Deep learning restores speech intelligibility in multi-talker interference for cochlear implant users.

Cochlear implants (CIs) do not offer the same level of effectiveness in noisy environments as in quiet settings. Current single-microphone noise reduction algorithms in hearing aids and CIs only remove predictable, stationary noise, and are ineffective against realistic, non-stationary noise such as multi-talker interference. Recent developments in deep neural network (DNN) algorithms have achieved noteworthy performance in speech enhancement and separation, especially in removing speech noise. However, more work is needed to investigate the potential of DNN algorithms in removing speech noise when tested with listeners fitted with CIs. Here, we implemented two DNN algorithms that are well suited for applications in speech audio processing: (1) recurrent neural network (RNN) and (2) SepFormer. The algorithms were trained with a customized dataset ( ∼ 30 h), and then tested with thirteen CI listeners. Both RNN and SepFormer algorithms significantly improved CI listener's speech intelligibility in noise without compromising the perceived quality of speech overall. These algorithms not only increased the intelligibility in stationary non-speech noise, but also introduced a substantial improvement in non-stationary noise, where conventional signal processing strategies fall short with little benefits. These results show the promise of using DNN algorithms as a solution for listening challenges in multi-talker noise interference.

Hidden hearing loss: Fifteen years at a glance.

Hearing loss affects approximately 18% of the population worldwide. Hearing difficulties in noisy environments without accompanying audiometric threshold shifts likely affect an even larger percentage of the global population. One of the potential causes of hidden hearing loss is cochlear synaptopathy, the loss of synapses between inner hair cells (IHC) and auditory nerve fibers (ANF). These synapses are the most vulnerable structures in the cochlea to noise exposure or aging. The loss of synapses causes auditory deafferentation, i.e., the loss of auditory afferent information, whose downstream effect is the loss of information that is sent to higher-order auditory processing stages. Understanding the physiological and perceptual effects of this early auditory deafferentation might inform interventions to prevent later, more severe hearing loss. In the past decade, a large body of work has been devoted to better understand hidden hearing loss, including the causes of hidden hearing loss, their corresponding impact on the auditory pathway, and the use of auditory physiological measures for clinical diagnosis of auditory deafferentation. This review synthesizes the findings from studies in humans and animals to answer some of the key questions in the field, and it points to gaps in knowledge that warrant more investigation. Specifically, recent studies suggest that some electrophysiological measures have the potential to function as indicators of hidden hearing loss in humans, but more research is needed for these measures to be included as part of a clinical test battery.

Quantifying the Impact of Auditory Deafferentation on Speech Perception.

The past decade has seen a wealth of research dedicated to determining which and how morphological changes in the auditory periphery contribute to people experiencing hearing difficulties in noise despite having clinically normal audiometric thresholds in quiet. Evidence from animal studies suggests that cochlear synaptopathy in the inner ear might lead to auditory nerve deafferentation, resulting in impoverished signal transmission to the brain. Here, we quantify the likely perceptual consequences of auditory deafferentation in humans via a physiologically inspired encoding-decoding model. The encoding stage simulates the processing of an acoustic input stimulus (e.g., speech) at the auditory periphery, while the decoding stage is trained to optimally regenerate the input stimulus from the simulated auditory nerve firing data. This allowed us to quantify the effect of different degrees of auditory deafferentation by measuring the extent to which the decoded signal supported the identification of speech in quiet and in noise. In a series of experiments, speech perception thresholds in quiet and in noise increased (worsened) significantly as a function of the degree of auditory deafferentation for modeled deafferentation greater than 90%. Importantly, this effect was significantly stronger in a noisy than in a quiet background. The encoding-decoding model thus captured the hallmark symptom of degraded speech perception in noise together with normal speech perception in quiet. As such, the model might function as a quantitative guide to evaluating the degree of auditory deafferentation in human listeners.

Psychometric properties of the English version of the Audio Processor Satisfaction Questionnaire (APSQ).

OBJECTIVE: The Audio Processor Satisfaction Questionnaire (APSQ) is a standardized tool to measure a user's satisfaction with their audio processor(s). It was first developed and validated in the German language. The purpose of the current study was to validate the English version of the APSQ. DESIGN: The 15 items of the APSQ were translated into English. Item and scale analyses assessed the quality of individual items and of the questionnaire in its entirety. STUDY SAMPLE: Sixty-seven adults with hearing implants participated. Forty-six of them completed the questionnaire twice within 2-4 weeks. RESULTS: High mean values were obtained with total scores and with scores of the comfort, social life, and usability domains, indicating that users are generally satisfied with their audio processors. The questionnaire achieved good test-retest reliability with high internal consistency. A significant positive correlation between time since implantation and user satisfaction was found. CONCLUSION: Results of the item and reliability analyses suggest that the English version of the APSQ is a valid and reliable tool to assess user satisfaction with their audio processor(s).

A state-of-the-art implementation of a binaural cochlear-implant sound coding strategy inspired by the medial olivocochlear reflex.

Cochlear implant (CI) users find it hard and effortful to understand speech in noise with current devices. Binaural CI sound processing inspired by the contralateral medial olivocochlear (MOC) reflex (an approach termed the 'MOC strategy') can improve speech-in-noise recognition for CI users. All reported evaluations of this strategy, however, disregarded automatic gain control (AGC) and fine-structure (FS) processing, two standard features in some current CI devices. To better assess the potential of implementing the MOC strategy in contemporary CIs, here, we compare intelligibility with and without MOC processing in combination with linked AGC and FS processing. Speech reception thresholds (SRTs) were compared for an FS and a MOC-FS strategy for sentences in steady and fluctuating noises, for various speech levels, in bilateral and unilateral listening modes, and for multiple spatial configurations of the speech and noise sources. Word recall scores and verbal response times in a word recognition test (two proxies for listening effort) were also compared for the two strategies in quiet and in steady noise at 5 dB signal-to-noise ratio (SNR) and the individual SRT. In steady noise, mean SRTs were always equal or better with the MOC-FS than with the standard FS strategy, both in bilateral (the mean and largest improvement across spatial configurations and speech levels were 0.8 and 2.2 dB, respectively) and unilateral listening (mean and largest improvement of 1.7 and 2.1 dB, respectively). In fluctuating noise and in bilateral listening, SRTs were equal for the two strategies. Word recall scores and verbal response times were not significantly affected by the test SNR or the processing strategy. Results show that MOC processing can be combined with linked AGC and FS processing. Compared to using FS processing alone, combined MOC-FS processing can improve speech intelligibility in noise without affecting word recall scores or verbal response times.

Optimized gain functions in ideal time-frequency masks and their application to dereverberation for cochlear implants.

The present study investigated three different reverberation suppression rules based on the parametric ideal ratio mask, which is a generalization of the classical Wiener filter with additional parameters controlling the threshold and slope. Automatic selection of parameter values for the ideal ratio mask was performed using particle swarm optimization. Three different parameter sets were tested using sentences corrupted by reverberation. The results demonstrated that when optimizing parameters based on an objective measure of speech quality rather than intelligibility, cochlear implant users were able to perform at a level equivalent to that attainable with anechoic stimuli.

Speech-in-Noise Recognition With More Realistic Implementations of a Binaural Cochlear-Implant Sound Coding Strategy Inspired by the Medial Olivocochlear Reflex.

OBJECTIVES: Cochlear implant (CI) users continue to struggle understanding speech in noisy environments with current clinical devices. We have previously shown that this outcome can be improved by using binaural sound processors inspired by the medial olivocochlear (MOC) reflex, which involve dynamic (contralaterally controlled) rather than fixed compressive acoustic-to-electric maps. The present study aimed at investigating the potential additional benefits of using more realistic implementations of MOC processing. DESIGN: Eight users of bilateral CIs and two users of unilateral CIs participated in the study. Speech reception thresholds (SRTs) for sentences in competition with steady state noise were measured in unilateral and bilateral listening modes. Stimuli were processed through two independently functioning sound processors (one per ear) with fixed compression, the current clinical standard (STD); the originally proposed MOC strategy with fast contralateral control of compression (MOC1); a MOC strategy with slower control of compression (MOC2); and a slower MOC strategy with comparatively greater contralateral inhibition in the lower-frequency than in the higher-frequency channels (MOC3). Performance with the four strategies was compared for multiple simulated spatial configurations of the speech and noise sources. Based on a previously published technical evaluation of these strategies, we hypothesized that SRTs would be overall better (lower) with the MOC3 strategy than with any of the other tested strategies. In addition, we hypothesized that the MOC3 strategy would be advantageous over the STD strategy in listening conditions and spatial configurations where the MOC1 strategy was not. RESULTS: In unilateral listening and when the implant ear had the worse acoustic signal-to-noise ratio, the mean SRT was 4 dB worse for the MOC1 than for the STD strategy (as expected), but it became equal or better for the MOC2 or MOC3 strategies than for the STD strategy. In bilateral listening, mean SRTs were 1.6 dB better for the MOC3 strategy than for the STD strategy across all spatial configurations tested, including a condition with speech and noise sources colocated at front where the MOC1 strategy was slightly disadvantageous relative to the STD strategy. All strategies produced significantly better SRTs for spatially separated than for colocated speech and noise sources. A statistically significant binaural advantage (i.e., better mean SRTs across spatial configurations and participants in bilateral than in unilateral listening) was found for the MOC2 and MOC3 strategies but not for the STD or MOC1 strategies. CONCLUSIONS: Overall, performance was best with the MOC3 strategy, which maintained the benefits of the originally proposed MOC1 strategy over the STD strategy for spatially separated speech and noise sources and extended those benefits to additional spatial configurations. In addition, the MOC3 strategy provided a significant binaural advantage, which did not occur with the STD or the original MOC1 strategies.

Lateralization of virtual sound sources with a binaural cochlear-implant sound coding strategy inspired by the medial olivocochlear reflex.

Many users of bilateral cochlear implants (BiCIs) localize sound sources less accurately than do people with normal hearing. This may be partly due to using two independently functioning CIs with fixed compression, which distorts and/or reduces interaural level differences (ILDs). Here, we investigate the potential benefits of using binaurally coupled, dynamic compression inspired by the medial olivocochlear reflex; an approach termed "the MOC strategy" (Lopez-Poveda et al., 2016, Ear Hear 37:e138-e148). Twelve BiCI users were asked to localize wideband (125-6000 Hz) noise tokens in a virtual horizontal plane. Stimuli were processed through a standard (STD) sound processing strategy (i.e., involving two independently functioning sound processors with fixed compression) and three different implementations of the MOC strategy: one with fast (MOC1) and two with slower contralateral control of compression (MOC2 and MOC3). The MOC1 and MOC2 strategies had effectively greater inhibition in the higher than in the lower frequency channels, while the MOC3 strategy had slightly greater inhibition in the lower than in the higher frequency channels. Localization was most accurate with the MOC1 strategy, presumably because it provided the largest and less ambiguous ILDs. The angle error improved slightly from 25.3° with the STD strategy to 22.7° with the MOC1 strategy. The improvement in localization ability over the STD strategy disappeared when the contralateral control of compression was made slower, presumably because stimuli were too short (200 ms) for the slower contralateral inhibition to enhance ILDs. Results suggest that some MOC implementations hold promise for improving not only speech-in-noise intelligibility, as shown elsewhere, but also sound source lateralization.

Intelligibility in speech maskers with a binaural cochlear implant sound coding strategy inspired by the contralateral medial olivocochlear reflex.

We have recently proposed a binaural cochlear implant (CI) sound processing strategy inspired by the contralateral medial olivocochlear reflex (the MOC strategy) and shown that it improves intelligibility in steady-state noise (Lopez-Poveda et al., 2016, Ear Hear 37:e138-e148). The aim here was to evaluate possible speech-reception benefits of the MOC strategy for speech maskers, a more natural type of interferer. Speech reception thresholds (SRTs) were measured in six bilateral and two single-sided deaf CI users with the MOC strategy and with a standard (STD) strategy. SRTs were measured in unilateral and bilateral listening conditions, and for target and masker stimuli located at azimuthal angles of (0°, 0°), (-15°, +15°), and (-90°, +90°). Mean SRTs were 2-5 dB better with the MOC than with the STD strategy for spatially separated target and masker sources. For bilateral CI users, the MOC strategy (1) facilitated the intelligibility of speech in competition with spatially separated speech maskers in both unilateral and bilateral listening conditions; and (2) led to an overall improvement in spatial release from masking in the two listening conditions. Insofar as speech is a more natural type of interferer than steady-state noise, the present results suggest that the MOC strategy holds potential for promising outcomes for CI users.

Roles of the Contralateral Efferent Reflex in Hearing Demonstrated with Cochlear Implants.

Our two ears do not function as fixed and independent sound receptors; their functioning is coupled and dynamically adjusted via the contralateral medial olivocochlear efferent reflex (MOCR). The MOCR possibly facilitates speech recognition in noisy environments. Such a role, however, is yet to be demonstrated because selective deactivation of the reflex during natural acoustic listening has not been possible for human subjects up until now. Here, we propose that this and other roles of the MOCR may be elucidated using the unique stimulus controls provided by cochlear implants (CIs). Pairs of sound processors were constructed to mimic or not mimic the effects of the contralateral MOCR with CIs. For the non-mimicking condition (STD strategy), the two processors in a pair functioned independently of each other. When configured to mimic the effects of the MOCR (MOC strategy), however, the two processors communicated with each other and the amount of compression in a given frequency channel of each processor in the pair decreased with increases in the output energy from the contralateral processor. The analysis of output signals from the STD and MOC strategies suggests that in natural binaural listening, the MOCR possibly causes a small reduction of audibility but enhances frequency-specific inter-aural level differences and the segregation of spatially non-overlapping sound sources. The proposed MOC strategy could improve the performance of CI and hearing-aid users.

A Binaural Cochlear Implant Sound Coding Strategy Inspired by the Contralateral Medial Olivocochlear Reflex.

OBJECTIVES: In natural hearing, cochlear mechanical compression is dynamically adjusted via the efferent medial olivocochlear reflex (MOCR). These adjustments probably help understanding speech in noisy environments and are not available to the users of current cochlear implants (CIs). The aims of the present study are to: (1) present a binaural CI sound processing strategy inspired by the control of cochlear compression provided by the contralateral MOCR in natural hearing; and (2) assess the benefits of the new strategy for understanding speech presented in competition with steady noise with a speech-like spectrum in various spatial configurations of the speech and noise sources. DESIGN: Pairs of CI sound processors (one per ear) were constructed to mimic or not mimic the effects of the contralateral MOCR on compression. For the nonmimicking condition (standard strategy or STD), the two processors in a pair functioned similarly to standard clinical processors (i.e., with fixed back-end compression and independently of each other). When configured to mimic the effects of the MOCR (MOC strategy), the two processors communicated with each other and the amount of back-end compression in a given frequency channel of each processor in the pair decreased/increased dynamically (so that output levels dropped/increased) with increases/decreases in the output energy from the corresponding frequency channel in the contralateral processor. Speech reception thresholds in speech-shaped noise were measured for 3 bilateral CI users and 2 single-sided deaf unilateral CI users. Thresholds were compared for the STD and MOC strategies in unilateral and bilateral listening conditions and for three spatial configurations of the speech and noise sources in simulated free-field conditions: speech and noise sources colocated in front of the listener, speech on the left ear with noise in front of the listener, and speech on the left ear with noise on the right ear. In both bilateral and unilateral listening, the electrical stimulus delivered to the test ear(s) was always calculated as if the listeners were wearing bilateral processors. RESULTS: In both unilateral and bilateral listening conditions, mean speech reception thresholds were comparable with the two strategies for colocated speech and noise sources, but were at least 2 dB lower (better) with the MOC than with the STD strategy for spatially separated speech and noise sources. In unilateral listening conditions, mean thresholds improved with increasing the spatial separation between the speech and noise sources regardless of the strategy but the improvement was significantly greater with the MOC strategy. In bilateral listening conditions, thresholds improved significantly with increasing the speech-noise spatial separation only with the MOC strategy. CONCLUSIONS: The MOC strategy (1) significantly improved the intelligibility of speech presented in competition with a spatially separated noise source, both in unilateral and bilateral listening conditions; (2) produced significant spatial release from masking in bilateral listening conditions, something that did not occur with fixed compression; and (3) enhanced spatial release from masking in unilateral listening conditions. The MOC strategy as implemented here, or a modified version of it, may be usefully applied in CIs and in hearing aids.

Investigating the effects of stimulus duration and context on pitch perception by cochlear implant users.

Cochlear implant sound processing strategies that use time-varying pulse rates to transmit fine structure information are one proposed method for improving the spectral representation of a sound with the eventual goal of improving speech recognition in noisy conditions, speech recognition in tonal languages, and music identification and appreciation. However, many of the perceptual phenomena associated with time-varying rates are not well understood. In this study, the effects of stimulus duration on both the place and rate-pitch percepts were investigated via psychophysical experiments. Four Nucleus CI24 cochlear implant users participated in these experiments, which included a short-duration pitch ranking task and three adaptive pulse rate discrimination tasks. When duration was fixed from trial-to-trial and rate was varied adaptively, results suggested that both the place-pitch and rate-pitch percepts may be independent of duration for durations above 10 and 20 ms, respectively. When duration was varied and pulse rates were fixed, performance was highly variable within and across subjects. Implications for multi-rate sound processing strategies are discussed.

Assessing the pitch structure associated with multiple rates and places for cochlear implant users.

Cochlear implant subjects continue to experience difficulty understanding speech in noise and performing pitch-based musical tasks. Acoustic model studies have suggested that transmitting additional fine structure via multiple stimulation rates is a potential mechanism for addressing these issues [Nie et al., IEEE Trans. Biomed. Eng. 52, 64-73 (2005); Throckmorton et al., Hear. Res. 218, 30-42 (2006)]; however, results from preliminary cochlear implant studies have been less compelling. Multirate speech processing algorithms previously assumed a place-dependent pitch structure in that a basal electrode would always elicit a higher pitch percept than an apical electrode, independent of stimulation rate. Some subjective evidence contradicts this assumption [H. J. McDermott and C. M. McKay, J. Acoust. Soc. Am. 101, 1622-1630 (1997); R. V. Shannon, Hear. Res. 11, 157-189 (1983)]. The purpose of this study is to test the hypothesis that the introduction of multiple rates may invalidate the tonotopic pitch structure resulting from place-pitch alone. The SPEAR3 developmental speech processor was used to collect psychophysical data from five cochlear implant users to assess the tonotopic structure for stimuli presented at two rates on all active electrodes. Pitch ranking data indicated many cases where pitch percepts overlapped across electrodes and rates. Thus, the results from this study suggest that pitch-based tuning across rate and electrode may be necessary to optimize performance of a multirate sound processing strategy in cochlear implant subjects.