Face Identification using HOG-PCA Feature Extraction and SVM Classifier

Face recognition is widely used in biometric applications, computer vision and surveillance systems. However, its effectiveness diminishes significantly when they are used under un-controlled environments such as illumination change situations, face position and expressions changes. Therefore, it become important to evaluate the performance of different feature extractions methods for its further incorporation to a Face Recognition Systems. This paper focuses on real-time face identification using Principal Component Analysis (PCA) and the Histograms of Oriented Gradients (HoG) descriptors combined with the Support Vector Machine (SVM) classifier. Feature extraction is a critical step inface recognition operations. in this paper, we compare between two feature extraction techniques. Our testing used Olivetti Dataset containing 40 persons; each person has 10 images in a different direction of the face, totally our dataset contains 400 images. The experimental result shows that PCA combined with SVM classifier provides better results in terms of training, features extraction time and the classification accuracy. The PCA is considered as an important feature method, particularly in Eigenfaces technique for almost all the face recognition algorithms.