Robust method to identify the speaker using lip motion features

Authentication system solely based on visual lip information is of advantage since the uttering characteristics/manner is unique to individual and difficult to imitate. A database containing the visual utterance of 50 speakers has been generated and each of the utterance is of duration 3 seconds. We consider three stages in this paper. In the first stage, the face region is obtained from image. In the second stage, mouth region extraction by back ground subtraction. In the third stage, considering the lip centroid as origin of coordinates, each pixel lip envelops is parameterized with polar (ordered from -π εα +π ) and Cartesian coordinates (ordered as widths). Finally, feature vector is stored in the database. For speaker authentication, feature vector is treated as template which is used to perform the analysis.