Automated Partitioning of Tonal Music

Most research related to automated analysis of music presupposes human partitioning of the input into segments corresponding to significant harmonic or melodic chunks. In this paper, we describe HarmAn, a system that partitions tonal music into harmonically significant segments corresponding to single chords and labels these segments with the proper chord labels. Segment labels are determined through template matching in the space of pitch class with conflict resolution between equally scoring templates resolved through simple preference roles. Our system’s results are compared with published results.