Note:  Laboratories spend far too much time reinventing the wheel.  In the interest of fighting this tendency, we make available any software and data that we have generated over the last few years.  In doing so, we just have two requests.

1.  Please use the appropriate citation where appropriate (e.g., publications, technical reports, etc)

2.  If you make additional modifications to any software that you think would be useful for others, or if you find errors in any of our work, please let us know so we can make these modifications available.

3.  Software is covered by copyleft provision of GPL.  Pay particular attention to warranty provision, basically there is no warranty.  You should assure yourself that software functions properly according to your own criteria.  

 

Copyright (C) 2000, Prateek Bansal, Zenzi Griffin, & Daniel Spieler
This program is free software; you can redistribute it and/or
modify it under the terms of the GNU General Public License
as published by the Free Software Foundation; either version 2
of the License, or (at your option) any later version.

This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
GNU General Public License for more details.

For copy of the GNU General Public License
along with this program; if not, write to the Free Software
Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA  02111-1307, USA.
or click here

Many experiments in our lab involve the digitization of spoken responses.  In most cases, the digitization begins with the onset of some stimulus.  The main dependent measures are the onset of the spoken response and the duration of the response.  This MATLAB code parses the wav file, and detects the onset of the speech signal and the offset of the signal.  The algorithm uses criteria for onset and offsets based on energy, zero crossings, and autocorrelations.  Specific published references will be added soon and are available by emailing daniel.spieler@psych.gatech.edu.

This software uses functions included in the Signal Processing Toolbox from MATLAB.  It was written using MATLAB version 5.2.1 but should run under most versions that are likely to be in use as long as the toolbox is installed.

epdone.m is used to parse and display the results of a single wav file.  Use this to adjust parsing criterion if interested in tweaking the parameters for your particular purposes.  Note that the line marking detected onset is placed at the end of the windowed segment so it actually appears later than the numerical value which is taken as the beginning of the windowed segment.

epdbatch.m allows the parsing of multiple wav files in batch mode.  It scans the current working directory, and parses all .wav files within this directory.  It prompts for the name of a new text file that will contain the results of this parse.  In this file, the name of the wav file and the time until onset and the duration of the speech signal is output.

If you experience difficulty in reading wave files, try converting to 11.5 kHz.

Initial tests of this software using digitized sentences and comparison with hand measurements gave correlations of over .9.  HOWEVER, you should not take this on faith because many factors many influence the quality of the measurements and the robustness of the algorithm, including sound quality, type of speech signal, speech duration, and many, many other factors.  Be sure to test this software on the range of responses that will be present in your empirical data.  It is up to the final user to determine if the accuracy of the measurements are adequate.  We make absolutely no warranty regarding the accuracy of this software, and all responsibility for its use are the sole responsibility of the final user.
 

Bansal, P., Griffin, Z. M., & Spieler, D. H. (2001).  Epd: Matlab wave file parsing software.  [On-line]. Available: http://psychology.gatech.edu/~spieler/software.html

Copyright (C) 2000, Daniel Spieler
This program is free software; you can redistribute it and/or
modify it under the terms of the GNU General Public License
as published by the Free Software Foundation; either version 2
of the License, or (at your option) any later version.

This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
GNU General Public License for more details.

For copy of the GNU General Public License
along with this program; if not, write to the Free Software
Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA  02111-1307, USA.
or click here

These are a set of Matlab routines for optimizing parameters for the exGaussian and Weibul distributions and evaluating these parameter estimates.  I have done some preliminary testing of this software but you should convince yourself of the accuracy rather than take my word for it.  Some comments are written in the code itself.  If you make modifications, find errors, or make changes that speed the routines up, I would appreciate hearing from you.

At present, fits are glacial but seem decent.  

Matlab routines

Input
Batch file - first line contains "exgauss" or "weibull", remaining lines contain file names with RT data.
RT files - RT data points, one per line.  Distributions are delimited by a -1 line.  Arbitrary number of distributions per file.

Output
PDF files - contain Matlab figures of equal area histograms overlaid with fitted function
CDF files - contains matlab figures of empirical CDF and fitted CDF
Parameter estimates - parameter files contain (of course) parameter estimates, and observed and critical values for Chi-square and Kolmogorov goodness of fit tests.

Copyright (C) 2000, Amit Mookerjee, Daniel Spieler, Zenzi Griffin
This program is free software; you can redistribute it and/or
modify it under the terms of the GNU General Public License
as published by the Free Software Foundation; either version 2
of the License, or (at your option) any later version.

This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
GNU General Public License for more details.

For copy of the GNU General Public License
along with this program; if not, write to the Free Software
Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA  02111-1307, USA.
or click here

This software provides the ability to present pictures, text, and/or audio on a PC and digitize an individuals response using a SoundBlaster 16 sound card.  Native DOS.

Included in this software is eyetracking code that is likely to be completely unusable to anyone unless they have exactly the same set up that we do.  Our main purpose is to make the other functions of the software available, particularly the ability to digitize spoken responses.  This works very nicely in conjunction with the MATLAB code above for parsing wave files.  Included in the zip file are executable code, along with all of the source code.  The software is compiled under DJGPP, a 32bit DOS environment.

Mookerjee, A., Griffin, Z. M., & Spieler, D. H. (2000).  Eye2:  DOS software for speech and eye movement recording.  [On-line] Available: http://psychology.gatech.edu/~spieler/software.html