SpeechEnhancement This is a speech enhancement project using LSTM recurrent neural network. The feature used here is the MFCC of the input sound signal. The model tries to learn a binary / ratio mask in the time-frequency domain (after short time Fourier transform).