Seminar on Computational Learning and Adaptation

Hierarchical Stochastic Models for Learning and Recognizing Human Activities

Hung Bui
AI Center
SRI International

An important problem in many applications, such as smart environments and intelligent assistants, involves understanding the typical patterns in a user’s daily activities from low-level sensory data. Due to the sequential nature of activities, dynamic frameworks such as hidden Markov models have been widely used for learning and classifying activities. However, as the space of activities becomes more complex, it will be necessary to make use of their known structural properties, especially their hierarchical organization and decomposition. In this talk, I will describe several hierarchical extensions of the basic hidden Markov model that describe activities at different levels of abstraction and resolution. Inference in these models is considerably more complex because of dependencies between the hidden layers. I will discuss techniques for offline parameter learning and online recognition, and I will show how they can be applied to activity modeling from human movement data. Experiments in this domain demonstrate the advantages of using these models over the basic framework.

Return to the seminar schedule