Visemes observed by hearing-impaired and normal-hearing adult viewers

Abstract

A series of VCV nonsense syllables formed with 23 consonants and the vowels /a/, /i/, /u/, and / / was presented on videotape without sound to 5 hearing-impaired adults and 5 adults with normal hearing. The two-fold purpose was (a) to determine whether the two groups would perform the same in their identification of visemes and (b) to observe whether the identification of visemes is influenced by vowel context. There were no differences between the two groups either with respect to the overall percentage of items correct or to the visemes identified. Noticeable differences occurred in viseme identification between the /u/ context and the other 3 vowel contexts; visemes with /a/ differed slightly from those with / / and /i/; and there were no differences in viseme identification for / / and /i/ contexts. Findings were in general agreement with other studies with respect to the visemes identified, provided it is acknowledged that changes can occur depending on variables such as talkers, stimuli, recording and viewing conditions, training procedures, and statistical criteria. A composite grouping consists of /p,b,m/; /f,v/; /theta sigma/; /w,r/; /tS,d3S,3/; and /t,d,s,k,n,g,l/.