[2K4-GS-9] Human interface

[2K4-GS-9-05] Development of a Method for Estimating Whole-body Posture during Walking from Gait Indices Measured Using Smart Shoes

〇Toshiaki Okamoto¹, Shunsuke Yamagata¹, Satoru Abe¹ (1. ASICS Corporation)

Keywords:Pose estimation, Time series analysis, Neural network, Principal component analysis, Gait

Pose estimation from a video is used in various applications such as form analysis in sports and estimation of fall risks in daily walking. However, it is not easy to record yourself continuously without the help of others or setting up a self-shooting environment. This study aims to develop a method to estimate the whole-body posture while walking with smart shoes which incorporate built-in motion sensors. Thirteen subjects wore the smart shoes, and walked straight for 5 meters while they were being filmed with a Kinect camera. Smart shoes data was regressed to principal component scores reflecting walking motion of the entire body, and the predicted principal component scores were reconstructed into anatomical landmarks. This technology enables estimation of walking posture in daily life by simply wearing smart shoes, without setting up a video recording environment.