Vector quantization device, vector inverse-quantization device, and methods of same
An LSP vector quantization device able to improve the precision of quantization in vector quantization where a codebook for first stage vector quantization is switched according to the type of a feature that has a correlation with the quantization target vector. In this device, a classifier ( 101 ) selects, from a plurality of code vectors for classifying, the code vector for classifying that indicates the type of the feature that has a correlation with the quantization target vector; a switch ( 102 ) selects, from among a plurality of first codebooks, the first codebook corresponding to the abovementioned type; an error minimizing unit ( 105 ) selects, from the plurality of first code vectors constituting the selected first codebook, the first code vector closest to the quantization target vector; a matrix determining unit ( 106 ) selects, from a plurality of matrices, the matrix corresponding to the abovementioned type; and the error minimizing unit ( 105 ) uses the selected matrix to select, from a plurality of second code vectors, the one that is closest to the residual vector between the selected first code vector and the quantization target vector.
1. A vector quantization apparatus comprising:
a first selector that selects a classification code vector indicating a type of characteristics having a correlation with a vector to be quantized from among a plurality of classification code vectors;
a second selector that selects a first codebook corresponding to the selected classification code vector from among a plurality of first codebooks;
a first quantizer, including a processor, that quantizes the vector to be quantized using a plurality of first code vectors making up the selected first codebook, to obtain a first code;
a third selector that selects a first rotation matrix corresponding to the selected classification code vector from among a plurality of first rotation matrices; and
a second quantizer that quantizes a first residual vector which is a difference between the first code vector indicated by the first code and the vector to be quantized using a plurality of second code vectors and the selected first rotation matrix, to obtain a second code.
2. The vector quantization apparatus according to claim 1 , wherein the second quantizer performs matrix calculation using the selected first rotation matrix on the plurality of second code vectors and quantizes the first residual vector using the plurality of second code vectors after the matrix calculation.
3. The vector quantization apparatus according to claim 1 , wherein the second quantizer performs matrix calculation using an inverse matrix of the selected first rotation matrix on the first residual vector and quantizes the first residual vector after the matrix calculation using the plurality of second code vectors.
4. The vector quantization apparatus according to claim 1 , wherein:
the third selector further selects a second rotation matrix corresponding to the selected classification code vector from among the plurality of second rotation matrices; and
the vector quantization apparatus further comprises a third quantizer that uses a plurality of third code vectors and the selected second rotation matrix, quantizes a second residual vector which is a difference between the first residual vector indicated by the second code and the second code vector and obtains a third code.
5. A vector dequantization apparatus comprising:
a receiver that receives a first code obtained by a vector quantization apparatus quantizing a vector to be quantized and a second code obtained by further quantizing a quantization error of the quantization;
a first selector that selects a classification code vector indicating a type of characteristics having a correlation with the vector to be quantized from among a plurality of classification code vectors;
a second selector that selects a first codebook corresponding to the selected classification code vector from among a plurality of first codebooks;
a first dequantizer, including a processor, that designates a first code vector corresponding to the first code from among a plurality of first code vectors making up the selected first codebook;
a third selector that selects a rotation matrix corresponding to the selected classification code vector from among a plurality of rotation matrices; and
a second dequantizer that designates a second code vector corresponding to the second code from among a plurality of second code vectors and obtains a quantization vector using the designated second code vector, the selected rotation matrix and the designated first code vector.
6. A vector quantization method performed by a processor, the vector quantization method comprising:
selecting a classification code vector indicating a type of characteristics having a correlation with a vector to be quantized from among a plurality of classification code vectors;
selecting a first codebook corresponding to the selected classification code vector from among a plurality of first codebooks;
quantizing the vector to be quantized using a plurality of st code vectors making up the selected first codebook, to obtain a first code;
selecting a rotation matrix corresponding to the selected classification code vector from among a plurality of rotation matrices; and
quantizing a first residual vector which is a difference between the first code vector indicated by the first code and the vector to be quantized using a plurality of second code vectors and the selected rotation matrix, to obtain a second code.
7. A vector dequantization method performed by a processor, the vector dequantization method comprising:
receiving a first code obtained by a vector quantization apparatus quantizing a vector to he quantized and a second code obtained by further quantizing a quantization error of the quantization;
selecting a classification code vector indicating a type of characteristics having a correlation with the vector to be quantized from among a plurality of classification code vectors;
selecting a first codebook corresponding to the selected classification code vector from among a plurality of first codebooks;
designating a first code vector corresponding to the first code from among a plurality of first code vectors making up the selected first codebook;
selecting a rotation matrix corresponding to the selected classification code vector from among a plurality of rotation matrices; and
designating a second code vector corresponding to the second code from among a plurality of second code vectors and obtaining a quantization vector using the designated second code vector, the selected rotation matrix and the designated first code vector.