I'm trying to reduce my magnitude spectra, which I need in octave bands. I have 5 measurements of a sound source, recorded with the same microphone at different times. I want to get an average model of my data. Therefore I do the following for each of my recordings:
- Divide my data in time windows
- Compute STFT
- Average STFT over time frames
- Sum bins that fall in a specific range, corresponding to different octave bands
Once I have my octave band spectra for my 5 files I do the average of all the 5 magnitudes. I can't find any information about averaging different magnitude spectra.
Am I doing this wrong? Any help would be appreciated.