Affect Recognition Using Dynamic Characteristics of Motion

Abstract

Recently, there has been a lot of work on affect recognition which is a relatively new research field. Here the detection and recognition of multiple emotions such as anger, joy, fear, etc. are being carried out. Furthermore, the research on affective computing is being conducted using various modalities, such as; facial expressions, bodily expressions, and speech, with facial expressions being the most researched modality. However, many researchers have lately highlighted the significance of bodily expressions in affect detection. In this paper, we have considered bodily expressions for recognizing affect and proposed a modified approach using dynamic characteristics of motion. The experiments were performed in MATLAB using the MPIIEmo dataset and results were compared with the existing research.