¹School of Information Security Engineering, Shanghai Jiao-tong University, Shanghai, China
²Shanghai Information Security Management and Technology Research Key Lab, Shanghai, China

---

Abstract—With the development of the multimedia technology, there are more and more video resources on the Internet, which are difficult to automatically recognize, classify and index. So solve these problems, we present a novel video content understanding scheme in this paper. This scheme is based on the combination strategy of different video features. To represent these video features, we use nine standard MPEG-7 descriptors, including color, texture, region and motion descriptors. We extract and combine these descriptors together to represent the whole video character. After that, we use an SVM as the classifier to train the model. The traditional 1-1 method of the SVM is modified by a Second-Prediction Strategy to gain higher classification accuracy. Finally, the videos are classified into five genres, including cartoons, commercial, music, news, and sports. We compare our classification results with some of the results in the recent papers, and demonstrate the effectiveness of our scheme.

Index Terms—Content understanding, video content classification, MPEG-7 descriptors, second-prediction, support vector machine.

[PDF]

Cite: Xinghao Jiang, Tanfeng Sun, Bin Chen, Rongjie Li, Bing Feng, "A Novel Video Content Understanding Scheme Based on Feature Combination Strategy," Journal of Computers vol. 4, no. 7, pp. 615-622, 2009.