Audio Classification and Segmentation for Sports Video Structure Extraction using Support Vector Machine

Video structure extraction is essential to automatic and content-based organization, retrieval and browsing of video. In this paper, we present a novel scheme for indexing and segmentation of video by analyzing the audio track using support vector machine. This analysis is then applied to structuring the sports video. Based on the attributes of sports video, we define three audio classes in sports video, namely play-audio, advertisement-audio and studio-audio. Support vector machine (SVM) is a valid statistic learning method. The work on audio classification using SVM is presented. Meanwhile, considering that it is highly impossible to change the audio types too suddenly, we apply smoothing rules in final segmentation of an audio sequence. Experimental results indicate that our framework can produce satisfactory results