Chromatic lip tracking using a connectivity based fuzzy thresholding technique

Abstract

This paper describes a new robust method for extracting visual labial information based on chromatic information. Using chromatic temporal information for a speaker and a special fuzzy based thresholding algorithm a stable technique of extracting a speaker's lips is proposed that overcomes many of the problems associated with previous adaptive threshold techniques. Experiments performed thus far show encouraging results while being computationally inexpensive thus lending the entire process well to a robust real time application.