Utilising wireless sensor networks towards establishing a method of sleep profiling

Based upon current sleep actigraphy techniques, this paper discussesan alternative non-contact method of sleep profiling that is potentially moresuitable for long term monitoring than current clinically approved techniques.The passive sleep actigraphy (PSA) platform presented here utilisesstrategically positioned accelerometers fixed on a mattress to quantify therecorded movements of a bed occupant. In this work, data captured from ayoung control group is decomposed into gravitational and inertial components.These components are then translated into activity counts using numerousquantification modalities and feature extraction techniques to isolate the mostdiscriminant attributes for optimal sleep/wake classification. These attributeswere then input into a random forest classifier to determine the sleep/wake stateof each subject based on their recoded actigraphy data with an accuracyof 89%. The findings suggest that the PSA platform is a potentially viablemethod of non-contact sleep profiling hence supporting further research intothis approach.