SIMON DURRANT

 

 Contents
Introduction
Research Interests
Publications
Teaching
Curriculum Vitae
Contact Details
 
Other Pages
Lateral Inhibition Software
Neural Networks Teaching

 

Introduction

I am Simon Durrant, currently Research Fellow at the University of Plymouth, where I am a member of the Interdisciplinary Centre for Computer Music Research.  I am also an Associate Tutor in the Department of Informatics at the University of Sussex where I am a member of the Bioinformatics and Machine Learning Laboratory. My current work is in the neuroscience of music, which combines my previous separate work in the psychology of music, and theoretical neuroscience.

 

Research Interests

Neuroscience of Music                                 

The neuroscience of music examines the relationship between music and the brain.  My interest is finding neural substrates for particular features of music, such as tonality, meter etc.., by using techniques such as fMRI and EEG.  By studying a rich, temporal and yet well understood art form (within a culture), we can gain considerable insight into how the brain works.  Listening to music involves both fundamental early auditory processes, and higher order cognitive functions, and therefore involve a considerable variety and complexity of brain areas.  In addition, by studying this, and comparing to activity in other domains such as language and pattern processing, we can also gain insight into how music is processed at a psychological level, which was the topic of some of my previous research.  Additionally, the motor skills involved in music performance, imagining music with no sound (music imagery), the therapeutic benefits of music and the educational benefits of music are all further topics that can benefit from, and in turn shed light on, the neuroscience of music.  Given this rich variety, it is not surprising that the neuroscience of music is a rapidly growing area of research.

I work on a project called LeStrum, with groups here at University of Plymouth (led by Professor Eduardo Miranda), University College London (led by Professor John Shawe-Taylor), the Leibniz Institute of Neurobiology (led by Professor Henning Scheich) and Johannes Kepler University, Linz (led by Professor Gerhard Widmer).  Currently, we are investigating the neural substrate for tonality via fMRI and EEG, building upon previous research by Janata et al (2002).  This is the first step towards establishing a more general and widespread framework for relating features of music to the neural substrate.

 

Theoretical Neuroscience

My interest in this area is concerned with developing models that help improve our understanding of information processing in the brain.  I have worked

I have investigated the link between independent component analysis (ICA) along with related techniques such as sparse coding, and receptive fields in the primary visual cortex.  I have also developed new algorithms which are generalisations of ICA to situations where the components may be correlated (CCA), or the basis functions may be correlated (CBA).  These algorithms have proven robust and useful in situations where the components are not known to be uncorrelated, which is true for most situations in the real world.

I have also looked at the role of negative correlation in neural processing (and beyond).  Negative correlation has a number of attractive statistical properties, such as space-filling and accelerated central limit convergence, which make it potentially useful in any system which has large numbers of noisy elements operating together, such as the brain.  Recent findings in data from the olfactory bulb and the inferior temporal cortex suggest that neural firing appears to be negatively correlated, and I've developed a model of how this can arise from lateral inhibitory connections in the neocortex, and demonstrated the superior performance of this model (to one with no lateral connections) on a stimulus tracking task.  The software for this model, along with a description, instructions and the accompanying paper, is available for download from the lateral inhibition software page.

In addition, I have looked at the phenomenon of suprathreshold stochastic resonance in the specific context of neural systems, and demonstrated how networks of spiking neurons can benefit from improved information transmission by the addition of noise, and that this noise can be controlled by the level of correlation in neural firing in cortical columns.

Finally, I have developed a non-threshold (non-neural) generalisation of the lateral inhibition technique, which allows de-noising of signals even when the noise is in the same frequency domain as the signal.  This technique, which is formally equivalent to a conditional maximum likelihood noise reduction approach, operates entirely in the time domain, and can be used in conjunction with any/all current frequency domain approaches.  In collaboration with Professor Keith Kendrick at the Babraham Institute and Professor Jianfeng Feng of the University of Warwick, I am currently developing practical real-world applications of this technique, which comes directly from research on neural systems.

 

Others

I have worked with artificial life models, including an examination and reappraisal of a well-known model of the Baldwin Effect, and retain an active research interest in the application of artificial life to understanding the origins of music.  This includes music as an evolutionary adaptation, the connection between music, evolution and language, and the evolution of particular musical styles.

Artificial neural network models also remain an active interest, both for their own theoretical properties and capabilities as AI techniques, and also in their specific application to music.  I have previously modelled the perception of musical meter using a modular neural network structure, demonstrating the interaction of pitch and rhythm in the perception of meter.  I have also shown the benefits of this modular approach more generally, and how it can be applied to other areas of music cognition.

 

Publications

Durrant, S.J. (2000):  Modelling the Perception of Musical Metre Using Neural Networks Proc. SRPMME 2000

Durrant, S.J. (2002): A Modular Neural Network Approach to Dynamic Music Perception Proc. ICMPC7 (pdf)

Durrant, S.J. and Huron, D. (2002): The Effect of Scale Degree on Melodic Accent: An Empirical Investigation CSML Internal Preprint (pdf)

Durrant, S.J. and Feng, J. (2006):  Negatively Correlated Firing: The Functional Meaning of Lateral Inhibition Within Cortical Columns Biological Cybernetics, 95:5, 431-453 (pdf)

Durrant, S.J., Kendrick, K. and Feng, J (2006):  Algorithms for Exploiting Negative Correlation Neural Computation (submitted) (pdf)

Durrant, S.J. and Feng, J. (2006):  Noise Reduction Though Lateral Inhibition Phy.Rev.Lett. (submitted) (pdf)

Durrant, S.J. and Feng, J. (2006):  Suprathreshold Stochastic Resonance in Neural Systems Tuned by Correlations Phy.Rev.E. (in revision) (pdf)

 

 

Teaching

I have taught on a variety of courses at both undergraduate and postgraduate level.  Teaching is at the University of Sussex unless indicated otherwise.

 

Curriculum Vitae

This is the abbreviated version; the full version can be downloaded in pdf format here: CV

Education

Employment

Publications

Skills

 

Contact Details

Mailing Address:

Dr Simon Durrant
Interdisciplinary Centre for Computer Music Research,
School of Computing, Communications and Electronics,
Portland Square,
University of Plymouth,
Drake Circus,
Plymouth PL4 8AA.
United Kingdom.

Telephone:  +44 (0)1752 232551

 

E-mail:  simon.durrant (at) plymouth.ac.uk

 

This page is maintained by Simon Durrant; last updated on January 29, 2007