This chapter describes two different approaches using the variogram in the context of Mel Frequency Cepstral Coefficients (MFCCs) and the evaluation of music similarity. The first approach is referred to as the full variogram approach; in this case, all the lags of the variogram of the second coefficient of the MFCC are employed. The second choice is referred to as the reduced variogram approach; in this case, a subset of the lags of the variogram of the MFCC matrix is considered. Thus, the usage of the variogram is proposed as a tool to synthesize the timbre information contained in the MFCCs. Also, four different weighting functions are tested for the calculation of the distance measure between songs. The performance of the methods proposed is evaluated by applying the pseudo-objective evaluation scheme of the MIREX AMS task. The results are compared against the scores obtained by other methods submitted to the MIREX AMS 2011. © 2013 Springer-Verlag.
CITATION STYLE
Tardón, L. J., & Barbancho, I. (2013). Music similarity evaluation using the variogram for MFCC modelling. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 7900 LNCS, pp. 313–332). https://doi.org/10.1007/978-3-642-41248-6_17
Mendeley helps you to discover research relevant for your work.