Memory And Melodic Density: A Model For Melody Segmentation
Miguel Ferrand, Peter C. Nelson, Geraínt A. Wiggins · 2003
We present a memory-based model for melodic segmentation based on the notion of melodic density. The model emphasises the role of short-term memory and time in music listening, by modelling the effects of recency in the perception of boundaries. We describe the model in detail and compare it with Cambouropoulos' Local Boundary Detection Model for a series of melody examples. First results indicate that this new model is more conservative, as it generates fewer total boundaries but preserves most boundaries that coincide with the limits of recurring patterns.