Inertial sensor location analysis in automatic balance assessment

This paper presents the study on the automatic computer-aided balance assessment system. The approach employs ambient assisted living architecture based on inertial sensors. The assessment relies on selected activities from the Berg Balance Scale test by means of the system supplied by a set of features extracted from the accelerometric signals. The feature space dimensionality reduction stage selects the feature vector for each activity and sensor combination using the Fisher´s linear discriminant. The system is trained and evaluated using expert scoring of examinations involving 52 patients with different balance abilities. The investigation on the system performance with various sets of available sensors is presented and discussed.