The Representation of Pitch in a Neural Net Model of Chord Classification

A fundamental concern in the construction of neural nets for musical applications is the representation of input to the system. The way in which input is ultimately represented is determined by several factors: (1) the theoretical viewpoint of the researcher, (2) the primary use of the net, and (3) available computational resources. Psychoacousticians may be interested in representations that capture aspects of the peripheral processing of musical signals by the basilar membrane as well as with central mechanisms. "Cognitivists" may be more concerned with representations at the abstract level of concepts, such as musical key. Researchers who want to use neural nets for musical signal processing may be interested in both lowand high-level representations in order to enable analysis of recorded musical sound and the subsequent conversion into the high-level abstraction of a musical