Performance optimization of a motor unit decomposition algorithm for real-time neural interfacing applications
Loading...
URL
Journal Title
Journal ISSN
Volume Title
School of Electrical Engineering |
Master's thesis
Unless otherwise stated, all rights belong to the author. You may download, display and print this publication for Your own personal use. Commercial use is prohibited.
Authors
Date
2024-11-24
Department
Major/Subject
Electronic and Digital Systems
Mcode
Degree programme
Master's Programme in Automation and Electrical Engineering
Language
en
Pages
67
Series
Abstract
The human neuromuscular system relies on muscles and their subcomponents, the motor units, to generate force and enable movement. The contraction of these motor units is triggered by electrical impulses, which can be measured on the surface of the skin via electromyography (EMG). This thesis focuses on optimizing a motor unit decomposition algorithm for real-time interfacing applications. While high-density EMG decomposition is well-established, it has not yet been integrated into clinical practice. Two main factors hinder this progression. The first is a lack of fundamental research into the characterization of neuromuscular disorders using motor unit decomposition. The second factor is the currently suboptimal runtime performance of existing decomposition algorithms, making real-time application impossible and slowing the iterative research process. To address these challenges, this work aims to improve the algorithm’s speed while maintaining or enhancing its accuracy. Additionally, a real-time extension is introduced to enable decomposition for applications such as real-time prosthetic control. To achieve these improvements, several strategies were explored, including guess count optimization, parallelization, and just-in-time compilation. Furthermore, transitioning the algorithm from matlab to Python was carried out to increase accessibility and integration with external tools. For the real-time extension, the algorithm was divided into a calibration phase that returns pre-calculated values. These values then enable a streamlined version of the algorithm to run in real time using a rolling window process. The results demonstrate a significant reduction in calibration time alongside a functional real-time extension. These advancements broaden the range of clinical applications beyond those of previous methods while also facilitating faster algorithm refinement. Furthermore, the findings indicate that the optimized algorithm achieves the processing speed required for real-time prosthetic control, paving the way for more intuitive and responsive neuroprosthetic devices.Description
Supervisor
Vujaklija, IvanThesis advisor
Taleshi, MansourKeywords
motor unit decomposition, real-time neural interface, high-density electromyography (HD-EMG), algorithm optimization, neuromuscular disorders, neuroprosthetic control