Facial expression recognition using krawtchouk moments and support vector machine classifier

In this paper, we have proposed an algorithm for facial expression recognition using Krawtchouk moments. In this method, useful facial patches are detected using Viola Jones algorithm. These detected patches are first preprocessed by performing Gaussian filtering and histogram equalization. As our facial expressions are mainly dependent on eyes and lips gestures so only these regions of interests are extracted. Computation of orthogonal moments is performed on extracted ROI at different spatial pyramid levels. These features are then concatenated so that both global and local features can be obtained. The dimensional space of the feature vector is reduced by using Neighbourhood Component Analysis (NCA) so that only relevant features with high expression information are selected. We use 10 fold cross validation to perform classification using binary Support Vector Machine (SVM) classifier. An efficiency of 95.62% and 95.31% on CK+ database and JAFFE database has been obtained respectively.