[ISEA2015] Paper: Evan Merz – Musical Structure Imitation using Segmentation, and k-Nearest Neighbors (kNN)

Abstract (Short paper)

Keywords: algorithmic music, machine learning, style imitation, segmentation, k nearest neighbors, midi, self similarity, structure, form.

Segmenting music is important in academic and commercial settings. Imitating musical structure requires interpretation and generalization of discovered structure. The program shown here is a work in progress that demonstrates an approach to structure imitation using a segmentation algorithm with a look back algorithm based on a probabilistic variant of kNN. A monophonic piece of music is segmented, then kNN is used to generate the structure of a new piece. This work shows that although the problem of structure generation is complex, it is not clear that a solution must be similarly complex.

  • Evan X. Merz, earned a Master’s Degree in electronic music from Northern Illinois University (USA) in 2010. He earned a doctorate in algorithmic composition from The University of California at Santa Cruz (USA) in 2013. Currently he teaches computer science at San Jose State University (USA).

Full text (PDF) p. 645-648