Looking for breakthrough ideas for innovation challenges? Try Patsnap Eureka!

A vector quantification method and vector quantifier

A vector quantization and vector technology, applied in the field of vector quantization methods and vector quantizers, can solve the problems of large codebook capacity, multiple quantization bits, and cost

Inactive Publication Date: 2009-08-19
HUAWEI TECH CO LTD
View PDF4 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0021] In summary, it can be seen that the 16-dimensional ISF vector quantization of the speech signal using the above quantization method requires a total of 42 bits, including 3 bits for the classification index and 39 bits for the quantization index, which requires more quantization bits, and the codebook capacity is correspondingly also larger

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • A vector quantification method and vector quantifier
  • A vector quantification method and vector quantifier
  • A vector quantification method and vector quantifier

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0077] Embodiment 1, a vector quantization method, the process is as follows figure 1 shown, including steps:

[0078] A1. Classify the vector to be quantized according to the distribution of jump values ​​between its dimension coefficients, obtain the trajectory vector corresponding to the classification and output the classification index;

[0079] The jump value between the vector dimension coefficients can reflect the change trend of the vector, and the distribution of the jump value shows the shape change of the vector. Generally speaking, the extreme value in the jump value has the most obvious influence on the shape, so in this embodiment, the maximum value and the second maximum value in the jump value are selected as the research object of the distribution, and the quantized vector is classified according to the position where it appears. The specific method is:

[0080] A11, calculate the jump value;

[0081] If the vector to be quantized is an n-dimensional vecto...

Embodiment 2

[0099] Embodiment 2, a vector quantization method, the process is as follows image 3 As shown, the main improvement of this embodiment compared with Embodiment 1 is that only the part of the jump value that plays a major role is considered in the classification, and the split quantization method is used when the difference vector is quantized; it includes steps:

[0100] B1. Calculate the vector to be quantized, that is, the 16-dimensional ISF vector f t =[f 0 , f 1 ,..., f 15 ], the jump value G between the coefficients of each dimension i :

[0101] G i = f i+1 -f i , i=0,1,...,14

[0102] There are 15 jump values ​​in total; the experimental analysis of the ISF coefficients in the frame found that there is a certain correlation between the ISF coefficients. 14 The value of is a negative number, in order to reduce the computational complexity, and it is proved by experiments that G 3 ~G 13 A total of 11 skip values ​​are good for classification, so only G is used...

Embodiment 3

[0116] Embodiment three, a vector quantization method, this embodiment is basically the same as embodiment two, the difference is:

[0117] 1. Use G when classifying 1 ~G 14 A total of 14 jump values ​​are used to classify ISF vectors; G 1 ~G 14 The maximum value M in max and the next largest value M max-1 There are 14×13=182 permutations and combinations in total, so the corresponding trajectory vectors are divided into 182 types, and the classification index needs to occupy 8 quantization bits;

[0118] 2. When quantizing the difference vector, the dimensions of the three split sub-vectors are: VQ1 is 5 dimensions, VQ2 is 5 dimensions, and VQ3 is 6 dimensions; each sub-vector is still quantized with 10 bits, and the entire quantization scheme A total of 38 bits are required.

[0119] Compared with the second embodiment, the classification in this embodiment is more detailed, and the accuracy of quantification is improved.

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

The invention discloses a vector quantization method. The core idea of the method is that classifying the vectors to be quantized in accordance with the jump value between different dimension coefficients of the vectors to be quantized, and quantizing the vectors to be quantized and the residual deviation of the classified fixed track vectors. The invention also provides a corresponding vector quantizer. The classification mode adopted in the invention can perform preliminary determination of vector dimension coefficient variation trend, so that the difference codebook corresponding to each type of vector has small variation range, and the capacity of codebook used in the following quantization procedure is effectively reduced under the prerequisite of assuring reproduction quality, thereby reducing the required quantization bit number.

Description

technical field [0001] The invention relates to the technical field of digital compression, in particular to a vector quantization method and a vector quantizer. Background technique [0002] With the rapid development of multimedia applications, digital compression technology has received extensive attention and research. Regardless of the transmission of voice, image or video, it relies on low bit rate and high quality compression coding. Currently widely used compression coding methods include run-length coding, prediction and interpolation coding, vector quantization (VQ: Vector Quantization) coding, and the like. Among them, the vector quantization coding method is a lossy coding method, which is widely used in speech and image compression technology because of its high compression ratio and simple decoding method. [0003] The core process of vector quantization coding is the process of quantizing the input vector. Before quantization, a series of standard vector mo...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
Patent Type & Authority Patents(China)
IPC IPC(8): H03M7/30G10L19/00H04N7/26G10L19/038
Inventor 李立雄张德军
Owner HUAWEI TECH CO LTD
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic, Popular Technical Reports.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com