Researching Audio Segmentation And Classification Methods Of The Broadcasting Repertoire

Audio classification is a base work for the audio information retrieval, and animportant part of many audio applications. This technique classifies the audio data intodifferent classes according to the feature of different type of audio.In this paper, various audio segmentation and classification method on broadcastradio are analyzed carefully. Base on the analysis on sequential audio according to timesequence, the point in which audio class is changed and the class of current audiosegment is identified.On the base of research, we realize the system for audio classification. Our systemcomprise of audio segmentation module and classification module. We use an approachbased on checking the changing trend of dividing entropy of every signal points in sliding windowsegmentation and Gaussian Mixture Model (GMM) classification scheme for audioclassification. Experiments show that our system is efficient with the averageclassification performance of over 89%. The application request is satisfied generally.