Predicting Fatigue and Psychophysiological Test Performance from Speech for Safety-Critical Environments

Area of Science:

• Aerospace Engineering
• Cognitive Science
• Speech Analysis

Background:

• Operator fatigue poses risks in safety-critical environments.
• Speech analysis offers a potential non-invasive method for fatigue detection.
• Previous research lacked objective comparisons and relied on subjective fatigue ratings.

Purpose of the Study:

• To assess the efficacy of voice features in predicting operator fatigue.
• To compare voice-based fatigue prediction with time-based and circadian rhythm factors.
• To evaluate the prediction accuracy for psychophysiological test scores and sleep latency.

Main Methods:

• Collected voice recordings and psychophysiological test data from seven aerospace personnel over 60 hours of sustained wakefulness.
• Analyzed voice features and their correlation with fatigue levels.
• Compared prediction accuracy of voice features versus time awake and circadian position.

Main Results:

• Both time awake and circadian position influenced voice features and test scores.
• Voice features were superior predictors of psychophysiological test scores and sleep latency compared to time-based metrics.
• Achieved mean absolute prediction errors of 5-12% for test scores and 17.5% for sleep latency.

Conclusions:

• Voice analysis provides a reliable method for estimating operator fatigue.
• Speech-based fatigue monitoring can be implemented in practical applications for safety-critical domains.
• Voice features offer a more accurate assessment of fatigue's impact on cognitive performance than traditional measures.