With the rapid growing amount of multimedia, content-based information retrieval has become more and more important. As a significant clue for video indexing and retrieval, audio detection and classification attracts much more attention and becomes a hot topic. On the basis of the priori model of news video structure, a selective ensemble support vector machines (SE-SVMs) is proposed to detect and classify the news audio into 4 types, i.e., silence, music, speech, and speech with music background. Experiments with news audio clips of 8514 seconds in total length illustrate that the average accuracy rate of the proposed audio classification method reaches to 98.9%, which is much better than that of the available SVM-based or traditional threshold-based method.
展开▼
