Computer Vision Based Real Time Lip Tracking for Person Authentication

Abstract

Automatic Person Recognition and Authentication faces a huge challenge especially in high risk environment. Multimodal Biometrics do address this issue up to a certain extent. Voice is the simplest and distinct single modality and lip movement is unique for each speaker. Dynamic movement of the lips if tracked online while a Speaker is speaking can authenticate a speaker in real time and the imposter can be identified thus making the biometric system robust and secure. This paper detects the lip movement of a Speaker, creates contours around the mouth region in real time and creates data base for a Speaker. Using the USB webcam, face detection was done from which the mouth region was segmented and contours were created around the lip area with the help of random points selected on both the upper and lower lips which were mathematically obtained. Based on these points of movement, a database was created for an individual. The time taken to create the speaker’s lip contour database takes hardly a few seconds which is the advantage and requirement for Real time applications which we have achieved.