MakiShell: Voice Interface with Neural Networks for Controlling a Hand Rehabilitation Exoskeleton with Hemiparesis

This article presents the design of “MakiShell,” a voice interface within a mobile application that uses a multilayer perceptron as a neural network to control a hand rehabilitation exoskeleton, targeting patients who have experienced hemiparesis following a stroke episode. The work demonstrates the feasibility of the multilayer perceptron in recognizing voice commands through an analysis of arrays based on the ASCII encoding of each word, with the purpose of individually controlling the finger actuators. Additionally, by enabling the creation of customized movement routines by therapists, more effective rehabilitation will be achieved by focusing on muscles that require greater attention. The research results will guide the development of more personalized rehabilitation tools, employing neural networks to significantly improve the quality of life of patients with multiple disabilities.