Sound Localization in Real-Time Vocoded Cochlear-Implant Simulations With Normal-Hearing Listeners

Abstract

Bilateral cochlear-implant (CI) users and single-sided deaf listeners with a CI are less effective at localizing sounds than normal-hearing (NH) listeners. This performance gap is due to the degradation of binaural and monaural sound localization cues, caused by a combination of device-related and patient-related issues. In this study, we targeted the device-related issues by measuring sound localization performance of 11 NH listeners, listening to free-field stimuli processed by a real-time CI vocoder. The use of a real-time vocoder is a new approach, which enables testing in a free-field environment. For the NH listening condition, all listeners accurately and precisely localized sounds according to a linear stimulus-response relationship with an optimal gain and a minimal bias both in the azimuth and in the elevation directions. In contrast, when listening with bilateral real-time vocoders, listeners tended to orient either to the left or to the right in azimuth and were unable to determine sound source elevation. When listening with an NH ear and a unilateral vocoder, localization was impoverished on the vocoder side but improved toward the NH side. Localization performance was also reflected by systematic variations in reaction times across listening conditions. We conclude that perturbation of interaural temporal cues, reduction of interaural level cues, and removal of spectral pinna cues by the vocoder impairs sound localization. Listeners seem to ignore cues that were made unreliable by the vocoder, leading to acute reweighting of available localization cues. We discuss how current CI processors prevent CI users from localizing sounds in everyday environments.

Keywords: acoustic stimulation; auditory perception; cochlear implant; psychoacoustics.

Figure 1.

(a) Diagram showing the processing blocks of one side of the CI simulation vocoder. (b) Input versus output acoustic levels for the 1 kHz electrode. Other electrodes differed slightly due to the frequency dependency contained in the ISO 226 standard. CI = cochlear implant; dB SPL = decibel of sound pressure level.

Figure 2.

Spectrograms of a BB stimulus and vocoder output. The acoustic stimulus (a) is processed by the vocoder and generated the left (b) and right (c) output (limited between 0.2 and 8 kHz). The cross-correlation of the acoustic stimulus (in blue) shows a peak at 0 ms (d), which will shift according to the applied ITD. The correlation between the right and left output (in red) shows no correlation between each other. BB = broadband.

Figure 3.

Localization paradigm. Target locations for (a) BB, (b) HP, and (c) LP filtered sounds for each sound level (represented by marker size). Vertical dashed lines indicate the ±15° boundaries of the central azimuth range. Head trajectories of example trials of listener S1 responding to a BB, 60 dB(A) sound in (d) azimuth (−72°), and (e) elevation (−28°), for the three listening conditions: normal hearing (NH—blue), bilateral CI vocoder (2CI—red), and unilateral CI vocoder with a contralateral NH ear (1CI—yellow). Horizontal dashed lines indicate target location; vertical dashed lines indicate target onset. BB = broadband; HP = high-pass; LP = low-pass.

Figure 4.

Azimuth stimulus–response plots for listener S1. (a–c) NH, (d–f) 2CI, and (g–i) 1CI hearing conditions. Gain (γ), bias (β), and response variability (σ) for the right (α > +15°) and left (α < −15°) sides are indicated in the lower-right and upper-left corners, respectively. BB = broadband; HP = high-pass; LP = low-pass; NH = normal hearing; 2CI = bilateral CI vocoder; 1CI: unilateral CI vocoder with a contralateral NH ear.

Figure 5.

Azimuth localization performance. The regression parameter gains (top row), bias (center row), and variability (bottom row) are determined per azimuth bin (left, center, and right on the abscissa) for the three listening conditions (colored lines, normal hearing, NH; bilateral vocoder, 2CI; unilateral vocoder in one ear, 1CI) and the three sound frequency bands BB (left column), HP (middle column), and LP (right column). Open circles, colored patches denote mean and standard error across listeners, respectively. Dashed lines (a–c): perfect localization (γ = 1) and no localization (γ = 0) for the gains; (d–f): perfect localization (β = 0°) and complete left/right tendency (β = ±90°) for the biases; (g–i): response variability for perfect localization (σ = 0°) and for a completely random response behavior (σ ∼ 51°). BB = broadband; HP = high-pass; LP = low-pass.

Figure 6.

Elevation performance for listener S1 for BB and HP sounds combined. Gain (γ), bias (β), and response variability (σ) are presented for target azimuth α < −15° and α > 15° for NH (top, blue), 2CI hearing (center, red), and 1CI (bottom, yellow) conditions. The vocoder in the 1CI condition was on the left side (bottom left graph). NH = normal hearing; 2CI = bilateral CI vocoder; 1CI: unilateral CI vocoder with a contralateral NH ear.

Figure 7.

(a) Elevation gain and (b) response variability as a function of target azimuth. Open circles connected by lines denote the binned mean across listeners, whereas patches denote the standard error of the mean across listeners. Colors indicate NH, 2CI, and 1CI conditions. NH = normal hearing; 2CI = bilateral CI vocoder; 1CI: unilateral CI vocoder with a contralateral NH ear.

Figure 8.

Promptness per spectral band (a–c) and listening condition (colors) for listener S1 on reciprobit scale. BB = broadband; HP = high-pass; LP = low-pass; NH = normal hearing; 2CI = bilateral CI vocoder; 1CI: unilateral CI vocoder with a contralateral NH ear.

Figure 9.

Promptness across azimuth for (a) BB, (b) HP, and (c) LP sounds. Open circles connected by lines denote binned mean promptness across listeners, whereas patches denote the standard error across listeners. Color indicates NH, 2CI, and 1CI conditions. BB = broadband; HP = high-pass; LP = low-pass; NH = normal hearing; 2CI = bilateral CI vocoder; 1CI: unilateral CI vocoder with a contralateral NH ear.