Authors: Andrew McDowell; Mark P. Donnelly; Chris D. Nugent; Michael J. McGrath
Addresses: School of Computing and Mathematics, University of Ulster, Shore Road, Newtownabbey, BT37 0QB, UK ' School of Computing and Mathematics, University of Ulster, Shore Road, Newtownabbey, BT37 0QB, UK ' School of Computing and Mathematics, University of Ulster, Shore Road, Newtownabbey, BT37 0QB, UK ' Intel Labs Europe, Collinstown Industrial Park, Leixlip, Co. Kildare, Ireland
Abstract: Based upon current sleep actigraphy techniques, this paper discusses an alternative non-contact method of sleep profiling that is potentially more suitable for long term monitoring than current clinically approved techniques. The passive sleep actigraphy (PSA) platform presented here utilises strategically positioned accelerometers fixed on a mattress to quantify the recorded movements of a bed occupant. In this work, data captured from a young control group is decomposed into gravitational and inertial components. These components are then translated into activity counts using numerous quantification modalities and feature extraction techniques to isolate the most discriminant attributes for optimal sleep/wake classification. These attributes were then input into a random forest classifier to determine the sleep/wake state of each subject based on their recoded actigraphy data with an accuracy of 89%. The findings suggest that the PSA platform is a potentially viable method of non-contact sleep profiling hence supporting further research into this approach.
Keywords: accelerometers; sleep actigraphy; sleep-wake classification; feature reduction; passive sensors; sleep profiling; smart homes; wireless senor networks; WSNs; non-contact methods; long term monitoring; recorded movements.
International Journal of Computers in Healthcare, 2012 Vol.1 No.4, pp.346 - 363
Received: 08 May 2021
Accepted: 12 May 2021
Published online: 30 Jan 2013 *