Analyzing and composing music from network motifs

In this paper we apply the concept of musical motifs and network motifs to artificially compose musc. We construct two different types of complex networks for music, namely, tonal network and rhythmic network, and identify tonal motifs and rhythmic motifs, respectively, from these networks. Then, from these two kinds of motifs, we define a set of composite motifs which represent basic groups of chronologically ordered notes. For example, t tonal motifs and r rhythmic motifs give a set of m composite motifs, where m > max(t, r). By combining these m motifs, a piece of music can be produced. We analyze a popular Japanese folk song, and identify its motifs. Music reconstructed from combining motifs has been found to “sound like” the original composition from which the motifs are constructed.