Real-time facial expressions recognition system for service robot based-on ASM and SVMs

Abstract

A real-time facial expressions recognition system is developed for human-robot interaction of service robot. The proposed system is mainly composed of two subsystems: one for Active shape model(ASM) motion extraction, and one for the classification of the estimated motion. The system first uses a cascade classifier to locate the potential face regions from video frame. Then, ASM is automatically initialized in the candidate regions. Based on the statistical property of deformable ASM, some facial features are extracted by using real-time pyramid ASM fitting method. The geometrical displacement between the estimated ASM feature node coordinates and mean shape of ASM is fed into the recognition subsystem in which, a Support Vector Machines (SVMs) is proposed to classify the calculated landmarks' relative motion. Experimental results show the potential performance of our facial expressions recognition system.