Camera-based Real-time Emotion Classification Using Support-Vector Machines

Abstract

Extracting facial emotions is a great challenge in the area of computer vision, unlike humans, machines have a lot of difficulties with interpreting emotions. Due to the work of Viola-Jones [1] emotion detection is already more feasible. Using a fast algorithm to extract the region of interest (ROI) for emotion detection brings the development of a real-time emotion classification algorithm one step closer. In this research an emotion detection algorithm aimed at real-time implementation on an embedded platform has been developed. Several algorithms have been implemented and tested to evaluate their results with respect to their real-time performance potential and classification efficiency. The developed algorithm relies on three fundamental findings, which relate to the way humans interpret emotions.