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

Codec method and system for source signal variable rate lattice vector quantization

technical field

The present invention relates to the field of source signal coding, more specifically, to a codec method and system for source signal variable-rate lattice vector quantization.

Background technique

The coding of existing digital source signals usually adopts transform coding, which divides the signal to be coded into sampling blocks of frames, and uses linear orthogonal transforms such as discrete Fourier transform and discrete cosine transform to process each frame signal, Calculate the transformation coefficient, and then quantize the variation coefficient to further improve the compression effect.

One method commonly used among quantization methods is the vector quantization method, in which several sampling coefficients are grouped together to form a vector, and each vector is approximated (quantized) with a codebook entry. The codebook entry chosen for quantizing the input vector is usually the nearest neighbor in the codebook according to the "minimum distance" criterion. Adding more codebooks in the codebook set increases the bit rate and complexity, but reduces the average distortion of quantization.

On the other hand, to adapt to the changing characteristics of the source, adaptive bit allocation is usually used. With adaptive bit allocation, different codebook sizes can be used to quantize the source vectors. In transform coding, the number of bits allocated to a source vector usually depends on the energy of that vector relative to other vectors in the same frame, without exceeding the maximum number of bits available to quantize all coefficients. Figure 1 and Figure 2 describe in detail the quantization block diagrams of common variable-rate quantization encoders and decoders. The variable rate quantization encoder and decoder shown in Figures 1 and 2 use multiple codebooks, usually with different bit rates, to quantize the source vector x. The source vector is usually obtained by transforming the signal and obtaining all or a subset of the transform coefficients.

A common variable-rate quantization encoder is shown in Figure 1, and its key component is a quantizer denoted by Q, which is used to select a codebook number n and a code vector index i to represent the quantized value y of the source vector x . The codebook number n designates the codebook selected by the encoder, while the index i represents the codevector selected in that particular codebook.

Typically, appropriate lossless coding techniques are applied to n and i in blocks _En and E _i respectively (i.e., _En and E _i in Fig. The average bit rate of the encoded codebook number n _E and index i _E is reduced before transmission over the communication channel.

Figure 2 shows a variable rate quantization decoder. The input of the decoder provides a demultiplexer DEMUX for separating the binary codes n _E and i _E ; the decoder also includes lossless decoding modules (i.e., D _n and D _i ), in which n _E and i _E is codebook number n and index i; the decoder also includes an inverse quantizer (denoted by Q ⁻¹ ) that receives codebook number n and index i and performs inverse quantization, which uses codebook number and index to recover the source vector The quantized value y of x. Different values of n usually result in different bit allocations resulting in different bit rates, and the number of bits required per dimension (i.e., the codebook bit rate) is defined as the ratio of the number of bits allocated to the source vector to the dimension of the source vector .

Usually, the construction of the codebook can adopt the following methods:

A popular approach is to employ a training algorithm such as the k-means algorithm to optimize the codebook entries according to the distribution of the sources. This approach results in an unstructured codebook, which typically must be stored and exhaustively searched for each source vector to be quantized. Therefore, the disadvantage of this method is the large memory requirement and the computational complexity, which grows exponentially with the bit rate of the codebook. If the variable-rate method is based on the above-mentioned unstructured codebook, the disadvantage of large memory requirement and computational complexity is further exacerbated, since a specific codebook usually needs to be assigned to each possible bit.

Another approach is to use a lattice vector quantizer, which reduces search complexity and, in many cases, can effectively reduce storage requirements. Lattice vector quantization is an algebraic vector quantizer, which is characterized by constructing a regular network in the multi-dimensional signal space. Divided into cells. Since the network is regular, the lattice points and cells are also regular. The main advantage of the lattice vector quantizer is that it is easy to construct a codebook, and it can perform high-dimensional quantization. Figure 3 shows an example in two-dimensional space, where the basic vectors are v1 and v2, the grid used in this example is a basic hexagonal lattice, represented by _Α2 , all points marked with crosses in this figure can be as follows get:

y=k1v1+k2v2 (1)

Wherein, y is a spatial grid point, and k1 and k2 can be any integers. Note that Figure 3 only represents a subset of the spatial grid points, because the spatial grid points themselves can be infinitely expanded.

When a certain spatial grid point is selected to construct a quantized codebook, a certain subset of the grid points is usually selected to obtain a codebook with a given (limited) number of bits. The advantage of using a grid point is that all When the nearest neighbor of the source vector x of the lattice point, there is a fast codebook search algorithm, and compared with other unstructured codebooks, the complexity can be greatly reduced. Furthermore, using lattice points eliminates the need to store a codebook, since the codebook can be obtained from the generator matrix.

即：The grid often used in lattice vector quantization is the D8 lattice. D8 is composed of Z8 lattice points v=(v ₁ ,…,v ₈ ) of 8-dimensional integer lattice, and satisfies

Right now:

Any 8-dimensional grid point y in the D8 grid can be generated by the following method:

y=[k ₁ k ₂ …k ₈ ]G _D8 (3)

where k ₁ , k ₂ , ..., k ₈ are signed integers, G _D8 is a generator matrix defined as:

$\mathrm{<span\; class="notranslate">\; G</span>}<span\; class="notranslate">\; =</span>\left[\begin{array}{cccccccc}\mathrm{<span\; class="notranslate">\; 2</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}\\ \mathrm{<span\; class="notranslate">\; 1</span>}& \mathrm{<span\; class="notranslate">\; 1</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}\\ \mathrm{<span\; class="notranslate">\; 1</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 1</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}\\ \mathrm{<span\; class="notranslate">\; 1</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 1</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}\\ \mathrm{<span\; class="notranslate">\; 1</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 1</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}\\ \mathrm{<span\; class="notranslate">\; 1</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 1</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}\\ \mathrm{<span\; class="notranslate">\; 1</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 1</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}\\ \mathrm{<span\; class="notranslate">\; 1</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 1</span>}\end{array}\right]<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 4</span>}<span\; class="notranslate">\; )</span>$

为：It is easy to verify that G _D8 is the inverse matrix of the generating matrix

for:

${\mathrm{<span\; class="notranslate">\; G</span>}}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; =</span>\left[\begin{array}{cccccccc}\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; 2</span>}}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}\\ <span\; class="notranslate">\; -</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; 2</span>}}& \mathrm{<span\; class="notranslate">\; 1</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}\\ <span\; class="notranslate">\; -</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; 2</span>}}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 1</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}\\ <span\; class="notranslate">\; -</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; 2</span>}}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 1</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}\\ <span\; class="notranslate">\; -</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; 2</span>}}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 1</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}\\ <span\; class="notranslate">\; -</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; 2</span>}}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 1</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}\\ <span\; class="notranslate">\; -</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; 2</span>}}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 1</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}\\ <span\; class="notranslate">\; -</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; 2</span>}}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 1</span>}\end{array}\right]<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 5</span>}<span\; class="notranslate">\; )</span>$

This inverse matrix is very useful in obtaining the coordinates of the D8 grid point y.

Contents of the invention

In order to solve the traditional variable-rate vector quantizer, the storage space increases with the increase of the number of codebooks, and the search operation is complicated because the codebook needs to be fully searched to obtain the best quantized vector during the quantization process. High degree of defects, the present invention provides the following technical solutions.

The first technical solution adopted by the present invention to solve the technical problems is to construct a coding method for source signal variable rate lattice vector quantization, including:

S1, transforming the input source signal from the time domain to the frequency domain to obtain spectral coefficients and control information;

S2. Perform grouping and bit allocation on the spectral coefficients to obtain bit allocation information;

S3. Based on the bit allocation information, lattice vector quantize the spectral coefficients;

S4, packing the quantization index, the bit allocation information, and the control information into an encoded bit stream.

In the encoding method of source signal variable rate lattice vector quantization according to the present invention, the step S3 further includes:

S31. For the spectral coefficients, calculate an offset vector;

S32. Scaling the offset vector to obtain a scaling vector;

S33, searching for a grid point closest to the scaling vector in the D8 grid space;

S34, calculating the coordinates of the nearest grid point;

S35, using the coordinates to calculate the D8 grid vector;

S36. Comparing whether the D8 grid vector is consistent with the nearest grid point, if they are consistent, the quantization ends, and the coordinates are output; if they are not consistent, approximate quantization is performed on the scaled vector.

In the encoding method of source signal variable rate lattice vector quantization described in the present invention, the approximation quantization in the step S36 further includes:

S361, re-scale the scaling vector to obtain a re-scaling vector, and use steps S33-S35 to calculate and obtain the second closest grid point, the second closest grid point coordinates, and the second D8 grid vector;

S362, comparing whether the second D8 grid vector is equal to the second nearest grid point, if not, repeat step S361 until the second D8 grid vector is equal to the second nearest grid point equal.

In the encoding method of source signal variable rate lattice vector quantization described in the present invention, the approximation quantization in the step S36 further includes:

S363, using steps S33-S35 to calculate the third D8 grid vector, the third nearest grid point and the third nearest grid point coordinates;

S364, comparing the third D8 grid vector with the third nearest grid point, if the two are not equal, the quantization ends, outputting the third nearest grid point coordinates and the number of quantization bits; if both If they are equal, repeat step S363 until they are not equal, and finally output the coordinates of the third nearest grid point and the number of quantization bits.

In the encoding method of source signal variable rate lattice vector quantization described in the present invention, in the step S31, the offset vector satisfies:

${\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; the\; y</span>}}}_{\mathrm{<span\; class="notranslate">\; p</span>}}<span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; p</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; a</span>}<span\; class="notranslate">\; ,</span>$

表示偏移矢量，y_p表示所述谱系数的子矢量，a=(2^-6 2^-6…2^-6)。in,

represents the offset vector, y _p represents the sub-vector of the spectral coefficient, a=(2 ⁻⁶ 2 ⁻⁶ …2 ⁻⁶ ).

In the encoding method of source signal variable rate lattice vector quantization according to the present invention, in the step S32, the scaling vector satisfies:

${\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; the\; y</span>}}}_{\mathrm{<span\; class="notranslate">\; p</span>}}<span\; class="notranslate">\; =</span>{\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; the\; y</span>}}}_{\mathrm{<span\; class="notranslate">\; p</span>}}<span\; class="notranslate">\; /</span>\mathrm{<span\; class="notranslate">\; \&beta;</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ;</span>$

表示所述缩放矢量，β(p)=2^R(p)/6表示缩放因子，R（p）表示每个所述谱系数的子矢量分配的量化比特数。in,

represents the scaling vector, β(p)=2 ^R(p) /6 represents the scaling factor, and R(p) represents the number of quantized bits allocated to each sub-vector of the spectral coefficient.

In the coding method of source signal variable rate lattice vector quantization described in the present invention, R (p) satisfies:

$\underset{\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; L</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\mathrm{<span\; class="notranslate">\; 8</span>}<span\; class="notranslate">\; \&times;</span>\mathrm{<span\; class="notranslate">\; R</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; \&Omega;</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; \&Psi;</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; 0</span>}<span\; class="notranslate">\; \&le;</span>\mathrm{<span\; class="notranslate">\; R</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \&le;</span>\mathrm{<span\; class="notranslate">\; 10</span>}$

Wherein, the number of spectral coefficients is N, and the N spectral coefficients are divided into L 8-dimensional sub-vectors, ψ represents the total number of quantized coding bits of a frame source signal, and Ω represents the frame source signal after the bit allocation algorithm The number of remaining bits is Ω.

The second technical solution adopted by the present invention to solve its technical problems is to construct a coding system for source signal variable rate lattice vector quantization, including:

Orthogonal transformation module for transforming the input source signal from the time domain to the frequency domain to obtain spectral coefficients and control information;

A spectral coefficient grouping and bit allocation module, configured to group and bit allocate the spectral coefficients to obtain bit allocation information;

A lattice vector quantization module, configured to perform lattice vector quantization on the spectral coefficients based on the bit allocation information;

The coded bit stream module is configured to pack the quantization index, the bit allocation information, and the control information into a coded bit stream.

The third technical solution adopted by the present invention to solve the technical problems is to construct a decoding method for source signal variable rate lattice vector quantization, including:

S1, receiving and decoding the coded bit stream to obtain the decoded bit stream;

S2. Perform bit allocation and quantization index decoding on the decoded bit stream;

S3, performing inverse lattice vector quantization based on the decoded quantization index to obtain a reconstructed quantized vector;

S4. Perform inverse orthogonal transformation on the reconstructed quantization vector based on the control information to obtain a reconstructed signal.

The fourth technical solution adopted by the present invention to solve the technical problems is to construct a decoding module of source signal variable rate lattice vector quantization, including:

An encoded bitstream decoding module, configured to receive the encoded bitstream for decoding to obtain the decoded bitstream;

A bit allocation and quantization index decoding module, configured to perform bit allocation and quantization index decoding on the decoded bit stream;

An inverse lattice vector quantization module, configured to perform inverse lattice vector quantization based on the decoded quantization index to obtain a reconstructed quantized vector;

An inverse orthogonal transform module, configured to perform inverse orthogonal transform on the reconstructed quantized vector based on the control information to obtain a reconstructed signal.

Compared with the traditional variable rate vector quantizer storing multiple vector codebooks, the method of the present invention does not need to store vector codebooks; in addition, there is a fast algorithm, and its computational complexity is greatly reduced compared with traditional vector quantization, that is, it has low computational complexity Third, it also has the advantage of being able to realize variable ratio rate quantization.

It should be appreciated by those skilled in the art that the foregoing summary is merely intended to provide a brief description of certain aspects of the invention. A more complete understanding of the present invention can be obtained by referring to the claims and the following detailed description of the preferred embodiments taken in conjunction with the accompanying drawings.

Description of drawings

The present invention will be further described below in conjunction with accompanying drawing and embodiment, in the accompanying drawing:

Fig. 1 is the quantization block diagram of traditional variable rate quantization coder;

Fig. 2 is the quantization block diagram of traditional variable rate quantization decoder;

Fig. 3 is a subset representing a certain spatial lattice point;

Figure 4A shows a block diagram of an encoder according to one embodiment of the invention;

Figure 4B shows a block diagram of a decoder according to one embodiment of the present invention;

Figure 5 shows a flowchart of a quantization method according to one embodiment of the present invention; and

Fig. 6 shows a flowchart of an inverse quantization method according to an embodiment of the present invention.

Detailed ways

The main purpose of the present invention is to provide a kind of source signal variable rate quantization technology (for example, including quantization method and quantization system etc.) based on space D8 grid point, this quantization technology can realize variable rate quantization compared with prior art , and requires no storage space and low computational complexity. The quantization technique can be applied to various variable rate coding systems and scalable coding systems. In order to simplify the description without causing confusion, the embodiments of the present invention may omit content known to those skilled in the art, such as bit allocation algorithm, spectral coefficient grouping algorithm, index encoding and decoding algorithm of lattice vector quantization, and the like.

In a preferred embodiment according to the present invention, the encoder of source signal variable rate lattice vector quantization mainly includes: orthogonal transformation module 101, spectral coefficient grouping and bit allocation module 102 and lattice vector quantization module 103 and encoding bit stream module 104 . The source signal variable rate lattice vector quantization decoder is the inverse system of the source signal variable rate lattice vector quantization encoder, which mainly includes a coded bit stream decoding module 204, a bit allocation and quantization index decoding module 201, and an inverse lattice vector quantization module 202 and an orthogonal transformation module 203 . Figure 4 schematically shows a block diagram of the entire codec.

Specifically, in FIG. 4A, according to an embodiment of the present invention, at the encoding end, in module 101, one of the orthogonal transforms such as discrete cosine transform (DCT), modified discrete cosine transform (MDCT), etc. is utilized The input original signal is transformed from the time domain to the frequency domain to obtain spectral coefficients and control information; next, in module 102, the spectral coefficients are grouped and bit allocated to obtain bit allocation information; then, in module 103, based on The bit allocation information is used to perform lattice vector quantization on the spectral coefficients; finally, the control information, bit allocation information and quantization index are packaged into a coded bit stream in module 104, which is input into a channel or stored.

In Fig. 4B, according to an embodiment of the present invention, at the decoding end, in module 204, the received encoded bit stream is decoded, and combined with module 201 to perform bit allocation and quantization index decoding; in module 202, according to module The quantized index decoded in 201 is subjected to inverse lattice vector quantization to obtain a reconstructed quantized vector; finally, in module 203, the reconstructed quantized vector is subjected to inverse orthogonal transformation under the control of the control information to obtain a reconstructed signal.

In Fig. 5, a more specific implementation process of the present invention is shown in detail. Assuming that the number of spectral coefficients obtained after a frame of signal is processed by one of linear orthogonal transforms such as discrete Fourier transform and discrete cosine transform is N, the above N spectral coefficients are divided into L 8-dimensional sub-vectors (that is, satisfying 8× L=N), assuming that the number of quantized bits allocated to each 8-dimensional sub-vector is R(p) bits/dimension (p=0,1,...,L-1), the total number of quantized coding bits of a frame signal is ψ, the number of remaining bits after the bit allocation algorithm is Ω, and the quantization steps at the encoding end are as follows:

In step 301, according to the total coding rate and the selected quantization bit allocation algorithm, the number of quantization bits allocated to each 8-dimensional sub-vector is determined as R(p) bits/dimension (p=0,1,...,L -1), R(p) should meet the following restrictions:

$\underset{\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; L</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\mathrm{<span\; class="notranslate">\; 8</span>}<span\; class="notranslate">\; \&times;</span>\mathrm{<span\; class="notranslate">\; R</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; \&Omega;</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; \&Psi;</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; 0</span>}<span\; class="notranslate">\; \&le;</span>\mathrm{<span\; class="notranslate">\; R</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \&le;</span>\mathrm{<span\; class="notranslate">\; 10</span>}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 6</span>}<span\; class="notranslate">\; )</span>$

p=0，···L-1，将其减去某一偏移矢量a=(2^-6 2^-6…2^-6)，得到偏移后的矢量

Next, in step 302, for an arbitrary 8-dimensional vector y _p =(y _{p, 1} y _{p, 2} ... y _{p, 8} ),

p=0,...L-1, subtract a certain offset vector a=(2 ^-6 2 ^-6 …2 ^-6 ) from it to get the offset vector

${\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; the\; y</span>}}}_{\mathrm{<span\; class="notranslate">\; p</span>}}<span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; p</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; a</span>}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 7</span>}<span\; class="notranslate">\; )</span>$

进行缩放得到缩放矢量

其中缩放因子β(p)=2^R(p)/6，则

In step 303, for the offset vector obtained in the above steps

Scale to get the scaled vector

where scaling factor β(p)=2 ^R(p) /6, then

最临近格点V，其满足：In step 304, search and scale vector in D8 grid space

The nearest grid point V, which satisfies:

$<span\; class="notranslate">\; \&ForAll;</span>\mathrm{<span\; class="notranslate">\; z</span>}<span\; class="notranslate">\; \&Element;</span>\mathrm{<span\; class="notranslate">\; D.</span>}\mathrm{<span\; class="notranslate">\; 8</span>}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; -</span>{\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; the\; y</span>}}}_{\mathrm{<span\; class="notranslate">\; p</span>}}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; |</span><span\; class="notranslate">\; \&le;</span><span\; class="notranslate">\; |</span><span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; z</span>}<span\; class="notranslate">\; -</span>{\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; the\; y</span>}}}_{\mathrm{<span\; class="notranslate">\; p</span>}}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; |</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 8</span>}<span\; class="notranslate">\; )</span>$

In step 305, calculate the Voronoi extended truncated coordinate k=(k ₁ k ₂ ...k ₈ ) of the D8 grid point V in R(p) bits/dimension, where 0≤k _i ≤2 ^R(p) -1, i =1, 2,...,8, the calculation of k is:

$\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; =</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; v</span>}{\mathrm{<span\; class="notranslate">\; G</span>}}_{\mathrm{<span\; class="notranslate">\; D.</span>}\mathrm{<span\; class="notranslate">\; 8</span>}}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \&CirclePlus;</span>\mathrm{<span\; class="notranslate">\; r</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; r</span>}<span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; 2</span>}}^{\mathrm{<span\; class="notranslate">\; R</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 9</span>}<span\; class="notranslate">\; )</span>$

是D8格的逆生成矩阵，见式(5)。in

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

In step 306, x=kG _D8 and z=r ^-1 x are calculated according to the given coordinates k=(k ₁ k ₂ ...k ₈ ) (where G _D8 is the generator matrix of D8 lattice, see formula 4), and in In the D8 grid space, search and scale the vector z closest to the grid point λ, and then calculate the D8 grid vector c:

c=x-rλ （10）

的最佳坐标，量化结束。如果c和V不一致，那么矢量

为局外点，此时需要逐次逼近的来量化。In step 307, compare the grid vector c and V, if c and V are consistent, then the coordinate k=(k ₁ k ₂ ...k ₈ ) is the scaling vector

The best coordinates, quantization ends. If c and V are inconsistent, then the vector

is an outlier point, which needs to be quantified by successive approximation.

缩放2，即

在步骤309、310中，在D8格空间中搜索与缩放矢量

最临近格点u，计算u的坐标j，即：Next, use the methods in steps 303-306 above to perform successive approximation. In step 308, the vector

Zoom 2, i.e.

In steps 309, 310, search and scale the vector in the D8 grid space

The nearest grid point u, calculate the coordinate j of u, that is:

$\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; =</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; u</span>}{\mathrm{<span\; class="notranslate">\; G</span>}}_{\mathrm{<span\; class="notranslate">\; D.</span>}\mathrm{<span\; class="notranslate">\; 8</span>}}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \&CirclePlus;</span>\mathrm{<span\; class="notranslate">\; r</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; r</span>}<span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; 2</span>}}^{\mathrm{<span\; class="notranslate">\; R</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 11</span>}<span\; class="notranslate">\; )</span>$

In steps 311 and 312, the D8 grid vector c' is calculated using the coordinate j, and the grid vector c' and u are compared. If c' and u are not equal, then steps 308 to 312 are repeated until c' and u are equal.

其中m=3、4、5或6；在步骤314中，计算

In step 313, the scaling vector is calculated

Where m=3, 4, 5 or 6; in step 314, calculate

最临近格点u′，计算u′的坐标j′，即：In step 315, search and scale vector in D8 grid space

The nearest grid point u', calculate the coordinate j' of u', that is:

${\mathrm{<span\; class="notranslate">\; j</span>}}^{<span\; class="notranslate">\; \&prime;</span>}<span\; class="notranslate">\; =</span><span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; u</span>}}^{<span\; class="notranslate">\; \&prime;</span>}{\mathrm{<span\; class="notranslate">\; G</span>}}_{\mathrm{<span\; class="notranslate">\; D.</span>}\mathrm{<span\; class="notranslate">\; 8</span>}}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \&CirclePlus;</span>\mathrm{<span\; class="notranslate">\; r</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; r</span>}<span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; 2</span>}}^{\mathrm{<span\; class="notranslate">\; R</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 12</span>}<span\; class="notranslate">\; )</span>$

In step 316, utilize coordinate j ' to calculate and obtain D8 grid vector c ", compare grid vector c " and u ', if c " and u ' are equal then k=j ', and repeat step 314 to step 316. If c " and u' are not equal, then stop the loop.

In step 318, the number of quantized bits allocated to each 8-dimensional sub-vector is R(p) bits/dimension (p=0,1....,L-1) and the D8 grid coordinate k _p is encoded and passed to decoder side.

Compared with the above quantization method at the encoding end, Fig. 6 shows a flow chart of inverse quantization at the decoding end, and the specific implementation steps are as follows:

In steps 401 and 402, the number of quantized bits allocated to each 8-dimensional sub-vector is R(p) bits/dimension (p=0,1,...,L-1) and D8 grid coordinate k _p .

Next, in step 403, x=kG _D8 is calculated for the given coordinates k _p =(k _p,1 k _p,2 ... k _p,8 ), and the nearest neighbor grid to the scaling vector z is searched in D8 grid space Point λ, then calculate the D8 grid vector

${\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; the\; y</span>}}}_{\mathrm{<span\; class="notranslate">\; p</span>}}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; r\&lambda;</span>}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 13</span>}<span\; class="notranslate">\; )</span>$

缩放因子为β(p)=2^R(p)/6：Then, in step 404, the vector Perform inverse scaling to get

The scaling factor is β(p)=2 ^R(p) /6:

${\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; the\; y</span>}}}_{\mathrm{<span\; class="notranslate">\; p</span>}}<span\; class="notranslate">\; =</span>{\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; the\; y</span>}}}_{\mathrm{<span\; class="notranslate">\; p</span>}}\mathrm{<span\; class="notranslate">\; \&beta;</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 14</span>}<span\; class="notranslate">\; )</span>$

加上偏移矢量a=(2^-6 2^-6…2^-6)，得到重建矢量

Finally, in step 405, the vector

Add the offset vector a=(2 ^-6 2 ^-6 …2 ^-6 ) to get the reconstruction vector

${\stackrel{<span\; class="notranslate">\; ~</span>}{\mathrm{<span\; class="notranslate">\; the\; y</span>}}}_{\mathrm{<span\; class="notranslate">\; p</span>}}<span\; class="notranslate">\; =</span>{\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; the\; y</span>}}}_{\mathrm{<span\; class="notranslate">\; p</span>}}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; a</span>}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 15</span>}<span\; class="notranslate">\; )</span>$

It should be noted that the present invention is not limited to the quantization of spectral coefficients, and is also applicable to the quantization of LPC coefficients in speech coding. In addition, the present invention can be applied to various variable-rate coding systems and scalable coding systems, and has wide applicability.

Claims (10)

1. A coding method for source signal variable rate lattice vector quantization, characterized in that, comprising: S1, transforming the input source signal from the time domain to the frequency domain to obtain spectral coefficients and control information; S2. Perform grouping and bit allocation on the spectral coefficients to obtain bit allocation information; S3. Based on the bit allocation information, lattice vector quantize the spectral coefficients; S4. Pack the quantization index, the bit allocation information, and the control information into an encoded bit stream. 2. The encoding method of source signal variable rate lattice vector quantization according to claim 1, wherein said step S3 further comprises: S31. For the spectral coefficients, calculate an offset vector; S32. Scaling the offset vector to obtain a scaling vector; S33, searching for a grid point closest to the scaling vector in the D8 grid space; S34, calculating the coordinates of the nearest grid point; S35, using the coordinates to calculate the D8 grid vector; S36. Comparing whether the D8 grid vector is consistent with the nearest grid point, if they are consistent, the quantization ends, and the coordinates are output; if they are not consistent, approximate quantization is performed on the scaled vector. 3. The encoding method of source signal variable rate lattice vector quantization according to claim 2, characterized in that, the approximation quantization in the step S36 further comprises: S361, re-scale the scaling vector to obtain a re-scaling vector, and use steps S33-S35 to calculate and obtain the second closest grid point, the second closest grid point coordinates, and the second D8 grid vector; S362, comparing whether the second D8 grid vector is equal to the second nearest grid point, if not, repeat step S361 until the second D8 grid vector is equal to the second nearest grid point equal. 4. The encoding method of source signal variable rate lattice vector quantization according to claim 3, characterized in that, the approximation quantization in the step S36 further comprises: S363, using steps S33-S35 to calculate the third D8 grid vector, the third nearest grid point and the third nearest grid point coordinates; S364, comparing the third D8 grid vector with the third nearest grid point, if the two are not equal, the quantization ends, outputting the third nearest grid point coordinates and the number of quantization bits; if both If they are equal, repeat step S363 until they are not equal, and finally output the coordinates of the third nearest grid point and the number of quantization bits. 5. The encoding method for source signal variable rate lattice vector quantization according to any one of claims 1-4, characterized in that, in the step S31, the offset vector satisfies: ${\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; the\; y</span>}}}_{\mathrm{<span\; class="notranslate">\; p</span>}}<span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; p</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; a</span>}<span\; class="notranslate">\; ,</span>$ in,

represents the offset vector, y _p represents the sub-vector of the spectral coefficient, a=(2 ⁻⁶ 2 ⁻⁶ …2 ⁻⁶ ). 6. The encoding method of source signal variable rate lattice vector quantization according to claim 5, characterized in that, in the step S32, the scaling vector satisfies: ${\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; the\; y</span>}}}_{\mathrm{<span\; class="notranslate">\; p</span>}}<span\; class="notranslate">\; =</span>{\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; the\; y</span>}}}_{\mathrm{<span\; class="notranslate">\; p</span>}}<span\; class="notranslate">\; /</span>\mathrm{<span\; class="notranslate">\; \&beta;</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ;</span>$ in,

represents the scaling vector, β(p)=2 ^R(p) /6 represents the scaling factor, and R(p) represents the number of quantized bits allocated to each sub-vector of the spectral coefficient. 7. the encoding method of source signal variable rate lattice vector quantization according to claim 6, is characterized in that, R (p) satisfies: $\underset{\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; L</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\mathrm{<span\; class="notranslate">\; 8</span>}<span\; class="notranslate">\; \&times;</span>\mathrm{<span\; class="notranslate">\; R</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; \&Omega;</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; \&Psi;</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; 0</span>}<span\; class="notranslate">\; \&le;</span>\mathrm{<span\; class="notranslate">\; R</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \&le;</span>\mathrm{<span\; class="notranslate">\; 10</span>}$ Wherein, the number of spectral coefficients is N, and the N spectral coefficients are divided into L 8-dimensional sub-vectors, ψ represents the total number of quantized coding bits of a frame source signal, and Ω represents the frame source signal after the bit allocation algorithm The number of remaining bits is Ω. 8. A coding system for source signal variable rate lattice vector quantization, characterized in that it comprises: Orthogonal transformation module for transforming the input source signal from the time domain to the frequency domain to obtain spectral coefficients and control information; A spectral coefficient grouping and bit allocation module, configured to group and bit allocate the spectral coefficients to obtain bit allocation information; A lattice vector quantization module, configured to perform lattice vector quantization on the spectral coefficients based on the bit allocation information; The coded bit stream module is configured to pack the quantization index, the bit allocation information, and the control information into a coded bit stream. 9. A decoding method for source signal variable rate lattice vector quantization, characterized in that it comprises: S1, receiving and decoding the coded bit stream to obtain the decoded bit stream; S2. Perform bit allocation and quantization index decoding on the decoded bit stream; S3, performing inverse lattice vector quantization based on the decoded quantization index to obtain a reconstructed quantized vector; S4. Perform inverse orthogonal transformation on the reconstructed quantization vector based on the control information to obtain a reconstructed signal. 10. A decoding module for source signal variable rate lattice vector quantization, characterized in that it comprises: An encoded bitstream decoding module, configured to receive the encoded bitstream for decoding to obtain the decoded bitstream; A bit allocation and quantization index decoding module, configured to perform bit allocation and quantization index decoding on the decoded bit stream; An inverse lattice vector quantization module, configured to perform inverse lattice vector quantization based on the decoded quantization index to obtain a reconstructed quantized vector; An inverse orthogonal transform module, configured to perform inverse orthogonal transform on the reconstructed quantized vector based on the control information to obtain a reconstructed signal.

Images