Hand motion recognition based on pressure distribution maps and LS-SVM

This paper presents a novel hand motion recognition method to achieve the volitional control of a multi-DOF prosthetic hand. Based on the principle that the pressure distributions on the surface of forearm caused by the muscular activities are different when different hand motions are performed, a human-machine interface (HMI) composed of 32 force sensitive resistor (FSR) sensors, measurement unit and recognition unit is designed to measure and classify the pressure distribution maps (PDM) of 9 typical hand motions. The least-squares support vector machine (LS-SVM) is adopted as the pattern recognition algorithm to classify the selected movement modes. The multi-subject experiment results show that the method can identify each motion with high accuracy.