A Connectionist Model for Chord Classification

Fabio Ghignatti Beckenkamp
Paulo Martins Engel

CPGCC, Instituto de Informática
Universidade Federal do Rio Grande do Sul - UFRGS
Caixa Postal 15064
91501-970 Porto Alegre - RS


This work is a contribution to the unification of the worlds of computer science and music. This is a first step in the direction of opportunities offered by connectionism on music research. This paper studies the application of a connectionist model on an actual problem of music domain and tries to demonstrate the efficiency of using connectionism in the chord classification problem. This work specifies a neural network model based on Backpropagation architecture to recognize four types of chords: major, minor, diminished and augmented. Simulations results have shown that the network recognized the four chord types for twelve major and twelve minor tonalities.