3D Human Face Analysis: Automatic Expression Recognition

A 3D automatic facial expression recognition procedure is presented in this work. The method is based on point-by-point mapping of seventeen Differential Geometry descriptors onto the probe facial depth map, which is then partitioned into seventy-nine regions. Then, features such as mean, median, mode, volumes, histograms are computed for each region and for each descriptor, to reach a varied large set of parameters representing the query face. Each set of parameters, given by a geometrical descriptor, a region, and a feature, form a trio, whose featuring numerical values are compared with appropriate thresholds, set via experimentation in a previous phase by processing a limited portion of the public facial Bosphorus database. This allows the identification of the emotion-based expression of the query 3D face among the six basic ones (anger, disgust, fear, joy, sadness, surprise). The algorithm was tested on the Bosphorus database and is suitable for applications in security, marketing, medical. The three-dimensional context has been preferred due to its invariance to different lightening/make-up/camouflage conditions.