Non-invasive neural interfacing for wearable electromyographic systems
File(s)
Author(s)
Mendez Guerra, Irene
Type
Thesis
Abstract
In our ever-growing digital world, our interaction with technology is constantly evolving, pursuing more intuitive and immersive experiences. This has led to the exploration of neural interfaces, a technology initially developed for rehabilitation purposes, for general consumer electronics. Neural interfaces promise a seamless interaction with technology by translating how our nervous system encodes intentions and movements. These interfaces can be established at various points along the neural pathway, with varying degrees of invasiveness. Unlike in clinical applications, non-invasive technologies are essential to address ethical concerns in widespread consumer use. One promising approach to meet these requirements involves detecting neural signals non-invasively at the terminal layer of the nervous system, where they are transformed into muscle forces and actions. This is achieved through surface electromyography (EMG) signals, which capture muscle electrical potentials. However, for advanced control paradigms, it’s essential to identify individual motor neuron activity rather than collective muscle signals. Recent break-throughs in decoding algorithms have made this possible, offering great potential for non-invasive wearable interfaces. However, current decoding algorithms have not been applied to the wrist, a prime location for wearable devices that shows unique electrical properties. In addition, these algorithms are designed for static conditions rather than natural dynamic movements. In this thesis, I address these challenges to develop an accurate, robust, non-invasive neural interface for the wrist, suitable for consumer wearables. Here I validate that motor neuron activity can be decoded from EMG signals recorded at the wrist, demonstrating its ability to accurately predict finger gestures in real-time. I also investigate how joint movements affect the decoding of motor neuron activity and introduce online learning metrics to adapt to dynamic conditions. These advancements enhance our understanding of non-invasive motor neuron interfacing at the wrist, bringing us closer to achieving seamless, intuitive human-computer interaction for general consumer applications.
Version
Open Access
Date Issued
2024-02-07
Date Awarded
01/05/2024
License URL
Advisor
Farina, Dario
Barsakcioglu, Deren Y.
Wetmore, Dan
Sponsor
Engineering and Physical Sciences Research Council
Meta (Firm)
Publisher Department
Department of Bioengineering
Publisher Institution
Imperial College London
Qualification Level
Doctoral
Qualification Name
Doctor of Philosophy (PhD)