Factors Contributing To Speech Perception Outcomes With Hearing Aid Signal Processing

Speech perception in hearing-impaired listeners can be adversely affected by various factors including the type of hearing aid signal processing, environmental limitations, and individual differences in the peripheral auditory physiology and higher-level cognitive processes. These factors together influence which aspects of the speech signal are potentially informative and beneficial for the hearing-impaired listener. Particular aspects were discussed in the context of two hearing aid signal processing techniques: amplitude compression and nonlinear frequency compression. Speech recognition in thirty-six hearing impaired and thirty-seven normal-hearing listeners were evaluated across generic and proprietary methods of amplitude compression in background noise and reverberation. Amplitude compression method had an overall effect on speech recognition, but did not vary across environmental conditions or listener-related differences. However, individual differences across hearing-impaired listeners influenced the use of audibility and temporal envelope information in noisy and reverberant environments. Improvements to existing acoustic models of speech perception were proposed with two neural models. The Neural-scaled entropy (NSE) model has potential to predict speech perception outcomes with nonlinear frequency compression technology in hearing aids, while the neural-signal-to-noise ratio envelope (neural-SNRENV) model has the potential to predict speech perception outcomes in noise due to individual differences in underlying physiology with hearing impairment. In general, results highlight the need to individualize the hearing aid fitting process in order to maximize useful speech information for a hearing-impaired listener. In order to achieve this goal, models based on auditory nerve output can be beneficial in designing measures to predict the nature of speech information being modified by hearing aid signal processing and the environment. These models can also provide insights into the encoding of processed and unprocessed signals in the impaired auditory system