University of Houston
Department of Computer Science

In partial fulfillment of the Requirements for the Degree of
Doctor of Philosophy in Computer Science

Dvijesh Shastri
will propose his Dissertation topic

Measurement of Facial Physiological Signals For Lie Detection

Abstract

This research effort focuses on localization of facial features on thermal infrared imagery. We propose a novel method that localizes the periorbital tissue on thermal infrared imagery. Specifically, the method segments consistently the tissue that is over the facial and ophthalmic arterial-venous complexes, which supply with blood the ocular muscle. This tissue area is used to extract a mean thermal signal over time (periorbital signal), which is a correlate of blood flow. Previous work demonstrated that the periorbital signal is associated to autonomic responses and it changes significantly upon the onset of instantaneous stress. The present method automates the measurement of the periorbital signal. This opens the way for applications of the technology in computational psychology and particularly in lie detection. Indeed, the method was tested as part of an imaging based lie detection system. The experimental set consisted of 39 subjects and the experimental design focused on realism. The results are very encouraging. As a byproduct of periorbital localization, quantification of blinking is now possible. The next step is the localization of the supra-orbital vessel, which supplies with blood the corrugator muscle. Recent research has demonstrated that blood flow in this vessel is associated with the onset of non-autonomic stress (mental load). The goal of this effort is to provide a complete set of stress measurements that are highly automated and can be performed at a distance. Rigorous and ubiquitous quantification of stress is expected to have an impact beyond lie detection and may revolutionize psychological practice.

Date: May 8^th, 2006
Time: 2:00PM
Place: PGH 218

Faculty, students, and the general public are invited.
Thesis Advisor: Ioannis Pavlidis