TR2004-063

Time Series Analysis and Segmentation Using Eigenvectors for Mining Semantic Audio Label Sequences


    •  Radhakrishnan, R.; Xiong, Z.; Divakaran, A.; Memon, N., "Time Series Analysis and Segmentation Using Eigenvectors for Mining Semantic Audio Label Sequences", IEEE International Conference on Multimedia and Expo (ICME), June 2004.
      BibTeX Download PDF
      • @inproceedings{Radhakrishnan2004jun,
      • author = {Radhakrishnan, R. and Xiong, Z. and Divakaran, A. and Memon, N.},
      • title = {Time Series Analysis and Segmentation Using Eigenvectors for Mining Semantic Audio Label Sequences},
      • booktitle = {IEEE International Conference on Multimedia and Expo (ICME)},
      • year = 2004,
      • month = jun,
      • url = {http://www.merl.com/publications/TR2004-063}
      • }
  • Research Areas:

    Multimedia, Speech & Audio


Pattern discovery from video has promising applications in summarizing different genre types including surveillance and sports. After pattern discovery, a summary of the video can be constructed from a combination of usual and unusual patterns depending on the application domain. In our previous work, we have used an unsupervised label mining approach to extract highlight moments from soccer video [1]. In this paper, we formulate the problem of pattern discovery from semantic audio labels as that of time series analysis and propose a new unsupervised mining framework based on segmentation theory using eigenvectors of the affinity matrix. We use synthetic label sequences generated from HMMs (Hidden Markov Models) to illustrate the effectiveness of the new scheme. We also present results of the proposed scheme for summarizing different sports video and compare its results with our earlier approach.