The Sound Quality of Cochlear Implants: Studies With Single-sided Deaf Patients

Abstract

Objective: The goal of the present study was to assess the sound quality of a cochlear implant for single-sided deaf (SSD) patients fit with a cochlear implant (CI).

Background: One of the fundamental, unanswered questions in CI research is "what does an implant sound like?" Conventional CI patients must use the memory of a clean signal, often decades old, to judge the sound quality of their CIs. In contrast, SSD-CI patients can rate the similarity of a clean signal presented to the CI ear and candidate, CI-like signals presented to the ear with normal hearing.

Methods: For Experiment 1 four types of stimuli were created for presentation to the normal hearing ear: noise vocoded signals, sine vocoded signals, frequency shifted, sine vocoded signals and band-pass filtered, natural speech signals. Listeners rated the similarity of these signals to unmodified signals sent to the CI on a scale of 0 to 10 with 10 being a complete match to the CI signal. For Experiment 2 multitrack signal mixing was used to create natural speech signals that varied along multiple dimensions.

Results: In Experiment 1 for eight adult SSD-CI listeners, the best median similarity rating to the sound of the CI for noise vocoded signals was 1.9; for sine vocoded signals 2.9; for frequency upshifted signals, 1.9; and for band pass filtered signals, 5.5. In Experiment 2 for three young listeners, combinations of band pass filtering and spectral smearing lead to ratings of 10.

Conclusion: The sound quality of noise and sine vocoders does not generally correspond to the sound quality of cochlear implants fit to SSD patients. Our preliminary conclusion is that natural speech signals that have been muffled to one degree or another by band pass filtering and/or spectral smearing provide a close, but incomplete, match to CI sound quality for some patients.

Figure 1

Rating scores (similarity to CI) for listeners in Experiment 1 in response to stimuli produced by sine and noise vocoders, a 5 channel sine vocoder with frequency up-shift as a function of insertion depth and band pass filtered, natural speech signals. Each open circle represents the judgment of one listener. The horizontal lines indicate the median score for a given condition.

Figure 2

Individual results for band pass filter conditions of Experiment 1.

Figure 3

Spectral profile of the vowel ‘eh’ following signal processing in the manner of Baer and Moore (20) illustrating the effects of spectral peak broadening or ‘smearing’.