Multistage Vector Quantization Using Correlated Codebooks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Multi-stage vector quantization methods waste storage space due to independent codebooks at each stage, leading to low resource usage efficiency.
Innovation Solution
Obtain the current stage codebook using the previous stage codebook, leveraging their correlation to eliminate the need for an independent codebook space, thereby reducing storage requirements and improving resource efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multi-stage vector quantization with independent codebooks at each stage is used, then quantization accuracy is improved, but storage space is wasted and resource usage efficiency is reduced
Solution Approach 1:
The patent merges the codebooks of multiple quantization stages by using the codebook from a reference stage (e.g., voice frame) to replace or supplement codebooks at other stages (e.g., noise frame). Instead of maintaining independent codebooks at each stage, the system combines them by selecting or reusing codebook indices from the reference stage, thereby reducing total storage requirements while preserving quantization accuracy.
Solution Approach 2:
The codebook from a single reference stage serves multiple functions across different quantization stages. The same codebook is reused for both voice frames and noise frames, or selectively applied based on frame type, making the codebook universal rather than stage-specific. This multi-functional usage eliminates redundant storage of identical or similar codebook data.
2Measurement precision
If multi-stage vector quantization with independent codebooks at each stage is used, then quantization accuracy is improved, but resource usage efficiency is reduced
Solution Approach 1:
The patent merges the codebooks of multiple quantization stages by using the codebook from a reference stage (e.g., voice frame) to replace or supplement codebooks at other stages (e.g., noise frame). Instead of maintaining independent codebooks at each stage, the system combines them by selecting or reusing codebook indices from the reference stage, thereby reducing total storage requirements while preserving quantization accuracy.
Solution Approach 2:
The codebook from a single reference stage serves multiple functions across different quantization stages. The same codebook is reused for both voice frames and noise frames, or selectively applied based on frame type, making the codebook universal rather than stage-specific. This multi-functional usage eliminates redundant storage of identical or similar codebook data.
Data Source
Figure 1
Figure 2~3
Figure 4
AI summary
The invention discloses a multi-stage quantization method, which includes the following steps: obtaining a reference codebook according to a previous stage codebook; obtaining a current stage codebook according to the reference codebook and a scaling factor; and quantizing an input vector by using the current stage codebook. The invention also discloses a multi-stage quantization device. With the invention, the current stage codebook may be obtained according to the previous stage codebook, by using the correlation between the current stage codebook and the previous stage codebook. As a result, it does not require an independent codebook space for the current stage codebook, which saves the storage space and improves the resource usage efficiency.