Automatic model parameter generation for the speech training of deaf children

Describes a new method which uses acoustic information derived from a text-to-speech system (TTS) to model voice parameters in a speech training system for deaf children. An earlier speech training system (Javkin et al., 1993) presented a series of speech parameters, derived from articulatory instruments and acoustic analysis, in a visual form. In that system, teacher's speech is input to the system and used as a model for the children to follow, and the children's speech is monitored to provide feedback. As with other computer-aided speech training systems, the teacher-assisted trainer is limited by the time (typically one or two hours per week) the students have with speech teachers. Additionally, students training without a teacher are restricted to learning pre-stored utterances.<
>