Dynamic Face and Upper-Body Emotion Recognition for Service Robots

Abstract

Dynamic face and upper-body emotional expressions are two of the main non-verbal communication modalities use by humans when communicating with each other. Service robots should be able to understand the emotions through these two modalities in-order to interact with humans more efficiently when providing services. This paper presents an approach to recognize the face and upper-body emotions through RGB-D (Kinect) sensor for service robots because it enables robust detection of the upper-body. Face and upper-body emotions are recognized by Linear Discriminant Analysis (LDA) and Pattern Recognition Neural Network (PRNN) classifiers.