Audio and video retrieval based on HCT

作者：Yang, Ji-chen; Yang, Zhuo-yan^*

来源：Journal of Convergence Information Technology, 2012, 7(13): 508-518.

DOI：10.4156/jcit.vol7.issue13.59

• Audio and video
• Euclidean distance
• Hierarchical Cellular Tree
• Histogram intersection
• Minimum distance

Structure and construction rule of Hierarchical Cellular Tree (HCT) is briefly introduced firstly; the algorithm of HCT is complemented and improved secondly; finally, the idea of minimum distance is proposed and made as be feature distance function in audio retrieval and histogram intersection and Euclidean distance is made as feature distance function in video retrieval. The experimental results show that: the recall of this method can reach 78%, which is much better than some common used distance formula in speaker retrieval; the number of cells gotten nearly equals speaker's number; saving retrieval time comparison with ergodic retrieval; in the seven kinds of audio sound retrieval, average recall is 65.08%, applause and laugh can all be retrieved, while the recall of cough and snap by camera is less than 30%. In video retrieval, the recall of using texture is 8.3% than using color histogram, which reaches 75%.

• 出版日期2012
• 单位华南理工大学