Weighted Markov Chain Model for Musical Composer Identification

Citation

Kaliakatsos-Papakostas MA, Epitropakis M & Vrahatis MN (2011) Weighted Markov Chain Model for Musical Composer Identification. In: Di Chio C, Brabazon A, Di Caro G, Drechsler R, Farooq M, Grahl J, Greenfield G, Prins C, Romero J, Squillero G, Tarantino E, Tettamanzi A, Urquhart N & Uyar A (eds.) Applications of Evolutionary Computation: EvoApplications 2011: EvoCOMNET, EvoFIN, EvoHOT, EvoMUSART, EvoSTIM, and EvoTRANSLOG, Torino, Italy, April 27-29, 2011, Proceedings, Part II. Lecture Notes in Computer Science, 6625. EVO 2011, Torino, Italy, 27.04.2011-29.04.2011. Berlin Heidelberg: Springer, pp. 334-343. http://link.springer.com/chapter/10.1007/978-3-642-20520-0_34; https://doi.org/10.1007/978-3-642-20520-0_34

Abstract
Several approaches based on the ‘Markov chain model' have been proposed to tackle the composer identification task. In the paper at hand, we propose to capture phrasing structural information from inter onset and pitch intervals of pairs of consecutive notes in a musical piece, by incorporating this information into a weighted variation of a first order Markov chain model. Additionally, we propose an evolutionary procedure that automatically tunes the introduced weights and exploits the full potential of the proposed model for tackling the composer identification task between two composers. Initial experimental results on string quartets of Haydn, Mozart and Beethoven suggest that the proposed model performs well and can provide insights on the inter onset and pitch intervals on the considered musical collection.