OBJECTIVE: This is a pared down implementation of my algorithm for
estimating utterance-to-utterance variability.  In this pared down
form, I have a single folder with recordings in it (./recordings).  

We first estimate the voiced portions of all recordings using the Perl
script 'prepareUtterances.pl'.  This will use Praat to generate a
"point process" file in ./ppfiles.  It will also produce a "raw"
version of each, with the suffix ".ppraw".

We then use matlab to do the estimation for each pair.

In the present case, you would use a terminal window, change directory
to the current one, and call 'prepareUtterances.pl'.

You would then use Matlab, and call the function trackSynch.m,
providing 3 parameters:

(1) The referent
(2) The comparator
(3) The ppraw file for the referent

I usually simply take the shorter of the two files as the referent.
But one can do it both ways, and take the average of the two
estimates, if desired.

Here is the Matlab code:

>> trackSynch('recordings/utterance2.wav','recordings/utterance1.wav','ppfiles/utterance2.ppraw');

For real work, you will need to automate the batch processing of many
files, setting up your filenames, deciding on which ppraw file to use,
calling trackSynch.m, and storing your results somewhere.  This can be
done with a simple wrapper function to trackSynch.m.
