Recovering speech intelligibility with deep learning and multiple microphones in noisy-reverberant situations for people using cochlear implants.

For cochlear implant (CI) listeners, holding a conversation in noisy and reverberant environments is often challenging. Deep-learning algorithms can potentially mitigate these difficulties by enhancing speech in everyday listening environments. This study compared several deep-learning algorithms with access to one, two unilateral, or six bilateral microphones that were trained to recover speech signals by jointly removing noise and reverberation. The noisy-reverberant speech and an ideal noise reduction algorithm served as lower and upper references, respectively. Objective signal metrics were compared with results from two listening tests, including 15 typical hearing listeners with CI simulations and 12 CI listeners. Large and statistically significant improvements in speech reception thresholds of 7.4 and 10.3 dB were found for the multi-microphone algorithms. For the single-microphone algorithm, there was an improvement of 2.3 dB but only for the CI listener group. The objective signal metrics correctly predicted the rank order of results for CI listeners, and there was an overall agreement for most effects and variances between results for CI simulations and CI listeners. These algorithms hold promise to improve speech intelligibility for CI listeners in environments with noise and reverberation and benefit from a boost in performance when using features extracted from multiple microphones.

Frequency importance functions in simulated bimodal cochlear-implant users with spectral holes.

Frequency importance functions (FIFs) for simulated bimodal hearing were derived using sentence perception scores measured in quiet and noise. Acoustic hearing was simulated using low-pass filtering. Electric hearing was simulated using a six-channel vocoder with three input frequency ranges, resulting in overlap, meet, and gap maps, relative to the acoustic cutoff frequency. Spectral holes present in the speech spectra were created within electric stimulation by setting amplitude(s) of channels to zero. FIFs were significantly different between frequency maps. In quiet, the three FIFs were similar with gradually increasing weights with channels 5 and 6 compared to the first three channels. However, the most and least weighted channels slightly varied depending on the maps. In noise, the patterns of the three FIFs were similar to those in quiet, with steeper increasing weights with channels 5 and 6 compared to the first four channels. Thus, channels 5 and 6 contributed to speech perception the most, while channels 1 and 2 contributed the least, regardless of frequency maps. Results suggest that the contribution of cochlear implant frequency bands for bimodal speech perception depends on the degree of frequency overlap between acoustic and electric stimulation and if noise is absent or present.

Some, but not all, cochlear implant users prefer music stimuli with congruent haptic stimulation.

Cochlear implant (CI) users often report being unsatisfied by music listening through their hearing device. Vibrotactile stimulation could help alleviate those challenges. Previous research has shown that musical stimuli was given higher preference ratings by normal-hearing listeners when concurrent vibrotactile stimulation was congruent in intensity and timing with the corresponding auditory signal compared to incongruent. However, it is not known whether this is also the case for CI users. Therefore, in this experiment, we presented 18 CI users and 24 normal-hearing listeners with five melodies and five different audio-to-tactile maps. Each map varied the congruence between the audio and tactile signals related to intensity, fundamental frequency, and timing. Participants were asked to rate the maps from zero to 100, based on preference. It was shown that almost all normal-hearing listeners, as well as a subset of the CI users, preferred tactile stimulation, which was congruent with the audio in intensity and timing. However, many CI users had no difference in preference between timing aligned and timing unaligned stimuli. The results provide evidence that vibrotactile music enjoyment enhancement could be a solution for some CI users; however, more research is needed to understand which CI users can benefit from it most.

[Online training for hearing implant surgery : A new approach to otological training. German version]. / Online-Training für Hörimplantatchirurgie : Ein neuer Ansatz in der otologischen Ausbildung.

OBJECTIVE: Education in microsurgery of the ear includes staged training to allow for mastering of the complex microsurgical procedures, particularly in the context of middle ear reconstruction and cochlear implantation. Traditional surgical training includes temporal bone preparations by cadaver dissection and supervised operating room practice. As these on-site trainings are limited, there is a need to broaden education facilities in an on-line format. Therefore, a first basic on-line training for otosurgery was developed. MATERIALS AND METHODS: The system consists of an artificial temporal bone model together with a set of basic surgical instruments and implant dummies. As an essential part of the training kit, a high-resolution camera set is included that allows for connection to a video streaming platform and enables remote supervision of the trainees' surgical steps by experienced otological surgeons. In addition, a pre-learning platform covering temporal bone anatomy and instrumentation and pre-recorded lectures and instructional videos has been developed to allow trainees to review and reinforce their understanding before hands-on practice. RESULTS: Over the three courses held to date, 28 participants with varying levels of prior surgical experience took part in this otological surgical training program. The immediate feedback of the participants was evaluated by means of a questionnaire. On this basis, the high value of the program became apparent and specific areas could by identified where further refinements could lead to an even more robust training experience. CONCLUSION: The presented program of an otosurgical online training allows for basal education in practical exercises on a remote system. In this way, trainees who have no direct access to on-site instruction facilities in ear surgery now have the chance to start their otosurgical training in an educational setting adapted to modern technologies.

Acceptance and Benefits of Electro-Acoustic Stimulation for Conventional-Length Electrode Arrays.

BACKGROUND: Prior studies have shown an advantage for electro-acoustic stimulation (EAS) in cochlear implant (CI) patients with residual hearing, but the degree of benefit can vary. The objective was to explore which factors relate to performance with and acceptance of EAS for CI users with conventional-length electrodes. METHODS: A retrospective chart review was conducted for adults with an average threshold of 75 dB hearing loss or better across 250 and 500 Hz preoperatively (n = 83). All patients underwent cochlear implantation with a conventional-length electrode. Low-frequency audiometric thresholds were measured at initial activation as well as 3 and 12 months postoperatively to determine who met the criteria for EAS. Speech perception for CNC words and AzBio sentences in quiet and +10 dB SNR noise was evaluated 3 and 12 months after activation. RESULTS: Speech perception in quiet and noise was similar regardless of whether or not the patient was eligible for EAS. Less than half of the patients who met the EAS criteria chose to use it, citing reasons such as physical discomfort or lack of perceived benefit. EAS users performed better on CNC words but not sentence recognition than EAS nonusers. CONCLUSIONS: EAS use is dependent on audiologic and nonaudiologic issues. Hearing preservation is possible with conventional electrodes, but hearing preservation alone does not guarantee superior speech perception.

Surgical anatomy of basal turn in relation to middle cranial fossa and round window as pertaining to middle fossa cochlear implant technique.

PURPOSE: The purpose of this investigation was to evaluate the relation of the basal turn of cochlea with middle cranial fossa and round window as pertaining to middle cranial fossa cochlear implant technique. METHODS: Fifty-four formalin preserved temporal bones were micro-dissected to expose the basal turn. The point (f) was marked on the basal turn of cochlea where there was minimum distance of basal turn of cochlea from the floor of middle cranial fossa (f1). The f-f1 distance, the angle (∠smf) and distance of point "f" from the round window (s-f) was measured. RESULTS: The mean minimum distance between basal turn of cochlea and floor of middle cranial fossa (f-f1) was 2.0 ± 0.7 mm.The point f was at mean distance (s-f) and angle (∠smf) of 14.7 ± 1.6 mm and 217° ± 13.7° from round window, respectively. CONCLUSIONS: The information may be helpful for designing appropriate electrode array for middle fossa technique especially for deciding the length of electrode array towards round window, as otherwise electrode may extend into vestibule of inner ear. In the cases where bone thickness between superior part of basal turn of cochlea and middle cranial fossa is < 1.5 mm, surgeon should be extra cautious.

Global Regulatory Review Needed for Cochlear Implants: A Call for FDA Leadership.

INTRODUCTION: Using the United States Food and Drug Administration (FDA) as example, we argue that regulatory agencies worldwide should review their guidance on cochlear implants (CIs). METHODS: This is a position paper, thus the methods are strictly argumentation. Here we give the motivation for our recommendation. The FDA's original approval of implantation in prelingually deaf children was granted without full benefit of information on language acquisition, on childcaregiver communication, and on the lived experience of being deaf. The CI clinical trials, accordingly, did not address risks of linguistic deprivation, especially when the caregiver's communication is not fully accessible to the prelingually deaf child. Wide variability in the effectiveness of CIs since initial and updated approval has been indicated but has not led to new guidance. Children need to be exposed frequently and regularly to accessible natural language while their brains are still plastic enough to become fluent in any language. For the youngest infants, who are not yet producing anything that could be called language although they might be producing salient social signals (Goldstein et al. Child Dev 80:636-644, 2009), good comprehension of communication from caregiver to infant is critical to the development of language. Sign languages are accessible natural languages that, because they are visual, allow full immersion for deaf infants, and they supply the necessary support for this comprehension. The main language contributor to health outcomes is this combination of natural visual language and comprehension in communication. Accordingly, in order to prevent possible language deprivation, all prelingually deaf children should be exposed to both sign and spoken languages when their auditory status is detected, with sign language being critical during infancy and early childhood. Additionally, all caregivers should be given support to learn a sign language if it is new to them so that they can comprehend their deaf children's language expressions fully. However, both languages should be made accessible in their own right, not combined in a simultaneous or total communication approach since speaking one language and signing the other at the same time is problematic. RESULTS: Again, because this is a position paper, our results are our recommendations. We call for the FDA (and similar agencies in other countries) to review its approval of cochlear implantation in prelingually deaf children who are within the sensitive period for language acquisition. In the meantime, the FDA should require manufacturers to add a highlighted warning to the effect that results with CI vary widely and CIs should not be relied upon to provide adequate auditory input for complete language development in all deaf children. Recent best information on users' experience with CIs (including abandonment) should be clearly provided so that informed decisions can be made. The FDA should require manufacturers' guidance and information materials to include encouragement to parents of deaf children to offer auditory input of a spoken language and visual input of a sign language and to have their child followed closely from birth by developmental specialists in language and cognition. In this way parents can align with providers to prioritize cognitive development and language access in both audio-vocal and visuo-gestural modalities. DISCUSSION: The arguments and recommendations in this paper are discussed at length as they come up.

Comparative study of two different perimodiolar and a straight cochlear implant electrode array: surgical and audiological outcomes.

PURPOSE: To compare the surgical and audiological outcomes with two perimodiolar electrode arrays (Nucleus 512-Contour Advance® y Nucleus 532-Slim Perimodiolar®) and a straight electrode array (Nucleus 422/522). METHODS: Patients were retrospectively selected from our cochlear implant program database. Only patients with a history of bilateral, sensorineural postlingually profound hearing loss who underwent cochlear implant surgery with either a N512, a N532 or a N422 were included. Throughout a year of follow-up, pure tone audiometry (PTA), speech perception, Impedances and T-C Thresholds levels were analyzed. Surgical data were also analyzed. RESULTS: 66 patients were included (19-CI532, 20-CI512 and 27-CI422). The most common type of cochlea access with the N532, N512 and N422 was through an extended round window, a promontorial cochleostomy and a pure round window, respectively. No significant differences were observed after 12 months in Mean PTA and Speech recognition. No significant differences were seen in the levels of hearing preservation at frequencies of 250 and 500. The average values of the impedances were significantly higher in the CI group N532 and N422 than in the N512. The mean values of the T and C levels were significantly lower in the CI groups N532 and N422 compared with the N512. CONCLUSIONS: No significant differences were observed after 12 months in Mean PTA and Speech recognition; however, a faster acquisition of auditory results were observed in the group of patients treated with the CI N532. The type of electrode array influences in the type of cochleostomy.

Preservation of residual hearing after cochlear implant surgery with slim modiolar electrode.

PURPOSE: To evaluate the insertion results and hearing preservation of a novel slim modiolar electrode (SME) in patients with residual hearing. METHODS: We retrospectively collected the data from the medical files of 17 patients (18 ears) implanted with a SME. All patients had functional low frequency hearing (PTA (0.125-0.5 kHz) ≤ 80 dB HL). The insertion results were re-examined from the postoperative cone-beam computed tomography scans. Postoperative thresholds were obtained at the time of switch-on of the sound processors (mean 43 days) and at latest follow-up (mean 582 days). The speech recognition in noise was measured with the Finnish matrix sentence test preoperatively and at follow-up. RESULTS: The mean insertion depth angle (IDA) was 395°. Neither scala dislocations nor tip fold over were detected. There were no total hearing losses. Functional low-frequency hearing was preserved in 15/18 (83%) ears at switch-on and in 14/17 (82%) ears at follow-up. According to HEARRING classification, 55% (10/18) had complete HP at switch-on and 41% (7/17) still at follow-up. Thirteen patients (14 ears) were initially fitted with electric-acoustic stimulation and seven patients (8 ears) continued to use it after follow-up. CONCLUSIONS: The preliminary hearing preservation results with the SME were more favorable than reported for other perimodiolar electrodes. The results show that the array may also be feasible for electro-acoustic stimulation; it is beneficial in that it provides adequate cochlear coverage for pure electrical stimulation in the event of postoperative or progressive hearing loss.

The Oticon Medical Neuro Zti cochlear implant and the Neuro 2 sound processor: multicentric evaluation of outcomes in adults and children.

Objective: This study evaluated the outcomes of the Oticon Medical Neuro Zti cochlear implant and the Neuro 2 sound processor.Design: Neuro One users were upgraded to Neuro 2. Monosyllabic word identification was evaluated in adults with Neuro One after ≥5 months, with Neuro 2 at upgrade, and with Neuro 2 after 3 months. Self-reported listening ability, satisfaction, and usability were measured in adults and children.Study sample: Participants were 44 adults and 26 children.Results: Speech identification scores in quiet and noise were 58% and 45% with Neuro One and 67% and 55% with Neuro 2 after 3 months, respectively. Hearing impairment duration and number of active electrodes significantly predicted speech identification in noise with Neuro 2. Significantly higher questionnaire ratings were obtained for Neuro 2 than Neuro One regarding listening ability in complex listening situations, comfort and music, as well as nine aspects of satisfaction and usability.Conclusion: This study demonstrates the clinical superiority of the Neuro 2 sound processor over Neuro One in terms of speech identification in quiet and in noise and reported patient benefit and satisfaction. Given the study design, sources of improvement may include factors unrelated to the sound processor itself.

Longitudinal effect of deactivating stimulation sites based on low-rate thresholds on speech recognition in cochlear implant users.

Objective: The objective of the current study was to examine the longitudinal effect of deactivating stimulation sites estimated to produce broad neural excitation on speech recognition. Design: Spatial patterns of neural excitation were estimated based on a previously established psychophysical measure, that is, detection threshold for low-rate pulse trains. Stimulation sites with relatively poor thresholds were deactivated in an experimental map. The acute effect was evaluated, in quiet and in noise, immediately after the experimental map was created (baseline), after the subjects practiced with the experimental map for two months (treatment), and after the subjects' daily map was switched back again to the clinical map for another two months (withdrawal). Study sample: Eight Cochlear Nucleus device users participated in the study. Results: For both listening in noise and in quiet, the greatest effect of deactivation was observed after the subjects were given time to adapt to the new frequency allocations. The effect was comparable for listening in fluctuating and steady-state noises. All subjects benefited from deactivation for listening in noise, but subjects with greater variability in thresholds were more likely to benefit from deactivation for listening in quiet. Conclusion: The benefit of electrode deactivation for speech recognition can increase with practice.

The influence of cochlear morphology on the final electrode array position.

OBJECTIVE: Preoperative information about cochlear morphology and size increasingly seems to be a defining factor of electrode choice in cochlear implant surgery. Different types of electrodes differ in length and diameter to accommodate individual cochlear anatomy. Smaller cochlear size results in increased insertion depth with a higher risk to dislocate and causes cochlear trauma with reduced postoperative outcome. The objective of the current study is to describe the three-dimensional size of the cochlea, to compare interindividual differences, to determine the relationship between cochlear size and insertion angle, and to define risk factors for dislocation during insertion. DESIGN: Four hundred and three patients implanted between 2003 and 2010 inserted via cochleostomy with a perimodiolar electrode array (Cochlear™ Contour Advance® electrode array) have been compared. CBCT (Cone beam computed tomography) was used to determine electrode array position (scala tympani versus scala vestibuli insertion, intracochlear dislocation, and insertion angle) and cochlear size (diameters and height). The trajectory of the electrode array and the lateral wall have been measured, and the position of the electrode array has been estimated. RESULTS: The mean value of the largest diameter was 9.95 mm and that of the perpendicular distance was 6.54 mm. There was a statistically significant correlation between those values. Mean height was 3.85 mm. The intracochlear relation of the electrode array and the modiolus showed a statistically significant relationship with the cochlear expanse. The electrode array was more likely to dislocate in cochleae with a smaller diameter and a lower height. Cochleae with insertions into scala vestibuli exhibited a smaller height compared to scala tympani insertions with statistical significance. CONCLUSION: Cochlear size and shape is variable, and the measured data of this study confirm the finding of other researchers. This study established two heights by two different planes to achieve a three-dimensional understanding of the cochlea. The electrode array was more likely to dislocate in cochleae with smaller diameter and smaller height. It can be assumed that the height established in this study seems to be a new preoperative parameter to underline the risk of scalar dislocation and not favored scala vestibuli insertion if using a cochleostomy approach. In conclusion, cochlear size, especially the height, is influencing the final position of the electrode array. Using preoperative scans of the cochlear diameters and cochlear height, a next step to custom-sized arrays is available.

On the accuracy of cochlear duct length measurement in computed tomographic images.

PURPOSE: Patient specific selection of cochlear implants would benefit from pre-operative knowledge of cochlear length. Several methods for its measurement or estimation have been described in literature. This study focused on the achievable accuracy in clinically available imaging. METHODS: Five simplified cochlea models milled into porcine bone were scanned in water using clinical cone beam computed tomography. Due to their well-known dimensions these phantoms served as gold standard for the length measurements. Each phantom was measured ten times using the custom software Comet. In addition, cochleae in ten image datasets taken indiscriminately from clinical routine were measured ten times each to test the precision under realistic conditions. The results were also compared to estimations based on the diameter of the basal turn (A value) as described in literature. RESULTS: Measurement accuracy of the phantoms' lengths was high (average error: - 0.2 mm; standard deviation: 0.3 mm). The pooled standard deviation for the measurements in clinical datasets was 0.6 mm. Errors resulted mainly from problems locating the helicotrema. The estimations differed on average - 1.7 to + 0.4 mm from the manual measurements and had standard deviations between 0.5 and 0.6 mm depending on the algorithm. CONCLUSIONS: The program Comet was successfully used to accurately measure the length of the cochlea models in clinically available imaging. The lower image quality of patient scans reduced the precision of the measurement. Estimations using the A value are a quicker alternative for averagely sized cochleae in cases where the lack of accuracy is tolerable.

Incomplete and false tract insertions in cochlear implantation: retrospective review of surgical and auditory outcomes.

OBJECTIVES: To evaluate incidence, demographics, surgical, and radiological correlates of incomplete and false tract electrode array insertions during cochlear implantation (CI). To evaluate outcomes in patients with incomplete electrode insertion (IEI). STUDY DESIGN: Retrospective analysis. SETTING: Otology and skull base center. PATIENTS AND METHODS: Charts of 18 patients (19 ears) with incomplete or false tract insertions of the electrode array were evaluated who underwent CI, with at least 1 year follow-up (from 470 cases). Demographic findings, etiologies, pre-operative radiologic findings, operative records, post-operative plain radiographic assessment for extent of electrode insertion, and switch-on mapping were evaluated. Audiological outcomes were evaluated using maximum and last recorded vowel, word, sentence, and comprehension scores for patients with IEI. RESULTS: Incidence of insertional abnormalities was 4.25% with 17 instances of incomplete and 2 cases of insertion into superior semicircular canal. Mean age and duration of deafness were 55.18 ± 4.62 and 22.12 ± 5.71 years. Etiologies in the IEI group were idiopathic, otosclerosis, meningitis, chronic otitis media (COM), temporal bone fractures, and Neurofibromatosis-2. 29.4% cases had cochlear luminal obstruction. Mean radiological and active electrophysiological length of insertion was 20.49 ± 0.66 and 19.49 ± 0.88 mm, respectively. No significant correlation was observed between audiological outcomes and insertional length except in time to achieve maximum word scores (p = 0.04). Age at implantation had significant correlations with last recorded word and comprehension scores at mean follow-up of 42.9 months, and with time to achieve maximum auditory scores. CONCLUSIONS: IEI during cochlear implantation using straight electrodes can occur with or without cochlear luminal obstruction. Age plays an important role in the auditory rehabilitation in this patient subset.

Deep learning models to remix music for cochlear implant users.

The severe hearing loss problems that some people suffer can be treated by providing them with a surgically implanted electrical device called cochlear implant (CI). CI users struggle to perceive complex audio signals such as music; however, previous studies show that CI recipients find music more enjoyable when the vocals are enhanced with respect to the background music. In this manuscript source separation (SS) algorithms are used to remix pop songs by applying gain to the lead singing voice. This work uses deep convolutional auto-encoders, a deep recurrent neural network, a multilayer perceptron (MLP), and non-negative matrix factorization to be evaluated objectively and subjectively through two different perceptual experiments which involve normal hearing subjects and CI recipients. The evaluation assesses the relevance of the artifacts introduced by the SS algorithms considering their computation time, as this study aims at proposing one of the algorithms for real-time implementation. Results show that the MLP performs in a robust way throughout the tested data while providing levels of distortions and artifacts which are not perceived by CI users. Thus, an MLP is proposed to be implemented for real-time monaural audio SS to remix music for CI users.

Effect of frequency mismatch and band partitioning on vocal tract length perception in vocoder simulations of cochlear implant processing.

The vocal tract length (VTL) of a speaker is an important voice cue that aids speech intelligibility in multi-talker situations. However, cochlear implant (CI) users demonstrate poor VTL sensitivity. This may be partially caused by the mismatch between frequencies received by the implant and those corresponding to places of stimulation along the cochlea. This mismatch can distort formant spacing, where VTL cues are encoded. In this study, the effects of frequency mismatch and band partitioning on VTL sensitivity were investigated in normal hearing listeners with vocoder simulations of CI processing. The hypotheses were that VTL sensitivity may be reduced by increased frequency mismatch and insufficient spectral resolution in how the frequency range is partitioned, specifically where formants lie. Moreover, optimal band partitioning might mitigate the detrimental effects of frequency mismatch on VTL sensitivity. Results showed that VTL sensitivity decreased with increased frequency mismatch and reduced spectral resolution near the low frequencies of the band partitioning map. Band partitioning was independent of mismatch, indicating that if a given partitioning is suboptimal, a better partitioning might improve VTL sensitivity despite the degree of mismatch. These findings suggest that customizing the frequency partitioning map may enhance VTL perception in individual CI users.

Rate modulation detection thresholds for cochlear implant users.

The perception of temporal amplitude modulations is critical for speech understanding by cochlear implant (CI) users. The present study compared the ability of CI users to detect sinusoidal modulations of the electrical stimulation rate and current level, at different presentation levels (80% and 40% of the dynamic range) and modulation frequencies (10 and 100 Hz). Rate modulation detection thresholds (RMDTs) and amplitude modulation detection thresholds (AMDTs) were measured and compared to assess whether there was a perceptual advantage to either modulation method. Both RMDTs and AMDTs improved with increasing presentation level and decreasing modulation frequency. RMDTs and AMDTs were correlated, indicating that a common processing mechanism may underlie the perception of rate modulation and amplitude modulation, or that some subject-dependent factors affect both types of modulation detection.

Modulation detection interference in cochlear implant listeners under forward masking conditions.

Little is known about cochlear implant (CI) users' ability to process amplitude modulation (AM) under conditions of forward masking (forward-modulation detection/discrimination interference, or F-MDI). In this study, F-MDI was investigated in adult CI listeners using direct electrical stimulation via research interface. The target was sinusoidally amplitude modulated at 50 Hz, and presented to a fixed electrode in the middle of the array. The forward masker was either amplitude modulated at the same rate (AM) or unmodulated and presented at the peak amplitude of its AM counterpart (steady-state peak, SSP). Results showed that the AM masker produced higher modulation thresholds in the target than the SSP masker. The difference (F-MDI) was estimated to be 4.6 dB on average, and did not change with masker-target delays up to 100 ms or with masker-target spatial electrode distances up to eight electrodes. Results with a coherent remote cue presented with the masker showed that confusion effects did not play a role in the observed F-MDI. Traditional recovery from forward masking using the same maskers and a 20-ms probe, measured in four of the subjects, confirmed the expected result: higher thresholds with the SSP masker than the AM masker. Collectively, the results indicate that significant F-MDI occurs in CI users.

Better-ear glimpsing with symmetrically-placed interferers in bilateral cochlear implant users.

For a frontal target in spatially symmetrically placed interferers, normal hearing (NH) listeners can use "better-ear glimpsing" to select time-frequency segments with favorable signal-to-noise ratio in either ear. With an ideal monaural better-ear mask (IMBM) processing, some studies showed that NH listeners can reach similar performance as in the natural binaural listening condition, although interaural phase differences at low frequencies can further improve performance. In principle, bilateral cochlear implant (BiCI) listeners could use the same better-ear glimpsing, albeit without exploiting interaural phase differences. Speech reception thresholds of NH and BiCI listeners were measured in three interferers (speech-shaped stationary noise, nonsense speech, or single talker) either co-located with the target, symmetrically placed at ±60°, or independently presented to each ear, with and without IMBM processing. Furthermore, a bilateral noise vocoder based on the BiCI electrodogram was used in the same NH listeners. Headphone presentation and direct stimulation with head-related transfer functions for spatialization were used in NH and BiCI listeners, respectively. Compared to NH listeners, both NH listeners with vocoder and BiCI listeners showed strongly reduced binaural benefit from spatial separation. However, both groups greatly benefited from IMBM processing as part of the stimulation strategy.

Speech enhancement for cochlear implant recipients.

In this study, a single microphone speech enhancement algorithm is proposed to improve speech intelligibility for cochlear implant recipients. The proposed algorithm combines harmonic structure estimation with a subsequent statistical based speech enhancement stage. Traditional minimum mean square error (MMSE) based speech enhancement methods typically focus on statistical characteristics of the noise and track the noise variance along time dimension. The MMSE method is usually effective for stationary noise, but not as useful for non-stationary noise. To address both stationary and non-stationary noise, the current proposed method not only tracks noise over time, but also estimates the noise structure along the frequency dimension by exploiting the harmonic structure of the target speech. Next, the estimated noise is employed in the traditional MMSE framework for speech enhancement. To evaluate the proposed speech enhancement solution, a formal listener evaluation was performed with 6 cochlear implant recipients. The results suggest that a substantial improvement in speech intelligibility performance can be gained for cochlear implant recipients in noisy environments.