Date of Award

6-22-2021

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Systems Engineering

First Advisor

Kamran Iqbal

Abstract

The upper limb is a complex part of the human body, and it is necessary to carry on daily tasks and maintain self-dependence. As a result, the daily life of people with upper limb amputation could not be as easy or as free as the more fortunate people without amputation. This work focuses on estimating the upper limb’s angular displacement, speed, acceleration, and torque at the elbow joint, it does so regardless of arm’s orientation and effort level. We do so to enable better control of the myoelectric prosthesis. This work complements the related existing literature, which usually does not count for variations in limb position or load. The developed estimator uses the neural command coming from the Central Nervous System (CNS) to the upper limb muscles for its estimation. This command is measured at the skin surface using the surface Electromyography (sEMG) Technique. The low signal-to-noise ratio (SNR) in sEMG makes it hard to translate it into useful controlling command, therefore we implement different preprocessing and filtering techniques before we compress the data. We discuss three approaches in dimension reduction and how they affect the speed and accuracy of the estimator. The estimator is an ANN-based Softmax classifier at its core, which has proven to have high accuracy in estimating the elbow joint angle. Mated with a proper dimension reduction technique, the estimator shows an estimation accuracy of 97% during offline estimation and 92.28% in online estimation with an average estimation speed of 20 microseconds. The high accuracy and short estimation time make the estimator suitable for prostheses control and other real-time applications, like exoskeletons, remote-controlled robotic arms, and other numerous Human Machine Interface (HMI) applications.

Download

Included in

Robotics Commons

Share

COinS