Facial expression recognition using wavelet based Support Vector Machine

The present work is an attempt to unravel the classical problem of Facial Expression Recognition (FER). In realization of the FER system the emphasis is given on preprocessing technique. The paper proposes the Gaussian mask for illumination correction pre-processing which when subtracted from histogram equalized illumination plane shows improvement in the image and the FER results. The present work shows the use of wavelet in feature extraction and Support Vector Machine (SVM) as a classifier will result in accurate and robust FER system. It proposes the quantization and encoding technique of the wavelet decomposed features that result in greater FER accuracy.