UbiSense - Ubiquitous Sensing and Behaviour Profiling

The UbiSense project aims to provide an unobtrusive health monitoring system for the elderly using embedded smart vision techniques to detect changes in posture, gait and activities.

To analyze the subjects activity, subject specific information called "personal metrics" are gathered by tracking the following features in image sequences:

  • position
  • gait
  • posture
  • motion

The personal metrics can be transmitted between sensors so that complex behaviour profiling can be performed by utilizing resources of other UbiSense nodes in the network.

A major challenge of vision-based systems is the apparent intrusion of privacy due to the way image data is transmitted and analyzed. To circumvent this problem, the captured images in UbiSense are immediately reduced at the device level into blobs, as shown in Fig. 1. No visual images are stored or transmitted any point making it impossible to reconstruct the original image information showing the visual appearance of the subject.



Fig 1:(a) The original video image, (b) the background subtracted image, and (c) the extracted blob.


A simulated house set up with a network of UbiSense system is illustrated in fig.2. Considering the fact that behaviour profiling involves a set of highly demanding computa-tional tasks that typically require expensive hardware and sophisticated setup procedure, the UbiSense pro-ject investigates the use of a networked low cost embedded smart video sensors to provide complex behav-iour profiling functions based on resource scavenging, dynamic clustering and self-configurations




Fig 2: A schematic diagram of UbiSense in a home environment

UbiSense Node

A prototype UbiSense node is designed with an em-bedded smart camera VCSBC50 (a DSP camera), a wireless link provided by an integrated BSN node, and a battery, as shown in Fig.3. The BSN node is a miniaturized (26mm) generic wireless sensing platform for context aware sensing environments. With its compact design (10x7x6cm), the UbiSense node can be installed easily by the user.




Fig 3: A prototype UbiSense node.

Contact

For more information regarding the UbiSense project please contact:

Mr Benny Lo
Department of Computing
180 Queens Gate
Imperial College of Science, Technology, and Medicine
London SW7 2BZ
United Kingdom

Email: benlo@doc.ic.ac.uk
Web: http://www.doc.ic.ac.uk/vip/