This proposal requests funds for developing an automated image processing system for measuring facial behavior. Measurement of the face in this project focuses on identifying facial behaviors in terms of muscular actions. based on the muscle-based scoring system most widely used in behavioral science. Image acquisition, preprocessing, analysis, and categorization are produced and coordinated by a computer system using computation such as digital signal processing and neural networks. This Phase I project has two major goals: a) fast and efficient image acquisition and preprocessing and b) accurate analysis and categorization. Three studies are conducted. One replicates and extends previous findings and develops three types of networks (standard backpropagation radial basis, and complex number feedforward networks), comparing their relative performance and accuracy. A second study develops a video preprocessing system using a digital signal processor and examines its speed. The last study integrates these two systems and tests the resulting system for accuracy and speed.Proposed commercial application:A facial measurement computer could be marketed in several markets, including law enforcement, security and intelligence forces for purposes of detecting deception and aiding interrogations, health and mental health care and treatment for purposes of identifying psychiatric disturbances: medical and scientific research for purposes of research on the face and emotion, personnel and human resources for purposes of evaluating applicants and market or consumer research for assessing consumer reactions.National Institute of Mental Health (NIMH)
