Face Recognition based on Histogram of Oriented Gradients, Local Binary Pattern and SVM/HMM Classifiers

Face recognition is one of the challenging biometric technologies which has widespread applications in many fields such as access to security systems, identification of a person in law enforcement, identifying the culprit during riots, breach of security etc. In many of the face recognition techniques, the unique features of the face image are extracted and compared with the images of the database to produce better success rates. In this paper we take into account both shape and texture information to derive feature vector based on Histogram of Oriented Gradients (HOG) and Local Binary Pattern. In both algorithms, the face is divided into small regions and features are extracted. The performance of different algorithms with Support Vector Machine (SVM) and Hidden Markov Model (HMM) classifier are compared. It is found that the concatenation of feature vector derived from HOG and LBP with SVM as classifier has produced better result of 99.4% of recognition rate.