Advanced feature extraction from speech

[yoshua]

Select and code up advanced feature extraction algorithms that may be relevant to the task of emotion classification (look at what the previous papers on the subject have reported).

[boulanni]

In the literature, people do pretty much the same as I have done: use libraries to extract a bunch of frame-level audio features and concatenate them.

The most commonly used ones seem to be MFCC + derivatives, autocorrelation and spectral statistics (moments, roll-off, centroid, flux, etc.) which I put into the 'minimal' set.

Features sometimes mentioned that I don't have yet (to add): pitch, harmonicity, LPC coefficients and zero crossings.