On the state definition for a trainable excitation model in HMM-based speech synthesis.

Ranniery Maia, Tomoki Toda, Keiichi Tokuda, Shinichi Sakai, Shun Nakamura

Browse the full ICASSP paper archive.