VSig: Hand-gestured signature recognition and authentication with wearable camera

Abstract

Wearable camera is gaining popularity not only as a recording device for law enforcement and hobbyists, but also as a human-computer interface for the next generation wearable technology. It provides a more convenient and portable platform for gesture input than stationary camera, but poses unique challenges due to user movement and scene variation. In this paper, we describe a robust wearable camera based system called VSig for hand-gestured signature recognition and authentication. The proposed method asks the user to virtually sign within the field of the view of the wearable camera. Fingertip is segmented out and tracked to reconstruct the signature. This is followed by signature matching for authentication with the pre-stored signatures of the individual. A dataset named SIGAIR comprising of hand-gestured signatures from 10 individuals has been created and used for testing. An average accuracy of 97.5% is achieved by the proposed method.