A semi-supervised single channel speech enhancement method using Variational Autoencoder and Non-negative Matrix Factorization
This repository remains the same strustures as A VARIANCE MODELING FRAMEWORK BASED ON VARIATIONAL AUTOENCODERS FOR SPEECH ENHANCEMENT. We replace MCMC-EM algorithm with backpropagation-EM algorithm for performing semi-supervised speech enhancement.
We use a variance modeling framework to model the speech and noise：
VAE for speech modeling
NMF for noise modeling