1. Introduction
Facial expression recognition has many applications in areas such as image understanding, psychological studies, and smarter human-computer interfaces. In recent years, a number of techniques have been presented in the literature of face expression recognition. According to the properties of images used, the methods can be divided into two classes: video frames based method [5], [6], [7], [11], [12] and still image based method [1], [2], [3], [4]. The former deals with a sequence of intensity image frames and extracts their dynamics for expression recognition. The successful methods include Optical Flow models [5] and Hidden Markov Models (HMM) [11], [12]. The latter analyzes single still image and gives a recognition result based on spatial analysis of features extracted. This class of methods can be grouped into two subclasses: holistic spatial analysis, such as Principle Components Analysis (PCA), and Fisher linear discriminates (FLD), and local spatial analysis, such as Gabor wavelet and local PCA [14].