Audio keywords detection in basketball video

This paper presents an audio keywords detection method for highlight retrieval in basketball video. The keywords contain shoes squeaking sound, speech, cheer, long whistle and short whistle, which correspond to basketball game events. After feature analysis, the Simple Excellent Feature Combination based on Pearson Correlation Coefficient (SEFC-PCC) is used to select efficient features, which contributes to a preferable performance and lower computational complexity. A novel multi-stage SVM classifier is proposed to do the final detection of the five audio keywords. There are 428 audio sequences about 704 seconds used in the validation experiment; it gives a performance evaluation with average detection accuracy of 92%~99%.