The development and evaluation of a sensor-fusion and adaptive algorithm for detecting real-time upper-trunk kinematics, phases and timing of the sit-to-stand movements in stroke survivors

Abstract

Low-cost wearable inertial sensors and balance plates offer great opportunities to provide body kinematic and spatial measurements of mobility-related activities, such as the sit-to-stand (STS) motion, a crucial movement to activities of daily living. This abstract presents the development of a Kalman-filter based sensor fusion algorithm with error compensation for detecting upper-trunk kinematics and a finite state machine based adaptive algorithm, which aims to analyze and detect crucial events, the transition of phases and timing of the movement. Both methods were tested on stroke survivors. The results show the sensor fusion algorithm has excellent correlation coefficients and contains very small errors in estimating rotation angles and velocities while the adaptive algorithm had a small bias and consistent delay in detecting the transition of phases.