Continuous identification of gait phase for robotics and rehabilitation using microsensors

Using microsensors for the robust and accurate analysis of human posture or gait is an interesting opportunity for rehabilitation and robotics applications. This paper describes a feasibility study in which the possibility of using a new type of embedded microsensors, based on the coupling of accelerometers and magnetometers, and developed by CEA/LETI is investigated. This study consists in identifying what part of the gait cycle is active by using a reconstruction of the knee joint angle by two microsensors fixed on tibia and thigh, during a steady-state sagittal walk. More than just an identification of a few gait states, this approach allows us to continuously extract the current position on the gait cycle. We compare the reconstructed knee joint angle with a stored reference taking into account uncertainties on the velocity and perturbations of the terrestrial magnetic field. To accurately identify the phase of the gait movement, we fuse different simple and complementary methods: morphomathematics, cyclogram analysis, wavelet transform, qualitative analysis, crosscorrelation. These results encourage us to extend this work to explore the possibility of recognition of a larger set of human movements using more sensors and improved algorithms of signal processing