Improved Duration Rules for Text-to-Speech Synthesis

The durations of phonemic segments in an utterance synthesized by rule are important because they affect both intelligibility and prosodic quality. There are many factors which influence segmental duration, including the identity of the segment, and extrinsic factors such as surrounding segments, lexical stress, syntax, and speaking rate. Several approaches to modeling the combined effects of the contributory factors have been proposed. Until recently, the duration rules of the experimental AT&T dyad synthesis-by-rule system have been primarily additive. Starting with a minimum duration for each segment, duration was augmented incrementally by each contributory extrinsic factor.