Real-time facial expression recognition for affective computing based on Kinect

Affective computing endows computers with the ability to observe and understand, thus generates a variety of emotional characteristics. Facial expression has continuous signal which is direct expression of emotion. Affective computing via facial expressions has aroused great interest in intelligent interactive and pattern recognition recently. RGB camera is sensitive to surroundings, especially to illumination conditions. 2D images are not robust to human faces which are 3D objects. In this paper, we propose a method that uses noisy depth data produced by the low resolution sensor for robust face recognition. We capture color and depth information by Kinect sensor, extract Facial Feature Points (FFPs) feature vector by Face Tracking SDK, and recognize facial expression by Random Forest (RF) algorithm. This method enables facial expression recognition implementation real-time in intelligent interactive.