MUSICAL MODELS FOR FOLK-SONG MELODY ALIGNMENT
Peter van Kranenburg, Anja Volk, Frans Wiering, Remco C. Veltkamp · 2009
In this paper we show that the modeling of musical knowledge within alignment algorithms results in a successful similarity approach to melodies. The score of the alignment of two melodies is taken as a measure of similarity. We introduce a number of scoring functions that model the influence of different musical parameters. The evaluation of their retrieval performance on a well-annotated set of 360 folk-song melodies with various kinds of melodic variation, shows that a combination of pitch, rhythm and segmentation-based scoring functions performs best, with a mean average precision of 0.83. 1.