Publications

Abstract : Many classifiers are in existence for handling linear and nonlinear type of data. In this paper, a method for handling nonlinear data for classification is proposed. The classifier is modeled for both two-class and multi-class problem. In this framework, the input samples are projected into a higher dimensional feature space by utilizing kernel trick. In order to avoid the difficulty of working in this higher dimension space, a dimensional reduction technique is used. This reduction technique results with the transformation matrix which is of lesser dimensions. Then the classification of data is carried out in this reduced subspace by using perceptron technique. The classifier is utilized for recognizing the emotion of a person based on the facial expressions. Experiments are carried on toy data set and JAFFE data set and also the results have shownthe effectiveness of this classifier. The classifier recognizes the 6 different Emotions with 98. 6% efficiency.