Unsupervised Activity Recognition:
Currently we are designing a novel unsupervised activity recognition system that uses in-home wireless sensors. This will enable passive monitoring of elderly people to assess their cognitive and physical capabilities. Existing activity recognition algorithms are mainly supervised and focus on determining which activity is currently being performed. We use data mining techniques (clustering and pattern mining) to recognize daily activities in an unsupervised way that will detect more details of daily activities (e.g., specific order of object uses during an activity, temporal characteristics of object uses).
AlarmNet:
AlarmNet is an assisted-living and residential monitoring network for smart healthcare research. Currently
we are implementing a Depression Monitoring System for elderly people. As part of this system I have designed and implemented a Sleeping Monitoring Subsystem
using Intel WISP's. We are also working on using the WISPs for early on-set detection of Alzheimer's.
PhysicalNet:
PhysicalNet is a generic paradigm for managing and programming world-wide distributed
heterogeneous sensor and actuator resources in a multi-user and multi-network environment. It is developed using Java programming language. As part of this
paradigm, we have designed an abstraction named Bundle for grouping wireless sensors and
actuators.
Real Time Requirements in WSNs:
Our main research goal is to design WSNs that have real time requirements, e.g., scheduling WSN streams that have application-defined end-to-end deadlines,
calculating capacity and throughput bounds of such scheduling algorithms. These problems are complicated by the presence of radio irregularity, interference,
burstiness, etc.