Desain, P., & Honing, H. (1994). Can music cognition benefit from computer music research? From foot-tapper systems to beat induction models. In Proceedings of the ICMPC 1994. 397-398. LiÀge: ESCOM.

Abstract

In music cognition research several theories have been proposed that are concerned with the perception of different types of temporal musical structure (e.g., meter, rhythm, and phrase structure). Often these models, and the models that deal with other aspects of music like melody and harmony, assume a successfully identified beat. For example, a beat can give a first rating of importance to musical events and suggest viable boundaries for chunking, it indicates the level of Ñfocusæ in the metrical structure (the level at which events pass at a moderate rate, i.e., tactus), it gives clues on whether the tempo changes and it often functions as a basis to distinguish between the canonical aspects of time and the continuously variable performance aspects (i.e., expressive timing). This makes beat induction Çthe process yielding a perceived pulse marking off durational units in the stimulus patternÇ a fundamental aspect of many models of music cognition. Beat induction or Foot-tapping is a example of a simple cognitive task that is nevertheless hard to model computationally. As result it has been an important issue in interactive computer music, with many different approaches explored. The systems that were developed all use a more or less pragmatic or technological approach (see, e.g., Boulanger, 1990; Chafe, Mont-Reynaud & Rush, 1989). Some of them derive a beat from a stream of musical events in a live, interactive performance situation (Dannenberg & Mont-Reynaud, 1987; Vercoe, 1985). Especially this use in a live performance forced the designers to think of solutions on problems that are rarely touched in cognitive models. In this paper we will give an overview of some aspects involved in this process and the most important models. We restrict ourselves to computational models and focus on the many theories that assume that beat induction is viable considering timed events alone (ignoring information about dynamics, pitch, etc.).

Full paper (only available in RTF format)