Compressed Sensing Techniques for EEG Signals

aut.embargoNoen_NZ
aut.filerelease.date2021-04-01
aut.thirdpc.containsNoen_NZ
dc.contributor.advisorLi, Jack Xuejun
dc.contributor.advisorMoir, Tom
dc.contributor.advisorGriffin, Anthony
dc.contributor.authorDao, Phuong Thi
dc.date.accessioned2019-04-01T00:29:46Z
dc.date.available2019-04-01T00:29:46Z
dc.date.copyright2019
dc.date.issued2019
dc.date.updated2019-03-31T23:00:35Z
dc.description.abstractRecent advancements in information and communication technologies (ICTs) along with the development of Wireless Body Area Networks (WBANs) have inspired telemedicine, which features a novel method in delivering the healthcare service. Telemedicine allows healthcare providers to monitor, evaluate, diagnose and treat patients without an inperson visit. Recent advancements in electronic circuits, computer science and software engineering have made it possible for us to achieve digitization, storage, synthesis and transmission of all kinds of analogue signals, including biomedical signals such as electroencephalogram (EEG), electrocardiogram (ECG), and electromyogram (EMG) for health monitoring of the brain, heart and muscles, respectively. Thus, it can improve the quality of medical treatment both in emergency cases and long-term care. In recent years, compressed sensing has emerged as a potential compression technique to allow for high compression with reasonable reconstruction errors. It is based on the principle that the sparsity of a signal can be optimally exploited to recover it from far fewer samples than that is required by the Shannon-Nyquist sampling theorem. Most of current works in compressed sensing have been applied for ECG and the event-related EEG signals. There are only few of them taking clinical EEG signal into account. Thus, it motivated us to study the application of compressed sensing in clinical EEG signals. We first perform a comprehensive survey on state-of-the-art EEG compressed sensing schemes, and then study the effect of epoch length on EEG compressed sensing. Moreover, we study the effect of time and frequency resolutions of EEG signals and propose a framework to design an optimal overcomplete Gabor dictionary. Finally, we propose a method to learn a dynamic dictionary based on the K-Singular Value Decomposition (K-SVD) algorithm for compressed sensing of clinical EEGs. The performance of proposed schemes is evaluated using the scalp EEG database from Children’s Hospital Boston and Massachusetts Institute of Technology (CHB–MIT). This comprehensive database was recorded from 23 paediatric subjects, including total of 989 hours EEG recording. The numerical results show the outperformance of K-SVD dictionary in both reconstruction error and time. We conclude that compressed sensing with K-SVD learned dictionaries is a promising candidate for EEG recording and transmission in telemedicine systems. The learned dictionary and the reconstruction framework can be applied directly in the reconstruction stage. Moreover, the learned procedure can be applied to learn the patient specific dictionaries in order to apply in monitoring and early detecting of seizure.en_NZ
dc.identifier.urihttps://hdl.handle.net/10292/12401
dc.language.isoenen_NZ
dc.publisherAuckland University of Technology
dc.rights.accessrightsOpenAccess
dc.subjectCompressed sensingen_NZ
dc.subjectEEGen_NZ
dc.subjectK-SVDen_NZ
dc.subjectGaboren_NZ
dc.titleCompressed Sensing Techniques for EEG Signalsen_NZ
dc.typeThesisen_NZ
thesis.degree.grantorAuckland University of Technology
thesis.degree.levelDoctoral Theses
thesis.degree.nameDoctor of Philosophyen_NZ
Files
License bundle
Now showing 1 - 1 of 1
Loading...
Thumbnail Image
Name:
license.txt
Size:
889 B
Format:
Item-specific license agreed upon to submission
Description:
Collections