Mario A. Svirsky
Speech Commun. Group, Res. Lab. of Electron., MIT, 50 Vassar St., Rm. 36-525, Cambridge, MA 02139
The Ineraid cochlear implant filters the speech spectrum into four frequency bands and delivers each signal to one of four electrodes in the cochlea. Higher frequency bands stimulate more basally placed electrodes. Ineraid users may recognize vowels using temporal cues, place-of-stimulation cues (such as amount of energy sent to each electrode) or a combination of place and temporal cues. A mathematical model is proposed to account for vowel perception by users of the Ineraid cochlear implant. The model is computationally similar to the one that was proposed for users of pulsatile cochlear implants [M. A. Svirsky and S. H. Svirsky, J. Acoust. Soc. Am. 92(4), 2415--6 (1992)], but the perceptual dimensions involved are different in the two cases. To test the model vowel identification data obtained with ``conflicting cue'' vowels was used [Dorman et al., J. Acoust. Soc. Am. 92(6), 3428--31 (1992)]. The model was run under three different assumptions: Ineraid users only employ temporal cues for vowel perception, they only employ place cues, or they combine temporal and place cues. Model output fit the data well only under the temporal/place assumption. This suggests that Ineraid users combine information from temporal and place cues in order to recognize vowels. [Work supported by NIH Grant No. R03-DC01721 and Contract No. N01-DC-2-2402.]