Maarten Versteegh http://lands.let.ru.nl/~versteegh/ maartenversteegh AT gmail DOT com Centre for Language Studies Radboud University Nijmegen Licensed under GPLv3 Extract frequency and acoustic similarity information from infant- and adult-directed speech. To run the example for extracting frequency counts I sent around: - make sure you have python2.7+ installed - install bidict from pypi (http://pypi.python.org/pypi/bidict/0.1.1) if you haven't already - set up your file paths as outlined in CORPUS_FILE_STRUCTURE_LAYOUT - run frequency_extractor.py script in /src directory To use any of the audio measurement, clustering or classification functions, make sure you have the following: - numpy and scipy (http://numpy.scipy.org/), linear algebra functions and scientific computing libraries - matplotlib (http://matplotlib.sourceforge.net/), plotting library - sklearn (http://scikit-learn.org/), machine learning library - for the formant measurements, I've used Praat. Make sure that python can call it, i.e. 'praat' should be in your PATH. (For OSX: the Praat program installs to /Applications/Audio/Praat.app/Contents/MacOS/Praat, which is not in the PATH by default. Easiest way to solve that is to put a symlink in /usr/bin to there, i.e. open a Terminal and do $ sudo ln -s /Applications/Audio/Praat.app/Contents/MacOS/Praat /usr/bin/praat For Linux: Praat installs to /usr/bin by default, so you don't have to do anything. For Windows: god knows.)