Fatigue Detection Using Voice Analysis

Abstract

Fatigue is harmful to human health as it impairs the maximal cognitive and physical performance. Fatigue is a critical element in many professions. Human voice has inevitable dependence on fatigue. The main objective of this study is to estimate the fatigue in an individual by speech analysis. Non-intrusive fatigue measurement systems are required to accurately examine the attentiveness and concentration of a person prior to and in an ongoing critical mission or during life threatening activities (e.g. for pilots, drivers, neurosurgeons etc., fatigue is a critical element in their profession). The speech based fatigue measurement system is non-intrusive and has many advantages over other measurements techniques. This study discusses fatigue and its affect on the human speech, factors affecting acquisition and analysis of speech data. In this study speech samples have been acquired individually from 14 healthy adults during 24 hours of sustained wakefulness. Speech samples were recorded using digital voice recorder in a quiet room. From each participant 14 samples were acquired and hence a database having a total of 196 samples was created. The different feature sets and different classifiers have been studied in order to improve the detection of fatigue.