Lip tracking using deformable models and geometric approaches

Abstract

Multimodal biometrics addresses the issue of recognizing and validating the identity of a person; however, the issue is for a single modality to be robust enough. Voice, being a simple biometric feature to acquire, and the accompanying movement of the lips being distinct for every person can stand up to this challenge. Tracking Lip Movement in real time can be an important biometric trait for Person Recognition and Authentication. A biometric system should be robust and secure especially when they must be deployed in vital domains. In this paper, we have used three different methods to draw the contours around the lips to detect the lip edges to establish lip movement. In the first method, dynamic lip edge patterns were drawn and simultaneously saved in a database created for each person. Secondly, Lip contours were created using Active Snakes models while in the third technique, the motion extraction of lip was implemented using edge detection algorithm. In the work presented here, we have segmented the lip region from the facial images, and have implemented and compared three different approaches to contour the lip region. However, no single model fits every application due to the various face poses and unpredicted lip movements. The target application should be the deciding factor for the consideration of lip movement as the biometric modality along with the voice biometric trait.