CN101010726A - Audio decoder, method and program - Google Patents

Abstract

An audio decoder for reproducing original signals from a bit stream including the downmix signals of a plurality of the original signals and auxiliary information representative of the gain ratio D and phase difference theta between the original signals, comprising a decoding part (100) for extracting the downmix signals form the bit stream; a transforming part (101) for transforming the extracted downmix signals to signals of frequency domain; a phase rotator determining part (102) for determining two phase rotators having their respective phase rotation angles a and ss that are obtained by dividing, with a diagonal line, a contained angle of a parallelogram, the length ratio of two adjacent lines of which is equal to the gain ratio D and the contained angle of which is equal to the phase difference theta; a separating part (103) for separating the signals of frequency domain to two separated signals whose phase differences than the decoded downmix signals are the angles alpha and beta, respectively; and an inverse transforming part (104) for inverse transforming the two separated signals into respective signals of time domain to reproduce the two audio signals.

Description

Audio decoder, method and program

Technical field

The present invention relates to according to expression the contract signal that mixes a plurality of original signals and the supplementary of the relation between the original signal, the decode demoder of original signal, and the invention particularly relates to, represent in above-mentioned supplementary original signal to be carried out the technology of high precision decoding under the situation of the phase differential between original signal and the ratio of gains.

Background technology

In recent years, developing the technology that is known as Spatial Codec (space encoding and decoding).Its purpose is and will comes telepresenc to multichannel to compress and encode with considerably less quantity of information, for example, AAC (advanced audio) mode of the multichannel code encoding/decoding mode that has been widely used at voice mode as Digital Television, 5.1 sound channel will have the bit rate of 512kbps or 384kbps, yet, Spatial Codec with 128kbps or 64kbps even the considerably less like this bit rate of 48kbps come multi-channel signal is compressed and is encoded to target.

As being used for above-mentioned technology, for example the record at patent documentation 1 is, by the phase differential between sound channel and the ratio of gains are encoded, thus can be with less quantity of information, handle keeping carrying out compressed encoding under the not impaired situation of telepresenc.

Yet, in the compress mode that has been widely used, the technology that some has adopted a part that phase differential between sound channel or the ratio of gains are encoded.For example, the AAC mode (ISO/IEC13818-7) above-mentioned is mounted with the technology that is known as Intensity Stereo (intensity stereo).

Patent documentation 1:UP2003/0236583A1 U.S. Patent bulletin

Yet, though in patent documentation 1, put down in writing the phase differential between sound channel or the ratio of gains are encoded,, do not disclose out according to such information, should the correct concrete decoding process that is separated into original multi-channel signal of how do i ability.The technology that relates to directivity information how to handle phase differential is not particularly disclosed out.

And in the IntensityStereo of MPEG mode AAC standard (ISO/IEC13818-7), disclosing with a plurality of frequency bands is unit, and quantizes phase differential with the two-value quantified precision.In the case, because do not need the directivity information of phase differential, phase differential can only show with 0 ° and 180 ° like this, so this will become the reason that causes sound quality deterioration.

Summary of the invention

The present invention is in view of above-mentioned problem points in the past, purpose is to provide a kind of audio decoder, it can basis, contracting of original signal is mixed phase information and ratio of gains information are that unit quantizes the resulting information in back with the frequency band between signal and sound channel, comes original signal is correctly regenerated.

In order to solve above-mentioned problem, audio decoder of the present invention is decoded to bit stream, and this bit stream comprises: the 1st coded data, the mixed signal that contracts that expression is contracted and mixed 2 sound signals and obtain; The 2nd coded data is represented the ratio of gains D between above-mentioned 2 sound signals; The 3rd coded data is represented the phase differential θ between above-mentioned 2 sound signals; And above-mentioned audio decoder above-mentioned 2 sound signals of regenerating, this audio decoder is characterised in that, comprising: decoding unit is decoded as the above-mentioned mixed signal that contracts with above-mentioned the 1st coded data; Converting unit will be a frequency-region signal in the mixed conversion of signals that contracts that above-mentioned decoding unit generated; The decision unit, 2 phase place rotation operators of decision phase place rotation angle, these 2 phase place rotation operators are the rotation operators that respectively angle [alpha] and β are made as the phase place rotation angle, above-mentioned angle [alpha] and β are respectively in a parallelogram, cut apart angle and the angle that obtains with diagonal line, the feature of above-mentioned parallelogram is: the ratio of gains D represented in abutting connection with the ratio of 2 length of sides and above-mentioned the 2nd coded data equate, and the represented phasing degree θ of its angle and above-mentioned the 3rd coded data equates; Separative element utilizes above-mentioned 2 phase place rotation operators and the represented ratio of gains D of above-mentioned the 2nd coded data, and above-mentioned frequency-region signal is separated into 2 separation signals, and these 2 separation signals are mixed signal for above-mentioned decoded contracting and held phase differential α and β respectively; And the inverse conversion unit, above-mentioned 2 separation signals are reversed respectively be changed to time-domain signal, and above-mentioned 2 sound signals of regenerating.

Constitute in view of the above, since can regenerate with angle [alpha] and β represented, will contract that to mix signal be the absolute phase of above-mentioned 2 sound signals of benchmark with above-mentioned, so, compare with the conventional art of the relative phase difference θ between above-mentioned 2 sound signals that only regenerates, improved the accuracy in reproduction of signal.

And, also can be above-mentioned decision unit, with 2 plural e ^{-j α}And e ^{J β}, or e ^{-j α}And e ^{J β}Conjugate complex number e ^{J α}And e- ^{J β}Decide as the phase place rotation operator; Above-mentioned separative element, will as above-mentioned phase place rotation operator respectively the decision plural number with multiply each other at the frequency-region signal that above-mentioned converting unit generated, generate above-mentioned 2 separation signals.

And, also can be that above-mentioned bit stream also comprises the 4th coded data, the 4th coded data is represented phase polarity information S, this phase polarity information S represents that the phase place of which signal in above-mentioned 2 sound signals is leading; Above-mentioned separative element, in the conjugate complex number with above-mentioned 2 plural numbers that determine and these 2 plural numbers, with corresponding one group of the represented phase polarity information S of above-mentioned the 4th coded data, multiply each other respectively with at the frequency-region signal that above-mentioned converting unit generated, generate above-mentioned 2 separation signals.

Constitute in view of the above, can correctly invest in order to obtain separation signal ground phase differential at frequency domain.Especially by importing phase polarity information S, the leading or phase lag of the phase place of 2 sound signals of can correctly regenerating.

And, also can be above-mentioned decision unit, according to

α＝arccos((1+Dcosθ)/((1+D ²+2Dcosθ) ^0.5))

β＝arccos((D+cosθ)/((1+D ²+2Dcosθ) ^0.5))

Ask above-mentioned angle [alpha] and angle beta, and decide above-mentioned 2 phase place rotation operators with α that obtains and β.And, also can be above-mentioned decision unit, according to

cosα＝(1+Dcosθ)/((1+D ²+2Dcosθ) ^0.5)

cosβ＝(D+cosθ)/((1+D ²+2Dcosθ) ^0.5)

Ask pairing cos α of above-mentioned angle [alpha] and the pairing cos β of above-mentioned angle beta, and decide above-mentioned 2 phase place rotation operators with cos α that obtains and cos β.

Constitute in view of the above, the absolute phase for above-mentioned above-mentioned 2 sound signals of mixing signal of contracting is obtaining tight regeneration geometrically.Generally speaking, the phase place rotation operator is not direct phase rotation angle, but considers and utilize the trigonometric function of phase rotation angle to represent, especially the formation by the latter, can not carry out the big arccos of calculated amount and calculate, just can determine the phase place rotation operator expeditiously.

And, also can be that above-mentioned the 3rd coded data is with the phase differential θ between above-mentioned 2 sound signals, utilize cos θ to represent in 0 ° to 180 ° scope, the value of the cos θ that above-mentioned the 3rd coded data of above-mentioned decision unit by using is represented decides above-mentioned 2 phase place rotation operators.

Constitute in view of the above, can carry out the calculating of cos θ, just can determine the phase place rotation operator expeditiously.

And, also can be the table that above-mentioned decision unit has the record functional value, this functional value is corresponding with a plurality of phase differential respectively, and is the functional value of representing with the trigonometric function of phase differential at least; Above-mentioned decision unit with reference to by the represented pairing functional value of phase differential θ of above-mentioned the 3rd coded data, decides above-mentioned phase place rotation operator in above-mentioned table.And, also can be that above-mentioned table record has, above-mentioned a plurality of phase differential θ distinguish the value of corresponding sin θ and the value of cos θ; Also can be preferably in above-mentioned table at this, the value of the identical pairing sin θ of phase differential θ and the value of cos θ are recorded in the adjacent areas.

Constitute in view of the above, when the above-mentioned phase place rotation operator of decision, can reduce the processing of trigonometric function at least.And, the value record of the value of sin θ and cos θ in the zone of adjacency, can be obtained functional value expeditiously.

And, also can be that above-mentioned table record has 4 functional values, these 4 functional values are corresponding with a plurality of combinations that ratio of gains D and phase differential θ form respectively, and become

W(D，θ)＝(1+Dcosθ)/((1+D ²+2Dcosθ) ^0.5)

X(D，θ)＝(Dsinθ)/((1+D ²+2Dcosθ) ^0.5)

Y(D，θ)＝(D+cosθ)/((1+D ²+2Dcosθ) ^0.5)

Z(D，θ)＝sinθ/((1+D ²+2Dcosθ) ^0.5)；

Above-mentioned decision unit, above-mentioned 4 functional values of reference in above-mentioned table, decide above-mentioned phase place rotation operator, above-mentioned 4 functional values are corresponding with the combination of forming by the represented ratio of gains D of above-mentioned the 2nd coded data with by the represented phase differential θ of above-mentioned the 3rd coded data respectively; Also can be preferably in above-mentioned table at this, pairing above-mentioned 4 functional values of the combination of identical ratio of gains D and phase differential θ are recorded in the adjacent areas.And, also can be that above-mentioned table record has, according to ratio of gains D above-mentioned 4 functions are carried out revised value.

Constitute in view of the above, when decision phase place rotation operator, all needed values can obtain from table.Especially,, be recorded in the zone of adjacency, just can obtain functional value expeditiously pairing above-mentioned 4 functional values of the combination of identical D and θ.

And, also can be that above-mentioned separative element generates reverb signal by carrying out the reverberation processing, and, with according to above-mentioned phase place rotation operator fixed ratio, above-mentioned frequency-region signal and the above-mentioned reverb signal that is generated are mixed, thereby generate above-mentioned 2 separation signals, above-mentioned reverberation is handled and is meant, to the additional reverberation of the frequency-region signal that above-mentioned converting unit generated.

Constitute in view of the above, by adding a certain amount of reverberation according to above-mentioned phase place rotation operator, just can separation signal and in the technology that acoustically produces simple and honest sense, show the effect of the above-mentioned signal phase of tight reproduction.

And, it also can be above-mentioned bit stream, contain the 2nd coded data and the 3rd coded data respectively at a plurality of frequency bands, the 2nd coded data is illustrated in the ratio of gains D of above-mentioned 2 sound signals of pairing frequency band separately, and the 3rd coded data is illustrated in the phase differential θ of above-mentioned 2 sound signals of pairing frequency band separately; Above-mentioned converting unit is a unit with above-mentioned frequency band, is frequency-region signal with the above-mentioned mixed conversion of signals that contracts; Above-mentioned decision unit is a unit with above-mentioned frequency band, determine 2 phase place rotation operators, these 2 phase place rotation operators are the rotation operators that respectively angle [alpha] and β are made as the phase place rotation angle, above-mentioned angle [alpha] and β are respectively in a parallelogram, cut apart angle and the angle that obtains with diagonal line, the feature of above-mentioned parallelogram is: the ratio of gains D represented in abutting connection with the ratio of 2 length of sides and above-mentioned the 2nd coded data equate, and the represented phasing degree θ of its angle and above-mentioned the 3rd coded data equates; Above-mentioned separative element is a unit with above-mentioned frequency band, utilizes above-mentioned 2 phase place rotation operators and the above-mentioned ratio of gains D that determines, generates 2 separation signals from above-mentioned frequency-region signal; Above-mentioned inverse conversion unit is a unit with above-mentioned frequency band, above-mentioned 2 separation signals reversed respectively is changed to time-domain signal, and from the resulting above-mentioned time-domain signal of all frequency domains, above-mentioned 2 sound signals of regenerating.

And, also can be that above-mentioned bit stream is at least for 1 in the above-mentioned frequency band or for the frequency band lower than the frequency that is predetermined, comprise the 4th coded data, the 4th coded data is represented phase polarity information S, this phase polarity information S is illustrated in the above-mentioned frequency band, and the phase place of which signal is leading in above-mentioned 2 sound signals; Above-mentioned decision unit is a unit with above-mentioned frequency domain, with 2 plural e ^{-j α}And e ^{J β}, or e ^{-j α}And e ^{J β}Conjugate complex number e ^{J α}And e ^{-j β}Decide as above-mentioned 2 phase place rotation operators; Above-mentioned separative element, for the frequency band that does not contain above-mentioned the 4th coded data, the plural number of above-mentioned decision respectively and the frequency-region signal that above-mentioned converting unit is generated are multiplied each other, for the frequency band that contains above-mentioned the 4th coded data, in the conjugate complex number with above-mentioned 2 plural numbers that determine and these 2 plural numbers, with corresponding one group of the represented phase polarity information S of above-mentioned the 4th coded data, multiply each other respectively with at the frequency-region signal that above-mentioned converting unit generated, generate above-mentioned 2 separation signals.

Constitute in view of the above, suitably rotation separates to phase place by coming according to the frequency content of signal, on the whole, can carry out signal regeneration accurately.Especially the sense of hearing of having considered the people is leading or phase lag for phase place, the low characteristics of sensitivity in higher frequency domain, so above-mentioned phase polarity information S is placed on than in the low frequency band of the frequency of predesignating, like this, both can subdue coding information quantity again acoustically not causing the deterioration of tonequality.

And, the present invention not only can be used as audio decoder and realizes, realize that but also can be used as the program that audio-frequency decoding method and computing machine carry out the processing that the above-mentioned distinctive unit that audio decoder had is performed of above-mentioned audio-frequency decoding method is as step.And, also can be used as the integrated circuit (IC) apparatus that is used for audio decoder and realize.

By audio decoder of the present invention, mix ratio of gains D and phase differential θ between signal and above-mentioned 2 sound signals owing to can mix contracting of obtaining according to 2 sound signals are contracted, regenerating with the above-mentioned mixed signal that contracts is the absolute phase of above-mentioned 2 sound signals of benchmark, therefore, compare with the conventional art of the relative phase difference θ between above-mentioned 2 sound signals that only regenerates, improved the accuracy in reproduction of signal.

Description of drawings

Fig. 1 is the figure that the formation of the audio decoder in the present embodiment 1 is shown.

Fig. 2 is the figure of formation that the bit stream of the input that becomes this audio decoder simply is shown.

Fig. 3 is that ratio of gains information, phase information and phase polarity information are shown is stored figure how.

Fig. 4 is the exemplary plot that the state between ratio of gains D and the phase differential θ is shown.

Fig. 5 is the figure that illustrates in the think of thought mode of asking phase differential α and β geometrically.

Fig. 6 (a) illustrates to contract to mix the figure of the relation between signal and original 2 sound channel signals, and Fig. 6 (b) illustrates the signal 1 that contracts when mixing signal and phase place rotation and finishing and the figure of the relation between the signal 2.

Fig. 7 is the figure that the formation of the audio coder in the present embodiment 2 is shown.

Fig. 8 is the figure that illustrates for the encoding book of phase differences.

Fig. 9 is the figure that the encoding book of the phase differential when encoding low bit rate is shown.

Figure 10 illustrates in order to ask other the think of figure of mode only of phase differential α and β geometrically.

Figure 11 is the figure that the formation of the audio decoder in the variation is shown.

Symbol description

100 decoding units

101 converting units

102 phase place rotation operator determining units

103 phase place rotary units

104 inverse conversion unit

200 the 1st coded data storage areas

201 the 2nd coded data storage areas

202 the 3rd coded data storage areas

203 the 4th coded data storage areas

700 the 1st coding units

701 the 1st converting units

702 the 2nd converting units

703 the 1st cutting units

704 the 2nd cutting units

705 the 3rd cutting units

706 the 4th cutting units

707 the 2nd coding units

708 the 3rd coding units

709 formatters

Embodiment

(embodiment 1)

Followingly audio decoder in the embodiments of the present invention 1 is described with reference to accompanying drawing.

Fig. 1 is the figure of formation that the audio decoder of present embodiment 1 is shown.Audio decoder decode bit stream shown in Figure 1, this bit stream comprises: the 1st coded data, the mixed signal that contracts that expression is contracted and mixed 2 sound signals and obtain; The 2nd coded data is represented the ratio of gains D between above-mentioned 2 sound signals; The 3rd coded data is represented the phase differential θ between above-mentioned 2 sound signals; And the 4th coded data, it is leading that expression phase polarity information S, this phase polarity information S show the phase place of which signal in above-mentioned 2 sound signals; And above-mentioned 2 sound signals of regenerating, above-mentioned audio decoder comprises: lsb decoder 100, converter section 101, phase place rotation operator determination section 102, separated part 103 and inverse conversion portion 104.

Lsb decoder 100 is decoded as the above-mentioned mixed signal that contracts with above-mentioned the 1st coded data, and the mixed conversion of signals that contracts that converter section 101 will be generated at above-mentioned lsb decoder 100 is the signal of frequency domain.

Phase place rotation operator determination section 102 determines 2 phase place rotation operators, these 2 phase place rotation operators are respectively the rotation operators that angle [alpha] and β is made as the phase place rotation angle, this angle [alpha] and β are in a parallelogram, cut apart the angle that obtains in abutting connection with the angle on 2 limits with diagonal line, the feature that above-mentioned parallelogram had is: the represented phase differential θ of the angle on adjacency 2 limits and above-mentioned the 3rd coded data equates, and the ratio of above-mentioned adjacency 2 length of sides ratio of gains D represented with representing above-mentioned the 2nd coded data equates.

Separated part 103 is utilized above-mentioned 2 phase place rotation operators and above-mentioned ratio of gains D, the signal of the frequency domain that is generated from above-mentioned converter section 101, isolate 2 separation signals, inverse conversion portion 104 reverses above-mentioned 2 separation signals and is changed to the signal of time domain, and above-mentioned 2 sound signals of regenerating.

Fig. 2 is the figure of formation that the bit stream of the input that becomes this audio decoder is shown simply.In this bit stream, in each frame that sets at interval with official hour, store the 1st to the 4th above-mentioned coded data, in Fig. 2, only show the example of 2 frames.

In Fig. 2, in the 1st coded data storage area the 200, the 2nd coded data storage area the 201, the 3rd coded data storage area 202 and the 4th coded data storage area 203, store the corresponding data about the 1st frame respectively, the formation of the 2nd frame and the 1st frame are same, so repeatedly.

In above-mentioned the 1st coded data storage area 200, for example store the signal that the signal that mixed 2 sound channel signals of will contracting has compressed in MPEG specification AAC mode.This so-called mixed being meant of contracting, signal is carried out the synthetic processing of vector.

In above-mentioned the 2nd coded data storage area 201, store, represent the value of the ratio of gains D between the sound signal of above-mentioned 2 sound channels.In above-mentioned the 3rd coded data storage area 202, store, represent the value of the phase differential θ between the sound signal of above-mentioned 2 sound channels.In above-mentioned the 4th coded data storage area 203, store, represent the phase place of which signal in the sound signal of above-mentioned 2 sound channels leading the value of phase polarity information S.

It should be noted that at this value of representing above-mentioned phase differential θ may not be the value of direct coding phase differential θ, for example also can be the data of the value of coding as cos θ.Under these circumstances, above-mentioned phase differential θ can be represented between 0 ° to 180 ° scope according to the value of cos θ.

Fig. 3 illustrates, and stores the figure of what kind of ratio of gains information, phase information and phase polarity information in above-mentioned the 2nd coded data storage area 201, above-mentioned the 3rd coded data storage area 202 and above-mentioned the 4th coded data storage area 203 respectively.It is stored according to 22 frequency bands that Fig. 3 shows ratio of gains information.For example, the information of the 1st ratio of gains is the information of the ratio of gains of frequency band from 0.000000kHz to 0.086133kHz; The information of the 2nd ratio of gains is the information of the ratio of gains of frequency band from 0.086133kHz to 0.172266kHz; So, coexistence contains 22 ratio of gains information.Equally, show and store 19 phase informations.Equally, show 11 the phase polarity information that store.Certainly, in the band segmentation method shown in Fig. 3 or to cut apart number etc. only be an example, also can be other value.

And, in Fig. 3, the number of phase information is lacked than the number of ratio of gains information, this is because in the characteristic of the sense of hearing, general sensitive cause for ratio of gains information, according to the bit rate of compression or the sample frequency of the sound signal of processing, make the number of phase information identical also harmless with the number of ratio of gains information.

And, also be same for phase polarity information.In the present embodiment, stored the phase polarity information about about 1kHz,, do not stored the polarity information of phase place for the frequency band more than about 1kHz.And, under the low situation of the bit rate of compression, for phase polarity information then 1 do not store yet.This is because on the characteristic of the sense of hearing, and phase polarity information sensitivity does not also reach so high cause.Certainly, have under the vacant situation in the bit rate of compression, the information of still storing the full range band can access better tonequality.

Work for the audio decoder of formation as above-mentioned will describe following.

At first, above-mentioned the 1st coded data of being stored in the above-mentioned bit rate of lsb decoder 100 decodings.As shown in Figure 2, owing to the 1st coded data is, the coded data after resulting 1 sound signal of sound signal of mixed 2 sound channels that will contract is encoded with AAC, therefore, lsb decoder 100 can be realized with the common AAC demoder of decoding bit stream.

Then, the conversion of signals that will decode at lsb decoder 100 of converter section 101 is a frequency-region signal.In the present embodiment, for example according to Fourier transform, the conversion of signals that will decode at lsb decoder 100 is the plural form of the fourier series of frequency domain.And the plural form of the fourier series that is converted is split into group according to shown 22 frequency bands of row on Fig. 3 left side.

At this, show Fourier transform as an example, but also nonessential like this, also can utilize QMF (Quadrature Mirror Filter: group quadrature mirror filter) according to plural number.

And phase place rotation operator determination section 102 according to above-mentioned the 2nd coded data and above-mentioned the 3rd coded data, is obtained the phase place rotation operator as phase place rotation angle α and β.

At this, above-mentioned the 2nd coded data is the value of the ratio of gains of each frequency band between the original signal of expression 2 sound channels, as shown in Figure 3, because ratio of gains D is stored in the bit stream by 22 frequency bands, therefore can obtain ratio of gains information by extracting ratio of gains D.And above-mentioned the 3rd coded data is the value of phase differential θ of each frequency band between the original signal of expression 2 sound channels, as shown in Figure 3, because phase differential is to be stored in the bit stream by 19 frequency bands, therefore, can obtain phase information by extracting phase differential θ.

Like this, how from resulting ratio of gains D and phase differential θ, obtaining contracts mixes the phase differential α and the β of signal and 2 sound channels original signal separately, will utilize Fig. 4 and Fig. 5 to describe.

Fig. 4 is the exemplary plot that the state of ratio of gains D and phase differential θ is shown.The mixed signal that contracts is will represent the signal of 2 arrows of original signal as the diagonal of the parallelogram on 2 limits, the phase differential α and the β of contract mixed signal and original signal separately, position as shown in Figure 4.

Fig. 5 is the figure that is illustrated in the think of thought mode of asking phase differential α and β geometrically.Triangle after Fig. 5 shows the parallelogram of Fig. 4 separated with diagonal line, when cornerwise length was made as X, this leg-of-mutton each length of side was 1, D, X, and folded respectively angle, each limit is α, 180-θ, β., then obtain if utilize the cosine law of trigonometric function at this

X ²=1+D ²-2Dcos (180-θ)=1+D ²+ 2Dcos θ (formula 1)

1=X ²+ D ²-2DXcos β (formula 2)

D ²=1+X ²-2Xcos α (formula 3).

By formula 1, draw X=(1+D ²+ 2Dcos θ) ^0.5

And, draw thus following formula substitution formula 2 and formula 3

α=arccos ((1+Dcos θ)/((1+D ²+ 2Dcos θ) ^0.5)) (formula 4)

β=arccos ((D+cos θ)/((1+D ²+ 2Dcos θ) ^0.5)) (formula 5)

That is,, obtain phase differential α, β, and obtain its corresponding phase place rotation operator according to above-mentioned formula 4 and formula 5 at above-mentioned phase place rotation operator determination section 102.Certainly, above-mentioned explanation is the explanation of carrying out according to the theory of mathematics, in the practice process, also can utilize approximate treatment or trigonometric function tabulation etc.

And, there is no need directly to use the cosine law.For example, untiing the problem of above-mentioned α, β, can utilize geometric problem shown in Figure 10 to solve, is to obtain with following formula naturally:

α＝atan(Dsin(θ)/(1+Dcos(θ)))

β＝atan(sin(θ)/(D+cos(θ)))。

In a word, when asking phase place rotation angle α and β phase differential θ between 2 original sound channel signals and the ratio of gains D, above-mentioned phase place rotation angle α, β asked as following such angle get final product, this angle is: will be D in abutting connection with the ratio on 2 limits, its angle is the above-mentioned angle of the parallelogram of θ, and cuts apart resulting angle with the diagonal line of this parallelogram.

And, in the above description, ask phase place rotation angle α, β with above-mentioned phase place rotation operator determination section 102, but in fact the value of phase place rotation angle α, β itself is unwanted, owing to need obtain the rotation operator e that makes the phase place rotation at this ^{J α}And e ^{-j β}, or the e of its conjugate complex number ^{-j α}And e ^{J β}, then above-mentioned phase place rotation operator determination section 102 need be obtained following trigonometric function value.Speech value on the contrary is as long as it is just passable to obtain following trigonometric function value.Needed trigonometric function value is:

Cos α (e ^{J α}Real part)

Sin α (e ^{J α}Imaginary part)

Cos β (e ^{J β}Real part)

Sin β (e ^{J β}Imaginary part).

That is, in the calculation of asking α, β shown in just now, it is unnecessary utilizing the arccos calculation specially to go to ask the value of α, β itself, as long as carry out the calculation on the right of the equal sign of following formula 6 and formula 7.

Cos α=(1+Dcos θ)/((1+D ²+ 2Dcos θ) ^0.5) (formula 6)

Cos β=(D+cos θ)/((1+D ²+ 2Dcos θ) ^0.5) (formula 7)

Nature can be used the quadratic relationship ((cosX) of trigonometric function about sin α, sin β ²+ (sinX) ²=1) etc. obtains simply.

And separated part 103 is utilized above-mentioned 2 phase place rotation angle α, β and above-mentioned the 4th coded data, will be 2 signals in the Signal Separation of the frequency domain of converter section 101 conversion.This process will be utilized Fig. 6 (a) and (b) describe.

Fig. 6 (a) is the figure that the relation between decoded signal and the original signal that should be separated is shown, and above-mentioned decoded signal promptly contracts the original signal of 2 sound channels and mixes the resulting mixed signal that contracts in back.The long arrow that is positioned at the center is a decoded signal, and in the present embodiment, because decoded signal is converted to fourier series, therefore, this arrow is the vector on the complex number plane.When this vector is made as C, the phase place rotation-α of C be made, just plural e need be established ^{-j α}, and carry out with C*e ^{-j α}The multiplying of represented plural number.The phase place rotation β of C to be made equally, just plural e need be established ^{J β}, and carry out with C*e ^{J β}The multiplying of represented plural number.

Like this when carrying out the multiplying of phase place rotation operator, make the vector C of expression decoded signal phase place rotation-α ,+the β angle, then consequently as Fig. 6 (b) shown in, obtained signal 1 when the expression phase place is rotated end and these 2 vectors of signal 2.The length of these vectors equals the length of vector C.

Then, in order to have carried out the correction with the corresponding gain of amplitude of separated signal, to have made rotating-vector and the modified value 1/ ((1+D of the signal 1 of α ²+ 2Dcos θ) ^0.5) multiply each other, make to have rotated+vector and modified value the D/ ((1+D of the signal 2 of β ²+ 2Dcos θ) ^0.5) multiply each other.The foundation of this correction is, is D in abutting connection with the ratio of the length on 2 limits, and its angle is that cornerwise length of the parallelogram of θ is ((1+D ²+ 2Dcos θ) ^0.5).

And, though told about in the above description, because cornerwise length is ((1+D ²+ 2Dcos θ) ^0.5), in view of the above, by making signal and 1/ ((1+D respectively ²+ 2Dcos θ) ^0.5) and D/ ((1+D ²+ 2Dcos θ) ^0.5) multiply each other, come modified gain, but in when coding, according to phase differential θ, mix signal itself to contracting and gain under the situation about adjusting, be not limit by this.For example, the situation of carrying out following processing is also arranged when coding.

That is, the gain of the 1st signal before the coding is that the gain of 1, the 2 signal is D, and its phase differential is under the situation of θ, and the energy of the signal before mixing that contracts is with (1+D ²) ^0.5Show.In addition, will contract the energy of the signal after mixing with (1+D ²+ 2Dcos θ) ^0.5When showing, according to above-mentioned θ, the energy of mixed signal and the original energy (1+D that signal had then contract ²) ^0.5Different.

Particularly, the contract energy (1+D of the signal after mixing ²+ 2Dcos θ) ^0.5With the original energy (1+D that signal had ²) ^0.5Compare, phase differential is consistent when 90 spend, yet it is also just big more to approach 0 degree phase differential more, and it is just more little to approach 180 degree phase differential more.That is, according to this performance, the energy of the resulting mixed signal that contracts is excessive from in-phase signal, or the energy of the resulting mixed signal that contracts is too small from inverse phase signal.

So the energy that making contracts mixes signal does not exist with ... phase differential, and with the original energy coincidence that signal had, will make contracts mixes signal times with (1+D ²) ^0.5/ (1+D ²+ 2Dcos θ) ^0.5Such adjustment.

When coding, carry out under the situation of such adjustment, then when decoding, at first above-mentioned contracting when encoding mixed the adjustment that signal itself is removed energy, and, will make (1+D in order to return original gain ²+ 2Dcos θ) ^0.5/ (1+D ²) ^0.5With the mixed signal multiplication that contracts, afterwards, when separating, make 1/ above-mentioned ((1+D according to the phasing degree ²+ 2Dcos θ) ^0.5) or D/ ((1+D ²+ 2Dcos θ) ^0.5) and each separated signal multiplication.

By continuous like this multiplying, (the 1+D of denominator, molecule ²+ 2Dcos θ) ^0.5Offset 1/ ((1+D ²) ^0.5) or D/ ((1+D ²) ^0.5) will be processed as the correction multiplier of the ratio of gains.In the case, shown in Fig. 6 (b), signal 1 and signal 2 when the phase place rotation is finished are with the multiplier 1/ ((1+D that only exists with ... ratio of gains D ²) ^0.5) or D/ ((1+D ²) ^0.5) multiply each other, come modified gain with this.

By the rotation of such vector and the correction of length, shown in Fig. 6 (a), can mix Signal Separation with contracting is signal 1 and signal 2 these 2 signals.

In separated part 103, carry out above-mentioned processing according to frequency band shown in Figure 3.At this, be noted that the frequency band in high frequency one side has, have only the situation of 1 phase information for 2 ratio of gains information, at this moment, this 1 phase information then becomes shared information.

And, in the above description, as an example, though make phase place rotation-α and+(that is, rotation operator has used e to β ^{-j α}And e ^{J β}), yet, the relation of advancing Yu retreating according to the original signal phase place, also can occur+α and-situation of β.Relation between the decoded signal of this moment and the original signal that should be separated, can obtain parallelogram (not shown) by the expression of reversing and represent that then the rotation operator that should use this moment is the conjugate complex number e of above-mentioned rotation operator the parallelogram shown in Fig. 6 (a) ^{J α}And e ^{-j β}

For the information of correctly carrying out above-mentioned processing is above-mentioned the 4th coded data, be above-mentioned phase polarity information, this information be present in as shown in Figure 3 in the bit stream of 11 frequency bands of lower frequency side.Utilize this information, can correctly determine the sense of rotation of phase place.Separated part 103 is utilized, and with the corresponding side of phase polarity information, separates 2 signals in 2 plural numbers that phase place rotation operator determination section 102 is determined and their conjugate complex number.

Because this phase polarity information is unnecessary information in the low frequency band of people's the sensitivity of phase polarity acoustically, therefore, can be present in all frequency bands.For the non-existent frequency band of phase polarity information, separated part 103 is directly utilized 2 plural numbers that determined by phase place rotation operator determination section 102, separates 2 signals.

Under the low situation of bit rate, can be used as 1 also non-existent variation of this phase polarity information and consider.A configuration example of the audio decoder that this variation is related is with shown in Figure 11.The related audio decoder of this variation is compared with the audio decoder (with reference to Fig. 1) of handling phase polarity information, difference is: omitted the 4th coded data (S), the direct utilization of separated part 103a by 2 plural numbers that phase place rotation operator determination section 102 is determined, separates 2 signals at all frequency bands.

Under the non-existent situation of above-mentioned phase polarity information, phase differential θ is 180 degree, be that 2 original signals are the situation of phase reversal or the situation that approaches phase reversal, can clearly know contracts mix the phase place that signal had state representation be, the state of the signal phase of the side that the energy in 2 signals originally is strong, therefore, above-mentioned α and β all can be spent as 0.At this moment, though be that 180 signals of spending phase places one side have become phase reversal originally, also can correctly guarantee the signal phase of the side that energy is strong at least.

At last, the signal of the frequency domain that inverse conversion portion 104 will be generated in separated part 103 reverses the signal that becomes time domain.In the present embodiment, with above-mentioned converter section 101 as asking the portion of the plural form of fourier series, 104 processing of carrying out inverse Fourier transform of above-mentioned inverse conversion portion according to Fourier transform.

As mentioned above, by present embodiment, bit stream is decoded, this bit stream comprises: the 1st coded data, expression contract and mix 2 resulting mixed signals that contract of sound signal; With the 2nd coded data, represent the ratio of gains D between above-mentioned 2 sound signals; And the 3rd coded data, represent the phase differential θ between above-mentioned 2 sound signals; In the audio decoder of above-mentioned 2 sound signals of regeneration, owing to have following unit, therefore can contract from sound signal 2 sound channels mix be 1 sound channel resulting contract mix signal and represent phase differential between above-mentioned sound signal and a small amount of supplementary of the ratio of gains, regeneration mixes the absolute phase of signal as above-mentioned 2 sound signals of benchmark with above-mentioned contracting, compare with the conventional art of the relative phase differential θ between above-mentioned 2 sound signals that only regenerates like this, improved the accuracy in reproduction of signal; The unit that above-mentioned audio decoder had is: decoding unit is decoded as the above-mentioned mixed signal that contracts with above-mentioned the 1st coded data; Converting unit, will above-mentioned decoding unit decoded the mixed signal that contracts, be converted to frequency-region signal; The decision unit, 2 phase place rotation operators of decision phase place rotation angle, these 2 phase place rotation operators are the rotation operators that respectively angle [alpha] and β are made as the phase place rotation angle, above-mentioned angle [alpha] and β are respectively in a parallelogram, cut apart angle and the angle that obtains with diagonal line, the feature of above-mentioned parallelogram is: the ratio of gains D represented in abutting connection with the ratio of 2 length of sides and above-mentioned the 2nd coded data equate, and the represented phasing degree θ of its angle and above-mentioned the 3rd coded data equates; Separative element, utilize above-mentioned 2 phase place rotation operators and the represented ratio of gains D of above-mentioned the 2nd coded data, separate above-mentioned frequency-region signal, above-mentioned frequency-region signal is separated into, mix 2 separation signals that signal is held phase differential α and β respectively at above-mentioned decoded contracting; And the inverse conversion unit, above-mentioned 2 separation signals are reversed respectively be changed to time-domain signal, and above-mentioned 2 sound signals of regenerating.

Put down in writing in the present embodiment, the signal of 2 sound channels contracted, and to mix be the Signal Processing of 1 sound channel, but be not so certain, for example contract when mixing at signal with a preceding left side, the preceding right side, a left side, back, right this 4 sound channel in back, mixed front left and rear left can contract earlier, it is right and the back is right before mixing to contract again, and the signal that will contract respectively again after mixing contracts mixed once more; Under the situation before and after separating respectively again about like this formerly separating, the invention that also can use among the application to be put down in writing.

And, in the phase place rotation operator determination section 102 and separated part 103 of present embodiment, owing to need carry out trigonometric function operation, yet in cheap processing such as processor than difficult, therefore, by carrying out following way, can handle very simply.

At first, at phase place rotation operator determination section 102, utilize phase differential θ and ratio of gains D, ask phase differential α and β, but when separated part 103 is carried out the phase place rotation processing, be not to utilize phase differential α and β itself, in fact employed value is e ^{(+/-) j α}And e ^{(-/+) j β}Value.That is, because

e ^(+/-)jα＝cosα(+/-)jsinα

e ^(-/+)jβ＝cosβ(-/+)jsinβ

So in fact needed value is cos α, sin α, cos β, reaches sin β, and passes through following formula

Cos α=(1+Dcos θ)/((1+D ²+ 2Dcos θ) ^0.5) (formula 8)

Sin α=(Dsin θ)/((1+D ²+ 2Dcos θ) ^0.5) (formula 9)

Cos β=(D+cos θ)/((1+D ²+ 2Dcos θ) ^0.5) (formula 10)

Sin β=sin θ/((1+D ²+ 2Dcos θ) ^0.5) (formula 11)

Phase information θ as the address, is added the table that cos θ and sin θ can references, then do not need the processing of trigonometric function, as long as add, multiplication and division, and square root calculation.And this moment is as long as the zone that cos θ and sin θ are write adjacency, just can obtain both sides' value by addressing simply.Especially in processor in recent years, the data transfer path (data bus) with 64 bit widths is more, as long as the zone that cos θ and sin θ are write adjacency, just can obtain both sides' value by a Machine cycle (machine cycle).

And, because cos α, sin α, cos β, and sin β can directly determine by phase information θ and ratio of gains information D, add the bivariate table of the address of setting phase information θ and ratio of gains information D, needed value cos α, sin α, cos β, and sin β in actual operation just can obtain by table access.Certainly in the case, also can be with cos α, the sin α relevant with the combination of ratio of gains information D, cos β, and the sin β zone of writing adjacency with identical phase information θ, and just can obtain all values by addressing simply.

And, in reality, as with reference to above-mentioned Fig. 6 (a) and the detailed description (b) the separating treatment process of signal carried out, the final value of using is in Signal Separation, be used in the phase place rotation processing cos α, sin α, cos β, and sin β multiply by value after the length of vector of the separated signal of expression, promptly multiply by the value after the modified value that is used for the corrected signal gain.

Therefore, preferably with functional value F1 (D, θ), F2 (D, θ) represent above-mentioned modified value, this be not with cos α, sin α, cos β, and the value that had of sin β itself store in the table, but revised value is stored in the table, this revised value is

cosα*F1(D，θ)

sinα*F1(D，θ)

cosβ*F2(D，θ)

sinβ*F2(D，θ)。

At this, convenience is, functional value F1 (D, θ), F2 (D, no matter θ) which all is the functional value of D and θ, also be the bivariate table of the address that constituted with D and θ at this table of considering, therefore do not causing memory size to increase or increasing on the basis of the numerous and diverse degree of accessing step, above-mentioned modified value can be stored in this table and can reference.

At this, in the explanation of above-mentioned signal separation process process, functional value F1 (D, θ), (D θ) is respectively F2

F1(D，θ)＝1/((1+D ²+2Dcosθ) ^0.5)

F2(D，θ)＝D/((1+D ²+2Dcosθ) ^0.5)

But in the coding specification of reality, also have with functional value as

F1(D，θ)＝1/((1+D ²) ^0.5)

F2(D，θ)＝D/((1+D ²) ^0.5)

Situation, as long as coding specification that therefore can be realistic is suitably to adjust above-mentioned modified value.

And, in disclosed MPEG mode Enhanced AAC+SBR mode (ISO14496-3:AMENDMENT2) in recent years, disclose and be not only 2 phase differential θ and ratio of gains D between sound signal, and utilize this is contracted and mix that signal adopts the method for all-pass wave filtering and the reverb signal that makes, will contract and mix the method that Signal Separation after 2 sound signals is original 2 sound signals.Yet at this, to phase place rotation angle α and β be divided into simply+θ/2 and-θ/2.

Described in this application method, owing to be to obtain the phase place rotation angle closely according to geometric principle, again because have more superior separating property than said method, so, the application's method is used in the installation of Enhanced AAC+SBR demoder, just any change can not take place on bit stream, i.e. the tonequality that obtains with the compatibility of relevant stream.In other words, the method described in the application's the embodiment, can with the use that combines of the method for utilizing reverb signal.

In MPEG mode Enhanced AAC+SBR mode (ISO 14496-3:AMENDMENT2), ratio of gains D is encoded as IID (Inter-channel Intensity Differences).And phase differential θ is encoded as IPD (Inter-channel Phase Differences) or ICC (Inter-channel Coherence).Particularly ICC is the index that 2 degrees of correlation between sound signal are shown, therefore, when this value for just and be worth when big, then correlativity is strong, that is to say that phase differential is little.And, when this value near 0 the time, then do not have correlationship, that is to say that phase differential approaches 90 degree.And when this value is negative and its absolute value when big, then Fu correlativity is strong, and the phase differential that is to say this moment approaches 180 and spends.Like this, the ICC parameter that can be used as the phase differential between 2 sound signals of expression is utilized.

And convenience is, because ICC has above-mentioned feature, so ICC is the value of expression for the cos θ of the phase differential θ between 2 sound signals.If ICC itself is exactly the value of cos θ, the value of the cos θ 11 from formula 6 to formula in the then above-mentioned explanation just can directly be used the value of ICC itself, therefore, can become very simple in calculating.

And, to utilize under the situation of above-mentioned reverb signal, the situation of the sharpness of sound can appear losing because of the different in kind of the sound signal of process object.For example, under the bigger situation of the phase differential between 2 sound signals of script, promptly approach the situation of anti-phase, or the big situation of the ratio of gains between 2 sound signals of script, or amplitude sharply changes the strong situation of composition etc. of attacking.Under these circumstances, as long as do not utilize reverb signal just passable.Or prepare the method for multiple generation reverb signal, according to the character of the sound signal of process object, come system of selection or switch.

At this moment, because the character of the sound signal of process object is judged, can carry out in decoding one side, come switching controls according to this judged result, thereby can not carry out any change on bit stream, promptly the compatibility with relevant stream just can obtain high tone quality.

Certainly, in the specification of new coded system, as long as on bit stream, set the sign whether utilize reverb signal, the judgement of the side that just do not need to decode, demoder can be simplified like this.Or which method to generate the sign of reverb signal with as long as set expression, and the judgement of the side that just do not need to decode, demoder can be simplified like this.

At this, set the multiple method that is used to generate reverb signal, for example set the multiple method that is used to generate the phase shift momentum of reverb signal.

And,, also can suitably switch according to the character of signal in the method for the described calculating angle of departure of the application or merely impartial method of distributing.And, also the sign that is used to switch can be added bit stream.

And, also the method for calculating the angle of departure can be fixed in a certain method, will represent whether to use the sign of reverb signal to add bit stream again.

(embodiment 2)

Hereinafter with reference to accompanying drawing the audio coder in the embodiments of the present invention 2 is described.

Fig. 7 is the figure that the formation of the audio coder in the present embodiment 2 is shown.This audio coder is the scrambler that generates bit stream, and this bit stream is an illustrated bit stream of having been decoded well by audio decoder in embodiment 1; This audio coder comprises: the 1st encoding section the 700, the 1st converter section the 701, the 2nd converter section the 702, the 1st cutting part the 703, the 2nd cutting part the 704, the 3rd cutting part the 705, the 4th cutting part the 706, the 2nd encoding section the 707, the 3rd encoding section 708 and formatter 709.

700 pairs of the 1st encoding section signal that has mixed after 2 sound signals that contracts is encoded.

The 1st converter section 701 is converted to frequency-region signal with the 1st sound signal, and the 2nd converter section 702 is converted to frequency-region signal with the 2nd sound signal.

The 1st cutting part 703 to the above-mentioned frequency-region signal that is generated at the 1st converter section 701, is that unit is cut apart with a plurality of frequency bands; The 2nd cutting part 704 to the above-mentioned frequency-region signal that is generated at the 1st converter section 701, is cut apart with the dividing method different with the 1st cutting part 703.

The 3rd cutting part 705 to the above-mentioned frequency-region signal that is generated at the 2nd converter section, is cut apart with the dividing method identical with the 1st cutting part 703; The 4th cutting part 706 to the above-mentioned frequency-region signal that is generated at the 2nd converter section 702, is cut apart with the dividing method identical with the 2nd cutting part 704.

The 2nd encoding section 707 detects the ratios of gains and also encodes, and this ratio of gains is the ratio of gains that the frequency-region signal of the frequency-region signal of each frequency band of being cut apart at the 1st cutting part 703 and each frequency band of being cut apart at the 3rd cutting part 705 corresponds respectively to each frequency band.

The 3rd encoding section 708 detected phase differences and represent the phase place of which signal leading information and encode, above-mentioned phase differential is the phase differential that the frequency-region signal of the frequency-region signal of each frequency band of being cut apart at the 2nd cutting part 704 and each frequency band of being cut apart at the 4th cutting part 706 corresponds respectively to each frequency band.

The formatter 709 multiplexed signals of exporting from above-mentioned the 1st to the 3rd encoding section.

Below will the work of above-mentioned audio coder be described.

At first, in the 1st encoding section 700, the signal that has mixed contracting after 2 sound signals is encoded.At this, for the mixing method that contracts, can be only to carry out additive operation, also can be the coefficient that after carrying out additive operation, multiply by regulation again.In a word, so long as to access the synthetic method of the vector of 2 sound signals just passable.Coding method also can be arbitrarily, adopts MPEG specification AAC mode to encode in the present embodiment.

Secondly, at the 1st converter section 701, the 1st sound signal is converted to frequency-region signal.In the present embodiment, adopt Fourier transform, the sound signal of importing is converted to plural fourier series.

At the 2nd converter section 702, the 2nd sound signal is converted to frequency-region signal.In the present embodiment, adopt Fourier transform, the sound signal of importing is converted to plural fourier series.

Secondly, at the 1st cutting part 703, the above-mentioned frequency-region signal that will be generated at the 1st converter section 701 is cut apart by a plurality of frequency bands.At this moment, the method for cutting apart will be in accordance with table shown in Figure 3.In Fig. 3, the beginning frequency of the frequency band of being cut apart is shown in a tabulation in left side, begins secondary series from a left side and represents the actual dividing method that carries out according to ratio of gains information.That is,, according to row of the leftmost side in the table of Fig. 3 with begin secondary series, and be segmented in the above-mentioned frequency-region signal that above-mentioned the 1st converter section is generated according to shown each frequency band respectively from a left side at the 1st cutting part 703.

The 2nd cutting part 704 also is same, is segmented in the above-mentioned frequency-region signal that the 1st converter section 701 is generated by each frequency band.The dividing method of this moment is deferred to table shown in Figure 3.In Fig. 3, the beginning frequency of the frequency band of being cut apart is shown in a tabulation in left side, begins the 3rd tabulation from a left side and shows the actual dividing method that carries out according to phase differential.That is,, according to row of the leftmost side in the table of Fig. 3 with begin the 3rd row from a left side, and be segmented in the above-mentioned frequency-region signal that above-mentioned the 1st converter section is generated according to shown each frequency band respectively at the 2nd cutting part 704.

Adopt and the identical dividing method of above-mentioned the 1st cutting part 703 at the 3rd cutting part 705, be segmented in the above-mentioned frequency-region signal that above-mentioned the 2nd converter section 702 is generated.

Adopt and the identical dividing method of above-mentioned the 2nd cutting part 704 at the 4th cutting part 706, be segmented in the above-mentioned frequency-region signal that above-mentioned the 2nd converter section 702 is generated.

Secondly, the 2nd encoding section 707 detects the ratios of gains and also encodes, and this ratio of gains is the band signal of being cut apart at the 1st cutting part 703 and the ratio of gains that corresponds respectively to each frequency band at the band signal that the 3rd cutting part 705 is cut apart.At this, for the detection method of the ratio of gains, the method that compares between the maximal value of pairing frequency band amplitude is arranged, or the method that energy level is compared etc., the whichever method can, like this ratio of gains that is detected is encoded in the 2nd encoding section 707.

Secondly, the 3rd encoding section 708 detected phase differences and represent the phase place of which signal leading information, be phase polarity information, and encode, above-mentioned phase differential is band signal of being cut apart at the 2nd cutting part 704 and the phase differential that corresponds respectively to each frequency band at the band signal that the 4th cutting part 706 is cut apart.At this, the method of detected phase difference has, and the typical value of in this frequency band, fourier series, real number value and imaginary value is asked the method for phase differential etc., the whichever method can, like this phase differential and the phase polarity information that are detected are encoded in the 3rd encoding section 708.

At this, hope can be noted the phase polarity Information (right side) of Fig. 3.Only detect the phase polarity information of 11 frequency bands that begin from the lower frequency region side, and coding.This is to have applied flexibly acoustically, and therefore the characteristic that the sensitivity of the phase polarity information of high-frequency domain is paused very late, can wait in expectation and subdue bit stream on the basis that does not reduce tonequality.

Under the low situation of bit stream, 1 of phase polarity information is not encoded yet.At last, in the output signal of formatter 709 multiplexed above-mentioned the 1st to the 3rd encoding section, and forming bit stream, can be arbitrary method as for the formation method.

Based on aforesaid present embodiment,, the signal after mixed 2 sound signals that contract is encoded owing to comprised the 1st encoding section; The 1st converter section is converted to frequency-region signal with above-mentioned the 1st sound signal; The 2nd converter section is converted to frequency-region signal with above-mentioned the 2nd sound signal; The 1st cutting part is that unit is segmented in the above-mentioned frequency-region signal that above-mentioned the 1st converter section is generated with a plurality of frequency bands; The 2nd cutting part, with the different dividing method of above-mentioned the 1st cutting part, be segmented in the above-mentioned frequency-region signal that above-mentioned the 1st converter section is generated; The 3rd cutting part, with the identical dividing method of above-mentioned the 1st cutting part, be segmented in the above-mentioned frequency-region signal that above-mentioned the 2nd converter section is generated; The 4th cutting part, with the identical dividing method of above-mentioned the 2nd cutting part, be segmented in the above-mentioned frequency-region signal that above-mentioned the 2nd converter section is generated; The 2nd encoding section, the detection ratio of gains is also encoded, and this ratio of gains is to cut apart band signal of being cut apart and the band signal of cutting apart at above-mentioned the 3rd cutting part the above-mentioned the 1st, corresponds respectively to the ratio of gains of each frequency band; The 3rd encoding section, the detected phase difference and represent the phase place of which signal leading information, and encode, this phase differential is band signal of cutting apart at above-mentioned the 2nd cutting part and the band signal of cutting apart at above-mentioned the 4th cutting part, corresponds respectively to the phase differential of each frequency band; And formatter, multiplexed output signal from above-mentioned the 1st to the 3rd encoding section; Therefore, can be for 2 original sound channel signals, form a bit stream, thereby realization high compression, comprise in the formation of above-mentioned formed bit stream: will contracting, to mix be the data that obtain after the signal encoding of 1 sound channel, and be the considerably less information of the 2 sound channels data that the back obtains of encoding to being used for this data separating.And this bit stream is suitable at the illustrated audio decoder of embodiment 1, and by above-mentioned audio decoder, is regenerated as 2 original sound channel signals with pinpoint accuracy.

Fig. 8 shows the encoding book of the phase differential that is used for the code book embodiment.When phase differential is θ, the θ that represents with cos θ then has been shown in the table of Fig. 8, and this table be used to encode value of cos θ.The row of the leftmost side show the threshold value when quantizing among Fig. 8.Be shown in Fig. 8 to be the table that is used for representing with the quantized value of 11 grades cos θ value, for example, the cos θ value between-1.000 to-0.969 is considered as identical quantification gradation and encodes.

As shown in Figure 8 this be set at, the value of cos θ is near 0 when (phase differential is near 90 °), with near (phase differential is near 0 °) or-1 near+1 (phase differential is at 180 °) compare, the precision of quantification is coarse.This is because when considering that phase differential is near 90 °, and it is low that the change of phase differential detects sensitivity, and phase differential is near 0 ° or near 180 ° the time, the cause of the auditory properties that the change detection senses degree of phase differential is high.

And as long as set such quantization threshold, near the true rate of generation of the quantized value of natural phase differential 90 ° will uprise, and adopting the variable length code like this is Hoffman code, just can improve code efficiency.One row of Fig. 8 central authorities show the length of the Hoffman code of different quantification gradations, and row on right side show corresponding Hoffman code.As we know from the figure, near the code length of the quantized value 90 ° is very short.

And,, in coding, want to reduce under the situation of bit rate if further utilize this character, as shown in Figure 9, near the quantized value of the quantized value that makes phase differential near 90 ° rate really increases, and therefore, near the quantification degree of accuracy 90 ° carried out coarse setting can raise the efficiency.Its reason is that sensitivity is acoustically paused later near 90 ° of phase differential, suppressed because of quantification causes acoustically deterioration, and rising is frequently spent in the appearance of the code that code length is short, thereby has reduced average bit rate.

Certainly, Fig. 8 only shows an example, for quantification gradation and nonessential is the quantification gradation of 11 values, also be same for the distribution method of Huffman code length, even be not also passable shown in the figure.

Audio decoder involved in the present invention can utilize at audio playback, is particularly useful for carrying out reception machine in music broadcast service or the music distribution service with low bit rate.

Claims (according to the modification of the 19th of treaty)

1. (modification) a kind of audio decoder is decoded to bit stream, and this bit stream comprises: the 1st coded data, the mixed signal that contracts that expression is contracted and mixed 2 sound signals and obtain; The 2nd coded data is represented the ratio of gains D between above-mentioned 2 sound signals; And the 3rd coded data, represent the phase differential θ between above-mentioned 2 sound signals; And above-mentioned audio decoder above-mentioned 2 sound signals of regenerating, this audio decoder is characterised in that, comprising:

Decoding unit is decoded as the above-mentioned mixed signal that contracts with above-mentioned the 1st coded data;

Converting unit will be a frequency-region signal in the mixed conversion of signals that contracts that above-mentioned decoding unit generated;

The decision unit, determine 2 phase place rotation operators, these 2 phase place rotation operators are the rotation operators that respectively angle [alpha] and β are made as the phase place rotation angle, above-mentioned angle [alpha] and β are respectively in a parallelogram, cut apart angle and the angle that obtains with diagonal line, the feature of above-mentioned parallelogram is: the ratio of gains D represented in abutting connection with the ratio of 2 length of sides and above-mentioned the 2nd coded data equate, and the represented phase differential θ of its angle and above-mentioned the 3rd coded data equates;

Separative element utilizes above-mentioned 2 phase place rotation operators and the represented ratio of gains D of above-mentioned the 2nd coded data, and above-mentioned frequency-region signal is separated into 2 separation signals; And

The inverse conversion unit reverses above-mentioned 2 separation signals respectively and to be changed to time-domain signal, and above-mentioned 2 sound signals of regenerating.

2. audio decoder according to claim 1 is characterized in that,

Above-mentioned decision unit is with 2 plural e ^{-j α}And e ^{J β}, or e ^{-j α}And e ^{J β}Conjugate complex number e ^{J α}And e ^{-j β}Decide as above-mentioned 2 phase place rotation operators;

Above-mentioned separative element, will as above-mentioned phase place rotation operator respectively the decision plural number with multiply each other at the frequency-region signal that above-mentioned converting unit generated, generate above-mentioned 2 separation signals.

3. audio decoder according to claim 2 is characterized in that,

Above-mentioned bit stream also comprises the 4th coded data, and the 4th coded data is represented phase polarity information S, and this phase polarity information S represents that the phase place of which signal in above-mentioned 2 sound signals is leading;

Above-mentioned separative element, in the conjugate complex number with above-mentioned 2 plural numbers that determine and these 2 plural numbers, with corresponding one group of the represented phase polarity information S of above-mentioned the 4th coded data, multiply each other respectively with at the frequency-region signal that above-mentioned converting unit generated, generate above-mentioned 2 separation signals.

4. audio decoder according to claim 1 is characterized in that,

Above-mentioned decision unit, according to

α＝arccos((1+Dcosθ)/((1+D ²+2Dcosθ) ^0.5))

β＝arccos((D+cosθ)/((1+D ²+2Dcosθ) ^0.5))

Ask above-mentioned angle [alpha] and angle beta, and decide above-mentioned 2 phase place rotation operators with α that obtains and β.

5. audio decoder according to claim 1 is characterized in that,

Above-mentioned decision unit, according to

cosα＝(1+Dcosθ)/((1+D ²+2Dcosθ) ^0.5)

cosβ＝(D+cosθ)/((1+D ²+2Dcosθ) ^0.5)

Ask pairing cos α of above-mentioned angle [alpha] and the pairing cos β of above-mentioned angle beta, and decide above-mentioned 2 phase place rotation operators with cos α that obtains and cos β.

6. (modification) audio decoder according to claim 1 is characterized in that,

Above-mentioned the 3rd coded data, the phase differential θ with between above-mentioned 2 sound signals represents with the value of cos θ;

Above-mentioned decision unit utilizes the value of the represented cos θ of above-mentioned the 3rd coded data, decides above-mentioned 2 phase place rotation operators.

7. (modification) audio decoder according to claim 6 is characterized in that,

With the value of above-mentioned cos θ, ask as the correlation between above-mentioned 2 sound signals.

8. (modification) audio decoder according to claim 1 is characterized in that,

Above-mentioned decision unit has the table of record functional value, and this functional value is corresponding with a plurality of phase differential respectively, and is the functional value of representing with the trigonometric function of phase differential at least;

Above-mentioned decision unit with reference to by the represented pairing functional value of phase differential θ of above-mentioned the 3rd coded data, decides above-mentioned phase place rotation operator in above-mentioned table.

9. (modification) audio decoder according to claim 8 is characterized in that,

Above-mentioned table record has, and above-mentioned a plurality of phase differential θ distinguish the value of corresponding sin θ and the value of cos θ.

10. (modification) audio decoder according to claim 9 is characterized in that,

In above-mentioned table, the value of the identical pairing sin θ of phase differential θ and the value of cos θ are recorded in the adjacent areas.

(11. modification) audio decoder according to claim 8 is characterized in that,

Above-mentioned table record has 4 functional values, and these 4 functional values are corresponding with a plurality of combinations that ratio of gains D and phase differential θ form respectively, and become

W(D，θ)＝(1+Dcosθ)/((1+D ²+2Dcosθ) ^0.5)

X(D，θ)＝(Dsinθ)/((1+D ²+2Dcosθ) ^0.5)

Y(D，θ)＝(D+cosθ)/((1+D ²+2Dcosθ) ^0.5)

Z(D，θ)＝sinθ/((1+D ²+2Dcosθ) ^0.5)；

Above-mentioned decision unit, above-mentioned 4 functional values of reference in above-mentioned table, decide above-mentioned phase place rotation operator, above-mentioned 4 functional values are corresponding with the combination of forming by the represented ratio of gains D of above-mentioned the 2nd coded data with by the represented phase differential θ of above-mentioned the 3rd coded data respectively.

(12. modification) audio decoder according to claim 11 is characterized in that,

In above-mentioned table, pairing above-mentioned 4 functional values of the combination of identical ratio of gains D and phase differential θ are recorded in adjacent areas

(13. modification) audio decoder according to claim 11 is characterized in that,

Above-mentioned table record has, and according to ratio of gains D above-mentioned 4 functions is carried out revised value.

(14. modification) audio decoder according to claim 1 is characterized in that,

Above-mentioned separative element generates reverb signal by carrying out the reverberation processing, and, with according to above-mentioned phase place rotation operator fixed ratio, above-mentioned frequency-region signal and the above-mentioned reverb signal that is generated are mixed, thereby generate above-mentioned 2 separation signals, above-mentioned reverberation is handled and is meant, to the additional reverberation of the frequency-region signal that above-mentioned converting unit generated.

(15. modification) audio decoder according to claim 1 is characterized in that,

Above-mentioned bit stream, contain the 2nd coded data and the 3rd coded data respectively at a plurality of frequency bands, the 2nd coded data is illustrated in the ratio of gains D of above-mentioned 2 sound signals of pairing frequency band separately, and the 3rd coded data is illustrated in the phase differential θ of above-mentioned 2 sound signals of pairing frequency band separately;

Above-mentioned converting unit is a unit with above-mentioned frequency band, is frequency-region signal with the above-mentioned mixed conversion of signals that contracts;

Above-mentioned decision unit is a unit with above-mentioned frequency band, determine 2 phase place rotation operators, these 2 phase place rotation operators are the rotation operators that respectively angle [alpha] and β are made as the phase place rotation angle, above-mentioned angle [alpha] and β are respectively in a parallelogram, cut apart angle and the angle that obtains with diagonal line, the feature of above-mentioned parallelogram is: the ratio of gains D represented in abutting connection with the ratio of 2 length of sides and above-mentioned second coded data equate, and the represented phasing degree θ of its angle and above-mentioned the 3rd coded data equates;

Above-mentioned separative element is a unit with above-mentioned frequency band, utilizes above-mentioned 2 phase place rotation operators and the above-mentioned ratio of gains D that determines, generates 2 separation signals from above-mentioned frequency-region signal;

The inverse conversion unit reverses above-mentioned 2 separation signals respectively and to be changed to time-domain signal, and above-mentioned 2 sound signals of regenerating.

(16. modification) audio decoder according to claim 15 is characterized in that,

Above-mentioned bit stream comprises the 4th coded data for 1 in the above-mentioned frequency band at least, and the 4th coded data is represented phase polarity information S, and this phase polarity information S is illustrated in the above-mentioned frequency band, and the phase place of which signal is leading in above-mentioned 2 sound signals;

Above-mentioned decision unit is a unit with above-mentioned frequency domain, with 2 plural e ^{-j α}And e ^{J β}, or e ^{-j α}And e ^{J β}Conjugate complex number e ^{J α}And e ^{-j β}Decide as above-mentioned 2 phase place rotation operators;

Above-mentioned separative element, for the frequency band that does not contain above-mentioned the 4th coded data, the plural number of above-mentioned decision respectively and the frequency-region signal that above-mentioned converting unit is generated are multiplied each other, for the frequency band that contains above-mentioned the 4th coded data, in the conjugate complex number with above-mentioned 2 plural numbers that determine and these 2 plural numbers, with corresponding one group of the represented phase polarity information S of above-mentioned the 4th coded data, multiply each other respectively with at the frequency-region signal that above-mentioned converting unit generated, generate above-mentioned 2 separation signals.

(17. modification) audio decoder according to claim 16 is characterized in that,

Above-mentioned bit stream only at the frequency band lower than the frequency of prior regulation, contains above-mentioned the 4th coded data.

18. (modification) a kind of audio-frequency decoding method is decoded to bit stream, this bit stream comprises: the 1st coded data, the mixed signal that contracts that expression is contracted and mixed 2 sound signals and obtain; The 2nd coded data is represented the ratio of gains D between above-mentioned 2 sound signals; And the 3rd coded data, represent the phase differential θ between above-mentioned 2 sound signals; And above-mentioned audio decoder above-mentioned 2 sound signals of regenerating, this audio-frequency decoding method is characterised in that, comprising:

Decoding step is decoded as the above-mentioned mixed signal that contracts with above-mentioned the 1st coded data;

Switch process will be a frequency-region signal in the mixed conversion of signals that contracts that above-mentioned decoding step generated;

Deciding step, determine 2 phase place rotation operators, these 2 phase place rotation operators are the rotation operators that respectively angle [alpha] and β are made as the phase place rotation angle, above-mentioned angle [alpha] and β are respectively in a parallelogram, cut apart angle and the angle that obtains with diagonal line, the feature of above-mentioned parallelogram is: the ratio of gains D represented in abutting connection with the ratio of 2 length of sides and above-mentioned second coded data equate, and the represented phase differential θ of its angle and above-mentioned the 3rd coded data equates;

Separating step utilizes above-mentioned 2 phase place rotation operators and the above-mentioned ratio of gains D that determines, from the frequency-region signal that above-mentioned switch process generated, generates 2 separation signals; And

The inverse conversion step reverses above-mentioned 2 separation signals respectively and to be changed to time-domain signal, and above-mentioned 2 sound signals of regenerating.

19. (increase) a kind of program is decoded to bit stream, this bit stream comprises: the 1st coded data, the mixed signal that contracts that expression is contracted and mixed 2 sound signals and obtain; The 2nd coded data is represented the ratio of gains D between above-mentioned 2 sound signals; And the 3rd coded data, represent the phase differential θ between above-mentioned 2 sound signals; And computing machine is carried out be used to the regenerate audio decoder of above-mentioned 2 sound signals to be handled, be it is characterized in that, make computing machine carry out following steps:

Decoding step is decoded as the above-mentioned mixed signal that contracts with above-mentioned the 1st coded data;

Switch process will be a frequency-region signal in the mixed conversion of signals that contracts that above-mentioned decoding step generated;

Deciding step, determine 2 phase place rotation operators, these 2 phase place rotation operators are the rotation operators that respectively angle [alpha] and β are made as the phase place rotation angle, above-mentioned angle [alpha] and β are respectively in a parallelogram, cut apart angle and the angle that obtains with diagonal line, the feature of above-mentioned parallelogram is: the ratio of gains D represented in abutting connection with the ratio of 2 length of sides and above-mentioned second coded data equate, and the represented phase differential θ of its angle and above-mentioned the 3rd coded data equates;

Separating step utilizes above-mentioned 2 phase place rotation operators and the above-mentioned ratio of gains D that determines, from the frequency-region signal that above-mentioned switch process generated, generates 2 separation signals; And

The inverse conversion step reverses above-mentioned 2 separation signals respectively and to be changed to time-domain signal, and above-mentioned 2 sound signals of regenerating.

Claims (18)

1. an audio decoder is decoded to bit stream, and this bit stream comprises: the 1st coded data, the mixed signal that contracts that expression is contracted and mixed 2 sound signals and obtain; The 2nd coded data is represented the ratio of gains D between above-mentioned 2 sound signals; And the 3rd coded data, represent the phase differential θ between above-mentioned 2 sound signals; And above-mentioned audio decoder above-mentioned 2 sound signals of regenerating, this audio decoder is characterised in that, comprising:

Decoding unit is decoded as the above-mentioned mixed signal that contracts with above-mentioned the 1st coded data;

Converting unit will be a frequency-region signal in the mixed conversion of signals that contracts that above-mentioned decoding unit generated;

The decision unit, determine 2 phase place rotation operators, these 2 phase place rotation operators are the rotation operators that respectively angle [alpha] and β are made as the phase place rotation angle, above-mentioned angle [alpha] and β are respectively in a parallelogram, cut apart angle and the angle that obtains with diagonal line, the feature of above-mentioned parallelogram is: the ratio of gains D represented in abutting connection with the ratio of 2 length of sides and above-mentioned the 2nd coded data equate, and the represented phase differential θ of its angle and above-mentioned the 3rd coded data equates;

Separative element, utilize above-mentioned 2 phase place rotation operators and the represented ratio of gains D of above-mentioned the 2nd coded data, above-mentioned frequency-region signal is separated into 2 separation signals, and these 2 separation signals and the above-mentioned decoded phase differential that mixes signal that contracts are respectively angle [alpha] and β; And

The inverse conversion unit reverses above-mentioned 2 separation signals respectively and to be changed to time-domain signal, and above-mentioned 2 sound signals of regenerating.

2. audio decoder according to claim 1 is characterized in that,

Above-mentioned decision unit is with 2 plural e ^{-j α}And e ^{J β}, or e ^{-j α}And e ^{J β}Conjugate complex number e ^{J α}And e ^{-j β}Decide as above-mentioned 2 phase place rotation operators;

Above-mentioned separative element, will as above-mentioned phase place rotation operator respectively the decision plural number with multiply each other at the frequency-region signal that above-mentioned converting unit generated, generate above-mentioned 2 separation signals.

3. audio decoder according to claim 2 is characterized in that,

Above-mentioned bit stream also comprises the 4th coded data, and the 4th coded data is represented phase polarity information S, and this phase polarity information S represents that the phase place of which signal in above-mentioned 2 sound signals is leading;

Above-mentioned separative element, in the conjugate complex number with above-mentioned 2 plural numbers that determine and these 2 plural numbers, with corresponding one group of the represented phase polarity information S of above-mentioned the 4th coded data, multiply each other respectively with at the frequency-region signal that above-mentioned converting unit generated, generate above-mentioned 2 separation signals.

4. audio decoder according to claim 1 is characterized in that,

Above-mentioned decision unit, according to

α＝arccos((1+Dcosθ)/((1+D ²+2Dcosθ) ^0.5))

β＝arccos((D+cosθ)/((1+D ²+2Dcosθ) ^0.5))

Ask above-mentioned angle [alpha] and angle beta, and decide above-mentioned 2 phase place rotation operators with α that obtains and β.

5. audio decoder according to claim 1 is characterized in that,

Above-mentioned decision unit, according to

cosα＝(1+Dcosθ)/((1+D ²+2Dcosθ) ^0.5)

cosβ?＝(D+cosθ)/((1+D ²+2Dcosθ) ^0.5)

Ask pairing cos α of above-mentioned angle [alpha] and the pairing cos β of above-mentioned angle beta, and decide above-mentioned 2 phase place rotation operators with cos α that obtains and cos β.

6. audio decoder according to claim 1 is characterized in that,

Above-mentioned the 3rd coded data, the phase differential θ with between above-mentioned 2 sound signals in 0 ° to 180 ° scope, represents with the value of cos θ;

Above-mentioned decision unit utilizes the value by the represented cos θ of above-mentioned the 3rd coded data, decides above-mentioned 2 phase place rotation operators.

7. audio decoder according to claim 1 is characterized in that,

Above-mentioned decision unit has the table of record functional value, and this functional value is corresponding with a plurality of phase differential respectively, and is the functional value of representing with the trigonometric function of phase differential at least;

Above-mentioned decision unit with reference to by the represented pairing functional value of phase differential θ of above-mentioned the 3rd coded data, decides above-mentioned phase place rotation operator in above-mentioned table.

8. audio decoder according to claim 7 is characterized in that,

Above-mentioned table record has, and above-mentioned a plurality of phase differential θ distinguish the value of corresponding sin θ and the value of cos θ.

9. audio decoder according to claim 8 is characterized in that,

In above-mentioned table, the value of the identical pairing sin θ of phase differential θ and the value of cos θ are recorded in the adjacent areas.

10. audio decoder according to claim 7 is characterized in that,

Above-mentioned table record has 4 functional values, and these 4 functional values are corresponding with a plurality of combinations that ratio of gains D and phase differential θ form respectively, and become

W(D，θ)＝(1+Dcosθ)/((1+D ²+2Dcosθ) ^0.5)

X(D，θ)＝(Dsinθ)/((1+D ²+2Dcosθ) ^0.5)

Y(D，θ)＝(D+cosθ)/((1+D ²+2Dcosθ) ^0.5)

Z(D，θ)＝sinθ/((1+D ²+2Dcosθ) ^0.5)；

Above-mentioned decision unit, above-mentioned 4 functional values of reference in above-mentioned table, decide above-mentioned phase place rotation operator, above-mentioned 4 functional values are corresponding with the combination of forming by the represented ratio of gains D of above-mentioned the 2nd coded data with by the represented phase differential θ of above-mentioned the 3rd coded data respectively.

11. audio decoder according to claim 10 is characterized in that,

In above-mentioned table, pairing above-mentioned 4 functional values of the combination of identical ratio of gains D and phase differential θ are recorded in the adjacent areas.

12. audio decoder according to claim 10 is characterized in that,

Above-mentioned table record has, and according to ratio of gains D above-mentioned 4 functions is carried out revised value.

13. audio decoder according to claim 1 is characterized in that,

Above-mentioned separative element generates reverb signal by carrying out the reverberation processing, and, with according to above-mentioned phase place rotation operator fixed ratio, above-mentioned frequency-region signal and the above-mentioned reverb signal that is generated are mixed, thereby generate above-mentioned 2 separation signals, above-mentioned reverberation is handled and is meant, to the additional reverberation of the frequency-region signal that above-mentioned converting unit generated.

14. audio decoder according to claim 1 is characterized in that,

Above-mentioned bit stream, contain the 2nd coded data and the 3rd coded data respectively at a plurality of frequency bands, the 2nd coded data is illustrated in the ratio of gains D of above-mentioned 2 sound signals of pairing frequency band separately, and the 3rd coded data is illustrated in the phase differential θ of above-mentioned 2 sound signals of pairing frequency band separately;

Above-mentioned converting unit is a unit with above-mentioned frequency band, is frequency-region signal with the above-mentioned mixed conversion of signals that contracts;

Above-mentioned decision unit is a unit with above-mentioned frequency band, determine 2 phase place rotation operators, these 2 phase place rotation operators are the rotation operators that respectively angle [alpha] and β are made as the phase place rotation angle, above-mentioned angle [alpha] and β are respectively in a parallelogram, cut apart angle and the angle that obtains with diagonal line, the feature of above-mentioned parallelogram is: the ratio of gains D represented in abutting connection with the ratio of 2 length of sides and above-mentioned the 2nd coded data equate, and the represented phasing degree θ of its angle and above-mentioned the 3rd coded data equates;

Above-mentioned separative element is a unit with above-mentioned frequency band, utilizes above-mentioned 2 phase place rotation operators and the above-mentioned ratio of gains D that determines, generates 2 separation signals from above-mentioned frequency-region signal;

Above-mentioned inverse conversion unit is a unit with above-mentioned frequency band, above-mentioned 2 separation signals reversed respectively is changed to time-domain signal, and from the resulting above-mentioned time-domain signal of all frequency domains, above-mentioned 2 sound signals of regenerating.

15. audio decoder according to claim 14 is characterized in that,

Above-mentioned bit stream comprises the 4th coded data for 1 in the above-mentioned frequency band at least, and the 4th coded data is represented phase polarity information S, and this phase polarity information S is illustrated in the above-mentioned frequency band, and the phase place of which signal is leading in above-mentioned 2 sound signals;

Above-mentioned decision unit is a unit with above-mentioned frequency domain, with 2 plural e ^{-j α}And e ^{J β}, or e ^{-j α}And e ^{J β}Conjugate complex number e ^{J α}And e ^{-j β}Decide as above-mentioned 2 phase place rotation operators;

Above-mentioned separative element, for the frequency band that does not contain above-mentioned the 4th coded data, the plural number of above-mentioned decision respectively and the frequency-region signal that above-mentioned converting unit is generated are multiplied each other, for the frequency band that contains above-mentioned the 4th coded data, in the conjugate complex number with above-mentioned 2 plural numbers that determine and these 2 plural numbers, with corresponding one group of the represented phase polarity information S of above-mentioned the 4th coded data, multiply each other respectively with at the frequency-region signal that above-mentioned converting unit generated, generate above-mentioned 2 separation signals.

16. audio decoder according to claim 15 is characterized in that,

Above-mentioned bit stream only at the frequency band lower than the frequency of prior regulation, contains above-mentioned the 4th coded data.

17. an audio-frequency decoding method is decoded to bit stream, this bit stream comprises: the 1st coded data, the mixed signal that contracts that expression is contracted and mixed 2 sound signals and obtain; The 2nd coded data is represented the ratio of gains D between above-mentioned 2 sound signals; And the 3rd coded data, represent the phase differential θ between above-mentioned 2 sound signals; And above-mentioned audio decoder above-mentioned 2 sound signals of regenerating, this audio-frequency decoding method is characterised in that, comprising:

Decoding step is decoded as the above-mentioned mixed signal that contracts with above-mentioned the 1st coded data;

Switch process will be a frequency-region signal in the mixed conversion of signals that contracts that above-mentioned decoding step generated;

Deciding step, determine 2 phase place rotation operators, these 2 phase place rotation operators are the rotation operators that respectively angle [alpha] and β are made as the phase place rotation angle, above-mentioned angle [alpha] and β are respectively in a parallelogram, cut apart angle and the angle that obtains with diagonal line, the feature of above-mentioned parallelogram is: the ratio of gains D represented in abutting connection with the ratio of 2 length of sides and above-mentioned the 2nd coded data equate, and the represented phase differential θ of its angle and above-mentioned the 3rd coded data equates;

Separating step utilizes above-mentioned 2 phase place rotation operators and the above-mentioned ratio of gains D that determines, from the frequency-region signal that above-mentioned switch process generated, generates 2 separation signals; And

The inverse conversion step reverses above-mentioned 2 separation signals respectively and to be changed to time-domain signal, and above-mentioned 2 sound signals of regenerating.

18. a program is decoded to bit stream, this bit stream comprises: the 1st coded data, the mixed signal that contracts that expression is contracted and mixed 2 sound signals and obtain; The 2nd coded data is represented the ratio of gains D between above-mentioned 2 sound signals; And the 3rd coded data, represent the phase differential θ between above-mentioned 2 sound signals; And computing machine is carried out be used to the regenerate audio decoder of above-mentioned 2 sound signals to be handled, be it is characterized in that, make computing machine carry out following steps:

Decoding step is decoded as the above-mentioned mixed signal that contracts with above-mentioned the 1st coded data;

Switch process will be a frequency-region signal in the mixed conversion of signals that contracts that above-mentioned decoding step generated;

Deciding step, determine 2 phase place rotation operators, these 2 phase place rotation operators are the rotation operators that respectively angle [alpha] and β are made as the phase place rotation angle, above-mentioned angle [alpha] and β are respectively in a parallelogram, cut apart angle and the angle that obtains with diagonal line, the feature of above-mentioned parallelogram is: the ratio of gains D represented in abutting connection with the ratio of 2 length of sides and above-mentioned the 2nd coded data equate, and the represented phase differential θ of its angle and above-mentioned the 3rd coded data equates;

Separating step utilizes above-mentioned 2 phase place rotation operators and the above-mentioned ratio of gains D that determines, from the frequency-region signal that above-mentioned switch process generated, generates 2 separation signals; And

The inverse conversion step reverses above-mentioned 2 separation signals respectively and to be changed to time-domain signal, and above-mentioned 2 sound signals of regenerating.

Images