irregularity confusion

[jamiebullock]

irregularity confusion

[q-depot]

I'm using the spectrum for both, below you can find a gist of my implementation(almost all features, some of them don't work), I ended up wrapping each feature into a class, after several refactoring and approaches this is what makes most sense to me.
mData is the input data, mDataN is set in the constructor of each class(not sure about this approach), in this case is the FFT_SIZE.

This is what you can find in the documentation

*data,: a pointer to the first element in an array of doubles representing the magnitude coefficients from the spectrum of an audio vector, (e.g. the first half of the array pointed to by *result from xtract_spectrum().

ciXtractFeature.cpp

Ah, right, missed that line (was right above the one I quoted :). In that case, FFT_SIZE / 2 (or my blocksize / 2) is correct, no?

[q-depot]

my FFT_SIZE is equal to BLOCKSIZE / 2, so the spectrum array is initialised with BLOCKSIZE but the bin numbers is FFT_SIZE, the other half of the array contains the bin frequencies.

[jamiebullock]

Yup, both irregularity functions take a block of spectral magnitude coefficients. Below shows a scaled
comparison between Irregularity (Krimphoff), Red, and Irregularity (Jensen), Green, for a linear cross-fade between a 440Hz sine wave and white noise. That is, we expect a noisy spectrum to be highly irregular. Smoothness measures more-or-less the same thing.

Thanks to both for helping me understand, cheers!