CN103700372B - A kind of parameter stereo coding based on orthogonal decorrelation technique, coding/decoding method - Google Patents

Abstract

The invention discloses a kind of parameter stereo coding based on orthogonal decorrelation technique, coding/decoding method.This coded method: 1) L channel of input, R channel time-domain signal are transformed to complex frequency domain respectively, obtain two complex frequency-domain signal L and R；2) L and R is output as mixed signal M under；3) utilize L and R to estimate side information, and carry out described side information merging quantization；Described side information includes cross-correlation coefficient ICC, Inter channel Intensity Difference IID and gyrobearing information RD in sound channel；4) to M time domain inverse transformation and encode output；Side information compressed encoding is exported.Coding/decoding method: signal mixed under the time domain of input is transformed to complex frequency domain output signal M；The side information of input is decoded；Output signal D equal and orthogonal with M amplitude is solved according to RD；Spin matrix H is calculated according to ICC, IID；Utilize H, M and D to obtain complex frequency-domain signal L and R changes to time domain output.The present invention substantially increases decoding speed.

Description

A kind of parameter stereo coding based on orthogonal decorrelation technique, coding/decoding method

Technical field

The invention belongs to audio compression field, relate to a kind of parameter stereo coding, coding/decoding method, particularly relate to a kind of base Parameter stereo coding, coding/decoding method in orthogonal decorrelation technique.

Background technology

Multimedia technology is risen the latter stage eighties, is one of technology popular in computer realm in recent years.Numeral sound Frequently encoding and decoding technique uses digital record and the transmission load mode of computer, processes various media, has widely Purposes.But it is limited to the network bandwidth and memory space, under ensureing high-quality sense of hearing, compressing original data is the fewest for trying one's best Bit number become study hotspot.

Why audio file can compress, and is because existing redundancy, and conventional compression algorithm mainly considers to eliminate in sound channel Redundancy, when not considering redundancy between sound channel, owing to code check and the channel number of multitrack audio file are linear relationships, channel number Increase can cause code check to be multiplied.From the beginning of MP3 in 1993, people gradually find it is also to there is redundancy and start between sound channel Attempt eliminating redundancy between sound channel, progressively solve the problem realizing ultralow code check under conditions of keeping higher subjective sense of hearing.Existing Between some elimination sound channels, the method for redundancy mainly has following 3 kinds:

1 > important component part during joint stereo is the audio coding decoding international standards such as MPEG-1, MEPG-2/4AAC, Including and difference transition coding and two kinds of stereo coding/decoding technology of intensity stereo.

(1) and difference transition coding

Also referred to as MS coding, is the abbreviation of Middle-Side coding.Original left right-channel signals is converted to by coding side And difference signal, decoding end does corresponding inverse transformation.The method utilizes inter-channel correlation, signal energy is concentrated on in sound channel, Redundancy between sound channel is removed with this.But the method compression efficiency heavy dependence signal self character, only in original left R channel phase Higher compression ratio just can be obtained in the case of closing property is very strong.

(2) intensity stereo

Intensity stereo basic thought is by being rotated by the coordinate axes in primary signal space, obtains the intensity of main shaft Signal and orthogonal residual signals.At coding side, abandon the residual signals orthogonal with main shaft and only to strength signal and seat Parameter anglec of rotation α carries out quantization encoding.In decoding end, according to angle [alpha] and strength signal, coordinate axes is made reverse rotation, reconstruct Go out the stereophonic signal of left and right acoustic channels.The left and right sound track signals that so reconstruct obtains is that amplitude is different, and phase information is consistent. But by being multiplied by the corresponding amplitude proportional factor, can preferably preserve original signal energy time envelope.Such process It is insensitive to phase difference and the fine structure thereof of each sound channel signal radio-frequency component that mode meets human auditory system, and in time The relatively important perception feature of energy envelope of change, thus without the subjective quality substantially reducing reconstructed stereoscopic acoustical signal.

2 > binaural cue coding

Within 2002, C.Faller proposes the encoding and decoding of entitled binaural cue coding (Binaural Cue Coding, BCC) Technology, its starting point is different with conventional stereo vocoding technique, is a kind of parametric coding technique theoretical based on spatial hearing.Vertical Monophonic signal is blended together, simultaneously according to spatial hearing characteristic, to each sound channel in conversion under body sound or multi-channel input signal Territory is analyzed, and extracts three kinds of spatial parameters: Inter channel Intensity Difference (Inter-channel Intensity Difference, IID), inter-channel time differences (Inter-channel Time Difference, ITD) and inter-channel correlation (Inter- Channel Coherence, ICC).IID and ITD and traditional spatial hearing cue abbreviation here is identical but implication is different, Corresponding abbreviation is all referring to parameter between sound channel rather than acoustic cue.The most mixed signal can use traditional audio coder to compile Code, parameter is embedded in bit stream as slave part after quantization encoding.In decoding end, decoded lower mixed signal utilizes Corresponding spatial parameter comes reconstructed stereoscopic sound or multi-channel audio signal.

The feature of binaural cue coding maximum is exactly that the compression that can provide high compression ratio and any channel version is compiled Code, is considered as a kind of development to intensity stereo, but is that it overcomes the defect of intensity-stereo encoding, have following several Individual obvious advantage: intensity-stereo encoding reality only utilizes IID to be reconstructed, only the most effective to high-frequency range, if by intensity The stereo low frequency that is extended to then can introduce serious noise, and binaural cue coding utilizes IID and ITD to be reconstructed, at full range Carry out in spectral limit processing and also will not produce critical noisy；Additionally intensity-stereo encoding can not reconstruct that to have width solid several The audio signal of picture, binaural cue coding then can utilize ICC well to reconstruct it.

3 > parameter stereo

J.Breebaart of PHILIPS Co. et al. proposed entitled parameter stereo (Parametric in 2004 Stereo, PS) encoding and decoding technique.Basic ideas are at coding side one monophonic signal of output and some spatial parameters, Decoding end utilizes monophonic signal and spatial parameter reconstruct two channel stereo signal, is that to eliminate redundancy between sound channel at present fitst water One of method.It is adopted as corresponding sound by MPEG-4 and 3GPP (3rd Generation Partnership Project) Frequently the ingredient in coding international standard.

At the coding side of parameter stereo, the left and right sound track signals of input is filtered respectively through hybrid orthogonal mirror image analysis Device group, the output of each wave filter is through down-sampled, it is thus achieved that the complex frequency-domain signal after point band.Utilize the multifrequency of each subband of left and right acoustic channels Territory signal estimation goes out corresponding spatial parameter and quantifies output, is added simultaneously and generates monophonic down-mix signal, by mixing Orthogonal mirror image synthesis filter banks switches back to time-domain signal again.The ultimate principle of parameter stereo coding is as shown in Figure 1.

Parameter stereo decoding is also, by hybrid orthogonal mirror image analysis filterbank, time-domain signal is transformed into complex frequency domain Carry out processing, therefore repeat no more the related content of hybrid orthogonal mirror filter group.Parameter stereo decoding be by under The mixed corresponding spatial parameter of signal combination carrys out synthetic environment sound (Synthetic Ambience), finally reconstructs stereophony The process of signal.Parameter stereo decoding principle is as shown in Figure 2.

In parameter stereo decoding end, the effect of de-correlation block is that the monophonic down-mix signal M utilizing input generates phase Decorrelation output signal D answered.In order to enable Exact recovery original signal, D should be orthogonal with M, in implementing, parameter stereo The decimation using frequency band different and to change postpones (Fractional Delay), all-pass filter (All-pass Filter) Module.Owing to phase place can not accurately be controlled by all-pass filter, therefore can affect the orthogonality of D and M, cause the letter of reconstruct Number there is bigger error.

Summary of the invention

For technical problem present in prior art, it is an object of the invention to provide a kind of accurately decorrelating method, The parameter stereo decoding method set up based on the method, recovers the double track letter of coding side input more accurately in decoding end Number.

The technical scheme is that

A kind of parameter stereo coding method based on orthogonal decorrelation technique, the steps include:

1) L channel of input, R channel time-domain signal are transformed to complex frequency domain respectively, obtain two complex frequency-domain signal L and R；

2) mixed signal M under is mixed and is output as under two complex frequency-domain signal L and R being carried out；

3) utilize complex frequency-domain signal L and R to estimate side information, and carry out described side information merging quantization；Described limit is believed Breath includes cross-correlation coefficient ICC, Inter channel Intensity Difference IID and gyrobearing information RD in sound channel；

4) lower mixed signal M is carried out time domain inverse transformation, obtain time-domain signal and encode output；The compression of described side information is compiled Code output.

Further, formula is utilizedCalculate cross-correlation coefficient ICC in described sound channel.

Further, formula is utilizedCalculate described Inter channel Intensity Difference IID.

Further, formula is utilized $RD=\left\{\begin{array}{cc}1& \arg \left(\frac{M}{L}\right)\∈\[0,\mathrm{\π})\\ 0& \arg \left(\frac{M}{L}\right)\∈\[\mathrm{\π},2\mathrm{\π})\end{array}\right.$ Calculate described gyrobearing information RD.

Further, using the energy of left and right acoustic channels complex frequency-domain signal and as weight to described side information RD, ICC, IID enter Row compressed encoding.

Further, according to audition equivalent rectangular bandwidth, described side information is merged on frequency domain；In described sound channel Cross-correlation coefficient ICC is quantified as 3bit, Inter channel Intensity Difference IID is quantified as 5bit, gyrobearing information RD is quantified as 1bit.

A kind of parameter stereo coding/decoding method based on orthogonal decorrelation technique, the steps include:

1) signal mixed under the time domain of input is transformed to complex frequency domain, export mixed signal M under a complex frequency domain；Limit letter to input Breath is decoded, and obtains cross-correlation coefficient ICC, Inter channel Intensity Difference IID and gyrobearing information RD in sound channel；

2) for signal M mixed under each complex frequency domain, one is solved according to gyrobearing information RD equal with its amplitude and orthogonal Output signal D；

3) calculate with signal D according to cross-correlation coefficient ICC in sound channel, Inter channel Intensity Difference IID, signal M recover L channel, The spin matrix H of R channel；

4) spin matrix H, lower mixed signal M and signal D is utilized to obtain complex frequency-domain signal L and R；

5) complex frequency-domain signal L and R are changed to time domain respectively, the left channel signals after being restored and right-channel signals.

Further, the method solving described signal D is:

81) set under complex frequency domain mixed signal M the vector corresponding to complex plane asA Yu b is respectively lower mixed letter Number corresponding vectorReal part and imaginary part；If solving orthogonal vectors corresponding for signal D in advance it isWherein, c Yu d is respectively VectorReal part and imaginary part；

82) according to formula x²+y²=a²+b²And ax+by=0, obtain

83) according to formula $\stackrel{\→}{D}=\left\{\begin{array}{cc}b-ai& RD=1\\ -b+ai& RD=0\end{array}\right.$ Solve described signal D.

Further, the method solving described spin matrix H is:

91) by vectorPlace axle and vectorPlace axle determines a coordinate system；By L channel sample value to this coordinate system The vector of initial point is designated asR channel sample value is designated as to the vector of this coordinate origin

92) two vectors are determined according to cross-correlation coefficient ICC in sound channelWithAngle 2 α；

93) vector is determined according to Inter channel Intensity Difference IIDMould length compare c；

94) determined than c with mould length by angle αWithMould length compare c₁WithWithMould length compare c₂；

95) c is compared according to mould length₁And c₂Determine angular bisector and the vector of angleAngle β；Obtain described spin matrix $H=\left[\begin{array}{cc}{c}_{2}cos(\mathrm{\β}+\mathrm{\α})& c_{2}sin(\mathrm{\β}+\mathrm{\α})\\ c_{1}cos(\mathrm{\β}-\mathrm{\α})& c_{1}cos(\mathrm{\β}-\mathrm{\α})\end{array}\right].$

Further, formula is utilized $\left[\begin{array}{c}\stackrel{\→}{L}\\ \stackrel{\→}{R}\end{array}\right]=H\left[\begin{array}{c}\stackrel{\→}{M}\\ \stackrel{\→}{D}\end{array}\right]$ Obtain complex frequency-domain signal L and R.

Compared with prior art, the positive effect of the present invention is:

As shown in Figure 5,6, under the conditions of identical code check, the present invention is in signal to noise ratio and two objective evaluation indexs of PEAQ It is better than HE-AAC+, and the present invention substantially increases decoding speed in the case of coding rate is in a slight decrease.

Accompanying drawing explanation

Fig. 1 is parameter stereo coding block diagram；

Fig. 2 is that parameter stereo decodes block diagram；

Fig. 3 is parametric stereo encoder；

Fig. 4 is parametric stereo decoder；

Fig. 5 is two kinds of method signal to noise ratio results under 32kbps and 64kbps code check；

Fig. 5 (a) is two kinds of method signal to noise ratio results under 32kbps code check, and Fig. 5 (b) is two kinds of method letters under 64kbps code check Make an uproar than result；

Fig. 6 is two kinds of method PEAQ scores under 32kbps and 64kbps code check；

Fig. 6 (a) is two kinds of method PEAQ scores under 32kbps code check, and Fig. 6 (b) is two kinds of methods under 64kbps code check PEAQ scores.

Detailed description of the invention

Describing the present invention below in conjunction with the accompanying drawings, the parameter stereo coding method of the present invention is as shown in Figure 3.

1, framing converts with complex frequency domain

The input of this module is respectively left channel signals and the right-channel signals of double track, and time-domain signal is through time-frequency conversion mould (such as: MDFT, QMF) transform to complex frequency domain after block, export complex signal L and R.

2, add and lower mixed module

The input of this module for left channel signals in double track and right-channel signals respectively through corresponding after time-frequency convert module Two complex frequency-domain signal L and R, then through under mix and be output as mixed signal M under with module,

M=g_lL+g_rR (1)

Wherein g_l=g_r=0.5

3, parameter extraction module

The input of this module, is output as respectively through 2 complex frequency-domain signal after time-frequency convert module for left and right sound track signals The side information after compression is quantified including ICC, IID, gyrobearing information RD.

Cross-correlation ICC (Inter-channel Cross-Correlation) in 3.1 sound channels

ICC calculates the cross-correlation coefficient of left and right acoustic channels, and computing formula is,

$ICC=\frac{\mathrm{Re}\left({\mathrm{LR}}^{*}\right)}{\sqrt{\left({\mathrm{LL}}^{*}\right)\left({\mathrm{RR}}^{*}\right)}}---\left(2\right)$

In formula, L is left channel signals, and R is right-channel signals, and * is conjugate operation, and Re () is the computing for the treatment of excess syndrome portion.

3.2 Inter channel Intensity Difference IID (Inter-channel Intensity Difference)

ICC calculates the intensity difference of left and right acoustic channels, and computing formula is,

$IID=10{\log }_{10}\left(\frac{{\mathrm{LL}}^{*}}{{\mathrm{RR}}^{*}}\right)---\left(3\right)$

In formula, L is left channel signals, and R is right-channel signals, and * is conjugate operation.

3.3 gyrobearing information outputs

In order to accurately solve the orthogonal signalling of lower mixed signal in decoding end, it is to be appreciated that lower mixed signal and L channel Angle, defines gyrobearing information according to this corner dimension, and when angle is less than π, gyrobearing value is 1；When angle is more than π Time, the gyrobearing value of information is 0.

$RD=\left\{\begin{array}{cc}1& \arg \left(\frac{M}{L}\right)\∈\[0,\mathrm{\π})\\ 0& \arg \left(\frac{M}{L}\right)\∈\[\mathrm{\π},2\mathrm{\π})\end{array}\right.---\left(4\right)$

In formula

$\arg \left(x\right)=arctan\left(\frac{im\left(x\right)}{re\left(x\right)}\right)---\left(5\right)$

In formula, x=M/L, im () are the computing for the treatment of excess syndrome portion, and re () is for taking imaginary-part operation.Gyrobearing information is each can be defeated Going out a 1bit information, merge when quantifying using the energy of left and right acoustic channels complex signal with as weight, by RD, ICC, IID enter frame by frame Row compressed encoding.

4, quantization is merged

Merge: according to audition equivalent rectangular bandwidth (ERB), spatial parameter is merged on frequency domain

BW=24.7 × (0.00437 × f_c+1)

Quantify: IID is quantified as 5bit, ICC and is quantified as 3bit, RD is 1bit.

The parameter stereo coding/decoding method of the present invention is as shown in Figure 4.

Parameter stereo decoding is also to be transformed at complex frequency domain by time-domain signal by mixing time-frequency conversion module Reason.Parameter stereo decoding be by lower mixed signal by Decorrelating algorithm to its orthogonal signalling, in conjunction with corresponding spatial parameter Reconstruct the process of two channel stereo signal.Hereinafter mainly illustrate that in decoder, decorrelation processes, upwards mixes two modules Ultimate principle and implementation method.

1, de-correlation block

The most mixed signal, after time-frequency conversion, becomes complex frequency-domain signal, does a vector from the point of view of signal correspondence complex plane. Solve an output vector equal and orthogonal with this vector magnitude, if certain puts in complex plane institute input signal after time-frequency convert Corresponding vector is

$\stackrel{\→}{M}=a+bi---\left(6\right)$

The corresponding vector of mixed signal under a with b is respectively in formulaReal part and imaginary part.

The orthogonal vectors solved in advance are

$\stackrel{\→}{D}=x+yi---\left(7\right)$

C Yu d orthogonal vectors respectively in formulaReal part and imaginary part.

Require that output vector is orthogonal with input and mould length is equal, therefore have

x²+y²=a²+b² (8)

Ax+by=0

Solve this binary quadratic equation group, have

${\stackrel{\→}{d}}_1=-b+ai$

${\stackrel{\→}{d}}_2=b-ai---\left(9\right)$

d₁、d₂For two solutions of de-correlated signals D, select correct solution, most important to accurately decoding.According to encoder The gyrobearing information being transmitted through, it may be determined that correctly solve D.

$\stackrel{\→}{D}=\left\{\begin{array}{cc}b-ai& RD=1\\ -b+ai& RD=0\end{array}\right.---\left(10\right)$

2, parameter mixing module

On complex frequency domain, plural number sampling point to the vector of the O of initial point isVertical with M and that mould is equal vector isBy WithThe axle at place may determine that the coordinate system of this complex plane.And wish that the L channel sample value recovered is designated as to the vector of initial pointSimilarly R channel sample value is designated as to the vector of initial pointSo problem translates into by known vectorHow to obtain UnknownWithVector.

Two vectors are determined by ICCWithAngle 2 α

2 α=arccos (icc) (11)

Determined by IIDWithThe mould length of complex signal compares c

$c={10}^{\frac{iid}{20}}---\left(12\right)$

Determined than c with mould length by angle αWithMould length compare c₁WithWithMould length compare c₂

$c_1=\frac{1}{\sqrt{1+c^2+2\·c\·cos\left(2\mathrm{\α}\right)}}---\left(13\right)$

$c_2=\frac{c}{\sqrt{1+c^2+2\·c\·cos\left(2\mathrm{\α}\right)}}$

ByAngle half α,WithMould length compare c₁WithWithMould length compare c₂DetermineWithAngle Angular bisector withAngle β, and spin matrix H.

$\mathrm{\β}=arctan\left(\frac{c_1-c_2}{c_1+c_2}tan\right(\mathrm{\α}\left)\right)---\left(14\right)$

$H=\left[\begin{array}{cc}{c}_{2}cos(\mathrm{\β}+\mathrm{\α})& c_{2}sin(\mathrm{\β}+\mathrm{\α})\\ c_{1}cos(\mathrm{\β}-\mathrm{\α})& c_{1}cos(\mathrm{\β}-\mathrm{\α})\end{array}\right]---\left(15\right)$

Utilize spin matrix H, lower mixed signalAnd de-correlated signalsThe vector that binaural signal is corresponding can be recovered.

$\left[\begin{array}{c}\stackrel{\→}{L}\\ \stackrel{\→}{R}\end{array}\right]=H\left[\begin{array}{c}\stackrel{\→}{M}\\ \stackrel{\→}{D}\end{array}\right]---\left(16\right)$

ByWithAvailable complex frequency-domain signal L and R, the more left sound after being restored respectively through time-frequency inverse transform block Road signal x_lWith right-channel signals x_r。

Method is evaluated and tested

We this patent propose method contrast with HE-AAC+, encoder code rate selection be 32kbps and 64kbps, test file chooses 12 cycle testss:

Table 1 is 12 cycle testss

Interpretational criteria chooses the objective difference score of signal to noise ratio (SNR), PEAQ.

Signal to noise ratio:

Signal to noise ratio is defined as the ratio of signal power and noise power, and unit is dB, and signal is original audio signal, and noise It it is then the error signal between audio signal and the original audio signal of recovery after parameter stereo encoding and decoding.Signal to noise ratio Audio signal and original audio signal after the highest then explanation reconstruct closer to, tonequality is the best；Otherwise then difference is the biggest, tonequality is more Difference.

$SNR=10{\log }_{10}\left(\frac{SignalPower}{NoisePower}\right)$

Signal to noise ratio result is as shown in Figure 5.

PEAQ score:

Reference signal and test signal, by imitating the auditory system of human ear, are analyzed and calculate by PEAQ algorithm To objective difference score ODG (Objective Difference Grade, ODG), represent and make people feel the interference value detested, take Value scope is [-4,0].PEAQ evaluating standard also uses Pyatyi tonequality decision metrics, and the tonequality of some subjective evaluation methods Decision metrics is similar.

Table 2 is the ODG score meaning of PEAQ

Tonequality	Tonequality damage describes	ODG score value
			Excellent	Cannot perception	0
Good	Can perception but noiseless	-1
			In	The most irritating	-2
Difference	Irritating	-3
			Unacceptable	The most irritating	-4

PEAQ scores is as shown in Figure 6.

Claims (8)

1. a parameter stereo coding method based on orthogonal decorrelation technique, the steps include:

1) L channel of input, R channel time-domain signal are transformed to complex frequency domain respectively, obtain two complex frequency-domain signal L and R；

2) mixed signal M under is mixed and is output as under two complex frequency-domain signal L and R being carried out；

3) utilize complex frequency-domain signal L and R to estimate side information, and carry out described side information merging quantization；Described side information includes sound channel Interior cross-correlation coefficient ICC, Inter channel Intensity Difference IID and gyrobearing information RD；Wherein, formula is utilized Calculate described gyrobearing information RD；

4) lower mixed signal M is carried out time domain inverse transformation, obtain time-domain signal and encode output；Described side information compressed encoding is exported.

2. coded method as claimed in claim 1, it is characterised in that utilize formulaCalculate described sound Cross-correlation coefficient ICC in road.

3. coded method as claimed in claim 1, it is characterised in that utilize formulaCalculate described sound Intensity difference IID between road.

4. the coded method as described in claims 1 to 3 is arbitrary, it is characterised in that with the energy of left and right acoustic channels complex frequency-domain signal and It is compressed encoding to described side information RD, ICC, IID as weight.

5. coded method as claimed in claim 4, it is characterised in that described side information is existed according to audition equivalent rectangular bandwidth Merge on frequency domain；In described sound channel, cross-correlation coefficient ICC is quantified as 3bit, Inter channel Intensity Difference IID is quantified as 5bit, rotation Turn azimuth information RD and be quantified as 1bit.

6. a parameter stereo coding/decoding method based on orthogonal decorrelation technique, the steps include:

1) signal mixed under the time domain of input is transformed to complex frequency domain, export mixed signal M under a complex frequency domain；The side information of input is entered Row decoding, obtains cross-correlation coefficient ICC, Inter channel Intensity Difference IID and gyrobearing information RD in sound channel；

2) for signal M mixed under each complex frequency domain, one is solved according to gyrobearing information RD equal and orthogonal with its amplitude defeated Go out signal D；Wherein, the method solving described signal D is:

21) set under complex frequency domain mixed signal M the vector corresponding to complex plane asUnder a with b is respectively, mixed signal is corresponding VectorReal part and imaginary part；If solving orthogonal vectors corresponding for signal D in advance it isWherein, x Yu y vector respectively Real part and imaginary part；

22) according to formula x²+y²=a²+b²And ax+by=0, obtain two solutions of signal D:

23) according to formulaSolve described signal D；

3) recovery L channel, right sound are calculated according to cross-correlation coefficient ICC, Inter channel Intensity Difference IID, signal M in sound channel with signal D The spin matrix H in road；

4) spin matrix H, lower mixed signal M and signal D is utilized to obtain complex frequency-domain signal L and R；

5) complex frequency-domain signal L and R are changed to time domain respectively, the left channel signals after being restored and right-channel signals.

7. coding/decoding method as claimed in claim 6, it is characterised in that the method solving described spin matrix H is:

71) by vectorPlace axle and vectorPlace axle determines a coordinate system；By L channel sample value to this coordinate origin Vector be designated asR channel sample value is designated as to the vector of this coordinate origin

72) two vectors are determined according to cross-correlation coefficient ICC in sound channelWithAngle 2 α；

73) vector is determined according to Inter channel Intensity Difference IIDMould length compare c；

74) determined than c with mould length by angle αWithMould length compare c₁WithWithMould length compare c₂；

75) c is compared according to mould length₁And c₂Determine angular bisector and the vector of angleAngle β；Obtain described spin matrix

$H=\left[\begin{array}{cc}{c}_2\cos (\mathrm{\β}+\mathrm{\α})& c_2\sin (\mathrm{\β}+\mathrm{\α})\\ c_1\cos (\mathrm{\β}-\mathrm{\α})& c_1\cos (\mathrm{\β}-\mathrm{\α})\end{array}\right].$

8. coding/decoding method as claimed in claim 7, it is characterised in that utilize formulaObtain complex frequency-domain signal L and R.