Human hand prosthesis simulation and control using electromyography signal classification by least-squares boosting ensemble

The human hand is an essential tool to human beings that allows the most effective interaction with the environment through its manipulation. Several conditions such as accidents and diseases can result in the loss of this limb, leading to limited motion and handling abilities. As a response, biomedical technology development of several devices has been proposed to contribute to the rehabilitation of patients who suffer the loss of the distal upper sections of their upper extremities. Because the active prosthesis is one of the most effective manners to recover the environment manipulation, it is necessary to include signal processing methods to use electrophysiological signals as the driving information for enforcing the hand mobilization. Hence, this research explores the development of a least-square boosting algorithm to categorize electromyographic signals that are used to define the controlled motion of a virtual active human hand prosthetic device. The proposed prosthetic hand was designed in Solidworks with motion characteristics that able it to perform five different movements associated with gasping. The input of the proposed algorithm takes five different kinds of electromyographic signals taken from the forearm that were pre-processed to complete the feature extraction task and the corresponding classification. The developed least-squares boosting ensemble was employed for the signal classification achieving an accuracy of 84.7 %. The results of the classification algorithm trigger one of the five possible trajectories that will be performed by a Proportional Integral Derivative controller in the Simulink environment.