Designing a circadian lighting and activity detection solution to enhance wellbeing for people with dementia​

Living with dementia can negatively impact wellbeing due to the onset of cognitive decline and associated behavioural and psychological symptoms. Prevalent of these symptoms is a heavily disrupted circadian rhythm which is amplified due to age-related factors and lack of daylight exposure. The circadian rhythm is responsible for aligning the internal body clock, calibrating sleep-wake cycles, maintaining hormone balance and regulating metabolism; all factors which influence daily activities for people with dementia. Research has shown that the best way to realign the circadian rhythm for people with dementia is through increased daylight exposure or through light therapies with outputs analogous to daylight. However, it is not enough to implement a lighting intervention for people with dementia without a robust solution for monitoring efficacy. Moreover, understanding the diverse behaviours observed in various types and states of progression of dementia can be limited due to the often invasive and therefore short-lived monitoring techniques deployed. In order to begin creating a solution which is both user-centred and fully representative of these person-specific behaviours, an intelligent, unobtrusive and scalable sensing paradigm is required. This study therefore presents a circadian-optimised luminaire with an integrated mm-wave sensor which supports bi-directional communications between an intelligent backend behaviour detection model. The behaviour model is initially trained on features extracted from various simulations of daily activities and falls gathered in a controlled environment. The real-time sensor data is transferred to the learned behaviour model, and depending on the observed activity, triggers changes to the lighting output from the luminaire, or triggers alerts to a caregiver’s mobile application. This study outlines a preliminary evaluation of the supervised machine learning algorithms used to detect and categorise daily activities and falls. This will lead to refinement of these algorithms and realisation of the hybrid lighting-sensing solution for real-environment deployment in care facilities.