Face recognition using BFSS features

摘要

Face recognition is a field which has risen to prominence because of its tremendous range of applications. Traditional biometric systems relied heavily on user interaction. Face recognition addresses this issue. It involves no user interaction. Modern security systems have seen an uncharacteristically rapid growth in the use of face recognition based systems seldom seen hitherto. Face recognition is used in other applications as well. Facebook makes use of face recognition for effective photo tagging. Gaming consoles like the PS4 and the Xbox360 use it as a means of authenticating users. Microsoft Kinect is another gaming console which uses face and image recognition as a part of its advanced motion sensing feature. Conventional techniques use SIFT (ScaleInvariant Feature Transform) or SURF (Speeded Up Robust Features) for feature detection and extraction. This is followed by matching of key points. The performances of these techniques are limited by a number of factors including lighting and resolution. This paper is an attempt to ameliorate the performance of these techniques using the FAST (Features from Accelerated Segment Test) detector, SVD (Singular Value Decomposition) for dimension reduction, SURF for feature extraction and FLANN (Fast Library for Approximate Nearest Neighbors) for matching. The results of the simulations against 6 datasets have been tabulated. It is observed that the technique used is more efficient in face recognition when compared with the classical techniques.