CN102801427A - 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 an input source signal from a time domain to a frequency domain to obtain a spectral coefficient and control information; s2, grouping and bit distribution are carried out on the spectral coefficients to obtain bit distribution information; s3, based on the bit distribution information, carrying out lattice vector quantization on the spectrum coefficient; and S4, packing the quantization index, the bit distribution 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; there is also an advantage in that variable rate quantization can be realized.

Description

Decoding method and system that source signal variable Rate grid vector quantizes

Technical field

The present invention relates to the source signal coding field, more specifically, relate to decoding method and system that source signal variable Rate grid vector quantizes.

Background technology

What the coding of existing digital source signal adopted usually is transition coding; It is divided into signal to be encoded the sampling block of frame; And adopt and every frame signal is handled such as linear orthogonal transformations such as DFT, discrete cosine transforms; Ask for conversion coefficient, then variation coefficient is quantized, with further raising compression effectiveness.

A kind of method commonly used in quantization method is a vector quantization method, therein, several sampling coefficient sets is formed a vector together, and with a code book item each vector is similar to (quantification).For quantizing nearest adjoint point in the code book that the selected code book item of input vector normally draws according to " distance is minimum " criterion.In the code book set, increase more code book and can increase bit rate and complexity, but can reduce the average distortion of quantification.

On the other hand, for the characteristic of the continuous variation that adapts to the source, use adaptive bit to distribute usually.Distribute through adaptive bit, can use different code book sizes to quantize the source vector.In transition coding, under the maximum situation that is no more than the available bit number that quantizes all coefficients, the bit number of distributing to the source vector depends on the energy of this vector with respect to other vectors in the same frame usually.Fig. 1 and Fig. 2 describe the quantification block diagram that common variable Rate quantizes the coding and decoding device in detail.Variable Rate quantizing encoder shown in Fig. 1 and Fig. 2 and decoder use a plurality of code books, and they have different bit usually, to quantize source vector x.Usually through signal being carried out conversion and obtaining all conversion coefficients or its subclass, obtain the source vector.

Common variable Rate quantizing encoder has been shown among Fig. 1, and its critical component is the quantizer of representing with Q, and this quantizer is used to select 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 is illustrated in the code vector of selecting in this specific code book.

Usually, suitable lossless coding technique is applied to piece E respectively _nAnd E _iIn n and i (that is E among Fig. 1, _nAnd E _i), so that they are being compounded among the multiplexer MUN to store or, to reduce the codebook number n that is encoded through before the traffic channel _EWith index i _EMean bit rate.

Fig. 2 shows variable Rate quantization decoder device.The input of this decoder provides and has been used for binary code n _EAnd i _ESeparation solution multiplexer DEMUX; This decoder also comprises losslessly encoding module (that is D, _nAnd D _i), n therein decodes _EAnd i _EBe codebook number n and index i; The inverse quantizer that this decoder also comprises n of receiving code this shop and index i and carries out re-quantization (is used Q ^-1Expression), it uses codebook number and index to recover the quantized value y of source vector x.Thereby different n values usually produces different Bit Allocation in Discrete and produces different bit, being defined as of the required bit number of every dimension (that is code book bit rate): the ratio of dimension of distributing to bit number and the source vector of source vector.

Usually, the structure of code book can adopt following several different methods:

A kind of popular method is the distribution according to the source, adopts training algorithm (like the k mean algorithm) to optimize the code book item.This method obtains the destructuring code book, and it must be stored and exhaustive search for each source vector to be quantified usually.Therefore, the shortcoming of this method is that memory requirements is big, and calculation of complex, and it forms exponential increase with the increase of code book bit rate.If the variable Rate method is based on above-mentioned non-structured code book, then memory requirements defective big and calculation of complex can further strengthen, because need distribute specific code book for each possible position usually.

Another kind method is to use the grid vector quantizer, and it has reduced search complexity, and in many cases, can reduce storage demand effectively.It is a kind of algebraically type vector quantizer that grid vector quantizes, and its feature is in the multidimensional signal space, constructs a kind of network clocklike, and the point in the network is called lattice point, and carries out vector quantization with lattice point, is divided into cell to signal space.Because network is clocklike, so lattice point and cell also are clocklike.The major advantage of grid vector quantizer is to construct code book easily, and can carry out the higher-dimension quantification.Fig. 3 shows the example in the two-dimensional space, and wherein basic vector is v1 and v2, and the lattice that use in this example are basic hexagonal-lattice, use Α ₂Expression, the institute with the cross sign among this figure a bit can obtain as follows:

y=k1v1+k2v2 （1）

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

When selecting a certain space lattice to construct the quantification code book; Usually a certain subclass of selection lattice point obtains to have the code book of given (limited) bit number; When the benefit of using lattice point is the nearest neighbor point of source vector x of all lattice points in confirming code book; There is quick code book searching algorithm, and compares, can greatly reduce complexity with other non-structured code books.In addition, use lattice point to need not to store code book, because code book can obtain from generator matrix.

The lattice point that often uses during grid vector quantizes is the D8 lattice.D8 is the Z8 lattice point v=(v by 8 dimension integer lattice ₁..., v ₈) form, and satisfy

That is:

Any 8 dimension lattice point y can generate through following method in the D8 lattice:

y=[k ₁?k ₂…k ₈]G _D8 （3）

K wherein ₁, k ₂..., k ₈Be the integer that symbol is arranged, G _D8Be generator matrix, be defined as:

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

Verify G easily _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 at the coordinate time that obtains D8 lattice point y.

Summary of the invention

For solve traditional variable Rate vector quantizer cause memory space to increase because of increase with the code book number and in quantizing process because of searching for entirely to obtain the defectives such as computational complexity height that best quantization vector makes its search to code book, spy of the present invention has provided following technical scheme.

The present invention solves first technical scheme that its technical problem adopts, and constructs the coding method that a kind of source signal variable Rate grid vector quantizes, and comprising:

S1 transforms from the time domain to frequency domain to obtain spectral coefficient and control information with the input source signal;

S2, to said spectral coefficient divide into groups with Bit Allocation in Discrete to obtain bit distribution information;

S3, based on said bit distribution information, grid vector quantizes said spectral coefficient;

S4 is packaged into coded bit stream with quantization index, said bit distribution information, said control information.

In the coding method that source signal variable Rate grid vector of the present invention quantizes, said step S3 further comprises:

S31 for said spectral coefficient, calculates offset vector;

S32 carries out convergent-divergent to said offset vector, obtains scaled vectors;

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

S34 calculates the said lattice point coordinate that closes on most;

S35 utilizes said coordinate Calculation D8 grid vector;

S36, whether more said D8 grid vector is consistent with the said lattice point that closes on most, if consistent, then quantizes to finish, and exports said coordinate; If inconsistent, then said scaled vectors carried out and approached quantification.

In the coding method that source signal variable Rate grid vector of the present invention quantizes, the quantification that approaches among the said step S36 further comprises:

S361 with said scaled vectors convergent-divergent once more, obtains scaled vectors once more, and utilization step S33-S35 calculates lattice point coordinate and the 2nd D8 grid vector that second lattice point, said second that close on most closes on most;

Whether S362, the lattice point that more said the 2nd D8 grid vector and said second closes on most equate, if unequal, repeating step S361 then, the lattice point that closes on most until said the 2nd D8 grid vector and said second equates.

In the coding method that source signal variable Rate grid vector of the present invention quantizes, the quantification that approaches among the said step S36 further comprises:

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

S364, the lattice point that more said the 3rd D8 grid vector and the said the 3rd closes on most if both are unequal, then quantize to finish, and exports the said the 3rd lattice point coordinate and the quantizing bit number that close on most; If both equate that then repeating step S363 is unequal until both, export the said the 3rd lattice point coordinate and the quantizing bit number that close on most at last.

In the coding method that source signal variable Rate grid vector of the present invention quantizes, in said step S31, said offset vector satisfies:

${\stackrel{\‾}{y}}_p=y_p-a,$

Wherein,

The expression offset vector, y _pThe sub-vector of representing said spectral coefficient, a=(2 ^-62 ^-62 ^-6).

In the coding method that source signal variable Rate grid vector of the present invention quantizes, in said step S32, said scaled vectors satisfies:

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

Wherein,

Represent said scaled vectors, β (p)=2 ^{R (p)}/ 6 expression zoom factors, R (p) representes the quantizing bit number that the sub-vector of each said spectral coefficient distributes.

In the coding method that source signal variable Rate grid vector of the present invention quantizes, R (p) satisfies:

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

Wherein, the number of said spectral coefficient is N, and a said N spectral coefficient is divided into L 8 dimension sub-vectors, and ψ representes the quantization encoding bit number that a frame source signal is total, and Ω representes that the frame source signal is Ω through remaining bits number behind the bit distribution algorithm.

The present invention solves second technical scheme that its technical problem adopts, and constructs the coded system that a kind of source signal variable Rate grid vector quantizes, and comprising:

The orthogonal transform module is used for the input source signal is transformed from the time domain to frequency domain to obtain spectral coefficient and control information;

Spectral coefficient divide into groups with the Bit Allocation in Discrete module, be used for to said spectral coefficient divide into groups and Bit Allocation in Discrete with the acquisition bit distribution information;

The grid vector quantization modules is used for based on said bit distribution information, and grid vector quantizes said spectral coefficient;

The coded-bit flow module is used for quantization index, said bit distribution information, said control information are packaged into coded bit stream.

The present invention solves the 3rd technical scheme that its technical problem adopts, and constructs the coding/decoding method that a kind of source signal variable Rate grid vector quantizes, and comprising:

S1, received code bit stream decode to obtain decoding bit stream;

S2 carries out Bit Allocation in Discrete and quantization index decoding to said decoding bit stream;

S3 carries out contrary grid vector based on the quantization index of decoding and quantizes to obtain rebuilding quantization vector;

S4 carries out inverse orthogonal transformation based on said control information to said reconstruction quantization vector and obtains reconstruction signal.

The present invention solves the 4th technical scheme that its technical problem adopts, and constructs the decoder module that a kind of source signal variable Rate grid vector quantizes, and comprising:

The coded bit stream decoder module is used for the received code bit stream and decodes to obtain decoding bit stream;

Bit Allocation in Discrete and quantization index decoder module are used for said decoding bit stream is carried out Bit Allocation in Discrete and quantization index decoding;

Contrary grid vector quantization modules is used for carrying out contrary grid vector based on the quantization index of decoding and quantizes to obtain rebuilding quantization vector;

The inverse orthogonal transformation module is used for based on said control information said reconstruction quantization vector being carried out inverse orthogonal transformation and obtains reconstruction signal.

Store a plurality of vector code books than traditional variable Rate vector quantizer, the inventive method need not to store the vector code book; In addition, have fast algorithm, its computational complexity quantizes to reduce significantly than conventional vector, promptly has the advantage of low computational complexity; The 3rd, also have the advantage that can realize that no-load voltage ratio speed quantizes.

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

Description of drawings

To combine accompanying drawing and embodiment that the present invention is described further below, in the 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 a sub-set of certain space lattice of expression;

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 provides a kind of based on the source signal variable Rate quantification technique of space D8 lattice point (for example being used for; Comprise quantization method and quantization system etc.); This quantification technique compared with prior art can realize that variable Rate quantizes, and need not memory space and computational complexity is low.This quantification technique can be applied to various variable rate coding systems and graduated encoding system.Easy unlikelyly obscure in order to narrate, the embodiment of the invention possibly omitted content known in those skilled in the art, for example the index code decode algorithm that quantizes of bit distribution algorithm, spectral coefficient grouping algorithm, grid vector etc.

In according to a preferred embodiment of the present invention, the encoder that source signal variable Rate grid vector quantizes mainly comprises: orthogonal transform module 101, spectral coefficient divide into groups and Bit Allocation in Discrete module 102 and grid vector quantization modules 103 and coded-bit flow module 104.And the decoder that source signal variable Rate grid vector quantizes is the inverse system of the encoder of source signal variable Rate grid vector quantification, mainly comprises coded bit stream decoder module 204, Bit Allocation in Discrete and quantization index decoder module 201, contrary grid vector quantization modules 202 and orthogonal transform module 203.Fig. 4 has schematically shown the block diagram of whole codec.

Particularly; In Fig. 4 A; According to one embodiment of present invention, at coding side, in module 101; One of orthogonal transform of utilization such as discrete cosine transform (DCT), improved discrete cosine transform (MDCT) etc. transforms from the time domain to frequency domain with the primary signal of input, obtains spectral coefficient and control information; Next, in module 102, spectral coefficient is divided into groups and Bit Allocation in Discrete, obtain bit distribution information; Then, in module 103,, spectral coefficient is carried out grid vector quantize based on bit distribution information; At last, control information, bit distribution information and quantization index are packaged into coded bit stream in module 104, input channel or storage.

Among Fig. 4 B, according to one embodiment of present invention,, in module 204, the coded bit stream that receives is decoded, and binding modules 201 carries out Bit Allocation in Discrete and quantization index decoding in decoding end; In module 202, carry out the quantization vector that contrary grid vector quantification obtains rebuilding according to the quantization index of decoding in the module 201; At last, in module 203, the quantization vector of reconstruction carries out inverse orthogonal transformation under the control of control information, obtain reconstruction signal.

In Fig. 5, show in detail an implementation procedure more specifically of the present invention.Suppose that a frame signal is N through the spectral coefficient number that obtains after handling such as one of linear orthogonal transformations such as DFT, discrete cosine transform, an above-mentioned N spectral coefficient is divided into L 8 dimension sub-vectors (that is, satisfies 8 * L=N); Suppose quantizing bit number that each 8 dimension sub-vector distributes be R (p) bit/dimension (p=0,1 ...; L-1); The total quantization encoding bit number of one frame signal is ψ, is Ω through remaining bits number behind the bit distribution algorithm, and the coding side quantization step is following:

In step 301, according to total coding code check and selected quantization bit allocation algorithm confirm quantizing bit number that each 8 dimension sub-vector distributes be R (p) bit/dimension (p=0,1 ..., L-1), R (p) should satisfy 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 8 n dimensional vector n y arbitrarily _p=(y _{P, 1}y _{P, 2}Y _{P, 8}),

P=0, L-1 deducts a certain offset vector a=(2 with it ^-62 ^-62 ^-6), the vector after obtaining squinting

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

In step 303, the offset vector that above-mentioned steps is drawn

Carry out convergent-divergent and obtain scaled vectors

Zoom factor β (p)=2 wherein ^{R (p)}/ 6, then

In step 304; Search closes on lattice point V most with scaled vectors

in the D8 grid space, and it satisfies:

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

In step 305, calculate D8 sound of laughing some V and block coordinate k=(k in the Voronoi of R (p) bit/dimension expansion ₁k ₂K ₈), 0≤k wherein _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

is the contrary generator matrix of D8 lattice, sees formula (5).

In step 306, according to giving position fixing k=(k ₁k ₂K ₈) calculating x=kG _D8And z=r ^-1X (G wherein _D8Be the generator matrix of D8 lattice, see formula 4), and search closes on lattice point λ most with scaled vectors z in the D8 grid space, calculates D8 grid vector c then:

c=x-rλ （10）

In step 307, comparative case vector C and V, if c is consistent with V, coordinate k=(k then ₁k ₂K ₈) be exactly scaled vectors

Best coordinates, quantize to finish.If c and V are inconsistent; Vector

is a point not in the know so, and what need approach one by one this moment quantizes.

Next, the method for utilization above-mentioned steps 303～306 is approached one by one.In step 308; With vector

convergent-divergent 2; Promptly

is in step 309,310; Search closes on lattice point u most with scaled vectors

in the D8 grid space; 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 ', the comparative case vector C ' and u, if c ' is unequal with u, then repeating step 308 equates until c ' and u to step 312.

In step 313, calculate wherein m=3,4,5 or 6 of scaled vectors

; In step 314, calculate

In step 315; Search closes on lattice point u ' most with scaled vectors

in the D8 grid space; 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, utilize coordinate j ' to calculate D8 grid vector c ", comparative case vector C " and u ' " equate then k=j ' with u ', and repeating step 314 is to step 316 if c." and u ' is unequal, then stops circulation for if c.

In step 318, the quantizing bit number that each 8 dimension sub-vector is distributed is that (p=0,1.... is L-1) with D8 lattice coordinate k for R (p) bit/dimension _pPass to decoding end behind the coding.

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

In step 401,402, from the code stream of coding decoding obtain the quantizing bit number that each 8 dimension sub-vector distributes 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}) calculating x=kG _D8, and search closes on lattice point λ most with scaled vectors z in the D8 grid space, calculates the D8 grid vector then

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

Then, in step 404, to vector Carrying out contrary convergent-divergent obtains

Zoom factor is β (p)=2 ^{R (p)}/ 6:

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

At last, in step 405, with vector

Add offset vector a=(2 ^-62 ^-62 ^-6), obtain rebuilding vector

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

Need to prove that the present invention is not limited to spectral coefficient is quantized, also be applicable to the quantification of LPC coefficient in the speech coding.In addition, the present invention can be applied to various variable rate coding systems and graduated encoding system, has extensive applicability.

Claims (10)

1. the coding method that source signal variable Rate grid vector quantizes is characterized in that, comprising:

S1 transforms from the time domain to frequency domain to obtain spectral coefficient and control information with the input source signal;

S2, to said spectral coefficient divide into groups with Bit Allocation in Discrete to obtain bit distribution information;

S3, based on said bit distribution information, grid vector quantizes said spectral coefficient;

S4 is packaged into coded bit stream with quantization index, said bit distribution information, said control information.

2. the coding method that source signal variable Rate grid vector according to claim 1 quantizes is characterized in that said step S3 further comprises:

S31 for said spectral coefficient, calculates offset vector;

S32 carries out convergent-divergent to said offset vector, obtains scaled vectors;

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

S34 calculates the said lattice point coordinate that closes on most;

S35 utilizes said coordinate Calculation D8 grid vector;

S36, whether more said D8 grid vector is consistent with the said lattice point that closes on most, if consistent, then quantizes to finish, and exports said coordinate; If inconsistent, then said scaled vectors carried out and approached quantification.

3. the coding method that source signal variable Rate grid vector according to claim 2 quantizes is characterized in that the quantification that approaches among the said step S36 further comprises:

S361 with said scaled vectors convergent-divergent once more, obtains scaled vectors once more, and utilization step S33-S35 calculates lattice point coordinate and the 2nd D8 grid vector that second lattice point, said second that close on most closes on most;

Whether S362, the lattice point that more said the 2nd D8 grid vector and said second closes on most equate, if unequal, repeating step S361 then, the lattice point that closes on most until said the 2nd D8 grid vector and said second equates.

4. the coding method that source signal variable Rate grid vector according to claim 3 quantizes is characterized in that the quantification that approaches among the said step S36 further comprises:

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

S364, the lattice point that more said the 3rd D8 grid vector and the said the 3rd closes on most if both are unequal, then quantize to finish, and exports the said the 3rd lattice point coordinate and the quantizing bit number that close on most; If both equate that then repeating step S363 is unequal until both, export the said the 3rd lattice point coordinate and the quantizing bit number that close on most at last.

5. the coding method that quantizes according to the described source signal variable Rate of arbitrary claim grid vector among the claim 1-4 is characterized in that, in said step S31, said offset vector satisfies:

${\stackrel{\‾}{y}}_p=y_p-a,$

Wherein,

The expression offset vector, y _pThe sub-vector of representing said spectral coefficient, a=(2 ^-62 ^-62 ^-6).

6. the coding method that source signal variable Rate grid vector according to claim 5 quantizes is characterized in that, in said step S32, said scaled vectors satisfies:

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

Wherein,

Represent said scaled vectors, β (p)=2 ^{R (p)}/ 6 expression zoom factors, R (p) representes the quantizing bit number that the sub-vector of each said spectral coefficient distributes.

7. the coding method that source signal variable Rate grid vector according to claim 6 quantizes is characterized in that, R (p) satisfies:

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

Wherein, the number of said spectral coefficient is N, and a said N spectral coefficient is divided into L 8 dimension sub-vectors, and ψ representes the quantization encoding bit number that a frame source signal is total, and Ω representes that the frame source signal is Ω through remaining bits number behind the bit distribution algorithm.

8. the coded system that source signal variable Rate grid vector quantizes is characterized in that, comprising:

The orthogonal transform module is used for the input source signal is transformed from the time domain to frequency domain to obtain spectral coefficient and control information;

Spectral coefficient divide into groups with the Bit Allocation in Discrete module, be used for to said spectral coefficient divide into groups and Bit Allocation in Discrete with the acquisition bit distribution information;

The grid vector quantization modules is used for based on said bit distribution information, and grid vector quantizes said spectral coefficient;

The coded-bit flow module is used for quantization index, said bit distribution information, said control information are packaged into coded bit stream.

9. the coding/decoding method that source signal variable Rate grid vector quantizes is characterized in that, comprising:

S1, received code bit stream decode to obtain decoding bit stream;

S2 carries out Bit Allocation in Discrete and quantization index decoding to said decoding bit stream;

S3 carries out contrary grid vector based on the quantization index of decoding and quantizes to obtain rebuilding quantization vector;

S4 carries out inverse orthogonal transformation based on said control information to said reconstruction quantization vector and obtains reconstruction signal.

10. the decoder module that source signal variable Rate grid vector quantizes is characterized in that, comprising:

The coded bit stream decoder module is used for the received code bit stream and decodes to obtain decoding bit stream;

Bit Allocation in Discrete and quantization index decoder module are used for said decoding bit stream is carried out Bit Allocation in Discrete and quantization index decoding;

Contrary grid vector quantization modules is used for carrying out contrary grid vector based on the quantization index of decoding and quantizes to obtain rebuilding quantization vector;

The inverse orthogonal transformation module is used for based on said control information said reconstruction quantization vector being carried out inverse orthogonal transformation and obtains reconstruction signal.

Images