Sequential streaming of speech sounds under normal and impaired hearing

Abstract

Segregating and understanding speech within complex auditory scenes remains a major challenge for hearing-impaired (HI) listeners. This study compared the ability of normal-hearing (NH) listeners and HI listeners with mild-to-moderate loss to segregate sequences of speech tokens, consisting of an unvoiced fricative consonant and a vowel (CV), based on a difference in fundamental frequency (F0) and/or vocal tract length (VTL). The CVs were amplified and spectrally shaped to ensure audibility for the HI listeners. In the streaming task, the CV tokens were concatenated into sequences that alternated in F0, VTL or both. The resulting interleaved sequences were preceded by a “word” consisting of two random syllables. The listeners were asked to indicate whether the word (which varied from trial to trial) was present in the interleaved sequence. The word, if present, occurred either within one sequence or across the alternating sequences. Preliminary results showed no difference in performance between the two groups, suggesting that the listeners with mild-to-moderate sensorineural hearing loss are able to use differences in F0 and VTL to segregate speech sounds in situations where there is no temporal overlap between the competing sounds.