Adaptive LPC Windowing for Better Voice Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current Linear Prediction Coding (LPC) analysis techniques face suboptimal performance due to the use of fixed window functions and increased complexity from multiple rounds of analysis with different window sizes, which affects the efficiency and accuracy of voice signal compression.
Innovation Solution
An adaptive windowing method is introduced, where the window function is dynamically selected based on the amplitude values of the input voice signal's sample points, allowing for improved linear prediction performance with reduced complexity by using specific cosine-based window functions for different segments of the signal.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a fixed window function is applied in the windowing process, then the device complexity is reduced, but the linear prediction performance deteriorates
Solution Approach 1:
The patent applies dynamics by making the window function adaptive rather than fixed. The system dynamically selects between different window functions (first window function when amplitude exceeds threshold, second window function when amplitude does not exceed threshold) based on the input signal characteristics, thereby optimizing linear prediction performance without significantly increasing complexity
Solution Approach 2:
The patent changes the parameter of the window function based on signal amplitude. By comparing the amplitude of the input signal with a threshold value and selecting different window functions accordingly, the system adapts the windowing parameters to match the signal characteristics, improving prediction accuracy while maintaining manageable complexity
2Measurement precision
If two rounds of LPC analysis are performed with different window sizes, then the linear prediction performance is improved, but the device complexity increases
Solution Approach 1:
The patent reduces complexity by making the analysis process dynamic and adaptive rather than performing fixed multiple rounds. The system dynamically determines whether to apply a first or second window function based on signal amplitude characteristics, achieving optimal performance with a single adaptive analysis pass instead of multiple fixed rounds
Solution Approach 2:
The patent changes the window function parameter based on signal characteristics (amplitude comparison with threshold) to optimize the single analysis round, eliminating the need for multiple analysis passes with different window sizes while maintaining improved prediction performance
Data Source
Figure 1~2
Figure 3~4
Figure 5~6
AI summary
The present invention relates to communication technologies and discloses a method, an apparatus and a system for Linear Prediction Coding (LPC) analysis to improve LPC prediction performance and simplify analysis operation. The method includes: obtaining signal feature information of at least one sample point of input signals; comparing and analyzing the signal feature information to obtain an analysis result; selecting a window function according to the analysis result to perform adaptive windowing for the input signals and obtain windowed signals; and processing the windowed signals to obtain an LPC coefficient for linear prediction. The embodiments of the present invention are applicable to LPC.