Training a vowel contrast using visual articulatory feedback

Abstract

The /i/–/I/ distinction can be difficult to make in speech for deaf, dysarthric, and many non-native English speakers. A computer program was designed that prompted speakers, with orthograpraphy, to produce a simple single syllable word containing either vowel and recorded the speakers response. The program provided feedback in one of two forms: either auditory (AF) or auditory and visual articulatory feedback (VAF). The auditory feedback was the actual and correct pronunciation of the word. VAF provided schematic representations of where the tongue was according the speaker’s actual production of the vowel in relation to where it would be placed for the best production of the target vowel. Articulatory representations were inferred automatically based on the measured formant frequencies of the speaker’s production. Four subjects who were particularly poor at making the vowel distinction were recruited to test VAF in a single subject case study design. After baseline scores for the vowel distinction were established in four sessions, each subject was given training using each of AF and VAF, with no words the same for each of the types of training. There was more improvement in the vowel distinction for the auditory and visual feedback compared to the auditory feedback alone.