International Journal of Applied Information Systems (IJAIS) – ISSN : 2249-0868 Foundation of Computer Science FCS, New York, USA Volume 5 – No. 9, July 2013 – www.ijais.org 9 Human Facial Expression Recognition using Gabor Filter Bank with Minimum Number of Feature Vectors Priya Sisodia M.Tech (Computer Science Engg.) Ajay Kumar Garg Engineering College, Ghaziabad, India. Akhilesh Verma Asst. Prof (Computer Science Engg.) Ajay Kumar Garg Engineering College, Ghaziabad, India. Sachin Kansal PhD (Mechanical Department) IIT Delhi, Delhi, India ABSTRACT The Human Facial Expression Recognition is used in many fields such as mood detection and Human Computer Interaction (HCI). Gabor Filters are used to extract features. Gabor has the useful property of robustness against slight object rotation, distortion and variation in illumination. In the present work the effort has been made to provide the modules of for Human facial expression recognition by reducing the number of parameters use to represent Gabor feature the space complexity can reduce. SVM classifier has multi-classes. SVM classifies the expression by comparing it with the trained data. Keywords Image Acquisition, Preprocessing, Feature Extraction, Classification. 1. INTRODUCTION Human facial expression recognition is defined as a change that happens in response to human internal emotional states. At the time of human communication, facial expression played a major role. Major component of human communication is facial expression which constitutes around 55% of the total communicated message. The basic facial expression that recognized by psychologists are: neutral, happiness, sadness, anger, fear, surprise and disgust. Applications of Human Facial Expression recognition system are psychological Studies, social interaction, Human Computer Interaction (HCI) (Ekman et al., 1997). There are many methods have been proposed for human facial expression recognition from static images (image database) to image sequence (video) (B. Fasel et al., 2003). Human facial expression analysis is a computer system that automatically analyzes and recognizes facial features and classifies those (Fernando De La Torre et al., 2011). Following steps involve in Human Facial expression recognition system are:- 1. Preprocessing: this step comprises of operations like image scaling, image brightness and contrast adjustment and other image enhancement operation. The Human Facial Expression system uses an image database to train and test the performance of the classifier. 2. Feature Extraction: the key parameters that efficiently represent the particular facial expression need to be extracted from the images. These parameters are used to discriminate between expressions. 3. Classification: Feature vector of test image is compared with feature vector of trained database and classify them accordingly (Shishir Bashyal et al., 2008). Techniques under Feature Extraction phase and Classification Phase are as follows (Ying-Li Tian et al., 2005). Feature Extraction Geometric Based Appearance/ texture Based 2D 3D PCA LBP ICA Gabor filter Classifiers SVM NN k-NN LDA HMM Appearance based approach outperforms Geometric based approach (Zhengyou Zhang et al., 1998). There are numbers of technique for facial expression analysis but no single method is efficient with respect to memory and time complexity. Figure 1: General Structure of Human Facial expression Analysis System Figure 2: Techniques for Feature Extraction and Classifier