Face Expression Detection on Kinect Using Active Appearance Model and Fuzzy Logic

One of interpersonal human communication is facial expression which is used to manifest emotions. If computers could recognize these emotional inputs, the interaction between human and computers will bemore natural. Currently, video game is one of the applications that use facial expression detection to increase natural interaction. Kinect is motion sensor for game controller that can track the detected face using Active Appearance Model (AAM). AAM is the method that adjusts shape and texture model in a new face, when there is variation of shape and texture comparing to the training result. The aim of this research is to detect facial expression by observing the change of key features in AAM using Fuzzy Logic. Fuzzy Logic is used to determine the current emotions based on prior knowledge derived from Facial Action Coding System (FACS). The experimental results on user-generated dataset show that: (i) FACS does not provide any information about the degree of muscle activation, thus the degree of the key features have to infer from specific dataset, (ii) detection accuracy of emotions depend on the kind of emotion itself.