Speech Fundamental Frequency Estimation Using Refined Cross-Correlation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for estimating the fundamental frequency of speech signals, particularly for male speakers, face challenges in accuracy and computational efficiency, especially when dealing with low fundamental frequencies.
Innovation Solution
A method that involves refining the signal spectrum, determining a cross-power spectral density, transforming it into a cross-correlation function, and estimating the fundamental frequency based on this function, which increases the amount of information available and improves estimation robustness for low frequencies.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If auto-correlation function methods are used to estimate fundamental frequency, then the estimation can be performed, but the accuracy deteriorates for low fundamental frequencies (male speakers)
Solution Approach 1:
The patent introduces a refined signal spectrum as an intermediary component. Instead of directly using the original signal spectrum in the auto-correlation function, the method first refines the spectrum (enhancing its quality and information content) and then uses this refined spectrum to compute the auto-correlation function. This intermediary refined spectrum acts as a mediator that improves the accuracy of fundamental frequency estimation, particularly for low frequencies, by providing enhanced spectral information before the correlation analysis.
2Measurement precision
If methods are used to overcome low fundamental frequency estimation problems, then accuracy may improve, but computational efficiency deteriorates or significant delay is introduced
Solution Approach 1:
The patent applies preliminary action by refining the signal spectrum before performing the auto-correlation function computation. The spectral refinement is performed as a preliminary step that prepares the spectrum data in advance, enhancing its information content and quality. This preliminary refinement allows the subsequent auto-correlation analysis to achieve higher accuracy for low fundamental frequencies without requiring computationally intensive methods or introducing significant delays, as the refinement process is efficiently integrated into the existing signal processing workflow.
Data Source
AI summary
The invention provides a method for estimating a fundamental frequency of a speech signal comprising the steps of receiving a signal spectrum of the speech signal, filtering the signal spectrum to obtain a refined signal spectrum, determining a cross-power spectral density using the refined signal spectrum and the signal spectrum, transforming the cross-power spectral density into the time domain to obtain a cross-correlation function, and estimating the fundamental frequency of the speech signal based on the cross-correlation function.