A machine learning approach for generating expressive musical transcriptions

In this thesis a method to obtain musical transcriptions annotated with expressive labels by means of machine learners is presented. To reduce the complexity of this problem the scope of the project has been narrowed and concretized to transcribing guitar performances. The transcription has been labeled with expressive labels concering the right hand playing position of the guitarist. This proof of concept has been executed in a dedicated environment called exprimulator that facilitates several tasks such als creating optimal classifiers, transcribing, annotating and visualizing resulting (intermediairy) data