SELF-CONFIGURING VIDEO-SENSOR NETWORKS Doros Agathangelou, Benny P.L. Lo, Jeffrey L. Wang and Guang-Zhong Yang * Abstract A growing demand on healthcare providers worldwide is the provision of long-term care for the elderly. Instead of relying on nursing homes, a better way to maintain and improve their well being is to provide managed care in their own dwellings. The use of video based sensing provides an effective means of detecting changes in activity, gait and posture but the installation and calibration of the sensors are major obstacles to overcome in practical applications. This paper presents a novel MDS (Multidimensional Scaling) based self-configuration technique for video sensor networks. It allows implicit estimation of the geometrical locations of the sensors and permits the optimal usage of re-sources under a distributed processing environment. 1. Introduction In almost all countries, longevity has given rise to expensive age-related disabilities and diseases. With the steady decline of the ratio of workers to retirees, a fundamental change of the way that we care for the aging population is required. Older adults of 65 and above already constitute one-fifth of the total population, and it is expected this will continue to grow. With the maturity of sensing and pervasive computing techniques, extensive research is being carried out in using sensor networks for home care environments. Much research is now directed towards the use of sensor networks for promoting healthy behaviour, early disease detection, improved treatment compliance, and support for informal care giving. These personal wellness systems are not meant to replace hospital, clinics, and physicians but rather to emphasis the activities of daily living as part of the healthcare mix. For the elderly, home-based healthcare encourages the maintenance of physical fitness, social activity and cognitive engagement to function independently in their own homes. Existing research has shown that when privacy and security issues are properly addressed, video based sensor networks provide an effective means of monitoring behaviour changes. For example, the UbiSense † system allows the captured image immediately turned into blobs at the device level that encapsulate shape outline and motion vectors of the body. No visual images are stored or transmitted at any stage of the processing. Furthermore, it is not possible to reconstruct this abstracted information into images. One of the major challenges of the video based system is the complexity of site installation and calibration of relative physical orientations and locations of the sensors, which are important for activity tracking and inferring abnormal behaviours. The purpose of this paper is to propose a novel self-configuration method based on multi-dimensional scaling (MDS) for estimating the spatial positions of the sensors. The proposed technique relies solely on * Department of Computing, Imperial College London, South Kensington Campus, 180 Queen’s Gate, London, SW7 2AZ, United Kingdom {da200e, benlo, lwang, gzy}@doc.ic.ac.uk http://www.doc.ic.ac.uk/vip/ubisense/ † The UbiSense project is a UK DTI funded project (It is not related to the company called Ubisense) 29