Machine learning for analysis of in-home sensory observations

People with dementia are more likely to develop several physical and psychiatric health issues, e.g., various infections, respiratory issues, and anxiety which lead to higherhospitalisation risks. Early prediction of these health issues is vital for timely interventions and reducing avoidable hospital admissions. The focus of this PhD project is on how machine learning techniques can be employed to explore continuously collected environmental and physiological data to identify changes in stress and agitation in subjects with dementia and generate relevant alerts for caregivers. This project will be carried out in close collaboration with the UKDementia Research Institute (UK DRI) Care Research & Technology Centre. As access to high quality labelled data is limited, unsupervised, self-supervised, or semi-supervised learning techniques will be great interest.