Development of a feedback training system with visualization of motion information using self-organizing map

Abstract

Motor skill trainings are widely used to augment human performance in various area such as sports, medical or welfare. We have previously proposed a visualization method to integrate significant amounts of information relative to human motion to facilitate convenient visual perception during motor learning. This motor learning system helps subjects acquire developed motor skills by referencing integrated information of optimized motion data using a visualized motor skill map. In the proposed method, a self-organizing map (SOM) is employed to visualize the integrated motion data. This map has an ability to express a difference between a teacher signals and measuring signals, as a distance between trajectories corresponding to various time-series motion data on the map. However, it is difficult to use for training system because drawn teacher trajectories are randomly arranged on the map by original algorithm of SOM. The aim of this study is to establish a novel training system using SOM by modifying its algorithm which is capable of controlling a drawn trajectory to compare between a teacher signals and measuring signals easily. To verify the validity of the proposed training system using modified SOM, we evaluated the difference between teacher and measured signals after proposed training. The results of experiments show that our visual feedback system enabled subjects to close whole teacher motion through trial and error training.