Vector Coding with Multi-Level Permutation to Cut Complexity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing multi-rate lattice vector quantization method in audio coding algorithms, such as the TCX part of the AMR-WB+ codec, suffers from high calculation complexity due to the need for multiple irregular factorial calculations during coding and decoding, which cannot be simplified effectively.
Innovation Solution
A vector coding method that decomposes an initial vector into a sign vector and an absolute vector, performs multi-level permutation-based coding on the absolute vector, and combines the sign code and absolute vector code to reduce calculation complexity, eliminating the need for complex factorial calculations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multi-rate lattice vector quantization is used for audio coding, then quantization accuracy is improved, but calculation complexity increases due to multiple irregular factorial calculations
Solution Approach 1:
The patent segments the vector quantization process into distinct stages: determining leader vectors, calculating distance metrics, and performing codebook selection. By dividing the complex quantization task into manageable segments, the method reduces calculation complexity while preserving quantization accuracy through systematic processing of each segment.
Solution Approach 2:
The patent changes the parameter representation by using simplified distance metrics and alternative mathematical formulations that avoid irregular factorial calculations. This parameter transformation maintains the essential quantization functionality while reducing computational burden through more efficient mathematical operations.
2Measurement precision
If multiple irregular factorial calculations are performed during coding and decoding, then quantization precision is maintained, but processing time increases
Solution Approach 1:
The patent performs preliminary calculations of distance metrics and leader vector determinations before the actual quantization process. By pre-computing these values and storing them in codebooks, the method eliminates the need for repeated irregular factorial calculations during real-time encoding and decoding, significantly reducing processing time while maintaining precision.
Solution Approach 2:
The patent creates pre-computed codebooks that contain leader vectors and distance metric information. These copied reference structures allow the system to avoid recalculating complex factorial expressions during operation, instead simply looking up pre-determined values that maintain quantization precision.
3Manufacturing precision
If complex factorial calculations are used in vector quantization, then coding accuracy is improved, but implementation complexity increases
Solution Approach 1:
The patent replaces complex mechanical-style factorial calculations with alternative mathematical approaches using distance metrics and linear algebra operations. This substitution maintains coding accuracy by preserving the essential geometric relationships while using more computationally efficient mathematical tools that are easier to implement.
Solution Approach 2:
The patent transforms the problem parameters from requiring factorial calculations to using distance-based metrics and vector operations. This parameter change maintains the accuracy of vector quantization by preserving the underlying mathematical relationships while enabling simpler implementation through standard linear algebra operations.
Data Source
Figure 1~2
Figure 3a~3b
Figure 4
AI summary
The present invention discloses a vector coding/decoding method, apparatus and a stream media player. The vector coding method includes: decomposing an initial vector to obtain a sign vector and an initial absolute vector; coding the sign vector to obtain a sign code; performing multi-level permutation-based coding on the initial absolute vector to obtain an absolute vector code; and combining the sign code and the absolute vector code to obtain a code of the initial vector. The vector decoding method includes: decomposing a code of an initial vector to obtain a sign code and an absolute vector code; decoding the sign code to obtain a sign vector; performing multi-level permutation-based decoding on the absolute vector code to obtain an initial absolute vector; and combining the sign vector and the initial absolute vector to obtain an initial vector. With the use of the solution, the calculation complexity is reduced.