ESCA Workshop on Audio-Visual Speech Processing (AVSP'97)
September 26-27, 1997
How are images of mouth shapes perceptually organised? We manipulated a 'vowel triangle' of photographic still images of the point English vowels ('oo, cee' and 'ah') and a similar triangle of consonants ('th,ff,mm'). Using a face-feature landmark algorithm, computer warping techniques controlled the physical step-size between images. Traditional two-stage categorical perception (CP) experiments (ABX discrimination and identification) with normal hearing subjects suggested that while the vowel series showed CP, consonant images did not. These findings are discussed in relation to recent findings of categoricity in visual image perception and implications for visual language processing.
Bibliographic reference. Campbell, Ruth / Benson, P. J. / Wallace, S. B. (1997): "The perception of mouthshape: photographic images of natural speech sounds can be perceived categorically", In AVSP-1997, 1-4.