Abstract
The purpose of this study is to define design variables of a detection algorithm of the loss of balance (LOB) with a balancing task in the seated posture. The design variables were determined through the sensitivity analysis to attain the highest success rate of the algorithm. The detection algorithm mainly consists of an internal model of the central nervous system and a control error anomaly detector. This study selected five time variables: two variables S1 and h1 were related to the internal model, another two S2 and h2 to the detector, and the other hd to both the model and the detector. S1 and S2 represent data selecting windows; h1 and h2 time shifts; and hd an operating period of the LOB detection algorithm. The success rate was increased by up to 10% comparing with the previously published case when S1 and S2 were set to 2.0〜8.0 sec, h2 to 0.01〜0.10 sec, and hd to 0.1〜0.34 sec. The results also showed that the success rate was insensitive to h1. This study suggested a scheme of improving the algorithm to detect the loss of balance.
