Audio Event Detection in Tennis Match Based on Contextual Information
QI Hong-zhu · Harbin Ligong Daxue xuebao · 2013
This paper sets its goal to audio event detection in tennis match and proposes an audio event detection method based on contextual information. This method detects the sounds of ball hit and line judge's shout by making use of the grunts of players when they hit the ball. In the construction of models for audio events,the unsupervised learning is employed to use the information of grunts. Compared with current methods,the proposed method does not need any labeled data,and can reduce the mismatch between training and testing. Experimental results further conform that the proposed method can improve the performance of audio event detection substantially.