A unified speaker-dependent speech separation and enhancement system based on deep neural networks | IEEE Conference Publication | IEEE Xplore