부산대학교 도서관

Inertial sensors are widely used to measure human movement. Although inertial sensors have been successfully applied to exergaming in the past, the problem of detecting foot motions to interact with stepping exergames is still largely understudied. In this work, we developed a new method to detect and classify step directions relying on inertial sensor data captured by two shoe-mounted inertial sensors. Drawing on previous results, we developed a single multiclass classifier to distinguish front, back, side, and center steps originating from any of these positions. Since some of these steps exhibit similar displacement patterns, the previous step position was also considered as an input to the classifier. The method was tested on a group of young and older adults, achieving an accuracy of 93.1%. Performance remained consistent throughout the acquisition time due to the introduction of a novel calibration approach designed to handle sensor orientation drift over time. This study provided the first insights into the potential of inertial sensors to detect the foot motions required to interact with stepping exergames. Experimental results support their application in a real scenario.