A Computer-based Articulation Training Aid for Short Words (cata)

A COMPUTER-BASED ARTICULATION TRAINING AID FOR SHORT WORDS (CATA) Mukund Devarajan Old Dominion University, 2003 Director: Dr. Stephen A. Zahorian Several improvements in the vowel articulation training aid (VATA) are described, as well as the efforts to extend the visual feedback system to operate with short words in the form of consonant, vowel and consonant (CVC). The extended version of the visual feedback system is referred to as CATA (Computer-based Articulation Training Aid); the vowel version of the aid (VATA) only operates with ten American English monopthong vowels. Improvements in VATA include the use of a neural network (NN) recognizer method to prune a large database of vowel recordings to eliminate noisy and/or mispronounced tokens. The spectral jitter problem, previously present in the VATA, has also been corrected. Initial steps in the development of CATA involved database preparation. The training methodologies and the step-by-step procedure for using Hidden Markov Modeling (HMM) for recognizing and segmenting a CVC database are described. The signal processing and recognition steps involved in building a real-time display system to provide visual feedback about the quality of pronunciation of the CVCs are described in detail. An attempt at using a time-delay neural network (TDNN) classifier for distinguishing phonemes present in the CVCs is described. Experiments conducted to improve the VATA and the initial results obtained with the CVC display system are reported.