Multi-modal analysis for person type classification in news video

Classifying the identities of people appearing in broadcast news video into anchor, reporter, or news subject is an im-portant topic in high-level video analysis. Given the visual resemblance of different types of people, this work explores multi-modal features derived from a variety of evidences, such as the speech identity, transcript clues, temporal video structure, named entities, and uses a statistical learning approach to combine all the features for person type classifica-tion. Experiments conducted on ABC World News Tonight video have demonstrated the effectiveness of the approach, and the contributions of different categories of features have been compared.