Takes a series of wav files of human speech and performs FFT on them, the frequency domain data is then passed into a feedforward NN and used to train it.
Minimum 100 epochs to a reasonable convergence for 3+ speakers
outputs array is used to define which speech sample corresponds to a person i.e Bob,John, Suzy
if the sample is from Bob , output = [1,0,0] John , output = [0,1,0] Suzy, output = [0,0,1]