Show simple item record

dc.contributor.advisorJones, Edward
dc.contributor.advisorGlavin, Martin
dc.contributor.advisorKilmartin, Liam
dc.contributor.authorCraven, Darren
dc.date.accessioned2016-11-24T11:37:43Z
dc.date.available2016-11-24T11:37:43Z
dc.date.issued2016-11-23
dc.identifier.urihttp://hdl.handle.net/10379/6168
dc.description.abstractAdvances in signal processing are enabling promising solutions for ambulatory monitoring in healthcare applications. A critical objective of such systems is low-power operation to enable long-term battery life in wearable/mobile devices. This project addresses the low-power objective through the use of lossy compression techniques, particularly focusing on Compressed Sensing (CS). CS is a recently-introduced paradigm for compression that has significant potential for deployment in Body Area Networks (BAN) compared to existing compression methods. A particular advantage of CS in this context is the possibility of very simple encoding implementations in the wearable sensors, with greater computational complexity in signal reconstruction. However, the performance of CS remains limited in terms of signal reconstruction quality when compared to existing approaches based on Nyquist sampling. In this thesis, novel encoding strategies for CS are developed and novel Dictionary Learning (DL) techniques are utilised for reconstruction to challenge the current state-of-the-art in this area. The proposed algorithms are analysed in terms of signal fidelity for a given Compression Ratio (CR) and results demonstrate their ability to outperform existing CS approaches. Furthermore, there is increasing interest in Computer Aided Diagnosis (CAD) systems that tolerate some loss in signal fidelity at higher CRs but where diagnostic integrity of the signal is maintained. The proposed algorithms are also evaluated in terms of their ability to maintain performance in CAD systems at higher CRs. Finally, this thesis analyses the power consumption of CS and compares it with a state-of-the-art lossy compression algorithm in terms of overall system power consumption (including power consumption of signal acquisition, digital processing, and wireless transmission) in a BAN. The results confirm the substantial benefits of employing CS as a low-energy encoding implementation.en_IE
dc.rightsAttribution-NonCommercial-NoDerivs 3.0 Ireland
dc.rights.urihttps://creativecommons.org/licenses/by-nc-nd/3.0/ie/
dc.subjectElectrical and electronic engineeringen_IE
dc.subjectBiomedical signal processingen_IE
dc.subjectCompressed sensingen_IE
dc.subjectWireless body area networksen_IE
dc.titleLow-power strategies for signal compression in ambulatory healthcareen_IE
dc.typeThesisen_IE
dc.contributor.funderProgramme for Research in Third Level Institutionsen_IE
dc.local.noteAn investigation of advanced signal processing techniques to reduce the power consumption in wearable ECG monitoring devices. In particular, novel algorithms for compressed sensing were proposed and demonstrated an ability to maintain diagnostic performance while extending battery life compared to existing approaches.en_IE
dc.local.finalYesen_IE
nui.item.downloads735


Files in this item

Thumbnail
Thumbnail

This item appears in the following Collection(s)

Show simple item record

Attribution-NonCommercial-NoDerivs 3.0 Ireland
Except where otherwise noted, this item's license is described as Attribution-NonCommercial-NoDerivs 3.0 Ireland