use wavelets to discover frequency at given times; wavelets are the windowing; compare with discreet frequency windows perhaps, for more precision?
wavelets will help decipher amplitude changes, envelope detection;
use fft to get major frequencies, and wavelets stretched to those wavelengths to detect ampltudes over time, which in turn will express rhythmic information
morphing overtone frequency amplitudes
with regard to samplerate sr
, frequency f
will switch f-1
times every sr/2
; additionally, a switch will occur every sr/f
a frequency is encoded by flipping a bit every ceil(sr/f) - 1 % 2
something about euclidian beats
where E(f,sr) where f is how many times the bit flips, and sr - f is how many bits it skips, at intervals floor(sr / f) plus split the mod over the intervals from the beginning
so E(5,16) yields 100100101010 5,16 1000100 100100 100
the minimum required length length of a symetric signal is also sr/2
, because the second half of the signal is the inverse of the first.
unless higher frequencies are encoded into the signal; this is possible to do; indeed, writing privately known frequencies greater than the nihquist limit into a signal could be a form of encryption. one could only detect them if one knows what they are; perhaps a range over a modulus would suffice for a large amount of data; fundamentally this may just be a phase shift of frequencies below the nihquist limit...