Hardware implementation of a voice stress detection algorithm using empirical mode decomposition

WCU Author/Contributor (non-WCU co-authors, if there are any, appear on document)
Joshua Schwartz (Creator)
Institution
Western Carolina University (WCU )
Web Site: http://library.wcu.edu/
Advisor
Robert Adams

Abstract: In previous research, using EMD for voice stress analysis showed promising results. The original algorithm was developed in Matlab, utilizing many of its functions. The algorithm decodes audio into a raw format sampling at 8k samples/sec. This signal array is then filtered and centered about the x-axis and the local maxima and minima are determined. Matlab’s built-in cubic spline interpolation function was then used to create the upper and lower envelopes. The first Intrinsic Mode Function (IMF) is determined by calculating the difference between the original signal and the mean of the envelopes. The algorithm continually extracts IMFs starting with the highest frequency IMF until it extracts the last IMF. The last IMF represents the frequency of the stress induced tremor in the subjects voice. The successfulness of the algorithm was measured using a series of questions designed to invoke a stress response concatenated with irrelevant questions designed not to invoke a stress response. The difference between the tremor frequency related to the stressful and non-stressful questions determines if a person may be being untruthful.

Additional Information

Publication
Thesis
Language: English
Date: 2011
Subjects
Voice -- Psychological stress analysis
Stress (Psychology) -- Measurement

Email this document to