Drum rhythm retrieval based on rhythm- and sound similarity

TitleDrum rhythm retrieval based on rhythm- and sound similarity
Publication TypeMaster Thesis
Year of Publication2011
AuthorsTutzer, F.
preprint/postprint documenthttp://mtg.upf.edu/system/files/publications/Tutzer-Felix-Master-thesis-2011.pdf
AbstractThe objective of this thesis is, to develop an application, similar to a midi drum sequencer, that is capable of automatically providing a user with rhythms in the context of a user generated reference rhythm. An underlying rhythmic similarity algorithm is used to compute the distance between the reference and all the possible candidate or target rhythms that are stored as rhythmic descriptions in a database. A rhythmic description contains several descriptors to describe the rhythmic quali- ties of the tracks of a rhythm such as event timing, sound class, syncopation degree and metrical weight histogram. A track represents the event sequence in a rhythm, where all the events belong to the same sound class. We de ne the distance between a reference and a target rhythm as the averaged di erences between the metrical weight histograms of the respective tracks that belong to the same sound class. The user can de ne how similar or dissimilar the retrieved target rhythms should be in the context of the reference rhythm by adjusting several settings in the user interface of the application. The quality of the rhythmic similarity measure was evaluated by comparing the results of listening tests, that were done with 6 electronic music producers includ- ing 1 professional percussionist, with the results we obtained from the algorithm. The evaluation shows that the rhythmic similarity algorithm works accurate for the purpose of this project. The system was implemented in Max/Msp and Java.
intranet