Commentator's Speech Extraction in Audio Stream of Sports Games

This paper proposes a method to deal with the problem of extracting commentator's speech in audio stream of live sports games. First, a two-pass metric-based audio segmentation module is developed to segment the audio stream into short ones with homogeneous acoustic features. Then a model-based classification module is adopted to extract the speech segments. For robust audio classification, various audio features have been used in this paper. Finally, a music scene analysis (Music-CASA) method is adopted to remove the speech in the advertisements with minimum loss of commentator's speech. By integrating all the techniques, an average F value of 94.79% is achieved in the commentator's speech extraction task evaluated on eleven games of six kinds of sports.