Segmentation of sensor data for Activity Monitoring

Within a smart environment, sensors have the ability to perceive changes of the environment itself and can therefore be used to infer higher levels of information such as activity behaviours. Sensor events collected over a period of time may contain several activities. Knowledge about how people perform their activities of daily living (ADL) has further inspired research related to the monitoring of interactions with domestic objects associated with the completion of each activity [1-3]. For example, using the telephone normally involves lifting the telephone handset, dialing numbers, having a conversation and hanging up the handset at the end of the conversation. A simple state-change sensor can be used to detect any change of state of an object which can subsequently be used to reflect the interactions of a human being with the object. For example, a state-change sensor attached to the handset of a telephone detects if the handset has been lifted from the telephone’s base station.  The fundamental process of any automatic activity monitoring system is therefore to process the data stream of sensor events and detect occurrences of activities.

The research challenge in this activity recognition process is how to extract high level activity information from low level sensor events. The problem is two-fold: (1) how to partition a stream of time-series sensor events into segments, each of which represents the duration of an activity; (2) how to recognise the activity associated within a segmented stream of sensor events.

The aim of this Project is to develop and evaluate a novel approach to segmentation to separate time-series sensor data into segments which may be furthered processed by an activity recognition algorithm.

This Project will be conducted within the Smart Environments Research Group within the School of Computing and Mathematics (http://serg.ulster.ac.uk/).

[1]     M. Philipose, K. Fishkin, M. Perkowitz, D. Patterson, D. Fox, H. Kautz and D. Hahnel, "Inferring activities from interactions with objects," IEEE Pervasive Computing, vol.3, pp. 50-57, 2004.

[2]     E. M. Tapia, S.S. Intille, K. Larson, “Activity recognition in the home using simple and ubiquitous sensors,” in Proc. 2nd Int. Conf. Pervasive Computing, IEEE 2004, LNCS 3001, pp. 158-175.

[3]     T. van Kasteren, A. Noulas, G. Englebienne, and B. Krose, “Accurate activity recognition in home setting,” Proc. 10th Int. Conf. Ubiquitous Computing, ACM 2008, pp. 1-9.

Personnel Involved

First Supervisor: Nugent, CD Professor
Second Supervisor: Chen, L Dr
Third Supervisor: Wang, H Dr

Collaboration: This project does not involve collaboration with another establishment

Synopsis:

Return to list of PhD Opportunities