IBM'S Text-To-Speech Technology by Maria Smith

This presentation gave an overview of IBM's text-to-speech technology. The two primary synthesis methods, formant and concatenative, were described. A high-level presentation of the full concatenative synthesis process was given, starting from text normalization to speech generation. A demonstration of both formant and concatenative synthesis was given at the end of the presentation.

Bio:

Maria Elena Smith received her Ph.D. in Computer Science from Cornell University. Upon completing her studies, she joined IBM where she has worked since 1990. Her career at IBM has involved work on various projects, most notably, Speech Technologies. She joined the speech recognition team in 1994, when the technology was in its infancy. She developed the first continuous speech recognition language model, a key contribution to the first continuous speech recognition product, ViaVoice. She is now the architect of the text-to-speech project, leading a world-wide team in the productization of text-to-speech technologies provided by IBM research as well as in the development of concatenative text-to-speech voices.