Event Location: NSH 1507

Abstract: To make computers proactive and assistive, we must enable them to
perceive, learn, and predict what is happening in their surroundings.
This presents us with the challenge of formalizing computational models
of everyday human activities. These mechanisms must perform well in the
face of data uncertainty and complex activity dynamics. Traditional
approaches to this end assume prior knowledge about the structure of
human activities, using which explicitly defined activity-models are
learned in a supervised manner. However, for a majority of everyday
environments such activity structure is generally not known a priori. In
this talk, I will discuss knowledge representations and manipulation
techniques that facilitate minimally supervised learning of activity
structure. In particular, I will present n-gram and Suffix Tree based
sequence representations for human activity analysis. I will discuss how
such a data-driven approach towards activity modeling can help discover
and characterize human activities, and learn typical behaviors crucial
for detecting irregular occurrences in an environment. I will provide
experimental validation of my approach for activity analysis in
environments such as a residential house, a loading dock area, and a
household kitchen.