Research seminar by Andrew McLeod

Research seminar by Andrew McLeod

September 7th, 15:00 at room 55.309 (Univesitat Pompeu Fabra)


Invited Research Seminar

Language Modelling for Automatic Music Transcription by Andrew McLeod


Automatic music transcription (AMT) is defined as the process of converting an acoustic music signal into some form of human- or machine-readable musical notation. Advances in automatic music transcription performance have been disappointingly slow, with accuracy still falling well below that of human experts at the task. In this presentation, I argue that this problem cannot be sufficiently solved without the use of some sort of music language model, trained on symbolic music data. To that end, I first present an HMM-based voice separation model which works on symbolic music data, both quantized and performed, and also show results of integrating it with an acoustic pitch-detection model for a cappella voice separation from an audio signal. I also discuss work on metrical structure (time signature) detection based on rhythmic analysis using a lexicalized PCFG, arguing that such a grammar can capture the long-range dependencies of musical rhythms. Finally, in future work, I look ahead to the integration of models such as these into a full AMT system with an acoustic model.



Andrew McLeod is a PhD student under the advisement of Mark Steedman at the University of Edinburgh, School of Informatics, Institute for Language, Cognition and Computation (ILCC), set to graduate in early 2018. He received a Bachelors and Masters in Mathematics and Computer Science from Emory University in the USA in 2013. His PhD subject is automatic music transcription, specifically working with symbolic music data to create a music language model based on music theory for the task.