The Bloomsbury Colleges | PhD Studentships | Studentships 2018 | Dynamic Bayesian Models for the Analysis of Music
Document Actions

Dynamic Bayesian Models for the Analysis of Music

Principal Supervisor: Dr Georgios Papageorgiou (Birkbeck, University of London)

Co-Supervisor: Professor Richard Widdess (SOAS, University of London)

Co-Supervisor: Dr Brad Baxter (Birkbeck, University of London)

Project description

Quantitative analysis of music is important in many areas of research and it has many meaningful applications. Research questions that can be addressed by such analyses include classification of musical works and characterization of the degree of similarity between them. Practical applications include building systems for aiding music theory study and teaching, designing music search and organization tools, and applications in machine listening, for instance for automatic speech recognition. While most studies have prioritised western popular or classical music, it is important to test the validity of models cross-culturally.

This PhD project will explore the application of Bayesian dynamic nonparametric models for addressing important research questions in the area of musicology. There are two important objectives. The first one is to develop new statistical methods that are nonparametric and dynamic in nature i.e. they rely on a minimal set of assumptions and they can be used for analysing sequential data, such as music data. The second one is to apply the developed methods to analyse North Indian music, and in classifying and characterizing the degree of similarity between North Indian musical works.

Important components of the project are the exploration of hidden Markov models (Rabiner 1989) and nonparametric mixtures of hidden Markov models (Ren et al. 2010), the development of Markov chain Monte Carlo algorithms and the implementation of these in a software, and applications of these models for the analysis of music data.

Subject areas/ keywords

Analysis of world musics; Bayesian statistics; Hidden Markov models; Markov chain Monte Carlo; North Indian classical music

Candidate requirements:

We invite applications from highly motivated students with a first class/upper second undergraduate degree or a Master’s degree in statistical sciences. Candidates must have experience in Bayesian statistics and computer programming and a strong interest in multi-disciplinary research. Candidates should also have an academic or professional qualification or advanced practical expertise in music, preferably Indian classical music.

Key References

  1. Mavromatis, P. (2005), A Hidden Markov Model of Melody Production in Greek Church Chant, Computing in Musicology, 14, 93-112.
  2. Rabiner, L. (1989), A Tutorial on Hidden Markov Models and Selected Applications in Speech Recognition, Proceedings of the IEEE, 77, 257–286.
  3. Rohrmeier, M. and Widdess, R. (2016), Incidental Learning of Melodic Structure of North Indian Music, Cognitive Science (accepted March 2016).
  4. Ren, L., Dunson, D., Lindroth, S. and Carin, L. (2010), Dynamic Nonparametric Bayesian Models for Analysis of Music, Journal of the American Statistical Association, 105, 458-472.
  5. Savage, P. and Brown, S. (2013), Toward a New Comparative Musicology, Analytical Approaches to World Music, 2, 148–198.

Further information

Further information about PhDs at Birkbeck is available here.

The successful student will be registered at: Birkbeck, University of London

How to apply

Please follow the online application process for the Full-Time PhD in Mathematics and Statistics at Birkbeck indicating your interest in working on this project.

Additionally, email your CV and a letter of interest, which sets out your qualifications for the studentship to Dr Papageorgiou.

Application forms and details about how to apply are available here:

Queries about the application process should be sent to:

The closing date for applications is 5pm, Wednesday 28 February 2018