Uncovering Neural Commands from Noisy Dermal Signals—an Experimental Verification of a Minimalistic Robotic Exoskeletal Hand Design for Medical Rehabilitation

Abstract

Exoskeletal medical rehabilitation devices require the interpretation of neural movement signals via surface electromyographic sensors. In many cases, the signals are imbedded in noise (due to underlying pathologies of the rehab patient) and must be processed such that movement commands can be processed instantly. This paper will report on a prototype design of a device that addresses many of the issues found with previous devices, including the size as well as providing insufficient assistance force, and associated signal processing logic. The prototype addresses these issues by using a tendon drive mechanism actuated by a power screw, which will have the ability to restore mobility to each finger of the user while at the same time providing a low profile and low operating power. The following features were integrated into the system: motor position control and characterization, a LabVIEW controller interface, and an EMG control scheme. The prototype was tested on five healthy participants and a cerebral palsy patient. Conducting computer simulations and testing of a physical prototype reveal that the exoskeleton design is viable. Furthermore, a viable controller was developed for the device and validated experimentally through multiple EMG filtering methods by using discrete wavelet transforms and bandpass filtering. This study also discovered that by using discrete wavelet transforms to filter EMG signals, it is possible for an individual with cerebral palsy to control the device. This paper is an application of data-enabled discovery tools in the field of medical orthotics. It provides guidance to readers relative to the use of signal processing techniques as applied to problems in injury rehabilitation.