Combining grammar-based and memory-based models of perception of time signature and phase

Abstract

The present study investigates the modeling of the perception of time signature and phase using grammar-based and memorybased approaches. It explores how far note-length can be used as the sole input to a model of perception of heard music. Two models are developed: one uses a rule-based grammar and the other uses a combination of a rule-based grammar and a memory-based approach. The relative success of the two models is compared and evaluated. This study explores one dialect, solo string music by Bach: The unaccompanied Suites for Violoncello, BWV 1007-1012 and unaccompanied Partitas and Sonatas for Violin BWV 1001-1006. It is shown that the results of two approaches improve over previous models such as those of Longuet-Higgins and Steedman (1971) and Temperley and Sleator (1999) and that the combination of the two approaches is most successful.