Sensing Social Behavior With Smart Trousers

Abstract

Nonverbal signals play an important role in social interaction. Body orientation, posture, hand, and leg movements all contribute to successful communication, though research has typically focused on cues transmitted from the torso alone. Here, we explore lower body movements and address two issues. First, the empirical question of what social signals they provide. Second, the technical question of how these movements could be sensed unintrusively and in situations where traditional methods prove challenging. To approach these issues, we propose a soft, wearable sensing system for clothing. Bespoke “smart” trousers with embedded textile pressure sensors are designed and deployed in seated, multiparty conversations. Using simple machine learning techniques and evaluating individual and community models, our results show that it is possible to distinguish basic conversational states. With the trousers picking up speaking, listening, and laughing, they present an appropriate modality to ubiquitously sense human behavior.