The effectiveness of speech therapy and training is limited by the time available for interaction between the therapist or teacher and the pupil. a device that could automatically compare a brief utterance to a standard or reference version of that utterance, and display a measure of the trial utterance on a speech quality scale, would serve as an important extension of speech therapy and training between lessons. Research into the nature of speech and rapid development of signal processing technology now permit construction of such a device at reasonable cost. We propose to determine the feasibility of applying phoneme-based speech recognition algorithms to evaluation of impaired speech.
We aim to display the result of evaluation on a scale that is consistent with the evaluations of trained speech scientists for the same examples of impaired speech.
