Bodily Expression for Automatic Affect Recognition

This chapter focuses on the why, what, and how of bodily expression analysis for automatic affect recognition. It first asks the question of ‘why bodily expression?’ and attempts to find answers by reviewing the latest bodily expression perception literature. The chapter then turns its attention to the question of ‘what are the bodily expressions recognized automatically?’ by providing an overview of the automatic bodily expression recognition literature. The chapter then provides representative answers to how bodily expression analysis can aid affect recognition by describing three case studies: (1) data acquisition and annotation of the first publicly available database of affective face-and-body displays (i.e., the FABO database); (2) a representative approach for affective state recognition from face-and-body display by detecting the space-time interest points in video and using Canonical Correlation Analysis (CCA) for fusion, and (3) a representative approach for explicit detection of the temporal phases (segments) of affective states (start/end of the expression and its subdivision into phases such as neutral, onset, apex, and offset) from bodily expressions. The chapter concludes by summarizing the main challenges faced and discussing how we can advance the state of the art in the field.