Hector Raul Javkin
Speech Technol. Lab., Panasonic Technol., Inc., 3888 State St., Santa Barbara, CA 93105
Speech synthesis by rule and its subsequent development into text-to-speech have progressed into both science and technology for use by the general population, particularly by persons with disabilities. This paper briefly reviews the advances in speech synthesis by rule and text-to-speech and then focuses on a new development---the use of text-to-speech in a speech training system for deaf children. The method uses the acoustic parameters that a text-to-speech system supplies to its formant synthesizer and converts them to pseudoarticulatory parameters equivalent to parameters measured from instruments monitoring the child's production. For example, the relative frequencies of the nasal pole and nasal zero are converted to a ``nasalization index'' equivalent to the output of a nasal sensor. The method enables a student to type any utterance she/he wants to learn and see a representation of the articulation of that utterance that corresponds to the feedback received from instruments. Preliminary testing of the use of the method will be reported.