Neuroflex: A Real-Time, Wearable Neuro-Muscular Rehabilitation System Using Soft Robotics

Abstract

Neuromuscular disabilities caused by strokes, paralysis, neurological damage, and muscle diseases could significantly decrease the functionality of the hand and upper limb and thus impede performing the activities of daily living. Even though the traditional rehabilitation method could help with restoring movement, the need for a therapist’s control and the frequent visits to the clinic could limit access to therapy. In this paper, the development of Neuroflex, an assistive wearable robotic rehabilitation device driven by EMG signals is discussed. The system presented uses EMG signals collected through hand-attached electrodes and the collection and pre-processing unit based on ESP32. The time-domain parameters such as RMS, MAV, and variance are calculated from the processed EMG signals. A Support Vector Machine (SVM) classifier is used to classify muscle activity at four distinct levels, which include: at rest, low activity level, medium activity level, and high activity level. The output from the classifier is converted into the parameters of the rehabilitation process and sent to the Arduino Uno to control the MG995 servo motor that works with the 3D printed hand exoskeleton through the Pulse Width Modulation (PWM). A web-based interface is developed to monitor and visualize EMG parameters and system parameters. Experiments have confirmed the operational functionality of the proposed sensing, classification, and robotic assistance scheme.