Audio-Based Event Detection in Videos - a Comprehensive Survey

Applications such as video classification, video summarization, video retrieval, highlight extraction, and so forth., need discriminating activities occurring in the video. Such activity or event detection in videos has significant consequences in security, surveillance, entertainment and personal archiving. Typical systems focus on the usage of visual cues. Audio cues, however, contains rich information that might be effectively used for event detection and Multimedia Event Detection (MED) could benefit from the attention of researchers in audio analysis. Many audio-based event detection methods have been proposed for specific applications, while others are generic. This survey presents an exhaustive review of efforts in the past years to address the issues of using audio-based cues in video event detection. Existing methods that are based on audio features and modeling techniques that have been used are summarized in this survey. We hope to provide a good understanding of different directions in this field of research. KeywordsEvent detection, Highlight Extraction, Gaussian Mixture Model, Support Vector Machine