Identifying Household Fingerprint Using Intelligent Passive Monitoring for People Living with Dementia

Monitoring and supporting daily routines in households with dementia is highly important. Here we analyse daily activity in 104 households with dementia across 24 months using passive motion sensors, door sensors and bed mat. We use Markov chain to extract the transitions around the house and we compare these transitions over time. We found that transitions around the house can be used to uniquely match the household as a distinctive fingerprint. We observed that the fingerprint of single-occupancy households is stronger than multiple occupancy and that predicting consecutive periods of time has a higher accuracy due to less change in routine. We focused the latter part of our study on the single occupancy households, specifically looking at the impact of the house floorplan. We observed that despite having similar or identical floorplans, the household fingerprint is still unique and related to the level of activity, and routine.