Neural harmonic cancellation model

[Not complete yet. Stay tuned!]

There are actually two models. First is a 'neural cancellation filter', that models the effect of removing spikes from a spike train to reduce effects of harmonic interference. The second is a 'concurrent vowel identification model' that adds just enough detail to predict identification rates, and the number of vowels reported per stimulus, in a double-vowel identification experiment.

This page points to software that can be used to reimplement the models. Edges are rather rough so it may not be trivial to set up. The software was written in ANSI C on a SUN under sunos4.1.1 using gcc. I originally used software from John Culling's excellent IWAVE package for peripheral filtering and hair-cell transduction. As that is not officially released yet, I rely here on the excellent Loughborough lutear package (1.6.1). Scripts are in Perl.

• Downloading and installing.
• Conventions.
• Making a summary autocorrelation function (SACF).
• Determining the dominant period.
• Using the cancellation filter to obtain an SACF representing the dominated vowel.
• Same, but with a place representation (rate vs CF).
• Making a set of template SACFs to match to.
• Comparing an SACF (filtered or not) to a template to identify the vowel
• Deciding whether to report one vowel or two.
• Predict the outcome of the double-vowel experiment.