Toward Acoustic-Based Normalization of Laryngeal EMG for Improved Interspeaker Consistency in Muscle-to-Acoustic...

Area of Science:

• Voice production research
• Laryngeal biomechanics
• Experimental phonetics

Background:

• Understanding intrinsic laryngeal muscle activation is crucial for voice research.
• Intrasubject variability in EMG recordings presents a significant challenge for inter-session comparisons.

Purpose of the Study:

• To develop and validate a normalization framework using acoustic targets to enhance the consistency of EMG-acoustic relationships across sessions.
• To reduce intrasubject variability in laryngeal muscle EMG data.

Main Methods:

• Collected EMG signals from cricothyroid (CT), thyroarytenoid (TA), and lateral cricoarytenoid (LCA) muscles across four sessions.
• Implemented an acoustic target-based normalization framework.
• Generated Muscle Activation Plots (MAPs) to visualize EMG-acoustic relationships (f0, SPL).

Main Results:

• The proposed normalization method significantly reduced intrasubject variability compared to traditional peak-based normalization.
• Inter-session consistency of the EMG-f0 relationship improved, with ICC increasing from 0.52 to 0.83.
• Mean Absolute Error (MAE) in the EMG-f0 relationship decreased from 84.87 Hz to 19.23 Hz.

Conclusions:

• The acoustic target-based normalization framework offers a more consistent approach for intramuscular EMG analysis in voice production.
• This method improves the reliability of interpreting muscle-specific control strategies in human voice.
• Findings advance experimental voice research by providing a more robust method for analyzing laryngeal muscle activity.