Classification of forearm movements based on kinematic parameters using artificial neural networks

Abstract

Human body motion tracking has been performed in the domain of diagnosis and therapy of movement disorders, facilitating human machine interaction and controlling wearable robots. Low-cost inertial measurement units are widely used in wearable robotics for tracking of upper/lower limb motions. Method for classification of forearm positions and movements based on training of feedforward artificial neural networks (ANN) is presented in this paper. The ANN input data were acquired using inertial sensors (3-axis accelerometer and gyroscope) placed on the forearm. The overall performance of forearm movement classification was 92.5%. The accuracy of classification of the forearm position was 81.2%.