Auditory Instantaneous Frequency Model

From Emcap

Contents

Auditory Instantaneous Frequency Model

by Susan Denham <denham@plymouth.ac.uk>

Centre for Theoretical and Computational Neuroscience, University of Plymouth

2008-03-11

Auditory Instantaneous Frequency Model code

What is the Module?

This is the Matlab implementation of the auditory instantaneous frequency model, reported in (Denham et al 2007).

Denham, S.L., Coath, M., Balaguer-Ballester, E. Sparse Time-Frequency respresentations in auditory processing, 19th Int. Congress on Acoustics, Madrid, September 2007.


What are the inputs and outputs?

results = auditoryInstantaneousFrequency(fName, nChans, resolution, binSize, doDisplay)

Inputs:

  • fName name of the sound file containing the stimulus (.wav or .au)
  • nChans number of cochlear channels; default 30.
  • resolution size of the instantaneous frequency bins; default 0.5 semitones
  • binSize time sampling bin size of the instantaneous frequency representation; default 0.001 s
  • doDisplay flag set for a plot of the cochlear, instantaneous frequency and transient responses; default 0.

Outputs:

  • results data structure in which the instantaneous frequency model responses are returned:
  • .eRespIF instantaneous frequency response (binned into spectrogram form)
  • .tRespIF transients of the instantaneous frequency response (binned into spectrogram form)
  • .tAxis time for each column in eRespIF, in seconds
  • .fAxis frequency corresponding to each row in eRespIF, in Hz

What are the parameters?

The auditory instantaneous frequency model has parameters which affect the spectral and temporal resolution of the output representation. The frequency bin size, expressed in semitones can be made arbitrarily small, at the cost of increased processing time, and time bin size is limited only by the stimulus sampling rate. The number of channels in the cochlear model can also be adjusted to increase resolution. There are also non-adjustable parameters which affect the smoothness of the frequency tracking; alpha, and beta. These have been adjusted to give a reasonable match to the FFT instantaneous frequency response.

What are the system requirements?

The only requirements are a copy of Matlab (developed in version R2007b), and the Matlab Signal Processing Toolbox. This model is in continuous time mode, and could be integrated into the continuous time version of the perceptual onsets code when it becomes available.

How to Cite

Denham, S.L., Coath, M., Balaguer-Ballester, E. Sparse Time-Frequency respresentations in auditory processing, 19th Int. Congress on Acoustics, Madrid, September 2007.

How to Use, an example

The function call has the following form: >> results = auditoryPOnsets(fName,nChans,resolution,binSize,doDisplay)

An example of the use of this function: results = auditoryInstantaneousFrequency('BKBQ0104.wav',30,.5,.005,1);

Results returns with the following fields and the display below.

>> results =

     tAxis: [1x342 double]
     fAxis: [168x1 double]
   eRespIF: [168x342 double]
   tRespIF: [168x342 double]
Speech example
Enlarge
Speech example