Human motion recognition and prediction using loose cloth

Abstract

Abstract Human motion analysis plays a crucial role in fields such as healthcare, human-robot interaction and virtual reality. Conventional approaches typically rely on tightly attached body sensors, which can prove uncomfortable and impractical. Here, we investigate motion recognition and prediction using garments incorporating embedded sensors. We analyse how the movement of loose-fitting clothing can predict body motion in both simulated and real-world scenarios. Results demonstrate that sensors attached to fabric can improve recognition accuracy by up to 40% improvement and require approximately 80% less movement history compared to sensors directly attached to the body. These findings indicate that garment motion provides valuable information for analysing human movement. The study additionally offers insights regarding the design of intelligent textiles with integrated sensing capabilities.