Voice conversion using deep neural networks with speaker-independent pre-training | IEEE Conference Publication | IEEE Xplore