Applying automatic phoneme analysis to the assessment of motor impaired speech

The feasibility of applying automatic phoneme analysis to articulation testing of motor-impaired speech has been investigated. An articulation test consisting of twenty-four monosyllabic words has been established. After the subject is prompted to say each word, the acoustic speech waveform is digitized and stored. Features are then extracted to facilitate segmentation of each word into individual phonemes. The phoneme sequence uttered by the subject is then compared to the expected sequence to determine the errors that have occurred.<>