MUMT 611 Presenation 4 Greg Eustace

Melodic Similarity: Annotated bibliography

  1. Aloupis, G., et al. 2003. Computing a geometric measure of the similarity between two melodies. Proceedings of the 15th Canadian Conference on Computational Geometry : 81-4.

    A geometric measure of the similarity between two melodies is determined by first representing melodies as functions of time and pitch and then superimposing these functions on to a cylinder, where time is measured in terms of the angle theta and pitch is measured in terms of z. The difference between two melodies is then calculated as the difference in area between the two functions.

  2. Grachten, M., J. Acros, R. de Mantaras. 2000. Melodic Similarity: Looking for a good abstraction level. Proceedings of the Fifth International Conference on Information Retrieval.

    A new measure of melodic similarity involving Implication/Realization (I/R) structures based on Narmour's theory of perception and cognition of melodies is presented. This is compared with other measures utilizing edit distances and dynamic programming.

  3. Hu, N., R. Dannenberg, A. Lewis. 2002. A probabilistic model of melodic similarity. Proceedings of the International Computer Music Conference.

    A probabilistic measure of melodic similarity is presented which uses dynamic programming techniques. This is compared with algorithms which measure edit distances.

  4. Kim, Y., et al. 2000. Analysis of a contour-based representation for melody. Proceedings of the International Symposium on Music Information Retrieval.

    Melodic representations are explored with focus on aspects of contour and rhythm. This leads to the development of melodic similarity algorithms to be implemented in a query by humming system.

  5. Typke, R., et al. 2003. Using transportation differences for measuring melodic similarity. Computer Science Department Technical Report, University of Utrecht.

    Melodic similarity is measured using the earth mover’s distance function and the proportional transportation distance. This follows from the representation of a melody as weigthed two-dimensional point set, defined by pitch and duration axes.

  6. Uitdenbgerd, A., J. Zobel. 1999. Melodic matching techniques for large music databases Proceedings of the Seventh ACM International Conference on Multimedia: 57 - 66.

    A three-stage framework is presented for melody matching, including melodic extraction which involves extracting the dominant tune from a piece of polyphonic music, melodic standardization which removes certain performance specific characteristics of a melody, and several algorithms for determining melodic similarity. The latter includes longest common substring and longest common subsequence as well as N-gram and dynamic programming techniques.