Facial Expression Recognition using 2D Stationary Wavelet Transform and Gray-Level Co-occurrence Matrix

Abstract

This paper presents an automated facial expression recognition (FER) system based on two dimensional stationary wavelet transform (2D-SWT) and gray-level co-occurrence matrix (GLCM). The proposed scheme employs 2D-SWT to decompose the image into a set of sub-bands. Then GLCM features are obtained from the 2D-SWT sub-bands. Subsequently, linear discriminant analysis (LDA) is harnessed to select the most relevant features. Finally, these features are used for classification of facial emotions using least squares variant of support vector machine (LS-SVM) with radial basis function (RBF) kernel. The performance of the pro-posed system is evaluated on two standard datasets namely, Extended Cohn-Kanade (CK+) and Japanese female facial expression (JAFFE). Experimental results based on 5-fold cross validation strategy indicate that the proposed scheme earns an accuracy of 96.72% and 99.79% over CK+ and JAFFE dataset respectively, which are superior to other competent schemes.