So I've been researching wavelet transforms and FFT. I want to feed wavelet transforms of a 1D signal in time into a neural net and train against a target variable at each time step. The idea being the WT/FFT can help the net pick up on data structures.
Only issue being I have no idea how to feed the coefficients of the FFT into the model because my approximation signal is length 87 while the original series had 44,000 samples.
I looked into zero-padding to restore the number of samples but that is apparently a technique used on the time-series itself before decomposing with FFT.
Any help appreciated, hard to find material on exactly this subject.
paper im recreating: https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0180944