Speech Pitch Estimation Using Mixture Gaussian Distributions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for estimating the pitch of a speech signal are affected by Formant frequency, leading to inaccurate pitch estimation due to the obscuration of the pitch by Formant frequencies and the quasi-periodic nature of speech signals.
Innovation Solution
A method and apparatus that compute a normalized autocorrelation function of a windowed speech signal, determine candidate pitches, interpolate their periods, generate Gaussian distributions, mix distributions within a certain threshold, select those with high likelihood, and use dynamic programming to estimate pitch, while also considering sub-harmonic frequency ranges to refine pitch estimation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional pitch extracting methods are used, then the process is simple, but the pitch cannot be accurately estimated due to Formant frequency interference
Solution Approach 1:
The patent segments the pitch estimation process into multiple stages: candidate pitch generation from autocorrelation, Gaussian distribution modeling for each candidate, mixture Gaussian distribution formation, and dynamic programming-based final estimation. This segmentation allows each stage to address specific aspects of the problem, improving overall accuracy while managing complexity through modular processing.
Solution Approach 2:
The patent introduces Gaussian distributions and mixture Gaussian distributions as intermediary representations between the raw autocorrelation function and the final pitch estimate. These intermediaries model the uncertainty and variability in pitch candidates, allowing the system to handle Formant frequency interference more robustly by evaluating multiple hypotheses simultaneously.
2Measurement precision
If multiple Gaussian distributions are mixed and dynamic programming is executed, then pitch estimation accuracy improves, but computational complexity increases
Solution Approach 1:
The patent performs preliminary actions by generating Gaussian distributions for candidate pitches and forming mixture Gaussian distributions before executing dynamic programming. This preliminary modeling of pitch uncertainties allows the dynamic programming stage to work with pre-processed, structurally organized data, reducing the computational burden during the final estimation phase.
Solution Approach 2:
The patent employs dynamic programming to adaptively select the most likely pitch trajectory across multiple frames, allowing the system to dynamically adjust to variations in speech signals while avoiding exhaustive search. This dynamic approach optimizes computational resources by focusing on promising pitch candidates rather than evaluating all possibilities equally.
Data Source
AI summary
A pitch estimating method and apparatus in which mixture Gaussian distributions based on candidate pitches having high period estimating values are generated, a mixture Gaussian distribution having a high likelihood is selected and dynamic programming is executed so that the pitch of the speech signal can be accurately estimated. The pitch estimating method comprises computing a normalized autocorrelation function of a windowed signal obtained by multiplying a frame of a speech signal by a window signal and determining candidate pitches from a peak value of the normalized autocorrelation function of the windowed signal, interpolating a period of the determined candidate pitches and a period estimating value representing a length of the period, generating Gaussian distributions for the candidate pitches for each frame for which the interpolated period estimating value is greater than a first threshold value, mixing the Gaussian distributions which are located at a distance less than a second threshold value to generate mixture Gaussian distributions and selecting at least one of the mixture Gaussian distributions that a likelihood exceeding a third threshold value, and executing dynamic programming for the frames to estimate the pitch of each frame, based on the candidate pitches of each of the frames and the selected mixture Gaussian distributions.