CAREER: Computer Recognition of Human Activity

Ohio State University Research Foundation -Do Not Use, Columbus OH

Investigators

Abstract

This is a 5-year continuing award. Intelligent machines capable of recognizing and analyzing human actions will play an important role in next generation human-computer interfaces. This research effort will address development of a computational vision framework for recognizing dynamic human movements. The goal is to model the perceptual dynamics associated with human activities to automatically identify various action categories and action efforts. The PI will examine a key-feature approach within a three-mode principal components framework to learn the expressive features needed for the model to produce results qualitatively similar to human judgments of the actions. Experiments will be conducted using motion capture and video data of multiple people performing common activities at different efforts. Results at each stage will be compared to perceptual experiments and alternative non-expressive methods. To demonstrate the approach, an automatic surveillance and monitoring application will be implemented. If successful the research should have broad impact on human-computer interaction in numerous domains as diverse as ergonomic evaluation, sports analysis, and computer animation. In conjunction with the research, the PI will create an educational plan that integrates topics on computer vision, human perception, human-computer interaction, and computer animation into a new cross-disciplinary curriculum that combines classroom discussions with practical laboratory experience, and which fosters undergraduate research-related opportunities.