Facial emotion detection using Guided Particle Swarm Optimization (GPSO)

Abstract

In this paper, we present a novel approach to human facial emotion detection by applying a modified version of the Particle Swarm Optimization (PSO) algorithm, which we called Guided Particle Swarm Optimization (GPSO). Our approach is based on tracking the movements of facial action units (AUs) that are placed on the face of a subject and captured in video clips. We defined particles that form swarms as vectors consisting of points from each domain of the AUs considered. Particles are allowed to move around the effectively n-dimensional search space in search of the emotion being expressed in each frame of a video clip (where n is the number of action units being tracked). Since there are more than one possible target emotions at any point in time, multiple swarms are used, with each swarm having a specific emotion as its target. We have implemented and tested the algorithm on video clips that contain all the six basic emotions, namely happy, sad, surprise, disgust, anger and fear. Our results show the algorithm to have a promising success rate.