A highlight scene detection and video summarization system using audio feature for a personal video recorder

The personal video recorder such as recordable-DVD recorder, Blu-ray disc recorder and/or hard disc recorder has become popular for a large volume storage device for video/audio content data and a browsing function that would quickly provide a desired scene to the user is required as an essential part of such a large capacity recording/playback system. We propose a highlight scene detection function by using only \´audio\´ features and realize a browsing function for the recorder that enables completely automatic detection of sports highlights. We detect sports highlights by identifying portions with "commentator\´s excited speech" using Gaussian mixture models (GMM\´s) trained using the MDL criterion. Our computation is carried out directly on the MDCT coefficients from the AC-3 coefficients thus giving us a tremendous speed advantage. Our accuracy of detection of sports highlights is high across a variety of sports.