An Expression Space Model for Facial Expression Analysis

Modeling of facial expression plays a very important role in the research of synthesis and recognition offacial expression. However the current facial expression model can not correctly represent the daily facial expressions. The nature of the facial expressions is analyzed. A qualitative description of the corresponding facial expression space is presented. And then a new facial expression space model is proposed with the characters of both discrete affective space model and continuous affective space model. To validate the rationality of the model, the experiments on a facial expression database JAFFE have been conducted using Gaussian mixture model based on the Gabor wavelet and principal component analysis. The distribution of different expressions is analyzed, and the qualitative and quantitative description of the facial expression space is accomplished. The experimental results show that the facial expression model proposed can rationally represent the daily facial expressions.