Author: Woojae Han

Publisher:

Published: 2011

Total Pages:

ISBN-13:

Individuals with {it sensorineural hearing loss} (SNHL) are prescribed hearing aids and/or a cochlear implant, based on their pure-tone threshold and speech perception scores. Although these assistive listening devices do help these individuals communicate in quiet surroundings, many still have difficulty understanding speech in noisy environments. Especially, listeners with mild-to-moderate SNHL have complained that their hearing aids do not provide enough benefit to facilitate understanding of normal speech. Why is it that the modern hearing aid, even with a high level of technology, does not produce one-hundred percent efficiency? We shall show that the current clinical measurements, which interpret the result as a mean score (e.g., pure-tone average, speech recognition threshold, AI-gram, etc.), do not deliver sufficient information about the characteristics of a SNHL listener's impairment when hearing speech, and thus, result in a poorly fitting hearing aid.\ This dissertation addressed three key questions, fundamental to clinical audiology and hearing science: (1) How well do the results of standard clinical tests predict the speech perception ability of SNHL patients? (2) Are the existing methods of hearing aid fitting (e.g., the half-gain rule, NAL-R, etc.) appropriate for modern hearing aid technology? (3) How useful are measured error patterns of speech perception in SNHL patients in addressing these perception errors?\ Four sub-studies were conducted for finding answers to the proposed questions: textbf{Study I} measured individual consonant errors to quantify how each hearing-impaired (HI) listener perceives speech sounds (e.g., high- vs. low-error consonants), and then compared the individual consonant errors to the results provided by currently used clinical measurements to ascertain the differences. The results of Study I showed that the HI ear had significant errors in receiving only a few consonants. There was a low correlation between the error rates of high-error consonants and either degree and configuration of pure-tone hearing threshold or average consonant scores. textbf{Study II} examined how reliably a CV listening test could measure a HI listener's consonant loss using only {it zero-error} (ZE) utterances (defined as utterances for which normal hearing (NH) listeners incur zero errors, citep{singh2011}) and having a statistically suitable number of presentations in CVs, in order to characterize unique HI consonant loss. We provided graphical as well as statistical analysis to see not only the error rate (%) of a target consonant but also its pattern of specific confusions. As we found in Study I, there was no measurable correlation between pure-tone threshold and the error rate, or no identification of high-error consonants in HI ears. As noise increased, the percentage of error and confusions of target consonants increased. Although some consonants showed significantly higher errors and resulted in more confusion than others, HI ears have a very different consonant confusion pattern than NH ears, which may not be either measured or analyzed by the use of average scores. Comparison between the two (separated) phases of the experiment (Exp.~II) showed a good internal consistency for all HI ears. textbf{Study III} investigated whether or not NAL-R amplification might offer a positive benefit to speech perception of each HI listener at the consonant level, i.e., differentiates consonants that are distorted with amplification from those that achieve a positive benefit from amplification. The results were then compared to the current clinical measurement to see a relation between consonants which have positive amplification benefit and hearing loss. Regardless of NAL-R amplification, HI listeners have their own consonant dependence and the dependence was not predicted by either pure-tone threshold or aided threshold. HI listeners who have symmetrical hearing loss do not have the same positive amplification benefit to the two ears. textbf{Study IV} characterized consonant perception errors of each HI listener by identifying missing critical features of misheard consonants as a function of signal-to-noise ratio (SNR), while following the same procedure (i.e., increasing the number of ZE utterance presentations up to 20) as in Study II, yet for the NAL-R amplification condition. As the noise increased, consonant error and confusions were significantly increased, although by applying gains provided by NAL-R amplification correction. The percentage of error and confusions of the target consonants were different across the HI ears, thus could not be averaged. When the results of Study IV were compared with those of Study II, a significant amplification effect is found. Generally, the percentage of error and confusions were decreased in the NAL-R condition as a function of SNRs. However, typical average analysis, using mean score and grouping the HI ears, failed to explain the idiosyncratic characteristics of HI speech perception.\ Overall, this series of studies concluded that current average measures and analyses have a serious, even fatal limitation in finding problems of HI speech perception. Therefore, we have explored the use of the nonsense CV test for as a more precise measure. We will show that this can make significant contributions to HI speech perception. We propose that this CV test and its application might be utilized in the clinical setting, to improve the diagnosis of HI speech perception. This research will help HI listeners hear day-to-day conversations more clearly, as well as aid in audiological diagnosis and successful rehabilitation to increase speech perception for HI listeners.