A PRINCIPLE OF CORRELATIVITY OF PERCEPTION AND ITS APPLICATION TO MUSIC RECOGNITION

Correlativity of perception is defined as a capacity to discover simil ar configurations of stimuli and to form high-level configurations fro m them. It is equivalent to describing information in terms of generat ive elements and their transformations. Such a representation saves me mory and reveals causality in data generation. This approach is implem ented in a model of artificial perception wherein data are self-organi zed in order to segregate patterns before recognizing them. Input info rmation is described as generative patterns and their transformations. The least complex data representation that leads to a causally relate d semantic description is chosen, with (Kolmogorov) complexity defined by the amount of memory store required. The model is applied to voice separation and to rhythm/tempo tracking. Chord spectra are described by generative subspectra, which correspond to tonal spectra, and by th eir translations, which coincide with the intervals of the chord. Time events are also described by generative rhythmic patterns. Tempo and rhythm interdependence is overcome by the optimal sharing of complexit y between representations of rhythmic patterns and tempo curve. The mo del explains the function of interval hearing, certain statements of m usic theory, and some effects of rhythm perception. Applications to im age processing and modeling of abstract thinking are also discussed.