Physiologically Based Speech Synthesis

Part of Advances in Neural Information Processing Systems 5 (NIPS 1992)

Bibtex Metadata Paper


Makoto Hirayama, Eric Vatikiotis-Bateson, Kiyoshi Honda, Yasuharu Koike, Mitsuo Kawato


This study demonstrates a paradigm for modeling speech produc(cid:173) tion based on neural networks. Using physiological data from speech utterances, a neural network learns the forward dynamics relating motor commands to muscles and the ensuing articulator behavior that allows articulator trajectories to be generated from motor commands constrained by phoneme input strings and global performance parameters. From these movement trajectories, a sec(cid:173) ond neural network generates PARCOR parameters that are then used to synthesize the speech acoustics.