Acoustic correlates of laryngeal-muscle fatigue: findings for a phonometric prevention of acquired voice pathologies

Victor J. BoucherInterspeech 2007

TL;DR

This paper investigates the acoustic markers of laryngeal muscle fatigue, a key factor in voice pathologies. By combining laryngeal EMG with acoustic analysis, the study reveals that traditional acoustic parameters don't directly correlate with fatigue, but voice tremor temporarily increases at critical fatigue points, suggesting a need for more sophisticated, compensatory-aware analysis for voice health monitoring.

What Problem Does This Paper Address?

Work-related voice injuries are a significant problem, particularly in professions like teaching, yet there's a lack of measurable phonometric standards for prevention. This is largely due to the difficulty in defining and detecting "voice fatigue," which is fundamentally a physiological condition of laryngeal muscle strain rather than a purely acoustic one. Previous attempts to link acoustic measures to fatigue have been inconsistent, partly because they haven't accounted for the underlying physiological state or the compensatory mechanisms the vocal system employs to maintain function under stress. This paper aims to bridge this gap by identifying reliable acoustic correlates of laryngeal muscle fatigue, paving the way for preventative measures.

Methods

The study employs a multi-faceted approach:

  • Electromyography (EMG) for Fatigue Estimation: A standard technique of analyzing EMG spectral compression was used. This involves applying Fourier analysis to laryngeal muscle contraction potentials. As muscles fatigue, frequency components shift towards lower values, quantified by "mean percentile frequency." This method was applied to the lateral cricoarytenoid (LCA) muscles of subjects over extended vocal effort periods (12-14 hours).
  • Acoustic Recording: During the EMG recording sessions, subjects produced sustained vowel sounds ([a:]) at regular intervals. These vocalizations were recorded using a high-quality microphone and DAT recorder.
  • Acoustic Parameter Analysis: A comprehensive set of 19 acoustic parameters, categorized into fundamental frequency, frequency perturbation, amplitude perturbation, tremor, and noise, were analyzed using the MDVP-Advanced software. Specific settings and procedures were standardized to ensure reliability.
  • Compensatory Effect Investigation: To explore how the vocal system compensates for fatigue, EMG was used to monitor the activity of both LCA (adductor) and cricothyroid (CT) muscles in one subject. The root-mean-square (RMS) values of their potentials were analyzed.

Datasets

The study utilized recordings from seven subjects who underwent prolonged vocal effort sessions (12-14 hours). For the compensatory effects investigation, data from one subject (M1) with implanted electrodes for both CT and LCA muscles was used. No new benchmark datasets were introduced; rather, existing recording and analysis protocols were applied.

Key Insights and Ideas

  • No Linear Correlation with Conventional Acoustics: The study found that standard acoustic parameters (like jitter, shimmer, etc.) do not show consistent linear correlations with laryngeal muscle fatigue as measured by EMG. This highlights a fundamental limitation in relying solely on these measures for fatigue detection.
  • Voice Tremor as a Potential Indicator: While not persistent, voice tremor parameters (specifically Fatr) showed temporary, systematic rises that coincided with critical shifts in muscle fatigue. This suggests tremor might be a transient acoustic manifestation of fatigue.
  • Compensatory Mechanisms are Crucial: The investigation into CT and LCA muscle activity revealed that as LCA muscles fatigue (showing decreased activity), CT muscles increase their activity. This compensatory action, likely to stabilize vocal folds, can influence acoustic output, potentially explaining why tremor doesn't always persist and why FO might rise.
  • Limitations of Current Voice Analysis Systems: The findings imply that current voice analysis systems, which often rely on comparing parameters to normative scales, are insufficient. They fail to capture the dynamic, compensatory strategies speakers employ, which are critical for understanding fatigue.

Key Results

  • Critical shifts in LCA muscle fatigue, identified by spectral compression in EMG, were observed at specific times of day across subjects.
  • Plots of voice tremor parameters (Fatr) showed systematic rises that corresponded to these critical fatigue points, particularly for six out of seven subjects.
  • In one subject, increased CT muscle activity was observed as LCA muscle activity decreased, indicating a compensatory strategy. This CT muscle activity was linked to a rise in fundamental frequency (FO).
  • The study found that none of the 19 conventional acoustic parameters analyzed presented consistent linear correlations with the observed spectral estimates of laryngeal muscle fatigue.

Limitations and Future Work

Limitations:

  • The study acknowledges that the results are preliminary, particularly regarding the compensatory effects.
  • The investigation into compensatory effects was conducted on only one subject.
  • The paper notes that current voice analysis systems are not adapted to capturing compensatory changes related to fatigue.

Future Work:

  • Further research is needed to validate the findings on compensatory effects with a larger sample size.
  • Developing more sophisticated acoustic analysis methods that can account for individual compensatory strategies is crucial for creating reliable phonometric signs of vocal fatigue.
  • The authors suggest that future systems should move beyond simple parameter comparisons to normative scales and instead focus on capturing relative changes and individual compensatory behaviors.

Summary updated

BibTeX
@inproceedings{boucher2007acoustic,
  author = {Victor J. Boucher},
  title = {{Acoustic correlates of laryngeal-muscle fatigue: findings for a phonometric prevention of acquired voice pathologies}},
  booktitle = {Interspeech},
  year = {2007},
}
Figure