Statistical Significance in Song-Spotting in Audio

We present some methods for improving the performance a system capable of automatically identifying audio titles by listening to broadcast radio. We outline how the techniques, placed in an identification system, allow us detect and isolate songs embedded in hours of unlabelled audio yielding over a 91% rate of recognition of the songs and no false alarms. The whole system is also able of working real-time in an off-the-shelf computer.

[1]  Dan Gusfield,et al.  Algorithms on Strings, Trees, and Sequences - Computer Science and Computational Biology , 1997 .

[2]  M S Waterman,et al.  Identification of common molecular subsequences. , 1981, Journal of molecular biology.

[3]  S. Karlin,et al.  Methods for assessing the statistical significance of molecular sequence features by using general scoring schemes. , 1990, Proceedings of the National Academy of Sciences of the United States of America.

[4]  D. Lipman,et al.  Improved tools for biological sequence comparison. , 1988, Proceedings of the National Academy of Sciences of the United States of America.

[5]  Pedro Cano,et al.  Automatic Segmentation for Music Classification using Competitive Hidden Markov Models , 2000, ISMIR.