I need to have my frequency spectrum divided in 1/3 octave bands, therefore I sum the magnitude in the bins that fall in the ranges of the bands. I understand from https://ccrma.stanford.edu/~jos/sasp/Summing_STFT_Bins.html that says that "Summing adjacent FFT bins in the STFT, therefore, synthesizes one sample from a wider pass-band implemented using an FFT." That is the only information I found about this topic, so I am doing the following:
- STFT of my signal, with a window size of 4096 and nfft of 4096 too
- Average over time
- Sum of the magnitude of frequency bins that fall in a certain range (corresponding to the low and upper boundaries of the octave bands)
I understand that this is the correct way to do it, however I have seen that other people sum the square magnitude to get the frequency spectrum in the octave bands and it does not seem clear to my why it is done that way.
Can anyone help me here or point to a reference that I can read?
Thanks in advance!