URL:

files/proceedings/2010/70.pdf

Abstract:

In this paper we present a system that learns rhythmic pat- terns from drum audio recording and synthesizes music variations from the learnt sequence. The procedure de- scribed is completely unsupervised and embodies the tran- scription of a percussion sequence into a fuzzy multilevel representation. Moreover, a tempo estimation procedure identifying the most regular subsequence is used to guaran- tee that the metrical structure is preserved in the generated sequence. The final synthesis is performed, recombining the audio material derived from the sample itself. Some examples of generations along with a descriptive evalua- tion are provided.