Effect of Data Selection on the Loss of Balance in the Seated Position

Abstract

A fast and accurate detection of the loss of balance (LOB) can reduce the elderly fall that is one of the most dangerous injuries. Real-time detection of body reactions against a fall will be used to improve the performance of devices for the elderly. The loss of balance is detected by control error anomaly (CEA) that is an unusually large value of the system control error. Three men took part in the experiment. Each volunteer was seated on a four-legged chair and he balanced the chair over its rear legs by his foot as long as possible. This system had single input and single output. The input was the foot's ground reaction force and the output was the angular acceleration of the chair. An internal model of balancing task was built of a linear equation of regression from the relationship between input and output. The output of the internal model was assumed as a control signal of the angular acceleration from the central nervous system. Data selection depends on the size of the foot reaction force data, S1, which were used to construct the internal model and the threshold, S2, which were used to detect the LOB. The size of the force data changed from 0.2 to 3 sec and also the threshold size changed from 0.02 to 0.2 sec. Therefore in the case of detecting the loss of balance under 3-sigma method, the detecting performance was improved in lager S1 and smaller S2. © 2009 International Federation of Medical and Biological Engineering.