Benefits to Speech Perception in Noise From the Binaural Integration of Electric and Acoustic Signals in Simulated Unilateral Deafness

Abstract

Objectives: This study used vocoder simulations with normal-hearing (NH) listeners to (1) measure their ability to integrate speech information from an NH ear and a simulated cochlear implant (CI), and (2) investigate whether binaural integration is disrupted by a mismatch in the delivery of spectral information between the ears arising from a misalignment in the mapping of frequency to place.

Design: Eight NH volunteers participated in the study and listened to sentences embedded in background noise via headphones. Stimuli presented to the left ear were unprocessed. Stimuli presented to the right ear (referred to as the CI-simulation ear) were processed using an eight-channel noise vocoder with one of the three processing strategies. An Ideal strategy simulated a frequency-to-place map across all channels that matched the delivery of spectral information between the ears. A Realistic strategy created a misalignment in the mapping of frequency to place in the CI-simulation ear where the size of the mismatch between the ears varied across channels. Finally, a Shifted strategy imposed a similar degree of misalignment in all channels, resulting in consistent mismatch between the ears across frequency. The ability to report key words in sentences was assessed under monaural and binaural listening conditions and at signal to noise ratios (SNRs) established by estimating speech-reception thresholds in each ear alone. The SNRs ensured that the monaural performance of the left ear never exceeded that of the CI-simulation ear. The advantages of binaural integration were calculated by comparing binaural performance with monaural performance using the CI-simulation ear alone. Thus, these advantages reflected the additional use of the experimentally constrained left ear and were not attributable to better-ear listening.

Results: Binaural performance was as accurate as, or more accurate than, monaural performance with the CI-simulation ear alone. When both ears supported a similar level of monaural performance (50%), binaural integration advantages were found regardless of whether a mismatch was simulated or not. When the CI-simulation ear supported a superior level of monaural performance (71%), evidence of binaural integration was absent when a mismatch was simulated using both the Realistic and the Ideal processing strategies. This absence of integration could not be accounted for by ceiling effects or by changes in SNR.

Conclusions: If generalizable to unilaterally deaf CI users, the results of the current simulation study would suggest that benefits to speech perception in noise can be obtained by integrating information from an implanted ear and an NH ear. A mismatch in the delivery of spectral information between the ears due to a misalignment in the mapping of frequency to place may disrupt binaural integration in situations where both ears cannot support a similar level of monaural speech understanding. Previous studies that have measured the speech perception of unilaterally deaf individuals after CI but with nonindividualized frequency-to-electrode allocations may therefore have underestimated the potential benefits of providing binaural hearing. However, it remains unclear whether the size and nature of the potential incremental benefits from individualized allocations are sufficient to justify the time and resources required to derive them based on cochlear imaging or pitch-matching tasks.

Fig. 1.

Graphical representation of the center frequencies (horizontal lines) and extent (vertical lines) of the output filters for the three processing strategies in terms of characteristic frequency (left panel) and insertion depth measured relative to the basal end of the basilar membrane (right panel).

Fig. 2.

Mean (bars) and individual (symbols) speech-reception thresholds for the NH ear alone at 50% correct (NH50), the CI-simulation ear alone at 50% correct (CI50), and the CI-simulation ear alone at 71% correct (CI71) in the main experiment. Thresholds for the CI-simulation ear alone are shown for the Ideal (light gray bars) and Realistic (white bars) processing strategies. Error bars indicate 95% confidence intervals, and standard deviations are shown above the graph. CI indicates cochlear implant; NH, normal hearing.

Fig. 3.

Psychometric functions showing the percentage of sentences for which all three key words were reported correctly as a function of SNR for the Ideal (solid gray line) and Realistic (solid black line) processing strategies. Data are extracted from the adaptive runs in the main experiment that estimated the Ideal (gray symbols) and Realistic (white symbols) CI71 thresholds. SNR indicates signal to noise ratio.

Fig. 4.

Mean binaural integration advantages for the Ideal (gray bars), Realistic (white bars), and Shifted (striped bars) processing strategies in the main experiment (left panel) and in the additional experiment (right panel). Binaural integration advantages were calculated as the change in the percentage of sentences recalled correctly when listening binaurally relative to listening monaurally using the CI-simulation ear alone (right panel). Error bars indicate 95% confidence intervals. CI indicates cochlear implant; SNR, signal to noise ratio.

Fig. 5.

Mean better-ear advantages for the Ideal (gray bars), Realistic (white bars), and Shifted (striped bars) processing strategies in the main experiment (left panel) and additional experiment (right panel). Better-ear advantages were calculated as the change in the percentage of sentences recalled correctly when listening binaurally relative to listening monaurally using the normal-hearing ear alone. Error bars indicate 95% confidence intervals. CI indicates cochlear implant; SNR, signal to noise ratio.