Perceptual adaptation to spectrally shifted vowels: experiment studies and an acoustic analysis framework

UnrestrictedMost currently available cochlear implant devices are designed to reflect the tonotopic representation of acoustic frequencies within the cochlea. Unfortunately, the electrode array cannot cover the entire cochlea due to physical limitations or patient-related factors. Therefore, CI patients generally listen to spectrally up-shifted and/or distorted speech. Acute studies suggest that speech performance is best when the acoustic input is spectrally matched to the cochlear place of stimulation; performance deteriorates as the spectral mismatch is increased. However, many CI users are able to somewhat adapt to spectrally shifted and distorted speech as they gain experience with their device. Motivated by both the theoretical and clinical implications of CI users’ perceptual adaptation, the present study explores perceptual adaptation to spectrally shifted vowels using behavioral studies and an acoustic analysis framework. Normal-hearing subjects are tested while listening to acoustic simulations of CI processing. These experiments are designed to address several issues related to perceptual adaptation to spectrally shifted speech. The first issue addressed by an experiment in this project is whether and where along the cochlea the lexically meaningful feedback is necessary for improving recognition performance of spectrally shifted vowels. Secondly, the effects of spectral shift extent on perceptual adaptation to spectrally shifted vowels are systematically investigated using an unsupervised learning protocol, including the effect of linearly spectral shift, the effect of different spectral envelope distortions, and the effect of spectral resolution on perceptual adaptation to severely-shifted vowels with and without concurrent exposures to moderately-shifted vowels. Finally, an acoustic analysis framework is developed to interpret experimental findings.; This framework is used to explore mechanisms that may underlie perceptual adaptation to spectrally shifted speech, as well as to functionally model speech recognition performance for spectrally shifted and/or degraded vowels in CI users. In terms of clinical relevance, the present study provides important guidelines for the development and refinement of fitting procedures and training protocols for post-lingually deafened CI users. In terms of theoretical relevance, the present study introduces some new perspectives regarding speech perception and adaptation. The result of the present study may also benefit currently available engineering techniques for automatic speech recognition