// lying discrete fourier transformation
// this has no mathematical proofs behind it
// key difference is that it operates in a sliding window mode without holding the window buffer itself
// instead it works in a faucet way, with a single scalar complex state for every frequency that you want to monitor stored and operates with this state one sample at a time
// absolute value of the state directly correlates with freqency domain ordinate value
// temelji na diskretni fourierjevi transformaciji in na weighted moving average
// namesto izhodnega vektorja ob množenju  ...
// nvm najprej bom poizkusil narediti dft po krščansko, potem se bomo šli glupe optimizacije