Structural boundary perception in popular music

The automatic extraction of musical structure from audio is an important aspect for many music information retrieval (MIR) systems. The criteria on which structural elements in music are defined in MIR systems is often not clearly stated but typically stem from (music) theoretical or signal-based properties. In many cases, however, perceptual-based criteria are the most relevant and systems need to be trained on or modeled after the perception of structural elements in music. Here, we investigate the perception of structural boundaries to Western popular music and examine the musical cues responsible for their perception. We make links to music theoretical descriptions of structural boundaries and to computational methods for extracting structure. The methods and data presented here are useful for developing and training systems for the automatic extraction of musical structure as it is perceived by listeners.