Behavior Recognition from Multiple Views Using Fused Hidden Markov Models

Abstract

In this work, we provide a framework for recognizing human behavior from multiple cameras in structured industrial environments. Since target recognition and tracking can be very challenging, we bypass these problems by employing an approach similar to Motion History Images for feature extraction. Modeling and recognition are performed through the use of Hidden Markov Models (HMMs) with Gaussian observation likelihoods. The problems of limited visibility and occlusions are addressed by showing how the framework can be extended for multiple cameras, both at the feature and at the state level. Finally, we evaluate the performance of the examined approaches under real-life visual behavior understanding scenarios and we discuss the obtained results.