Using Hidden Markov Models to Build Behavioural Models to Detect the Onset of Dementia

摘要

Innovative methodologies to provide care for the elderly people in their homes form an emerging and evolving field of research. Proactive care for Dementia is an important challenge that should be researched. Using Ambient Intelligence (AmI) solutions, different data modalities can be collected from home settings. Suggested solutions are concentrating on providing behaviour monitoring or telemonitoring solutions that are apt to support and help the clinicians' and carers' decision making in addition to helping family members to receive assurance about their relatives. We are using Hidden Markov Models in order to build a behavioural model based on raw sensor data. Although binary simple sensors are used, the resulting model can detect abnormalities, sudden and gradual, in elderly people's behaviour, which may be considered an indicator of dementia. The role of the suggested system is to raise an alarm whenever a behavioural change is detected and to leave decision making to the carer.