Repository | Book | Chapter
(1997) Music, Gestalt, and computing, Dordrecht, Springer.
Emulating Gestalt mechanisms by combining symbolic and subsymbolic information processing procedures
Udo Mattusch
pp. 469-481
Pattern matching based on Gestalt principles can be understood as a process of discovering an abstract logical function Knowledge based processes are needed in order to handle musical information on higher levels and to describe the relations between patterns in special musical contexts, e.g. in harmonic perception. The article sketches a computational approach to a cognitive model of Gestalt perception integrating subsymbolic and symbolic methods of information processing. The model is based on the combination of a neural network (backpropagation net) with methods of inductive learning in artificial intelligence. For the special case of musical transposition, the transfer of information between subsymbolic and symbolic based methods is demonstrated. From this point of view, the recognition of musical transposition, being a special case of Gestalt perception, depends on the inference of logical functions which can not be achieved by standard methods of pattern matching.
Publication details
DOI: 10.1007/BFb0034134
Full citation:
Mattusch, U. (1997)., Emulating Gestalt mechanisms by combining symbolic and subsymbolic information processing procedures, in M. Leman (ed.), Music, Gestalt, and computing, Dordrecht, Springer, pp. 469-481.
This document is unfortunately not available for download at the moment.