Automatic Estimation of Articulatory Controls in a Text-to- Speech System
- Parthasarathy S.
The transformation of a linguistic description of English text to realistic speech waveforms is guided by a model of human speech production. The formulation of rules for the model parameters has always involved human feedback. The quality of speech synthesized using hypothesized rules is typically judged by a human listener, and perhaps by comparing the features of synthetic and natural speech. Rules are then modified to make the synthetic speech sound natural. This procedure is tedious, time- consuming, and may not result in optimal solutions.