Event detection models using 2d-BN and CRFs

Abstract

In this paper, we propose two novel semantic event detection models, i.e., Two-dependence Bayesian Network (2d-BN) and Conditional Random Fields (CRFs). 2d-BN is a simplified Bayesian Network classifier which can characterize the feature relationships well and be trained more efficiently than traditional complex Bayesian Networks. CRFs are undirected probabilistic graphical models which offer several particular advantages including the abilities to relax strong independence assumptions in the state transition and avoid a fundamental limitation of directed probability graphical models. Based on multi-modality fusion and mid-level keywords representation, we use a three-level framework to detect semantic events. The first level extracts audiovisual features, the mid-level detects semantic keywords, and the highlevel infers events using 2d-BN and CRFs models. Compared with state of the art, extensive experimental results demonstrate the effectiveness of the proposed two models. © Springer-Verlag Berlin Heidelberg 2007.