Measurement of Daily Human Activity Using Ultrasonic 3D Tag System: Robust Estimation of Position by Redundant Sensor Data

摘要

This paper describes methods for accurately and robustly estimating 3D position based on highly redundant sensor data from an ultrasonic 3D tag system that the authors have developed for measuring daily human activities. This paper compares a least median of squares method and a least square method, and clarifies that the former is effective for accurately estimating 3D position by eliminating an effect of outliners, while both are effective for robustly estimating 3D position even when occlusion occurs. The results of the experiment conducted in a daily living space for evaluating the effectiveness of the applied estimation methods are reported.