[an error occurred while processing this directive]

Skip to Content

Computer Science Colloquia

Friday, October 28, 2011
Timothy Hnat
Advisor: Kamin Whitehouse
Attending Faculty: Jack Stankovic (Chair), Wes Weimer, Kevin Skadron, and Scott Acton

Olsson Hall, Room 228E, 1:00 PM

Ph.D. Defense Presentation
Infrastructure-based Occupant Localization for Indoor Home Environments

From the days of the sextant and reading of stars for trans-atlantic boat navigation, to a compass and topology map, and our modern Global Position System (GPS), the desire of people to know where they are located and how they will get to their destination. Technology has steadily improved our localization ability but has so far failed to adequately address the problem of or tracking a person within a building. Nearly all solutions rely on a technological device to be carried or operated by an individual. Our infrastructure-based solution removes the requirement of a person carrying a device or performing a task and still provides tracking capabilities for multiple people at a room-level resolution. Our approach utilizes sensors that measure the height of a person as he/she walks through a doorway. This solves the two major problems with prior work. First, our sensors do not depend on an occupant to carry any device while he/she is within the house. This means he/she cannot forget to change batteries or leave the device at some location within the house. Second, because our sensors do not have video cameras on them, there is no concern about privacy violations. We needed to overcome a variety of technical and computational challenges to construct our tracking solution. First, a new sensing platform needed to be built that was capable of sensing the direction and height of a person as he/she walks through a doorway. Second, a signal processing stage is utilized to isolate events and their features from each data stream provided by the height sensors. Next, our tracking algorithm is based on a particle filter and is used to help avoid the errors generated by the prior stage. The single largest challenge here is handling false positives and false negatives along with the amount of variation in height measurements obtained from signal processing. Our contributions include a new hardware design for measuring the height and direction of a person as he/she walk through a doorway, a signal processing algorithm which handles detecting events and fusing multiple data streams together, and a tracking algorithm to determine where each occupant is located within a house. This technology will enable interesting future applications and research in smart environments.