Facial Expression Recognition Using PCA and Various Distance Classifiers

Abstract

Information Technology is playing a very big role in today’s world. Our interaction with IT is mainly through advanced human computer interfaces. In this regard we seek to enhance the interface in a way such that it can take into account the human facial expression and respond according to a person’s feelings in a broader sense. We propose here a simple yet efficient way of facial expression recognition using Eigenspaces and dimensionality reduction techniques and multiple classifiers. It is a modified approach to the original Eigenface method by Turk and Pentland (J Cogn Neurosci, 1991) [1] for face recognition, where using the standard JAFFE database we classify each test image as belonging to one of the six basic expression classes—anger, disgust, fear, happiness, sadness or surprise. In the process we put to test four different classifiers—Euclidean distance, Manhattan distance and Cosine distance and Mahalanobis distance classifiers. In this paper we present with experimental evidence the accuracy of our method and the comparative results yielded by each of the classifiers.