Detecting and monitoring behavioural change through personalised ambient monitoring

摘要

Bipolar disorder (BD) is one form of mental illness and is estimated to affect around0.4{1.6% of the population. The disorder is characterised by recurrent episodes of maniaand depression and is estimated to cost the UK economy £5.21 billion a year. Manypeople with BD self-monitor their behaviour to help them identify the early warningsigns of an affective episode. The Personalised Ambient Monitoring (PAM) projecthas been conceived take ideas from existing telehealth systems and apply them to BD.By using a distributed network of discreet, unobtrusive sensors, the user's behaviouralpatterns can be monitored and deviations in their behaviour can be detected. In doingso it is hoped that the early warning signs can be detected and that this can be used toassist them in their self-monitoring.The PAM system is being developed by a multi-disciplinary team based at the ISVR andthe School of Management at the University of Southampton, the School of Electricaland Electronic Engineering at the University of Nottingham and the Department ofComputing Science and Mathematics at the University of Stirling.This thesis presents the background and motivations for the PAM project, the approachthe project will take, a review of appropriate data analysis techniques and the experimentalwork that has been undertaken in the investigation of accelerometry for activitymonitoring, the use of a wireless camera to monitor a complex environment and the useof multiple sensors to capture behaviour patterns in a technical trial.Results from the technical trial show that it is possible to process information froma variety of sensors to identity activity signatures and behavioural patterns in normalcontrols. When these activity patterns are trained on week-days, the results presentedshow that it is possible to identify weekend days as being behaviourally different.