Two-Stage Vector Quantization for Low-Bitrate Audio Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current speech and audio coding technologies, such as the Enhanced Voice Service (EVS) codec, face inefficiencies in memory usage and computational complexity, particularly at lower bitrates, especially when dealing with voiced, unvoiced, generic, transition, or comfort noise generation parts of audio signals.
Innovation Solution
A method involving a two-stage quantization process where a first quantized representation is determined using a codebook, and a second quantized representation is derived based on this initial encoding, allowing for adaptive codebook selection and normalization to improve coding efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If lattice based codebooks are used for quantization, then memory requirements and computational complexity are reduced, but quantization efficiency deteriorates at lower bitrates
Solution Approach 1:
The codebook is segmented into multiple subsets, each optimized for different quantization scenarios. The method divides the quantization process into stages, where a first codebook subset is used for initial quantization and a second codebook subset is used for refinement, allowing efficient operation at lower bitrates while maintaining overall system complexity manageable
Solution Approach 2:
The codebook selection is made dynamic and adaptive based on the quantization results. The system selectively chooses between different codebook subsets depending on the specific quantization scenario and bitrate requirements, optimizing performance for each condition rather than using a fixed codebook structure
2Quantity of substance
If structured codebooks are used for quantization, then memory requirements are reduced, but quantization quality deteriorates for different coding modes
Solution Approach 1:
Different codebook subsets are designed with specialized structures optimized for specific coding modes (voiced, unvoiced, generic, transition, or comfort noise generation). Each subset provides locally optimized quantization quality for its target coding mode while maintaining overall memory efficiency through the structured organization of multiple subsets
Solution Approach 2:
The plurality of codebook subsets collectively provides universal coverage for all coding modes. The system can selectively activate appropriate subsets based on the current coding mode, allowing a single memory-efficient codebook structure to serve multiple functions and maintain high quantization quality across diverse audio signal types
Data Source
AI summary
It is inter alia disclosed to determine a first quantized representation of an input vector, and to determine a second quantized representation of the input vector based on a codebook depending on the first quantized representation.