

$w2\left[i\right]=\left\{\begin{array}{cc}0.26+0.74\·\cos (2\·\mathrm{\π}(31-4\·i)/127),& i=\mathrm{0,1},...,7\\ 1,& i=\mathrm{8,9},...,71\\ 0.26+0.74\·\cos (2\·\mathrm{\π}\·(4\·i-285)/127),& i=\mathrm{72,73},...,79\end{array}\right.$

When frame length N=40,

$w1\left[i\right]=\left\{\begin{array}{cc}0.28+0.72\·\cos (2\·\mathrm{\π}\·(31-8\·i)/127),& i=\mathrm{0,1},...,3\\ 1,& i=\mathrm{4,5},...,35\\ 0.28+0.72\·\cos (2\·\mathrm{\π}\·(8\·i-281)/127),& i=\mathrm{36,37},...,39\end{array}\right.$

$w1\left[i\right]=\left\{\begin{array}{cc}0.22+0.78\·\cos (2\·\mathrm{\π}\·(31-8\·i)/127),& i=\mathrm{0,1},...,3\\ 1,& i=\mathrm{4,5},...,35\\ 0.22+0.78\·\cos (2\·\mathrm{\π}\·(8\·i-281)/127),& i=\mathrm{36,37},...,39\end{array}\right.$

S703, signal after windowing is processed, obtained linear forecast coding coefficient, for linear prediction.

The Linear prediction analysis method that the embodiment of the present invention six provides, by obtaining the average energy of zero-crossing rate and front (or rear) M sampling point of input signal, and according to this zero-crossing rate and average energy, input signal is carried out to self-adaptation windowing, therefore, can be in the situation that the less encoder complexity of increase have improved the estimated performance of linear predictive coding.

Embodiment seven:

The Linear prediction analysis method that the embodiment of the present invention seven provides, as shown in Figure 8, the method step comprises:

S801, obtain input signal coded system, input signal is signal G.711, may be A-law signal, may be also mu-law signal; Input signal is changed, obtained PCM signal;

S802, input signal coded system is analyzed, and according to analysis result to PCM signal carry out self-adaptation windowing as:

If coded system is A-law, with the first window function, PCM signal is carried out to windowing, even xd[i]=x[i] w1[i], i=0,1 ..., N-1, wherein xd[i] be the signal after windowing, w1[i] be the first window function, x[i] be PCM signal.

Otherwise, with the second window function, PCM signal is carried out to windowing, even xd[i] and=x[i] w2[i], i=0,1 ..., N-1, wherein w2[i] be the second window function.

Window function w1[i] and w2[i] can by great many of experiments, select according to different audio encoding devices, be applicable to respectively different signals, for example w1[i] be sinusoidal windows, w2[i] be Hamming window; Or w1[i] be hamming window, w2[i] be sinusoidal windows;

In a concrete realization, when coded system is A-law or mu-law, when frame length N=80,

$w1\left[i\right]=\left\{\begin{array}{cc}0.16+0.84\·\cos (2\·\mathrm{\π}\·(31-4\·i)/127),& i=\mathrm{0,1},...,7\\ 1,& i=\mathrm{8,9},...,71\\ 0.16+0.84\·\cos (2\·\mathrm{\π}(4\·i-285)/127),& i=\mathrm{72,73},...,79\end{array}\right.$

$w2\left[i\right]=\left\{\begin{array}{cc}0.26+0.74\·\cos (2\·\mathrm{\π}(31-4\·i)/127),& i=\mathrm{0,1},...,7\\ 1,& i=\mathrm{8,9},...,71\\ 0.26+0.74\·\cos (2\·\mathrm{\π}\·(4\·i-285)/127),& i=\mathrm{72,73},...,79\end{array}\right.$

When frame length N=40,

$w1\left[i\right]=\left\{\begin{array}{cc}0.28+0.72\·\cos (2\·\mathrm{\π}\·(31-8\·i)/127),& i=\mathrm{0,1},...,3\\ 1,& i=\mathrm{4,5},...,35\\ 0.28+0.72\·\cos (2\·\mathrm{\π}\·(8\·i-281)/127),& i=\mathrm{36,37},...,39\end{array}\right.$

$w1\left[i\right]=\left\{\begin{array}{cc}0.22+0.78\·\cos (2\·\mathrm{\π}\·(31-8\·i)/127),& i=\mathrm{0,1},...,3\\ 1,& i=\mathrm{4,5},...,35\\ 0.22+0.78\·\cos (2\·\mathrm{\π}\·(8\·i-281)/127),& i=\mathrm{36,37},...,39\end{array}\right.$

S803, signal after windowing is processed, obtained linear forecast coding coefficient, for linear prediction.

The Linear prediction analysis method that the embodiment of the present invention seven provides, obtain input signal coded system, and input signal is changed, obtain PCM signal, according to this Signal coding mode, input signal is carried out to self-adaptation windowing, therefore, can be in the situation that the less encoder complexity of increase have improved the estimated performance of linear predictive coding.

The linear prediction analysis device that the embodiment of the present invention provides, as shown in Figure 9, comprising:

Acquiring unit 901, for obtaining the characteristics of signals information of at least one sampling point of input signal;

Analytic unit 902, for characteristics of signals information is compared to analysis, obtains analysis result;

Add window unit 903, for selecting window function to carry out self-adaptation windowing to input signal according to analysis result, obtain signal after windowing;

Processing unit 904, processes for signal after stating windowing, obtains linear forecast coding coefficient, for linear prediction.

The linear prediction analysis device that the embodiment of the present invention provides, by input signal is analyzed, obtains result, and distribute the required window function of windowing according to analysis result self-adaptation, therefore, can be in the situation that the less encoder complexity of increase have improved the estimated performance of linear predictive coding.

Wherein, in another embodiment of the present invention, as shown in figure 10, analytic unit 902 comprises:

Computing module 902A, for calculating the value of the characteristics of signals information that acquiring unit 901 obtains, the value of characteristics of signals information comprises the mean value of the value of characteristics of signals information of some sampling points and/or the value of the characteristics of signals information of certain a plurality of sampling point;

Judge module 902B, for judging whether the value of the characteristics of signals information that computing module 902A draws is greater than or is more than or equal to a certain threshold value; Or for judging signal type and/or the coded system of the input signal that acquiring unit 901 obtains.

Further, in above-mentioned analytic unit 902, also comprise:

Modular converter 902C, is converted to pulse code modulation signal for the input signal that acquiring unit 901 is obtained.

The linear prediction analysis device that the embodiment of the present invention provides, by input signal is analyzed, obtains result, and distribute the required window function of windowing according to analysis result self-adaptation, therefore, can be in the situation that the less encoder complexity of increase have improved the estimated performance of linear predictive coding.

The linear predictive coding system that the embodiment of the present invention provides, as shown in figure 11, comprising:

Linear prediction analysis device 1101, for obtaining the characteristics of signals information of at least one sampling point of input signal; Characteristics of signals information is compared to analysis, obtain analysis result; According to analysis result, select window function to carry out self-adaptation windowing to input signal, obtain signal after windowing; Signal after windowing is processed, obtained linear forecast coding coefficient;

Code device 1102, encodes for the linear forecast coding coefficient obtaining according to linear prediction analysis device 1101.

The linear predictive coding system that the embodiment of the present invention provides, can obtain result first by input signal is analyzed, and distributes the required window function of windowing according to analysis result self-adaptation, and then obtains linear forecast coding coefficient; And then encode according to this linear forecast coding coefficient.Therefore, can be in the situation that the less encoder complexity of increase have improved the estimated performance of linear predictive coding.

Wherein, in another embodiment of the present invention, as shown in figure 12, linear prediction analysis device 1101 is identical with the linear prediction analysis device structure in above-described embodiment, at this, just repeats no more.

The linear predictive coding system that the embodiment of the present invention provides, can obtain result first by input signal is analyzed, and distributes the required window function of windowing according to analysis result self-adaptation, and then obtains linear forecast coding coefficient; And then encode according to this linear forecast coding coefficient.Therefore, can be in the situation that the less encoder complexity of increase have improved the estimated performance of linear predictive coding.

One of ordinary skill in the art will appreciate that all or part of flow process realizing in above-described embodiment method, to come the hardware that instruction is relevant to complete by computer program, described program can be stored in a computer read/write memory medium, this program, when carrying out, can comprise as the flow process of the embodiment of above-mentioned each side method.Wherein, described storage medium can be magnetic disc, CD, read-only store-memory body (Read-Only Memory, ROM) or random store-memory body (Random Access Memory, RAM) etc.

The above; be only the specific embodiment of the present invention, but protection scope of the present invention is not limited to this, is anyly familiar with those skilled in the art in the technical scope that the present invention discloses; can expect easily changing or replacing, within all should being encompassed in protection scope of the present invention.Therefore, protection scope of the present invention should be as the criterion by the described protection domain with claim.

Claims (11)

1. a Linear prediction analysis method, is characterized in that, comprising:

The amplitude of first sampling point and the amplitude of last sampling point of obtaining input signal, described input signal comprises N sampling point, and N is positive integer;

Amplitude to the amplitude of described first sampling point and last sampling point is analyzed, and according to analysis result, input signal is carried out to self-adaptation windowing, obtains signal after windowing;

When N=40, the amplitude of the amplitude of described first sampling point and last sampling point is analyzed, and according to analysis result, input signal is carried out to self-adaptation windowing, obtain signal after windowing, comprising:

When if the amplitude of described first sampling point is greater than or is more than or equal to a certain threshold value, with the first window function, 4 of described input signal foremosts sampling point is carried out to self-adaptation windowing, obtain signal after windowing; Otherwise, with the second window function, 4 of described input signal foremosts sampling point is carried out to self-adaptation windowing, obtain signal after windowing;

To the 5th to the 36th sampling point of input signal, window function is all made as 1;

When if the amplitude of described last sampling point is greater than or is more than or equal to a certain threshold value, with the 3rd window function to described input signal backmost 4 sampling points carry out self-adaptation windowing, obtain signal after windowing; Otherwise, with four-light function to described input signal backmost 4 sampling points carry out self-adaptation windowing, obtain signal after windowing;

Signal after described windowing is processed, obtained linear forecast coding coefficient for linear prediction.

2. Linear prediction analysis method according to claim 1, is characterized in that,

Described the first window function w (n) is:

w(n)＝0.26+0.74·cos(2·π·(31-8·n)/127)，n＝0,1,2,3

Described the second window function w (n) is:

w(n)＝0.23+0.77·cos(2·π·(31-8·n)/127),n＝0,1,2,3

Described the 3rd window function is:

w(n)＝0.26+0.74·cos(2·π·(8·n-281)/127),n＝36,37,38,39

Described four-light function is:

w(n)＝0.23+0.77·cos(2·π·(8·n-281)/127),n＝36,37,38,39。

3. Linear prediction analysis method according to claim 1 and 2, is characterized in that,

Described threshold value is 128 or 157.

4. Linear prediction analysis method according to claim 1, is characterized in that, the amplitude of first sampling point and the amplitude of last sampling point of obtaining input signal comprise:

Described input signal is converted to pulse code modulation signal;

Obtain the amplitude of first sampling point and the amplitude of last sampling point of the described input signal after conversion.

5. Linear prediction analysis method according to claim 1, is characterized in that,

The amplitude of first sampling point of described input signal is | x[0] |, the amplitude of last sampling point is | x[N-1] |; Wherein, x[i], i=0,1 ..., N-1 is input signal.

6. a Linear prediction analysis method, is characterized in that, comprising:

The amplitude of first sampling point and the amplitude of last sampling point of obtaining input signal, described input signal comprises N sampling point, and N is positive integer;

Amplitude to the amplitude of described first sampling point and last sampling point is analyzed, and according to analysis result, input signal is carried out to self-adaptation windowing, obtains signal after windowing;

When N=80, the amplitude of the amplitude of described first sampling point and last sampling point is analyzed, and according to analysis result, input signal is carried out to self-adaptation windowing, obtain signal after windowing, comprising:

When if the amplitude of described first sampling point is greater than or is more than or equal to a certain threshold value, with the 5th window function, 8 of described input signal foremosts sampling point is carried out to self-adaptation windowing, obtain signal after windowing; Otherwise, with the 6th window function, 8 of described input signal foremosts sampling point is carried out to self-adaptation windowing, obtain signal after windowing;

To the 9th to the 72nd sampling point of input signal, window function is all made as 1;

When if the amplitude of described last sampling point is greater than or is more than or equal to a certain threshold value, with the 7th window function to described input signal backmost 8 sampling points carry out self-adaptation windowing, obtain signal after windowing; Otherwise, with the 8th window function to described input signal backmost 8 sampling points carry out self-adaptation windowing, obtain signal after windowing;

Signal after described windowing is processed, obtained linear forecast coding coefficient for linear prediction.

7. Linear prediction analysis method according to claim 6, is characterized in that,

Described the 5th window function w (n) is:

w(n)＝0.16+0.84·cos(2·π·(31-4·n)/127)，n＝0,1,2,...,7

Described the 6th window function w (n) is:

w(n)＝0.26+0.74·cos(2·π·(31-4·n)/127),n＝0,1,2,...,7

Described the 7th window function is:

w(n)＝0.16+0.84·cos(2·π·(4·n-285)/127),n＝72,73,74,...,79

Described the 8th window function is:

w(n)＝0.26+0.74·cos(2·π·(4·n-285)/127),n＝72,73,74,...,79。

8. according to the Linear prediction analysis method described in claim 6 or 7, it is characterized in that,

Described threshold value is 128 or 157.

9. a Linear prediction analysis method, is characterized in that, comprising:

Obtain input signal coded system, described input signal is signal G.711;

Input signal is changed, obtained PCM signal;

Input signal coded system is analyzed, and according to analysis result, PCM signal is carried out to self-adaptation windowing, obtain signal after windowing;

Signal after described windowing is processed, obtained linear forecast coding coefficient for linear prediction.

10. Linear prediction analysis method according to claim 9, is characterized in that, described input signal coded system is analyzed, and according to analysis result, PCM signal is carried out to self-adaptation windowing, obtains signal after windowing, comprising:

If coded system is A-law, with the first window function, PCM signal is carried out to windowing;

Otherwise, with the second window function, PCM signal is carried out to windowing.

11. according to the Linear prediction analysis method described in claim 9 or 10, it is characterized in that, described the first window function is sinusoidal windows, and the second window function is for being Hamming window; Or the first window function is hamming window, the second window function is sinusoidal windows.