Location

E14-633

Description

The continued miniaturization of computational devices is set to revolutionize the field of healthcare and health monitoring for the elderly; a multitude of varied and unobtrusive sensors will be able to continually assess the mobility and well-being of elders in their own homes.
To this end, the fusion of multiple sensors and modalities can be used to provide greater accuracy and coverage than with any single sensor. In this talk, we investigate how fusion can be leveraged in a privacy-respectful environment–an environment in which sensitive information, such as appearance, should not be extracted, stored, transmitted, or used.
Specifically, we introduce and discuss an extension to a Receiver Operating Characteristic (ROC)-based technique known as the Bayesian Framework for Feature Selection (BFFS). This allows us to choose optimal feature sets from disparate sensors which, when used to classify elder activity, have maximal overlap. This provides the groundwork for a probabilistic framework to temporally match data streams without the requirement of operating at the level of identity.
With a privacy-respectful method for the matching of data in place, we discuss the recognition of activities through the fusion of an ambient, vision-based “silhouette/blob” sensor and a wearable device. We demonstrate that naïve sensor fusion is advantageous for monitoring several activities representative of Activities of Daily Living (ADLs), before extending our approach to utilize more detailed information regarding subject pose and motion. Our technique is then presented which extracts a convex hull probabilistically using multiple ambient devices. We demonstrate how a canonical descriptor can be extracted through cylindrical re-projection of the convex hull and subsequently used to characterize the evolution of posture.

Biographies

Douglas McIlwraith earned his MA and MSc degrees from Trinity Hall, Cambridge University and Imperial College, University of London. He completed his PhD on the topic of sensor fusion at the latter, and now works in the area of mobile learning for Sharp Laboratories Europe. Douglas has made research contributions to the fields of distributed systems, ubiquitous computing, pervasive sensing, robotics, and security.

Host/Chair: Joseph A. Paradiso