Real-time Behavior Interpretation
Abstract
The performer proposes a new research effort aimed at advancing technologies for understanding the actions, intentions, and social situations that are exhibited in the observable behavior of moving shapes. Specifically, they aim to address three technical challenges in the development of robust real-time behavior interpretation systems. First, they aim to overcome combinatorial explosions in the search space of possible interpretations by developing new techniques for incremental probabilistic abductive reasoning, enabling real-time resourceconstrained analysis of continuous observations. Second, the performer will address the problem of ambiguity in behavior signals by development new techniques for joint action perception and interpretation within a probabilistic reasoning framework, where actions are disambiguated by the contexts in which they appear. Third, they aim to provide our algorithms with the knowledge necessary to interpret observations of social interaction, particularly where these interactions are defined by movement of location and spatial configurations. They will evaluate their research efforts using visual materials in the style of a film created by Fritz Heider and Marianne Simmel in their famous social psychology experiments.This three-year research effort aims to significantly advance fundamental technologies for automated behavior interpretation, with direct relevance to current and future operational needs of the Department of the Navy.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Aug 12, 2016
- Source ID
- N000141612435
Entities
People
- Andrew Gordon
Organizations
- Office of Naval Research
- United States Navy
- University of Southern California