Doctor of Philosophy (PhD)
This research addresses the development of machine learning techniques used to create musical scores and performances that are inspired by the intersection of speech and music. Machine learning models are created from MIDI files that are transcribed from datasets of musical audio recordings and human speech audio recordings. Through the creation of succinct models, model based cross synthesis is possible. Models trained on musical MIDI data are asked to replicate MIDI data that approximate human speech. Alternatively, models that have been trained on MIDI data that approximate speech are asked to replicate musical MIDI data. The product of these developed techniques is a collection of piano music, Seven Piano Etudes Speaks the Moody Machine. These etudes are intended to be performed on one Yamaha Disklavier piano with two performers, one human pianist and one machine player piano.
Thompson, William Alexander IV, "Creating Musical Scores Inspired by the Intersection of Human Speech and Music Through Model-Based Cross Synthesis" (2022). LSU Doctoral Dissertations. 5876.