Musical style identification using grammatical inference: The encoding problem

Abstract

A Musical Style Identification model based on Grammatical Inference (GI) is presented. Under this model, regular grammars are used for modeling Musical Style. Style Classification can be used to implement or improve content based retrieval in multimedia databases, musicology or music education. In this work, several GI Techniques are used to learn, from examples of melodies, a stochastic grammar for each of three different musical styles. Then, each of the learned grammars provides a confidence value of a composition belonging to that grammar, which can be used to classify test melodies. A very important issue in this case is the use of a proper music coding scheme, so different coding schemes are presented and compared, achieving a 3 % classification error rate. © Springer-Verlag Berlin Heidelberg 2003.