Weighting Function for LPC Coefficient Quantization

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing linear predictive coding (LPC) technologies face challenges in accurately quantizing LPC coefficients, leading to a narrowing dynamic range and instability, which affects the quality of synthesized signals due to unequal significance of LPC coefficients during quantization.

Innovation Solution

A weighting function determination apparatus and method that converts LPC coefficients to line spectral frequency (LSF) or immitance spectral frequency (ISF) coefficients, using a combined weighting function based on spectral magnitudes and frequency information to enhance quantization efficiency and reflect the significance of LPC coefficients.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Device complexity

If all LPC coefficients are quantized with the same significance, then the quantization process is simple, but the quality of the synthesized input signal deteriorates

Engineering Contradiction:
Improvequantization process complexityVSAvoidsynthesized signal quality
Core Design Contradiction:
Device complexityVSManufacturing precision

Solution Approach 1:

The patent applies different quantization significance levels to different LPC coefficients based on their individual importance. Each coefficient is assigned a weight reflecting its significance, allowing critical coefficients to be quantized with higher precision while less important ones use coarser quantization. This resolves the contradiction by making the quantization process adaptive to local coefficient characteristics rather than applying uniform quantization.

Inventive Principle:
Principle #3Local quality

Solution Approach 2:

The patent introduces weighting parameters that change the quantization step size or precision for each LPC coefficient. By dynamically adjusting these parameters based on coefficient importance, the system achieves both simplified processing (through parameterized quantization) and improved signal quality (through adaptive precision), resolving the contradiction between simplicity and quality.

Inventive Principle:
Principle #35Parameter changes

2Ease of operation

If LPC coefficient quantization is performed without considering coefficient significance, then the quantization process is straightforward, but the dynamic range narrows and stability verification becomes difficult

Engineering Contradiction:
Improvequantization operation simplicityVSAvoidquantization stability
Core Design Contradiction:
Ease of operationVSReliability

Solution Approach 1:

The patent performs preliminary analysis to determine the significance of each LPC coefficient before quantization. By pre-calculating weights or importance metrics for each coefficient, the system prepares the quantization process in advance, ensuring that stability-critical coefficients receive appropriate attention. This preliminary action maintains operational simplicity while embedding stability considerations into the quantization design.

Inventive Principle:
Principle #10Preliminary action

3Device complexity

If uniform quantization significance is applied to all LPC coefficients, then the coding complexity is reduced, but the coding performance deteriorates

Engineering Contradiction:
Improvecoding complexityVSAvoidcoding performance
Core Design Contradiction:
Device complexityVSLoss of information

Solution Approach 1:

The patent implements local quality assessment by evaluating each LPC coefficient's contribution to the overall signal representation. Coefficients with higher impact on signal fidelity are assigned greater quantization precision, while less critical ones use coarser levels. This localized approach optimizes coding performance by concentrating bits where they provide maximum benefit, resolving the contradiction between coding simplicity and performance.

Inventive Principle:
Principle #3Local quality

Data Source

PatentEP4095854B1Weight function determination device and method for quantizing linear prediction coding coefficient
Publication Date: 2024.08.07 SAMSUNG ELECTRONICS CO LTD
  • EP4095854B1 patent drawingFigure 1
  • EP4095854B1 patent drawingFigure 2
  • EP4095854B1 patent drawingFigure 3

AI summary

A weighting function determination method includes obtaining a line spectral frequency (LSF) coefficient or an immitance spectral frequency (ISF) coefficient from a linear predictive coding (LPC) coefficient of an input signal and determining a weighting function by combining a first weighting function based on spectral analysis information and a second weighting function based on position information of the LSF coefficient or the ISF coefficient.