Personalized ambient monitoring: accelerometry for activity level classification

Amor, J.D. and James, C.J. (2008) Personalized ambient monitoring: accelerometry for activity level classification. In 4th European Conference of the International Federation for Medical and Biological Engineering. Springer. pp. 866-870 . (doi:10.1007/978-3-540-89208-3_207).

Abstract

Bipolar disorder (BD) is a mental disorder characterized by recurrent episodes of mania and depression. The disorder can be very disruptive and relapses often result in hospitalization. With adequate training, sufferers are able to control their symptoms and reduce the disruption to their daily lives. As an aid to this self-control process the Personalized Ambient Monitoring (PAM) project is being developed. The PAM project aims to allow patients with BD to monitor their condition and obtain indications of their mental state. This will be achieved through the use of multiple discreet sensors, personalized for each patient’s needs. The sensors will detect the correlates of mania and depression, which will be used to derive trends in the mental health state of the patient. The major symptoms of BD center on the patient’s activity level and circadian rhythm. Manic episodes are typified by increased energy and activity, often with a decreased need for sleep. Depressive episodes however often present with diminished activity. It is our aim that by measuring the patient’s activity levels and circadian rhythm we can provide information that the patient can use to help control their symptoms. Here we present some preliminary work aimed at distinguishing different activities and activity levels in normal controls, based on a small, body-mounted triaxial accelerometer. A number of participants were asked to complete some basic activities whilst wearing the accelerometer. The data was preprocessed to extract a number of salient features, which were used to train a Neuroscale algorithm. Neuroscale produces a generative mapping that visualizes high-dimensional data in a lower-dimensional space, which, with the addition of a clustering algorithm, can be used to classify unknown data points. It is expected that this approach, combined with data from other sensor types will form the backbone of the PAM approach applied to BD.

This record has no associated files available for download.

Contributors

Download statistics