Абстрактный

Facial Expression and Visual Speech Based Person Authentication Using Local Binary Pattern Histogram

S. Saravanan, S. Palanivel, M. Balasubramanian

This paper is an important milestone towards automatic person authentication. In this work, from a video source, video frames with pose free face is automatically sensed. Then mouth region is localized automatically. Then person identification is done using local binary pattern histogram of mouth region. Person identification performance while using neutral face, smile expression and visual speech are compared. A big custom dataset of 180 videos of 30 persons are used. The result gives the conclusion that for person identification visual speech performs well with F1 score of 0.95, followed by neutral face, 0.93 and then smile expression, 0.87. It also concludes that the mouth component of the face itself is efficient enough to identify persons.