Line Spectral Frequency Quantization With Predictive Subvectors
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing speech and audio encoders that employ Line Spectral Frequencies (LSFs) for Linear Predictive Coding (LPC) coefficients require significant bit resources for quantization, leading to computational inefficiencies.
Innovation Solution
A method involving a two-stage quantization process where a first sub-vector is quantized with a single-stage quantizer, a residual vector is formed, and a second sub-vector is quantized with a multiple-scale lattice quantizer, followed by combining these to predict higher-order coefficients using predictor coefficients.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a single-stage quantizer is used for LSF coefficients, then the device complexity is low, but the bit rate requirement is high
Solution Approach 1:
The LSF coefficient vector is divided into two parts: a first part containing the most significant coefficients quantized with a first quantizer, and a second part containing the remaining coefficients quantized with a second quantizer. This segmentation allows different quantization strategies to be applied to different parts of the signal, reducing overall bit rate while maintaining quality.
Solution Approach 2:
Different quantization precision and methods are applied to different parts of the LSF coefficient vector. The first part receives higher precision quantization while the second part receives lower precision quantization, optimizing the trade-off between bit rate and reconstruction quality based on the local importance of each coefficient.
2Manufacturing precision
If a multiple-scale lattice quantizer is used for the second sub-vector, then the manufacturing precision is improved, but the device complexity increases
Solution Approach 1:
The quantization process is segmented into two stages: a first quantizer handling the most significant coefficients with simpler processing, and a multiple-scale lattice quantizer handling the remaining coefficients with more sophisticated processing. This segmentation allows high precision where needed while limiting complexity to manageable levels.
Solution Approach 2:
The multiple-scale lattice quantizer is applied only to the second part of the coefficient vector rather than the entire vector, providing enhanced precision where it is most beneficial while avoiding the excessive complexity that would result from applying it universally.
3Quantity of substance
If predictor coefficients are used to predict higher-order coefficients, then the bit rate is reduced, but the reliability may be affected
Solution Approach 1:
Predictor coefficients are pre-calculated and stored for different speech conditions. During encoding, the appropriate predictor coefficients are selected and applied to predict higher-order LSF coefficients from lower-order ones, reducing the number of bits needed to transmit these coefficients while maintaining accuracy through the use of pre-optimized predictors.
Solution Approach 2:
The quantized LSF coefficients from the first part are used as input to the prediction process, creating a feedback loop where the quantized values influence the prediction of remaining coefficients. This feedback mechanism ensures consistency between quantized and predicted values while minimizing the bit rate required.
Data Source
AI summary
It is disclosed inter alia a method comprising: determining a first sub vector comprising a first plurality of audio parameter coefficients of an audio parameter vector, quantise the first sub vector with a first quantizer to; determining a residual vector, determine a second sub vector comprising a second plurality of coefficients of the residual vector, quantising the second sub vector with a second quantizer to give a quantised second sub vector; combining the quantised second sub vector and the quantised first sub vector to give a quantized audio parameter sub vector comprising a second plurality of quantised audio parameter coefficients; and predict at least one audio parameter coefficient for a quantized audio parameter vector using the second plurality of quantised audio parameter coefficients, wherein the at least one audio parameter coefficient is a higher order audio parameter coefficient than an order of the second plurality of quantised audio parameter coefficients.