Grasp control of a prosthetic hand through peripheral neural signals

Abstract

The use of neural electrodes to stimulate the Peripheral Nervous System (PNS) of upper limb amputees is giving promising results in restoring tactile feedback. The same interfaces could be used to record the motor activity originated from the brain and transferred to the muscles. In this paper, the possibility to control a prosthetic hand by means of neural signals acquired through tf-LIFE4 electrodes implanted in a human subject was investigated. A Support Vector Machine (SVM) algorithm was adopted to classify two common demanded grasps. The obtained classes were converted into reference positions for a position-and-slippage control strategy that guarantees to perform stable grasps with a prosthetic hand avoiding slippage events. The achieved results showed an accuracy of the classifier higher than 90% and a success rate of the control strategy equal to 100%.