Improved watershed lips detection and modified H8 tracking system based on Lyapunov stability theory

Abstract

This paper presents an automatic lips detection and tracking system using the improved watershed algorithm and a new modified H8 tracking system based on Lyapunov stability theory (LST). A direct lips detection approach using watershed segmentation without preliminary face localization is proposed. The watershed lips detection system is improved to get rid of the face rotation circumstances. Furthermore, the conventional H8 filter is modified according to the LST to ensure that the lips tracking algorithm is always in the stable condition and the tracking error would be asymptotically converge to zero as the time approaching to infinity. The proposed system is evaluated and the simulation results reveal an appreciable performance. © 2009 IEEE.