Automatic Facial Feature Points Extraction and Expression Recognition Based on Video Database

Abstract

In this paper, we propose a facial expression recognition method using Support Vector Machine on mapped coordinate sequence features. The proposed system is almost fully automatic, in which landmark initialization is based on general knowledge with edge information, and missing information compensation is done by ASM. The geometric features of facial expressions were extracted from sequences of facial landmarks. Validation experiments were conducted using facial expression sequences extracted from a video based facial expression database.