摘要：This paper proposed a method of audio event chage detection and clustering in movies. Three steps criterion method is used to detect audio event change in movies ,non silece segment is gotton from audio events by using energy firstly, potential audio event change point is gotton by calculating the distance of two sliding windows secondly , penalty distance is used to judge whether a potential audio enent change point is real or not lastly. Spectural clustering is used to estimate the number of audio events and k-means is used to cluster the same audio events . Experimental results have demonstrated that the proposed method can reach a satisfactory results.
