Live Face Recognition and Its Application in Automatic Door Locking System

Abstract

This paper proposes an approach for live face recognition on Rasperbery Pi, and used for face recognition based door locking system. The face images are acquired by an camera and fed to live face detection and face recognition system. The face detection is based on Haar cascades to extract face features. Then the face is analyzed to check whether it is a live face or a fake face image fed from a printed picture or video containing face image sequences. In this study, we consider the information from the changes of pixel intensity in a series of acquired face images that reflect the chages of heart rate during the rhythmic contraction and relaxion of the heart. Once verified as a live face, the face image is fed to the face recognition process using eigenfaces and principal component analyis. The input image is compared with the database to check whether it belongs to a person in the database or not. If yes, the door is controlled to allow an entrance. The live face detection and recognition was validated on a database including 25 people with 400 training images. The average accuracy obtained approximately 80%. The hardware including camera and actuactors are with low expense but meet real time performance.