A new approach of facial expression recognition based on Contourlet Transform

Abstract

Contourlet Transform provides a flexible multiresolution, local and directional image expansion and a sparse representation for two dimensional piecewise smooth signal resembling images. It overcomes the weakness of wavelet transform in dealing with high dimensional signals. In this paper, the theory of Contourlet Transform is introduced and a new approach of facial expression recognition based on Contourlet Transform is proposed. Locally Linear Embedding is then applied for feature dimensionality reduction, and Vector Support Machine is used to classify the seven expressions (anger, disgust, fear, happiness, neutral, sadness and surprise) of JAFFE database. Both Wavelet Transform and Principal Component Analysis (PCA) algorithm are carried out to compare with the approach proposed. Experimental results show that the proposed approach gets better recognition rate than both Wavelet Transform and Principal Component Analysis. The facial expression recognition based on Contourlet Transform is an effective and feasible algorithm.