A model for the perception of temporal patterns

F. Lerdahl and R. Jackendoff (1983) suggested a two-stage model as part of a general theory of rhythmic organization. In the first-stage, well-formedness rules impose `grammatical´ constraints on the set of possible rhythmic groupings. In the second stage, preference rules select the best rhythmic grouping based on listeners´ perceptual tendencies. A preliminary model, motivated by the work of Lerdahl and Jackendoff, that attempts to account for some of experimental results in rhythm perception. The model was realized by a recurrent neural network composed of several modules. Simulations explored the model´s response to a number of temporal patterns chosen to illustrate various problematic aspects of rhythmic organization