In music similarity and in the related task of genre classification, a distance measure between Gaussian mixture models is frequently needed. We present a comparison of the Kullback-Leibler distance, the earth movers distance and the normalized L2 distance for this application. Although the normalized L2 distance was slightly inferior to the Kullback-Leibler distance with respect to classification performance, it has the advantage of obeying the triangle inequality, which allows for efficient searching.
|Title of host publication||Proceedings of the 8th International Conference on Music Information Retrieval|
|Number of pages||2|
|Publisher||Austrian Computer Society|
|Publication status||Published - 2007|
|Event||International Conference on Music Information Retrieval - Vienna, Austria|
Duration: 23 Sep 2007 → 27 Sep 2007
Conference number: 8
|Conference||International Conference on Music Information Retrieval|
|Period||23/09/2007 → 27/09/2007|
Jensen, J. H., Ellis, D. P. W., Christensen, M. G., & Jensen, S. H. (2007). Evaluation of Distance Measures Between Gaussian Mixture Models of MFCCs. In Proceedings of the 8th International Conference on Music Information Retrieval (pp. 107-108). Austrian Computer Society.