In this paper we introduce stochastic models that characterize the structure of typical television program genres. We show how video sequences can be represented using discrete-symbol sequences derived from shot features. We then use these sequences to build HMM and hybrid HMM-SCFG models which are used to automatically classify the sequences into genres. In contrast to previous methods for using SCGFs for video processing, we use unsupervised training without an a priori grammar. © Springer-Verlag Berlin Heidelberg 2003.
CITATION STYLE
Taskiran, C. M., Pollak, I., Bouman, C. A., & Delp, E. J. (2003). Stochastic models of video structure for program genre detection. Lecture Notes in Computer Science (Including Subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 2849, 84–92. https://doi.org/10.1007/978-3-540-39798-4_13
Mendeley helps you to discover research relevant for your work.