Directly modeling voiced and unvoiced components in speech waveforms by neural networks | IEEE Conference Publication | IEEE Xplore