Facial Image Based Expression Classification System Using Committee Neural Networks
Gayatri Suhas Paknikar · OhioLink ETD Center (Ohio Library and Information Network) · 2008
Human communication has two main aspects: verbal (auditory) and non-verbal (visual).Facial expressions, body movements and physiological reactions are the basic units of the non-verbal communication.Facial expressions and related changes in facial patterns give us information about the emotional state of the person.Psychopathology, stress detection, human-computer interface, and terror deterrence etc., are some of the applications of facial expression detection.The goal of this study was to classify different facial expressions of individuals from static facial images from a large database with an improved accuracy over previously presented systems.Two classification approaches were used.First, a parameter based classification system was developed which classified the expressions based on the actual parameter values directly.Evaluation of the parameter based system revealed that it could accurately classify only one expression.In the second approach, the committee neural network system was used to classify seven basic emotion types from facial images.Two types of committees, viz.primary committee and secondary committee were trained and evaluated.Committee performance was better than performance of individual networks.The integrated committee system, which incorporated both primary and secondary committees, accurately classified the expressions in 94.73% of the cases.