I've been working on extracting data from an extremely noisy signal. The signal itself is the 1st derivative of raw mean squared (RMS) of an audio that may contain segments with some single low frequency (LF). The RMS window size I'm using is comparable to the range of the wavelengths of these LFs, hence RMS sequence and its derivative has amplitude modulation (AM) artifacts where a LF is present.
Now if I build a set of sequences averaged from the original one with different kernel size, one can see that each such average "absorbs" AM artifacts of a certain LF. (On these graphs the frequency rises from left to right).
Then I multiply the values of these sequences and get something that shows a more prominent structure, more or less close to what I'm trying to extract.
The idea itself is quite obvious and I'm sure it has a name. So I would appreciate if anybody could tell me the name and anything else related to this method :)
