Abstract
We propose the concept of Active Human Interface (AHI) that makes the machine (computer and/or robot) respond to human being more actively and for establishing the new paradigm to realize the AHI, as the first step of this study, we investigate the method of machine recognition of human emotions.
This paper deals with the neural network method of human emotion recognition from facial expressions. Facial expressions were categorized into 6 groups (Surprise, Fear, Disgust, Anger, Happiness and Sadness), and obtained CCD camera-acquired data with respect to facial characteristic points relating to 3 components of face (Eyebrows, Eyes and Mouth). Then we generated the position information and shape information about the 6 basic facial expressions for 30 clients. These information were input into the Input units of the 4-layered neural network and network learning was carried out by back propagation algorithm. The neural network recognition system of facial expressions showed a high recognition rate up to 80% to 6 basic facial expressions for both the position and shape information and particularly the system showed a smaller rate of mis-recognition between some of 6 basic expressions.
