We propose a new approach of person identification using lip motion features for access control. This approach focuses on the information of the lip motion, whereas conventional approaches utilize still images which is containing the whole human face and are based on methods like template matching.
The testee is asked to make a specific utterance,and the image sequence containing the mouth motion is taken by a video camera. For each frame of the image sequence, lip color area is extracted by HSL transform, and the circumrectangle of the mouth is determined.
As a result of the above process, the opening ratio of the mouth (width versus height of the mouth) is estimated from the obtained bounding box of the mouth for each frame, and the sequence of the opening ratio of the mouth is registered as visual password.
A person is identified by comparing the acquired sequence of the opening ratio of the mouth with the visual passwords in a database by DP matching. The experimental result shows that this method realizes good person identification when length of utterance is not short.
more infomation of this research
To Home Page of Chihara-Lab