Public defence in Bioelectronics and Instrumentation, M.Sc.(Tech.) Dennis Yeung

The title of the thesis: Algorithms for robust human-machine interfacing via surface electromyography

Doctoral student: Dennis Yeung
Opponent: Prof. Cristoano De Marchis, University of Messina, Italy 
Custos: Prof. Ivan Vujaklija, Aalto University School of Electrical Engineering, Department of Electrical Engineering and Automation

The surface electromyographic (sEMG) signal is a measurement of the electrical activity associated with muscular contraction. This technology has been extensively used in the study of human motor control strategies, muscular force modulation, and motor dysfunction. As a medium for human-machine interfacing, sEMG has primarily been utilised in the control of robotic prostheses and exoskeletons, with recent applications emerging in the domain of wearable consumer electronics that support gesture recognition. Thus far, a major cause of performance degradation in such interfaces arises from signal non-stationarities induced by various activities of daily living. As such, algorithms designed to overcome these challenges are investigated in this thesis work.

This thesis presents extensions to regression-based algorithms that facilitate continuous control across multiple degrees-of-freedom. Namely, a directional forgetting scheme for applying recursive least squares updates on model parameters and an adaptive extension to control models based on non-negative matrix factorisation are presented. Validation of these methods includes evaluations using real-time virtual control tasks and standardised clinical tests conducted with able-bodied and differently-limbed subjects.

Furthermore, contributions toward the extraction of motor unit spike trains from high-density sEMG via blind-source separation techniques are presented. Motor unit firing times represent the finest encoding of muscle force generation and access to such information may enable more intuitive interfacing systems. Here, methods for integrating decomposed motor unit activity into dexterous interfacing systems are described, alongside approaches for adapting decomposition parameters and respective experimental validations. These contributions thus aim to advance the development of sEMG-driven human-machine interfaces that are both robust and highly functional.

Keywords: algorithms, electromyography, human-machine interfaces

Thesis available for public display 10 days prior to the defence at: https://aaltodoc.aalto.fi/doc_public/eonly/riiputus/

Contact:

Doctoral theses in the School of Electrical Engineering: https://aaltodoc.aalto.fi/handle/123456789/53