Composer Classification in Symbolic Data Using PPM

The aim of this work is to propose four methods for composer classification in symbolic data based on melodies making use of the Prediction by Partial Matching (PPM) algorithm, and also to propose data modeling inspired on psycho physiological aspects. Rhythmic and melodic elements are combined instead of using only melody or rhythm alone. The models consider the perception of pitch changing and note durations articulations then the models are used to classify melodies. On the evaluation of our approach, we applied the PPM method on a small set of monophonic violin melodies of five composers in order to create models for each composer. The best accuracy achieved was of 86%, which is relevant for a problem domain that by now can be considered classic in MIR.