Face Recognition Technique Using PCA, Wavelet and SVM

Biometric-based technologies include the identification based on physiological characteristics such as face, fingerprints, hand geometry, hand veins, palm, iris, retina, ear, voice and behavioral traits such as gait, signature and keystroke dynamics [1]. These biometric technologies require some voluntary action by the user. However, face recognition can be done passively without any explicit action or participation of the user, since face images can be acquired from a distance by a camera. The face reorganization system is more appropriate for security and surveillance purposes. Facial images can be easily obtained with a couple of inexpensive fixed cameras; they cannot be modified or forged. Face images are not affected by background sound noise. Face recognition algorithms with preprocessing of the images may compensate for noise, slight variations in orientation, scale and illumination. Face recognition by computer can be divided into two approaches [2], namely, constituent-based and face-based. In constituent-based approach, recognition is based on the relationship between human facial features such as eyes, mouth, nose, profile silhouettes and face boundary [3]. The success of this approach highly depends on the accuracy of the facial feature extraction schemes. Every human face has similar facial features; a small deviation in the extraction may introduce a large classification error. Face-based approach [4] uses the face as a whole for recognition. Many face based recognition algorithms have been developed and each has its strength. Principal Component Analysis (PCA) [5] has been proven to be an effective face-based approach. Sirovich and Kirby [6] first proposed a method using Karhunen-Loeve (KL) transform to represent human faces. In their method, faces are represented by a linear combination of weighted eigenvectors, known as eigenfaces. Turk and Pentland [7] developed a face recognition system using PCA. However, common PCAbased methods suffer from two limitations, namely, poor discriminatory power and large computational load. In view of the limitations of the existing PCA-based approach, we applied a PCA on wavelet subband of face image. Face image is decomposed into a number of subbands with different frequency components using the wavelet transform (WT). Low frequency subband at 4 level is used for PCA. After each decomposition level resolution of image decreases which decreases computation and increases speed. In this study, we used ORL database to test the performance of proposed method. Wavelet and PCA are used for feature extraction whereas support vector machine (SVM) and nearest distance classifier are used for classification of images. This paper is organized as follows. Section 2 introduces Wavelet and PCA which were used for feature extraction. In section 3 classification method, SVM is explained. Section 4 shows the experimental results and section 5 presents conclusion.