An effective method for video genre classification

As a common media type, video is closely bound up with our life. Since the number and the kinds of videos increase steadily, how to organize the enormous amount of videos and obtain the content of interest has become an important research issue. And the video analysis system emerges, also the research of video gene classification has become an important topic. This paper focuses on classification on video genres of cartoons, movies, advertisements, news, and sports. It can be served for video organization, retrieval, etc. Based on the analysis on different video genres, we fuse video's time feature and color feature from shots together. Specifically, there are seven features including gradient and color features and each one could be an expert for some genre of video. We select these expert features and let them collaborate to improve the accuracy of classification. Then support vector machine (SVM) is used for classification. Experimental results on large amount of video demonstrate the effectiveness of the proposed method.