Estimation of daily forearm and wrist motion from shoulder and elbow kinematics by using artificial neural networks

Multifunctional arm prostheses have been developed since last several decades. One of the major problems that cause the loss of interest in current prostheses is the inadequate control interface between the patient and the prosthesis. The purpose of this research is to investigate the effectiveness of applying the kinematic data of the shoulder and elbow joint to control the arm prosthesis. In this paper, we propose an artificial neural network (ANN) technique to estimate the forearm and wrist motion pattern from shoulder and elbow kinematics for the control of arm prostheses. A number of activities which are essential and frequently performed in daily living are considered here and the proposed multilayer ANN is applied to classify them using the shoulder and elbow kinematics.