An autoencoder neural-network based low-dimensionality approach to excitation modeling for HMM-based text-to-speech.

Srikanth Vishnubhotla, Raul Fernandez, Bhuvana Ramabhadran

Browse the full ICASSP paper archive.