CN102801427B - Encoding and decoding method and system for variable-rate lattice vector quantization of source signal - Google Patents

Abstract

The invention relates to a coding and decoding method and system for variable rate lattice vector quantization of a source signal. The coding method comprises the following steps: s1, transforming the input source signal from the time domain to the frequency domain to obtain spectral coefficients and control information; s2, grouping and bit allocating the spectral coefficients to obtain bit allocation information; s3, lattice vector quantizing the spectral coefficients based on the bit allocation information; s4, packing the quantization index, the bit allocation information and the control information into a coded bit stream. Compared with the traditional variable rate vector quantizer which stores a plurality of vector codebooks, the method of the invention does not need to store the vector codebooks; in addition, a fast algorithm exists, and the operation complexity of the algorithm is greatly reduced compared with the traditional vector quantization, namely the algorithm has the advantage of low operation complexity; third, there is an advantage in that variable rate quantization can be realized.

Description

The decoding method of source signal variable Rate triangular norm over lattice and system

Technical field

The present invention relates to source signal coding field, more specifically, relate to decoding method and the system of source signal variable Rate triangular norm over lattice.

Background technology

What the coding of existing digital source signal adopted usually is transition coding, signal to be encoded is divided into the sampling block of frame by it, and adopt the such as linear orthogonal transformation such as discrete Fourier transform (DFT), discrete cosine transform to process every frame signal, ask for conversion coefficient, then variation coefficient is quantized, to improve compression effectiveness further.

A kind of method conventional in quantization method is vector quantization method, several downsampling factor group is formed a vector wherein together, and is similar to (quantification) each vector with a code book item.For quantizing adjoint point nearest in the code book item code book that normally basis " apart from minimum " criterion draws selected by input vector.In codebook set, increase more code book can increase bit rate and complexity, but the average distortion of quantification can be reduced.

On the other hand, in order to adapt to the feature of the continuous change in source, usually self-adaptive background updatemodel is used.By self-adaptive background updatemodel, different code book sizes can be used to quantize source vector.In transition coding, in the maximum situation being no more than the available bit number quantizing all coefficients, the bit number distributing to source vector depends on the energy of this vector relative to other vectors in same frame usually.Fig. 1 and Fig. 2 describes the quantification block diagram that common variable Rate quantizes coding and decoding device in detail.Variable Rate quantizing encoder shown in Fig. 1 and Fig. 2 and decoder use multiple code book, and they have different bit rates usually, to quantize source vector x.Usually by converting signal and obtaining all conversion coefficients or its subset, source vector is obtained.

Common variable Rate quantizing encoder has been shown in Fig. 1, and its critical component is the quantizer represented with Q, and this quantizer is for selecting an a codebook number n and code vector index i to characterize the quantized value y of source vector x.Codebook number n indicates the code book that encoder is selected, and index i represents the code vector selected in this specific code book.

Usually, suitable lossless coding technique is applied to block E respectively _nand E _iin n and i (that is, the E in Fig. 1 _nand E _i), so that they are compounded in multiplexer MUN with store or by traffic channel before, reduce by the codebook number n encoded _ewith index i _emean bit rate.

Fig. 2 shows variable Rate quantization decoder device.The input of this decoder provides for by binary code n _eand i _ebe separated demodulation multiplexer DEMUX; This decoder also comprises losslessly encoding module (that is, D _nand D _i), decode n wherein _eand i _efor codebook number n and index i; This decoder also comprises and receives codebook number n and index i and the inverse quantizer carrying out re-quantization (uses Q ^-1represent), it uses codebook number and index to carry out the quantized value y of Restorer varieties vector x.Different n values usually produces different bits and distributes thus produce different bit rates, and often the required bit number (that is, codebook bit rate) of dimension is defined as: the ratio distributing to the bit number of source vector and the dimension of source vector.

Usually, the structure of code book can adopt following multiple method:

Popular method is the distribution according to source, adopts training algorithm (as k mean algorithm) to optimize code book item.The method obtains destructuring code book, and it must carry out storing and exhaustive search for each source vector to be quantified usually.Therefore, the shortcoming of the method is that memory requirements is large, and calculation of complex, it forms exponential increase with the increase of codebook bit rate.If variable Rate method is based on above-mentioned non-structured code book, then memory requirements defect that is large and calculation of complex can strengthen, further because usually need for specific code book is distributed in each possible position.

Another kind method uses triangular norm over lattice device, which reduces search complexity, and in many cases, effectively can reduce storage demand.Triangular norm over lattice is a kind of algebraically type vector quantizer, and its feature is in Multidimensional signal space, constructs a kind of regular network, and the point in network is called lattice point, and carries out vector quantization with lattice point, and signal space is divided into cell.Because network is regular, therefore lattice point and cell are also regular.The major advantage of triangular norm over lattice device easily constructs code book, and can carry out higher-dimension quantification.Fig. 3 shows the example in two-dimensional space, and wherein basic vector is v1 and v2, and the lattice used in this example are basic hexagonal-lattice, use Α ₂represent, a little can obtain as follows with the institute of cross mark in this figure:

y＝k1v1+k2v2 (1)

Wherein, y is space lattice, and k1 and k2 can be any integer.Notice that Fig. 3 is a subset of representation space lattice point, because this space lattice itself can infinitely be expanded.

When selecting a certain space lattice to construct quantification code book, the a certain subset of usual selection lattice point obtains the code book with given (limited) bit number, the benefit using lattice point is when determining the nearest neighbor point of source vector x of all lattice points in code book, there is fast codebook search algorithms, and compared with other non-structured code books, greatly can reduce complexity.In addition, use lattice point without the need to storing code book, because code book can obtain from generator matrix.

The lattice point often used in triangular norm over lattice is D8 lattice.D8 is by the Z8 lattice point v=(v of 8 dimension integer lattice ₁..., v ₈) composition, and meet that is:

In D8 lattice, any 8 dimension lattice point y can be generated by following method _:

y＝[k ₁k ₂… k ₈]G _D8(3)

Wherein k ₁, k ₂..., k ₈the integer having symbol, G _d8be generator matrix, be defined as:

$G=\left[\begin{array}{cccccccc}2& 0& 0& 0& 0& 0& 0& 0\\ 1& 1& 0& 0& 0& 0& 0& 0\\ 1& 0& 1& 0& 0& 0& 0& 0\\ 1& 0& 0& 1& 0& 0& 0& 0\\ 1& 0& 0& 0& 1& 0& 0& 0\\ 1& 0& 0& 0& 0& 1& 0& 0\\ 1& 0& 0& 0& 0& 0& 1& 0\\ 1& 0& 0& 0& 0& 0& 0& 1\end{array}\right]---\left(4\right)$

Easy checking G _d8it is the inverse matrix of generator matrix for:

$G^{-1}=\left[\begin{array}{cccccccc}\frac{1}{2}& 0& 0& 0& 0& 0& 0& 0\\ -\frac{1}{2}& 1& 0& 0& 0& 0& 0& 0\\ -\frac{1}{2}& 0& 1& 0& 0& 0& 0& 0\\ -\frac{1}{2}& 0& 0& 1& 0& 0& 0& 0\\ -\frac{1}{2}& 0& 0& 0& 1& 0& 0& 0\\ -\frac{1}{2}& 0& 0& 0& 0& 1& 0& 0\\ -\frac{1}{2}& 0& 0& 0& 0& 0& 1& 0\\ -\frac{1}{2}& 0& 0& 0& 0& 0& 0& 1\end{array}\right]---\left(5\right)$

This inverse matrix is very useful when obtaining the coordinate of D8 lattice point y.

Summary of the invention

Memory space is caused to increase because of the increase with code book number and because needing to carry out entirely searching for the defects such as the computational complexity height making it search for obtain best quantization vector to code book in quantizing process, spy of the present invention gives following technical scheme to solve traditional variable Rate vector quantizer.

The first technical scheme that the present invention solves the employing of its technical problem constructs the coding method of a source signals variable Rate triangular norm over lattice, comprises the steps:

S1, input source signal is transformed from the time domain to frequency domain to obtain spectral coefficient and control information;

S2, described spectral coefficient to be divided into groups and bit divides and is equipped with acquisition bit distribution information;

S3, based on described bit distribution information, spectral coefficient described in triangular norm over lattice;

S4, quantization index, described bit distribution information, described control information are packaged into coded bit stream.

Wherein, described step S3 comprises the steps: further

S31, for described spectral coefficient, calculate skew after vector;

S32, convergent-divergent is carried out to the vector after described skew, obtain scaled vectors;

S33, the lattice point that search and described scaled vectors are closed on most in D8 grid space;

The lattice point coordinate closed on most described in S34, calculating;

S35, described coordinate is utilized to calculate D8 grid vector;

Whether S36, more described D8 grid vector be consistent with the described lattice point closed on most, if unanimously, then quantizes to terminate, export described coordinate; If inconsistent, then described scaled vectors performed and approach quantification.

In the coding method of source signal variable Rate triangular norm over lattice of the present invention, the quantification that approaches in described step S36 comprises further:

S361, by described scaled vectors convergent-divergent again, obtain scaled vectors again, use step S33-S35 to calculate the second lattice point closed on most, the second lattice point coordinate closed on most, and the 2nd D8 grid vector;

Whether S362, more described 2nd D8 grid vector be equal with described second lattice point closed on most, if unequal, then repeats step S361, until described 2nd D8 grid vector is equal with described second lattice point closed on most.

In the coding method of source signal variable Rate triangular norm over lattice of the present invention, the quantification that approaches in described step S36 comprises further:

S363, utilization step S33-S35 calculate the 3rd D8 grid vector, the 3rd lattice point closed on most and the 3rd lattice point coordinate closed on most;

S364, more described 3rd D8 grid vector and described 3rd lattice point closed on most, if both are unequal, then quantize to terminate, export the described 3rd lattice point coordinate closed on most and quantizing bit number; If both are equal, then repeat step S363 until both are unequal, finally export the described 3rd lattice point coordinate closed on most and quantizing bit number.

In the coding method of source signal variable Rate triangular norm over lattice of the present invention, in described step S31, the vector after described skew meets:

${\stackrel{\‾}{y}}_p=y_p-a,$

Wherein, represent the vector after skew, y _prepresent the sub-vector of described spectral coefficient, a=(2 ^-62 ^-62 ^-6) be offset vector.

In the coding method of source signal variable Rate triangular norm over lattice of the present invention, in described step S32, described scaled vectors meets:

${\hat{y}}_p={\stackrel{\‾}{y}}_p/\mathrm{\β}\left(p\right);$

Wherein, represent described scaled vectors, β (p)=2 ^{r (p)}/ 6 represent zoom factor, and R (p) represents the quantizing bit number that the sub-vector of each described spectral coefficient distributes.

In the coding method of source signal variable Rate triangular norm over lattice of the present invention, R (p) meets:

$\underset{p=1}{\overset{L}{\mathrm{\Σ}}}8\×R\left(p\right)+\mathrm{\Ω}=\mathrm{\Ψ},0\≤R\left(p\right)\≤10$

Wherein, the number of described spectral coefficient is N, and N number of described spectral coefficient is divided into L 8 dimension sub-vectors, Ψ represents that the quantization encoding bit number that a frame source signal is total, Ω represent that frame source signal remaining bits number after bit distribution algorithm is Ω.

The second technical scheme that the present invention solves the employing of its technical problem constructs the coded system of a source signals variable Rate triangular norm over lattice, comprising:

Orthogonal transform module, for transforming from the time domain to frequency domain to obtain spectral coefficient and control information by input source signal;

Spectral coefficient grouping and bit distribution module, for divide into groups to described spectral coefficient and bit divides and is equipped with acquisition bit distribution information;

Triangular norm over lattice module, for based on described bit distribution information, spectral coefficient described in triangular norm over lattice;

Coded-bit flow module, for being packaged into coded bit stream by quantization index, described bit distribution information, described control information;

Wherein, described triangular norm over lattice module comprises based on spectral coefficient described in described bit distribution information triangular norm over lattice:

For described spectral coefficient, calculate the vector after skew;

Convergent-divergent is carried out to the vector after described skew, obtains scaled vectors;

The lattice point that search and described scaled vectors are closed on most in D8 grid space;

The lattice point coordinate closed on most described in calculating;

Described coordinate is utilized to calculate D8 grid vector;

Whether more described D8 grid vector is consistent with the described lattice point closed on most, if unanimously, then quantizes to terminate, exports described coordinate; If inconsistent, then described scaled vectors performed and approach quantification.

The 3rd technical scheme that the present invention solves the employing of its technical problem constructs the coding/decoding method of a source signals variable Rate triangular norm over lattice, comprises the steps:

S1, received code bit stream carry out decoding to obtain decoding bit stream;

S2, described decoding bit stream carried out to bit distribution and quantization index decoding;

S3, carry out inverse triangular norm over lattice based on the quantization index of decoding and obtain rebuilding quantization vector;

S4, based on described control information, inverse orthogonal transformation is carried out to described reconstruction quantization vector and obtain reconstruction signal;

Wherein, described step S3 comprises the steps: further

The quantizing bit number that the sub-vector that S31, decoding obtain each spectral coefficient distributes and D8 lattice coordinate;

S32, described D8 lattice coordinate is utilized to calculate D8 grid vector;

S33, inverse convergent-divergent is carried out to described D8 grid vector, obtain scaled vectors;

S34, described scaled vectors added offset vector obtains rebuilding quantization vector.

The 4th technical scheme that the present invention solves the employing of its technical problem constructs the decode system of a source signals variable Rate triangular norm over lattice, comprising:

Coded bit stream decoder module, carries out decoding to obtain decoding bit stream for received code bit stream;

Bit distributes and quantization index decoder module, for carrying out bit distribution and quantization index decoding to described decoding bit stream;

Inverse triangular norm over lattice module, obtains rebuilding quantization vector for carrying out inverse triangular norm over lattice based on the quantization index of decoding;

Inverse orthogonal transformation module, obtains reconstruction signal for carrying out inverse orthogonal transformation based on described control information to described reconstruction quantization vector;

Wherein, described inverse triangular norm over lattice module is carried out inverse triangular norm over lattice based on the quantization index of decoding and is obtained rebuilding quantization vector and comprise:

The quantizing bit number that the sub-vector that decoding obtains each spectral coefficient distributes and D8 lattice coordinate;

Described D8 lattice coordinate is utilized to calculate D8 grid vector;

Inverse convergent-divergent is carried out to described D8 grid vector, obtains scaled vectors;

Described scaled vectors is added offset vector obtains rebuilding quantization vector.

Store multiple vector code book compared to traditional variable Rate vector quantizer, the inventive method is without the need to storing vector code book; In addition, there is fast algorithm, its computational complexity comparatively classical Vector quantization significantly reduces, and namely has the advantage of low computational complexity; 3rd, also there is the advantage that can realize no-load voltage ratio rate quantization.

It will be appreciated by those skilled in the art that foregoing general is only used to provide the simple description of particular aspects of the present invention.In conjunction with the drawings also with reference to claim and following detailed description of preferred embodiment, can obtain and understand more completely of the present invention.

Accompanying drawing explanation

Below in conjunction with drawings and Examples, the invention will be further described, in accompanying drawing:

Fig. 1 is the quantification block diagram of traditional variable Rate quantizing encoder;

Fig. 2 is the quantification block diagram of traditional variable Rate quantization decoder device;

Fig. 3 is the subset representing certain space lattice;

Fig. 4 A shows the block diagram of encoder according to an embodiment of the invention;

Fig. 4 B shows the block diagram of decoder according to an embodiment of the invention;

Fig. 5 shows the flow chart of quantization method according to an embodiment of the invention; And

Fig. 6 shows the flow chart of inverse quantization method according to an embodiment of the invention.

Embodiment

Main purpose of the present invention be to provide a kind of for the source signal variable Rate quantification technique based on space D8 lattice point (such as, comprise quantization method and quantization system etc.), this quantification technique compared with prior art, can realize variable Rate quantize, and without the need to memory space and computational complexity low.This quantification technique can be applied to various variable rate coding system and graduated encoding system.Easy unlikelyly to obscure to describe, the embodiment of the present invention may eliminate content known in those skilled in the art, such as the index code decode algorithm etc. of bit distribution algorithm, spectral coefficient grouping algorithm, triangular norm over lattice.

According to a preferred embodiment of the present invention, the encoder of source signal variable Rate triangular norm over lattice mainly comprises: the grouping of orthogonal transform module 101, spectral coefficient and bit distribution module 102 and triangular norm over lattice module 103 and coded-bit flow module 104.And the decoder of source signal variable Rate triangular norm over lattice is the inverse system of the encoder of source signal variable Rate triangular norm over lattice, mainly comprises coded bit stream decoder module 204, bit distributes and quantization index decoder module 201, inverse triangular norm over lattice module 202 and orthogonal transform module 203.Fig. 4 diagrammatically illustrates the block diagram of whole codec.

Specifically, in Figure 4 A, according to one embodiment of present invention, at coding side, in module 101, utilize one of such as discrete cosine transform (DCT), the orthogonal transform of discrete cosine transform (MDCT) etc. that improves the primary signal of input to be transformed from the time domain to frequency domain, obtain spectral coefficient and control information; Next, in module 102, spectral coefficient is divided into groups and bit distribution, obtain bit distribution information; Then, in module 103, based on bit distribution information, triangular norm over lattice is carried out to spectral coefficient; Finally, control information, bit distribution information and quantization index are packaged into coded bit stream in module 104, input channel or storage.

In Fig. 4 B, according to one embodiment of present invention, in decoding end, in module 204, the coded bit stream received is decoded, and binding modules 201 carries out bit distribution and quantization index decoding; In module 202, carry out according to the quantization index of decoding in module 201 quantization vector that inverse triangular norm over lattice obtains reconstruction; Finally, in the module 203, the quantization vector of reconstruction carries out inverse orthogonal transformation under the control of control information, obtains reconstruction signal.

In Figure 5, an implementation procedure more specifically of the present invention is shown in detail.Assuming that the spectral coefficient number that a frame signal obtains after such as one of linear orthogonal transformation such as discrete Fourier transform (DFT), discrete cosine transform process is N, above-mentioned N number of spectral coefficient is divided into L 8 dimension sub-vectors (namely, meet 8 × L=N), assuming that the quantizing bit number that each 8 dimension sub-vector distributes is R (p) bit/dimension (p=0,1, L-1), the total quantization encoding bit number of one frame signal is Ψ, after bit distribution algorithm, remaining bits number is Ω, and coding side quantization step is as follows:

In step 301, according to total coding code check and selected quantization bit allocation algorithm determine each 8 dimension sub-vectors distribute quantizing bit numbers be R (p) bit/dimension (p=0,1 ... L-1), R (p) should meet following restriction:

$\underset{p=1}{\overset{L}{\mathrm{\Σ}}}8\×R\left(p\right)+\mathrm{\Ω}=\mathrm{\Ψ},0\≤R\left(p\right)\≤10---\left(6\right)$

Next, in step 302, to a certain arbitrary 8 n dimensional vector n y _p=(y _{p, 1}y _{p, 2}y _{p, 8}), p=0 ... L-1, is deducted a certain offset vector a=(2 ^-62 ^-62 ^-6), obtain the vector after offseting

${\stackrel{\‾}{y}}_p=y_p-a---\left(7\right)$

In step 303, to the offset vector that above-mentioned steps draws carry out convergent-divergent and obtain scaled vectors wherein zoom factor β (p)=2 ^{r (p)}/ 6, then

In step 304, search and scaled vectors in D8 grid space close on most lattice point V, it meets:

$\begin{array}{cc}\∀z\∈D8& \left|\right|v-{\hat{y}}_p\left|\right|\end{array}\≤\left|\right|z-{\hat{y}}_p\left|\right|---\left(8\right)$

In step 305, calculate D8 sound of laughing some V and block coordinate k=(k in the Voronoi expansion of R (p) bit/dimension ₁k ₂k ₈), wherein 0≤k _i≤ 2 ^{r (p)}-1, i=1,2 ..., 8, k is calculated as:

$k=\left(v{G}_{D8}^{-1}\right)\⊕r,r=2^{R\left(p\right)}---\left(9\right)$

Wherein be the inverse generator matrix of D8 lattice, see formula (5).

Within step 306, according to giving position fixing k=(k ₁k ₂k ₈) calculate x=kG _d8and z=r ^-1x (wherein G _d8be the generator matrix of D8 lattice, see formula 4), and in D8 grid space, search and scaled vectors z close on lattice point λ most, then calculate D8 grid vector c:

c＝x-rλ (10)

In step 307, comparative case vector C and V, if c with V is consistent, then coordinate k=(k ₁k ₂k ₈) be exactly scaled vectors best coordinates, quantize terminate.If c and V is inconsistent, so vector for point not in the know, now need quantizing of Approach by inchmeal.

Next, the method for above-mentioned steps 303 ~ 306 is used, Approach by inchmeal.In step 308, by vector convergent-divergent 2, namely in step 309,310, search and scaled vectors in D8 grid space close on most lattice point u, calculate the coordinate j of u, that is:

$j=\left(u{G}_{D8}^{-1}\right)\⊕r,r=2^{R\left(p\right)}---\left(11\right)$

In step 311,312, utilize coordinate j to calculate D8 grid vector c ', comparative case vector C ' and u, if c ' and u is unequal, then repeat step 308 to step 312, until c ' is equal with u.

In step 313, scaled vectors is calculated wherein m=3,4,5 or 6; In a step 314, calculate

In step 315, search and scaled vectors in D8 grid space close on most lattice point u ', calculate the coordinate j ' of u ', that is:

$j^{\′}=\left(u^{\′}{G}_{D8}^{-1}\right)\⊕r,r=2^{R\left(p\right)}---\left(12\right)$

In step 316, coordinate j ' is utilized to calculate D8 grid vector c ", comparative case vector C " and u ', if c " equal with u ', k=j ', and repeat step 314 to step 316.If c " and u ' is unequal, then stop circulation.

In step 318, by each 8 dimension sub-vectors distribute quantizing bit numbers be R (p) bit/dimension (p=0,1. ..., L-1) and D8 lattice coordinate k _pdecoding end is passed to after coding.

The method that relatively above-mentioned coding side quantizes, Fig. 6 shows the flow chart of decoding end re-quantization, and concrete implementation step is as follows:

In step 401,402, from coding code stream decoding obtain each 8 dimension sub-vectors distribute quantizing bit numbers be R (p) bit/dimension (p=0,1 ..., L-1) and D8 lattice coordinate k _p.

Next, in step 403, to giving position fixing k _p=(k _{p, 1}k _{p, 2}k _{p, 8}) calculate x=kG _d8, and in D8 grid space, search and scaled vectors z close on lattice point λ most, then calculate D8 grid vector

${\hat{y}}_p=x-\mathrm{r\λ}---\left(13\right)$

Then, in step 404, to vector carry out inverse convergent-divergent to obtain zoom factor is β (p)=2 ^{r (p)}/ 6:

${\hat{y}}_p={\stackrel{\‾}{y}}_p/\mathrm{\β}\left(p\right)---\left(14\right)$

Finally, in step 405, by vector add offset vector a=(2 ^-62 ^-62 ^-6), obtain rebuilding vector

${\tilde{y}}_p={\stackrel{\‾}{y}}_p+a---\left(15\right)$

It should be noted that, the present invention is not limited to and quantizes spectral coefficient, is also applicable to the quantification of LPC coefficient in speech coding.In addition, the present invention can be applied to various variable rate coding system and graduated encoding system, has applicability widely.

Claims (9)

1. the coding method of a source signals variable Rate triangular norm over lattice, is characterized in that, comprise the steps:

S1, the source signal of input is transformed from the time domain to frequency domain to obtain spectral coefficient and control information;

S2, described spectral coefficient to be divided into groups and bit divides and is equipped with acquisition bit distribution information;

S3, based on described bit distribution information, spectral coefficient described in triangular norm over lattice;

S4, quantization index, described bit distribution information, described control information are packaged into coded bit stream;

Wherein, described step S3 comprises the steps: further

S31, for described spectral coefficient, calculate skew after vector;

S32, convergent-divergent is carried out to the vector after described skew, obtain scaled vectors;

S33, the lattice point that search and described scaled vectors are closed on most in D8 grid space;

The lattice point coordinate closed on most described in S34, calculating;

S35, described coordinate is utilized to calculate D8 grid vector;

Whether S36, more described D8 grid vector be consistent with the described lattice point closed on most, if unanimously, then quantizes to terminate, export described coordinate; If inconsistent, then described scaled vectors performed and approach quantification.

2. the coding method of source signal variable Rate triangular norm over lattice according to claim 1, is characterized in that, the quantification that approaches in described step S36 comprises the steps: further

S361, by described scaled vectors convergent-divergent again, obtain scaled vectors again, use step S33-S35 to calculate the second lattice point closed on most, the second lattice point coordinate closed on most and the 2nd D8 grid vector;

Whether S362, more described 2nd D8 grid vector be equal with described second lattice point closed on most, if unequal, then repeats step S361, until described 2nd D8 grid vector is equal with described second lattice point closed on most.

3. the coding method of source signal variable Rate triangular norm over lattice according to claim 2, is characterized in that, the quantification that approaches in described step S36 comprises the steps: further

S363, utilization step S33-S35 calculate the 3rd D8 grid vector, the 3rd lattice point closed on most and the 3rd lattice point coordinate closed on most;

S364, more described 3rd D8 grid vector and described 3rd lattice point closed on most, if both are unequal, then quantize to terminate, export the described 3rd lattice point coordinate closed on most and quantizing bit number; If both are equal, then repeat step S363 until both are unequal, finally export the described 3rd lattice point coordinate closed on most and quantizing bit number.

4. the coding method of the source signal variable Rate triangular norm over lattice according to claim arbitrary in claim 1-3, is characterized in that, in described step S31, the vector after described skew meets:

${\stackrel{\‾}{y}}_p=y_p-a,$

Wherein, represent the vector after skew, y _prepresent the sub-vector of described spectral coefficient, a=(2 ^-62 ^-62 ^-6) be offset vector.

5. the coding method of source signal variable Rate triangular norm over lattice according to claim 4, is characterized in that, in described step S32, described scaled vectors meets:

${\hat{y}}_p={\stackrel{\‾}{y}}_p/\mathrm{\β}\left(p\right);$

Wherein, represent described scaled vectors, β (p)=2 ^{r (p)}/ 6 represent zoom factor, and R (p) represents the quantizing bit number that the sub-vector of each described spectral coefficient distributes.

6. the coding method of source signal variable Rate triangular norm over lattice according to claim 5, is characterized in that, R (p) meets:

$\begin{array}{cc}\underset{p=1}{\overset{L}{\mathrm{\Σ}}}8\×R\left(p\right)+\mathrm{\Ω}=\mathrm{\Ψ}& 0\≤R\left(p\right)\≤10\end{array}$

Wherein, the number of described spectral coefficient is N, and N number of described spectral coefficient is divided into L 8 dimension sub-vectors, Ψ represents that the quantization encoding bit number that a frame source signal is total, Ω represent that frame source signal remaining bits number after bit distribution algorithm is Ω.

7. the coded system of a source signals variable Rate triangular norm over lattice, is characterized in that, comprising:

Orthogonal transform module, for transforming from the time domain to frequency domain to obtain spectral coefficient and control information by the source signal of input;

Spectral coefficient grouping and bit distribution module, for divide into groups to described spectral coefficient and bit divides and is equipped with acquisition bit distribution information;

Triangular norm over lattice module, for based on described bit distribution information, spectral coefficient described in triangular norm over lattice;

Coded-bit flow module, for being packaged into coded bit stream by quantization index, described bit distribution information, described control information;

Wherein, described triangular norm over lattice module comprises based on spectral coefficient described in described bit distribution information triangular norm over lattice:

For described spectral coefficient, calculate the vector after skew;

Convergent-divergent is carried out to the vector after described skew, obtains scaled vectors;

The lattice point that search and described scaled vectors are closed on most in D8 grid space;

The lattice point coordinate closed on most described in calculating;

Described coordinate is utilized to calculate D8 grid vector;

Whether more described D8 grid vector is consistent with the described lattice point closed on most, if unanimously, then quantizes to terminate, exports described coordinate; If inconsistent, then described scaled vectors performed and approach quantification.

8. the coding/decoding method of a source signals variable Rate triangular norm over lattice, is characterized in that, comprise the steps:

S1, received code bit stream carry out decoding to obtain decoding bit stream;

S2, described decoding bit stream carried out to bit distribution and quantization index decoding;

S3, carry out inverse triangular norm over lattice based on the quantization index of decoding and obtain rebuilding quantization vector;

S4, based on control information, inverse orthogonal transformation is carried out to described reconstruction quantization vector and obtain reconstruction signal;

Wherein, described step S3 comprises the steps: further

The quantizing bit number that the sub-vector that S31, decoding obtain each spectral coefficient distributes and D8 lattice coordinate;

S32, described D8 lattice coordinate is utilized to calculate D8 grid vector;

S33, inverse convergent-divergent is carried out to described D8 grid vector, obtain scaled vectors;

S34, described scaled vectors added offset vector obtains rebuilding quantization vector.

9. the decode system of a source signals variable Rate triangular norm over lattice, is characterized in that, comprising:

Coded bit stream decoder module, carries out decoding to obtain decoding bit stream for received code bit stream;

Bit distributes and quantization index decoder module, for carrying out bit distribution and quantization index decoding to described decoding bit stream;

Inverse triangular norm over lattice module, obtains rebuilding quantization vector for carrying out inverse triangular norm over lattice based on the quantization index of decoding;

Inverse orthogonal transformation module, obtains reconstruction signal for carrying out inverse orthogonal transformation based on control information to described reconstruction quantization vector;

Wherein, described inverse triangular norm over lattice module is carried out inverse triangular norm over lattice based on the quantization index of decoding and is obtained rebuilding quantization vector and comprise:

The quantizing bit number that the sub-vector that decoding obtains each spectral coefficient distributes and D8 lattice coordinate;

Described D8 lattice coordinate is utilized to calculate D8 grid vector;

Inverse convergent-divergent is carried out to described D8 grid vector, obtains scaled vectors;

Described scaled vectors is added offset vector obtains rebuilding quantization vector.