Classification of melodies by composer with hidden Markov models

Probabilistic-Logical Modeling of Music

“…Another group of researchers has focussed on probabilities and HMMs as a basis for developing music analysis and synthesis applications (e.g. [7,6]). An important disadvantage of that approach is that it is not very flexible: it is hard to incorporate expert knowledge in such models or to experiment with variants of HMMs.…”

“…values(tr_ss(_PrevS), [1,2,3,4,5,6,7] For every note, it is first determined whether it is a rest or a "real" note. The note value of the "real" note is checked in check real note.…”

“…Pollastri and Simoncelli [7] have used HMMs for classification of melodies by composer. In a similar way, we will use our model to classify fragments of polyphonic music.…”

Musical Style Identification Using Grammatical Inference: The Encoding Problem

“…A 3% error in Style Identification is obtained, comparing with a 5% using k-TSI and a 7.66 % using ECGI techniques. Although comparisons with the success rates of other style identification models is not very meaningful unless the same datasets are used, if we look to other similar studies [11] [12] [17], these average rates of success can be considered as quite good. It is worth noting that GI techniques tend to need a larger quantity of training samples to get good results.…”

“…But other applications can be musicology (finding authors for anonymous pieces) or music education. Some AI techniques that have been employed are Hidden Markov Models [11], SelfOrganising Maps [12] and Neural Networks [17]. This paper is focused in our MSI work [4] [5] that have been extended by adding new coding schemes for music.…”