Landmark identification in 3D image for facial expression recognition

摘要

Facial expression recognition plays a major role in non verbal communication. Recognition by machine is still a challenging problem. To automate the recognition for human machine interaction, a system is proposed in this paper. The proposed system uses shape descriptors to identify twelve land marks which mainly contribute to the facial expression recognition. From the location and the size or boundary of the land marks by matching with Facial Landmark Model (FLM), basic expressions are identified. The experimental results show that the shape descriptors and post processing correctly identifies landmarks automatically. The architectural distortion of action units is used to identify the basic facial expressions and tested on Bosphorous data set.