A Novel Real-Time Commercial Detection Scheme

Nowadays, millions of people´s living and work habits are affected by TV commercials. A feature-based real-time TV commercial detection algorithm is proposed in this work. In terms of the combination of the visual and acoustic features, and the temporal information, we detect the end and the start boundary of the commercial separately. Then, in order to refine the detecting result further, we address a set of basic features that is easy to distinguish the commercial from general programs. Based on these features, a finite automation is build simultaneously, which well illustrates our detection method clearly. The experimental results show that our algorithm can yield better recall (96.47%) and precision (97.27%) by comparing with current main approaches.