Encoding Null Vector Regions in Spectral Coefficient Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In audio/speech encoding, when bits are limited or the spectrum concentrates energy in certain frequency bands, many vectors are quantized as null vectors, leading to a sparse spectrum and inefficient bit usage, resulting in wasted bits indicating these null vectors.
Innovation Solution
The proposed method performs spectral cluster analysis to identify null vectors and neighboring non-null vectors, converting codebook indications for null vectors into efficient indices, reducing the number of bits required for encoding by representing null vectors regions and quantizing the ending index of these regions adaptively.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If codebook indications for all vectors are transmitted directly, then complete spectral information is preserved, but bit consumption increases when many vectors are null
Solution Approach 1:
The invention extracts and identifies null vectors from the spectral data, separating them from non-null vectors. By detecting regions where consecutive vectors are null and extracting only the essential information (starting position and count of null vectors), the patent eliminates redundant bit transmission while preserving complete spectral information at the decoder side through appropriate handling of null vector regions.
2Measurement precision
If null vectors are encoded individually with codebook indications, then each vector is accurately represented, but encoding efficiency decreases due to redundant bits
Solution Approach 1:
The invention merges consecutive null vectors into a single encoded representation by identifying null vector regions and encoding them with a unified structure (starting position index + count). This combining approach maintains accurate representation of each null vector's contribution to the spectrum while dramatically improving encoding efficiency by eliminating redundant individual encodings of consecutive null vectors.
Solution Approach 2:
The patent introduces dynamic adaptation in the encoding process by detecting the actual distribution of null vectors in the spectral data and adjusting the encoding strategy accordingly. The encoder dynamically identifies null vector regions and applies optimized encoding only where needed, while using standard encoding for non-null regions, thereby adapting the encoding precision and bit allocation to the actual signal characteristics.
Data Source
AI summary
This invention introduces apparatus and methods to efficiently encode the quantization parameters of split multi-rate lattice vector quantization. In this invention, by doing spectral analysis on the split multi-rate vector quantized spectrum, the spectrum is split to null vectors region and non-null vectors region. For the null vectors region, instead of transmitting series of indication for null vectors, an indication of null vectors region and the quantized value of index of the ending vector in the null vectors region (or the number of the null vectors in the null vectors region) are transmitted. The indication of null vectors region can be designed in many ways, the only requirement is the indication should be distinguishable in the decoder side. The ending index or the number of null vectors can be quantized by an adaptively designed codebook. By applying of the invented method, some bits can be saved from the codebook indications.