Assessing the contribution of spectral cues to recognition of frequency-lowered consonants

Abstract

Commercially available strategies for restoring audibility of critical high frequency cues to patients with severe high frequency hearing loss translate information from high-frequency regions with unaidable hearing to lower-frequency regions with aidable hearing. Methods for synthesizing lowered spectral features, rather than generating them from the signal, have been proposed, though no commercially available hearing aid uses such a method. We assessed consonant discrimination under three configurations of a spectral feature synthesis method intended for use in frequency lowering. Lowered consonants were rendered using one or two narrowband noise components presented in a low frequency region with aidable hearing. Different configurations conveyed different spectral cues intended to distinguish among lowered consonants. In a short pilot study, preliminary analysis found no significant difference in consonant matching accuracy between the different configurations, suggesting that listeners were not making use of additional spectral cues when they were available. However, there were some observable trends in the data, as well as open questions about the possible impact of training and acclimatization on listener performance. We will present the findings of a more extensive study to determine whether listeners with training can make use of enhanced spectral cues to distinguish among frequency-lowered consonants.