Significant Event Detection in Sports Video Using Audio Cues
K. Pradeep, M. Tech · 2013
This work presents audio cue based approaches for event detection from sports video for construction of sports video summary. The approach has been proved effective applied to soccer and cricket videos. The approach extracts the event based on audio level, the sports matches always have mass spectators who show there response in the form of loud cheering, applause and referees signal also. These audio symbols can be detected by measure audio level and pattern which is normally higher than regular audio level. The co-relation between audio and video frames will help to extract frames and summarize these audio samples to combine event, all such events expected together can be used for summary. Proposed Methodology detects the peak values of the audio samples present in the sports video streams and looks for the video frame corresponding to the audio peak detected. The video event detection is generated as AVI streams using the video frames around the audio peaks detected. The experimental results found to match the manual examination and ground truth of the various sports videos.