Division of Computing Science and Mathematics University of Stirling

Statistics of Natural Sound

Statistics of Natural Sound

There's lots of work on the natural statistics: workshops (Gordon 1997, 2000, 2002), whole journal issues (Network: Computation in neural Systems, 14, 3, 2003). Nearly all of this work is on images: there is much less work on the natural statistics on sound (or should that be sounds?). Of course, there is some interest: see the bibliography at the end. (and if you would like to add other papers to this, please tell me). The two areaes are not independent of each other - one might consider, for example, this in terms of certain types of optimal coding (see Bibliography 2).

There are a number of reasons why researchers might be interested in this area: these range from seeking better coding systems (for storage or transmission of sound), to understanding how sound has shaped animal auditory systems over both evolutionary time, and during development. My own interest in this area stems from trying to comprehend what the early auditory system is doing and why. I have been working on biologically inspired auditory processing for the what and where tasks for years (see elsewhere on this website!), and had become interested in why the auditory system is like it is. It was only a short step from there to looking at natural statistics of sound. I gave a seminar in Edinburgh (IMNS, School of Electrical Engineering) on May 4 2004, and the slides from this are here (in pdf). The sounds referred to in the slides are to be found here (these are sounds from PCAs, ICAs from some TIMIT utterances, and an extract from Coltrane's "Equinox").

I have had a presentation on this accepted for the forthcoming Gordon Conference on Sensory Coding and the Natural Environment, to be held in Oxford, September 2004.

Bibliography: 1: natural statistics of sound


H. Attias, C.E. Schreiner. Low-order temporal statistics of natural sounds, Advances in Neural Information Processing Systems 9, 27-33, 1997
S. A. Abdallah and M. D. Plumbley. If edges are the independent components of natural images, what are the independent components of natural sounds? In Proceedings of the International Conference on Independent Component Analysis and Blind Signal Separation (ICA2001), San Diego, California, December 9-13, 2001. Pages 534-539, 2001.
M.S. Lewicki, "Efficient coding of natural sounds", Nature Neuroscience 5(4) 354-363,2002.
D.J. Klein, P. Koenig, K.P. Koerding, "Sparse spectrotemporal coding of sounds", EURASIP J. Appl Sig Processing, 7, 659-667, 2003.

Bibliography: 2: natural statistics relevant to sound


M.S. Lewicki and T.J. Sejnowski, "Learning Overcomplete Representations", Neural Computation, 12, 2, 261-266, 2000.
B. J. Vincent, R.J. Baddeley "Synaptic energy efficiency in retinal processing", Vision Research, 43, 1283-1290, 2003.

Back to Professor Smith's home page.

Last updated: Friday, 23-Dec-2005 13:53:38 GMT

If you have any difficulties accessing this page, or you have any queries/suggestions arising from this page, please email:
Prof Leslie S Smith (lss(nospam_please)@cs.stir.ac.uk)

computing logos