The Effect of Moving Window on Acoustic Analysis

Area of Science:

• Speech Science
• Acoustic Analysis
• Biomedical Engineering

Background:

• Accurate voice analysis is crucial for diagnosing voice disorders.
• Traditional methods for selecting voice segments may lack stability, impacting analysis reliability.
• The moving window method offers a novel approach to segment selection.

Purpose of the Study:

• To evaluate the efficacy of the moving window method for voice segment selection.
• To compare the acoustic stability of segments identified by the moving window method versus traditional techniques.
• To assess the impact of different segment selection methods on the discrimination of normal and disordered voices.

Main Methods:

• Compared three segment selection methods: moving window, mid-vowel, and whole vowel.
• Analyzed acoustic parameters including jitter, shimmer, signal-to-noise ratio (SNR), cepstral peak prominence (CPP), and correlation dimension (D2).
• Utilized an artificial neural network to evaluate voice discrimination capabilities.

Main Results:

• The moving window method yielded more stable voice segments, indicated by lower perturbation and nonlinear dynamics, and higher SNR and CPP.
• Achieved a discrimination accuracy rate of 91.90% for disordered voices using the moving window method.
• Outperformed mid-vowel (72.34%) and whole vowel (70.34%) methods in voice discrimination.

Conclusions:

• The moving window method is superior for obtaining stable voice segments.
• This method significantly enhances the accuracy of distinguishing between normal and disordered voices.