DRED: Deep REDundancy Coding of Speech Using a Rate-Distortion-Optimized Variational Autoencoder | IEEE Journals & Magazine | IEEE Xplore