Multimodal indoor social interaction sensing and real-time feedback for behavioural intervention

Abstract

Social interactions play an important role in people's personal as well as working life. Interactions come in various forms, identifiable mainly by duration and proximity. The ability to detect and distinguish interactions can often shed light over worktask performance, epidemic spreading, personal relationship development, use of space and more. Questionnaires and direct observations have often been used as mechanisms to identify interactions, however, these are either very expensive in terms of staff time, yield very coarse grained information or do not scale. Technology has started cutting costs by allowing automatic detection, however precise interaction identification often requires individuals to wear custom hardware. The aim of my work is to exploit the capabilities of off-the-shelf wearable devices (i.e. smart watches and fitness trackers) to build a social interactions sensing platform which offers accuracy and scalability. To this end, non-verbal behaviours, such as, body language, will be considered in addition to the occurrence of the interactions (individuals involved, duration and location) with the objective of providing unobtrusive real-time feedback.