A Two-Stage Deep Representation Learning-Based Speech Enhancement Method Using Variational Autoencoder and Adversarial Training | IEEE Journals & Magazine | IEEE Xplore