Indian Journal of Science and Technology

Abstract

In day to day, the Object Liveness Detection and Genuine Face recognition became important in many real time applications such as security systems with user authentication, live video processing, Object Identification, Object Recognition and many. Most of the existing systems for Face recognition and anti-spoofing classifier is accomplished to detect object features and finds spoofing attacks on all subjects. However, by considering the individual differences among several objects, the basic classifier cannot simplify well to all subjects. In this work, the proposed system allows to select specific object based on Region of Interest (ROI) and extract features of ROI, then recognises face and later check for spoofing attacks using Fuz-SVM classifier specifically trained for each object, which avoids the interferences between several objects. Moreover, by considering all possible rare and uncommon fake samples for training, we propose a combined Histogram of Oriented Gradients with Local Phase Quantization (HOG-LPQ), which makes it practical to train well performed individual Object to its certain face with liveness detection and, the proposed system includes not only processing of specific selected object and also extracts features of blurred images. We performed experiments on various real time objects with exiting data base; the details are discussed in the prospect of the proposed approach.

Keywords: Authentication, Fuz-SVM, Genuine, Histogram of Oriented Gradients, Local Phase Quantization, Liveness Detection, Region of Interest (ROI),