Linear Predictive Analysis Using Dynamic Coefficients for Spectral Precision
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional linear predictive analysis methods for audio and acoustic signals suffer from degraded precision in approximating spectral envelopes due to the use of fixed coefficients, especially when spectral peaks are not high, leading to reduced analysis accuracy.
Innovation Solution
A linear predictive analysis method that calculates autocorrelation and determines coefficients from tables based on fundamental frequency and pitch gain, adjusting the coefficient selection based on the period and correlation values to improve precision by selecting appropriate coefficients from multiple tables.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a fixed coefficient is used for modifying autocorrelation, then the device complexity is reduced, but the manufacturing precision of spectral envelope approximation deteriorates
Solution Approach 1:
The patent transitions from fixed coefficients to dynamic coefficients that adapt to signal characteristics. The coefficient is determined based on the fundamental frequency and pitch gain of the input signal, allowing the system to adjust the autocorrelation modification dynamically according to the actual signal properties, thereby improving spectral envelope approximation precision while maintaining reasonable complexity through algorithmic determination.
Solution Approach 2:
The patent changes the parameters used for coefficient determination from fixed values to variable parameters derived from signal analysis. Specifically, the coefficient is determined using the fundamental frequency and pitch gain as input parameters, allowing the system to adapt to different signal conditions and improve approximation precision across various spectral characteristics.
2Measurement precision
If autocorrelation is modified by multiplying with a coefficient, then spectral peak suppression is improved, but the loss of information in the autocorrelation function increases
Solution Approach 1:
The patent uses parameter-based coefficient determination where the modification coefficient is derived from fundamental frequency and pitch gain. This ensures that the autocorrelation is modified with a scientifically justified parameter that targets spectral peak suppression while preserving essential signal characteristics, thereby reducing information loss compared to arbitrary fixed coefficients.
Solution Approach 2:
The system uses pitch gain as a feedback parameter to control the degree of autocorrelation modification. The pitch gain, which reflects the strength of periodic components in the signal, provides feedback information that guides the coefficient selection, ensuring that spectral peaks are suppressed appropriately without过度 modifying the autocorrelation function and losing critical information.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
An autocorrelation calculating part 21 calculates autocorrelation Ro(i) from an input signal. A predictive coefficient calculating part 23 performs linear predictive analysis using modified autocorrelation R'o(i) obtained by multiplying the autocorrelation Ro(i) by a coefficient wo(i). Here, it is assumed that a case where, for at least part of each order i, the coefficient wo(i) corresponding to each order i monotonically increases as a value having negative correlation with a fundamental frequency of an input signal in a current frame or a past frame increases and a case where the coefficient wo(i) monotonically decreases as a value having positive correlation with a pitch gain in a current frame or a past frame increases, are comprised.