Bouncing Seat: An Immersive Virtual Locomotion Interface with LSTM Based Body Gesture Estimation

A pilot study of the bouncing seat system is presented. This system consists of an air cushion and gravicorder. The former ena-bles users to move easier and feel lively motion. The latter and its computing units detect users' body sway and recognize its gestures. Four commands (walk, right turn, left turn, and jump) and associated intuitive body gestures were designed. The estimator based on the multi-timescale LSTM was trained using newly cre-ated body gesture dataset and achieved 88% inference accuracy. Results suggest that our system has potential for providing a high-er sense of immersion and enjoyment than the joysticks.