Human lower limb motion recognition based on translation invariance wavelet transform and RBF neural networks

The effective de-noising of gait kinematic signals is the prerequisite and guarantee for correct recognition and diagnose. Traditional Fourier transform and wavelet analysis can introduce the additional disturbance during de-noising process named pseudo-Gibbs phenomenon. In this paper, translation invariance wavelet de-noising method is proposed to process the kinematics information acquired from inertial sensors mounted on the lower limb of human. This way, pseudo-Gibbs phenomenon was inhibited effectively and high precision classification of human lower limb motion pattern was achieved by combining the propose de-noising method with radial-based function (RBF) neural network. Experimental results demonstrated the effectiveness and correctness of the proposed system.