The Effect of Frequency Resolution on Intelligibility Sentence and its Relevance to Cochlear Implant Design

Authors

Seth H. Roy, University of Nebraska-LincolnFollow

First Advisor

Thomas Carrell

Second Advisor

Ashok Samal

Date of this Version

12-2016

Citation

Roy, Seth. (2016). The Effect of Frequency Resolution on Intelligibility Sentence and its Relevance to Cochlear Implant Design (Masters Thesis). University of Nebraska-Lincoln, NE.

Comments

A THESIS Present to the Faculty of The Graduate College at the University of Nebraska In Partial Fulfillment of Requirements For the Degree of Master of Science, Major: Computer Science, Under the Supervision of Professors Thomas Carrell and Ashok Samal. Lincoln, Nebraska: December, 2016

Copyright 2016 Seth H. Roy

Abstract

The purpose of this study is to understand how electrical stimulation (as opposed to acoustical stimulation) of the auditory nerve is used in cochlear implants. Speech is a complex signal that changes rapidly in time and frequency domains. Since phonemes (the smallest unit of speech that distinguishes words) depend on nuanced differences in frequency patterns, it would be expected that a signal with drastically reduced frequency information would be of limited value for conveying speech. Such a frequency-poor signal is the object to be investigated in the present work. It is also the basis of the way speech is represented in cochlear implants. How could sound in which most frequency information has been discarded by successfully used by so many thousands of individuals? There must be more information in the signal such as timing and amplitude that are important for the speech signal. In addition, semantic context and visual information play a significant role in speech intelligibility. It is the goal of this thesis is to examine how this information aggregates into the perception of speech signals limited by poor frequency resolution, such as cochlear implants.

To accomplish this goal, sentence lists were created with systematically varying levels of frequency resolution. Normally, hearing listeners were asked to identify the last word of each sentence presented to them at the different levels of frequency resolution. To examine the effect of context, half of the sentences ended with predictable words and half ended with unpredictable words. The intelligibility of predictable and unpredictable words was compared at six different frequency resolutions. For this study, we used the standard R-SPIN sentences because each list was constructed to be equally intelligible with each of the other lists. The overall pattern of results showed that there were large effects of predictability and frequency resolution. There was an interaction between these two main effects that will be discussed below.

Advisors: Thomas Carrell and Ashok Samal