The long-term objectives of the study are to 1) determine how vowel context affects the perception of an adjacent consonant for normal-hearing (NH) and hearing-impaired (HI) listeners, 2) determine the predicting factors of vowel masking on an adjacent consonant for people with hearing impairment, and 3) delineate implications for the design of hearing aid algorithms. This application adopts the framework that successful speech signal processing depends on the auditory system's ability to recover the temporal and frequency properties present in the signal, despite the assimilation between signal components.
The specific aims of the study are to describe whether and how the three aspects of vowel masking 1) spectral characteristics, 2) temporal characteristics, and 3) syllable presentation level, affect the audibility of an adjacent consonant in NH and HI listeners. Concise description of design: First, a series of CV, VC, and VCV nonsense speech syllables spoken by two speakers will be recorded. Second, consonant identification using the recorded syllables will be tested at three presentation levels (55 dB, 70 dB, and 90 dB SPL). A set of syllables will be selected for each subject for masking pattern measurement based on the identification results. Third, masking patterns of the selected syllables as a function of temporal position and presentation level will be acquired. Correspondence analysis will be used to search for the relationship between the consonant identification performance and vowel context, presentation level, and syllabic position of the consonant. A series of metrics will be calculated to compare the similarity and difference between masking patterns under different conditions and between NH and HI listeners. Relevance to public health: The proposed research increases our understanding of the effects of hearing loss on consonant perception and how it varies with vowel context. The knowledge may contribute to the design of hearing aid algorithms, such as adjusting the gain of a weak sound according to the spectral components of its neighboring sound.
