Correcting low-frequency phase distortion in electroglottograph waveforms

Area of Science:

• Speech Science
• Bioacoustics
• Signal Processing

Background:

• Electromyography (EGG) waveforms are crucial for voice analysis.
• High-pass filtering is used to reduce noise but can distort EGG signals.
• Acoustic signal acquisition systems may unintentionally introduce high-pass filtering.

Purpose of the Study:

• To analyze the impact of dynamic high-pass filtering on EGG waveforms.
• To investigate methods for correcting phase distortion caused by high-pass filters.
• To assess the effectiveness of digital and analog correction techniques.

Main Methods:

• Simulated dynamic high-pass filtering with varying cutoff frequencies and filter orders.
• Analysis of Fourier components (amplitude and phase) of EGG waveforms.
• Development and testing of digital and analog compensation networks.
• Utilizing square wave responses to characterize unknown filter parameters.

Main Results:

• High-pass filters with cutoff frequencies significantly below the voice fundamental frequency (>=10x) have minimal amplitude impact.
• Phase distortion significantly increases with lower cutoff frequencies and higher filter orders.
• Both digital filter techniques and a simple analog network can effectively reduce phase distortion.
• Square wave response allows for parameter adjustment of compensators for unknown filters.

Conclusions:

• Dynamic high-pass filtering below the voice fundamental frequency primarily causes phase distortion, not amplitude changes, in EGG waveforms.
• Phase distortion is exacerbated by high-order filters and breathy voice quality.
• Effective correction strategies, including analog methods, exist for high-pass filter-induced distortion in EGG signals.
• A practical approach is presented for correcting distortion even when filter characteristics are not precisely known.