Musical Style Classification Using Low-Level Features

In this paper we address the problem of musical style classification. This problem has several applications like indexing in musical databases or development of automatic composition systems. Starting from MIDI files of real-world improvisations, we extract the melody track and cut it into overlapping segments of equal length. From these fragments, numerical features are extracted as descriptors of style samples. Then a cascade correlation neural network is adopted to build an effective musical style classifier. Preliminary experimental results show the effectiveness of the developed classifier that represents the first component of a musical audio retrieval system.