Fatigue Estimation Using Voice Analysis

Abstract

In the present article, we present a means to remotely and transparently estimate an individual's level of fatigue by quantifying changes in his or her voice characteristics. Using Voice analysis to estimate fatigue is unique from established cognitive measures in a number of ways: (1) speaking is a natural activity requiring no initial training or learning curve, (2) voice recording is a unobtrusive operation allowing the speakers to go about their normal work activities, (3) using telecommunication infrastructure (radio, telephone, etc.) a diffuse set of remote populations can be monitored at a central location, and (4) often, previously recorded voice data are available for post hoc analysis. By quantifying changes in the mathematical coefficients that describe the human speech production process, we were able to demonstrate that for speech sounds requiring a large average air flow, a speaker's voice changes in synchrony with both direct measures of fatigue and with changes predicted by the length of time awake.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Aug 01, 2006
Accession Number
ADA542363

Entities

People

  • Eric Friets
  • Glen Greenough
  • Harold P. Greeley
  • Jeffrey Whitmore
  • Joel Berg
  • John A. Wilson
  • Joseph Picone
  • Thomas Nesthus

Organizations

  • Creare, Inc.

Tags

Communities of Interest

  • C4I
  • Human Systems
  • Space

DTIC Thesaurus Topics

  • Air Flow
  • Air Force
  • Air Force Research Laboratories
  • Automated Speech Recognition
  • Coefficients
  • Communication Systems
  • Control Systems
  • Flow
  • Frequency
  • Frequency Domain
  • Infrastructure
  • Measurement
  • Military Research
  • New Hampshire
  • Production
  • Recognition
  • United States

Readers

  • Aerospace Test and Evaluation
  • Computational Linguistics
  • Structural Health Monitoring of Composite Structures.