Speech Understanding With Various Maskers in Cochlear-Implant and Simulated Cochlear-Implant Hearing: Effects of Spectral Resolution and Implications for Masking Release

Interaural asymmetry of dynamic range: Abnormal fusion, bilateral interference, and shifts in attention

Speech information in the better ear interferes with the poorer ear in patients with bilateral cochlear implants (BiCIs) who have large asymmetries in speech intelligibility between ears. The goal of the present study was to assess how each ear impacts, and whether one dominates, speech perception using simulated CI processing in older and younger normal-hearing (ONH and YNH) listeners. Dynamic range (DR) was manipulated symmetrically or asymmetrically across spectral bands in a vocoder. We hypothesized that if abnormal integration of speech information occurs with asymmetrical speech understanding, listeners would demonstrate an atypical preference in accuracy when reporting speech presented to the better ear and fusion of speech between the ears (i.e., an increased number of one-word responses when two words were presented). Results from three speech conditions showed that: (1) When the same word was presented to both ears, speech identification accuracy decreased if one or both ears decreased in DR, but listeners usually reported hearing one word. (2) When two words with different vowels were presented to both ears, speech identification accuracy and percentage of two-word responses decreased consistently as DR decreased in one or both ears. (3) When two rhyming words (e.g., bed and led) previously shown to phonologically fuse between ears (e.g., bled) were presented, listeners instead demonstrated interference as DR decreased. The word responded in (2) and (3) came from the right (symmetric) or better (asymmetric) ear, especially in (3) and for ONH listeners in (2). These results suggest that the ear with poorer dynamic range is downweighted by the auditory system, resulting in abnormal fusion and interference, especially for older listeners.

Effects of better-ear glimpsing, binaural unmasking, and spectral resolution on spatial release from masking in cochlear-implant users

Bilateral cochlear-implant (BICI) listeners obtain less spatial release from masking (SRM; speech-recognition improvement for spatially separated vs co-located conditions) than normal-hearing (NH) listeners, especially for symmetrically placed maskers that produce similar long-term target-to-masker ratios at the two ears. Two experiments examined possible causes of this deficit, including limited better-ear glimpsing (using speech information from the more advantageous ear in each time-frequency unit), limited binaural unmasking (using interaural differences to improve signal-in-noise detection), or limited spectral resolution. Listeners had NH (presented with unprocessed or vocoded stimuli) or BICIs. Experiment 1 compared natural symmetric maskers, idealized monaural better-ear masker (IMBM) stimuli that automatically performed better-ear glimpsing, and hybrid stimuli that added worse-ear information, potentially restoring binaural cues. BICI and NH-vocoded SRM was comparable to NH-unprocessed SRM for idealized stimuli but was 14%-22% lower for symmetric stimuli, suggesting limited better-ear glimpsing ability. Hybrid stimuli improved SRM for NH-unprocessed listeners but degraded SRM for BICI and NH-vocoded listeners, suggesting they experienced across-ear interference instead of binaural unmasking. In experiment 2, increasing the number of vocoder channels did not change NH-vocoded SRM. BICI SRM deficits likely reflect a combination of across-ear interference, limited better-ear glimpsing, and poorer binaural unmasking that stems from cochlear-implant-processing limitations other than reduced spectral resolution.

The P300 Auditory Event-Related Potential May Predict Segregation of Competing Speech by Bimodal Cochlear Implant Listeners

Compared to normal-hearing (NH) listeners, cochlear implant (CI) listeners have greater difficulty segregating competing speech. Neurophysiological studies have largely investigated the neural foundations for CI listeners' speech recognition in quiet, mainly using the P300 component of event-related potentials (ERPs). P300 is closely related to cognitive processes involving auditory discrimination, selective attention, and working memory. Different from speech perception in quiet, little is known about the neurophysiological foundations for segregation of competing speech by CI listeners. In this study, ERPs were measured for a 1 vs. 2 kHz contrast in 11 Mandarin-speaking bimodal CI listeners and 11 NH listeners. Speech reception thresholds (SRTs) for a male target talker were measured in steady noise or with a male or female masker. Results showed that P300 amplitudes were significantly larger and latencies were significantly shorter for the NH than for the CI group. Similarly, SRTs were significantly better for the NH than for the CI group. Across all participants, P300 amplitude was significantly correlated with SRTs in steady noise (r = -0.65, p = 0.001) and with the competing male (r = -0.62, p = 0.002) and female maskers (r = -0.60, p = 0.003). Within the CI group, there was a significant correlation between P300 amplitude and SRTs with the male masker (r = -0.78, p = 0.005), which produced the most informational masking. The results suggest that P300 amplitude may be a clinically useful neural correlate of central auditory processing capabilities (e.g., susceptibility to informational masking) in bimodal CI patients.

No interaction between fundamental-frequency differences and spectral region when perceiving speech in a speech background

Differences in fundamental frequency (F0) or pitch between competing voices facilitate our ability to segregate a target voice from interferers, thereby enhancing speech intelligibility. Although lower-numbered harmonics elicit a stronger and more accurate pitch sensation than higher-numbered harmonics, it is unclear whether the stronger pitch leads to an increased benefit of pitch differences when segregating competing talkers. To answer this question, sentence recognition was tested in young normal-hearing listeners in the presence of a single competing talker. The stimuli were presented in a broadband condition or were highpass or lowpass filtered to manipulate the pitch accuracy of the voicing, while maintaining roughly equal speech intelligibility in the highpass and lowpass regions. Performance was measured with average F0 differences (ΔF0) between the target and single-talker masker of 0, 2, and 4 semitones. Pitch discrimination abilities were also measured to confirm that the lowpass-filtered stimuli elicited greater pitch accuracy than the highpass-filtered stimuli. No interaction was found between filter type and ΔF0 in the sentence recognition task, suggesting little or no effect of harmonic rank or pitch accuracy on the ability to use F0 to segregate natural voices, even when the average ΔF0 is relatively small. The results suggest that listeners are able to obtain some benefit of pitch differences between competing voices, even when pitch salience and accuracy is low.

Using Spectral Blurring to Assess Effects of Channel Interaction on Speech-in-Noise Perception with Cochlear Implants

Cochlear implant (CI) listeners struggle to understand speech in background noise. Interactions between electrode channels due to current spread increase the masking of speech by noise and lead to difficulties with speech perception. Strategies that reduce channel interaction therefore have the potential to improve speech-in-noise perception by CI listeners, but previous results have been mixed. We investigated the effects of channel interaction on speech-in-noise perception and its association with spectro-temporal acuity in a listening study with 12 experienced CI users. Instead of attempting to reduce channel interaction, we introduced spectral blurring to simulate some of the effects of channel interaction by adjusting the overlap between electrode channels at the input level of the analysis filters or at the output by using several simultaneously stimulated electrodes per channel. We measured speech reception thresholds in noise as a function of the amount of blurring applied to either all 15 electrode channels or to 5 evenly spaced channels. Performance remained roughly constant as the amount of blurring applied to all channels increased up to some knee point, above which it deteriorated. This knee point differed across listeners in a way that correlated with performance on a non-speech spectro-temporal task, and is proposed here as an individual measure of channel interaction. Surprisingly, even extreme amounts of blurring applied to 5 channels did not affect performance. The effects on speech perception in noise were similar for blurring at the input and at the output of the CI. The results are in line with the assumption that experienced CI users can make use of a limited number of effective channels of information and tolerate some deviations from their everyday settings when identifying speech in the presence of a masker. Furthermore, these findings may explain the mixed results by strategies that optimized or deactivated a small number of electrodes evenly distributed along the array by showing that blurring or deactivating one-third of the electrodes did not harm speech-in-noise performance.

Speech masking release in Hybrid cochlear implant users: Roles of spectral and temporal cues in electric-acoustic hearing

When compared with cochlear implant (CI) users utilizing electric-only (E-Only) stimulation, CI users utilizing electric-acoustic stimulation (EAS) in the implanted ear show improved speech recognition in modulated noise relative to steady-state noise (i.e., speech masking release). It has been hypothesized, but not shown, that masking release is attributed to spectral resolution and temporal fine structure (TFS) provided by acoustic hearing. To address this question, speech masking release, spectral ripple density discrimination thresholds, and fundamental frequency difference limens (f₀DLs) were evaluated in the acoustic-only (A-Only), E-Only, and EAS listening modes in EAS CI users. The spectral ripple and f₀DL tasks are thought to reflect access to spectral and TFS cues, which could impact speech masking release. Performance in all three measures was poorest when EAS CI users were tested using the E-Only listening mode, with significant improvements in A-Only and EAS listening modes. f₀DLs, but not spectral ripple density discrimination thresholds, significantly correlated with speech masking release when assessed in the EAS listening mode. Additionally, speech masking release correlated with AzBio sentence recognition in noise. The correlation between speech masking release and f₀DLs likely indicates that TFS cues provided by residual hearing were used to obtain speech masking release, which aided sentence recognition in noise.

Factors Affecting Bimodal Benefit in Pediatric Mandarin-Speaking Chinese Cochlear Implant Users

OBJECTIVES

While fundamental frequency (F0) cues are important to both lexical tone perception and multitalker segregation, F0 cues are poorly perceived by cochlear implant (CI) users. Adding low-frequency acoustic hearing via a hearing aid in the contralateral ear may improve CI users' F0 perception. For English-speaking CI users, contralateral acoustic hearing has been shown to improve perception of target speech in noise and in competing talkers. For tonal languages such as Mandarin Chinese, F0 information is lexically meaningful. Given competing F0 information from multiple talkers and lexical tones, contralateral acoustic hearing may be especially beneficial for Mandarin-speaking CI users' perception of competing speech.

DESIGN

Bimodal benefit (CI+hearing aid - CI-only) was evaluated in 11 pediatric Mandarin-speaking Chinese CI users. In experiment 1, speech recognition thresholds (SRTs) were adaptively measured using a modified coordinated response measure test; subjects were required to correctly identify 2 keywords from among 10 choices in each category. SRTs were measured with CI-only or bimodal listening in the presence of steady state noise (SSN) or competing speech with the same (M+M) or different voice gender (M+F). Unaided thresholds in the non-CI ear and demographic factors were compared with speech performance. In experiment 2, SRTs were adaptively measured in SSN for recognition of 5 keywords, a more difficult listening task than the 2-keyword recognition task in experiment 1.

RESULTS

In experiment 1, SRTs were significantly lower for SSN than for competing speech in both the CI-only and bimodal listening conditions. There was no significant difference between CI-only and bimodal listening for SSN and M+F (p > 0.05); SRTs were significantly lower for CI-only than for bimodal listening for M+M (p < 0.05), suggesting bimodal interference. Subjects were able to make use of voice gender differences for bimodal listening (p < 0.05) but not for CI-only listening (p > 0.05). Unaided thresholds in the non-CI ear were positively correlated with bimodal SRTs for M+M (p < 0.006) but not for SSN or M+F. No significant correlations were observed between any demographic variables and SRTs (p > 0.05 in all cases). In experiment 2, SRTs were significantly lower with two than with five keywords (p < 0.05). A significant bimodal benefit was observed only for the 5-keyword condition (p < 0.05).

CONCLUSIONS

With the CI alone, subjects experienced greater interference with competing speech than with SSN and were unable to use voice gender difference to segregate talkers. For the coordinated response measure task, subjects experienced no bimodal benefit and even bimodal interference when competing talkers were the same voice gender. A bimodal benefit in SSN was observed for the five-keyword condition but not for the two-keyword condition, suggesting that bimodal listening may be more beneficial as the difficulty of the listening task increased. The present data suggest that bimodal benefit may depend on the type of masker and/or the difficulty of the listening task.

Using recurrent neural networks to improve the perception of speech in non-stationary noise by people with cochlear implants

Speech-in-noise perception is a major problem for users of cochlear implants (CIs), especially with non-stationary background noise. Noise-reduction algorithms have produced benefits but relied on a priori information about the target speaker and/or background noise. A recurrent neural network (RNN) algorithm was developed for enhancing speech in non-stationary noise and its benefits were evaluated for speech perception, using both objective measures and experiments with CI simulations and CI users. The RNN was trained using speech from many talkers mixed with multi-talker or traffic noise recordings. Its performance was evaluated using speech from an unseen talker mixed with different noise recordings of the same class, either babble or traffic noise. Objective measures indicated benefits of using a recurrent over a feed-forward architecture, and predicted better speech intelligibility with than without the processing. The experimental results showed significantly improved intelligibility of speech in babble noise but not in traffic noise. CI subjects rated the processed stimuli as significantly better in terms of speech distortions, noise intrusiveness, and overall quality than unprocessed stimuli for both babble and traffic noise. These results extend previous findings for CI users to mostly unseen acoustic conditions with non-stationary noise.

The effect of presentation level on spectrotemporal modulation detection

The understanding of speech in noise relies (at least partially) on spectrotemporal modulation sensitivity. This sensitivity can be measured by spectral ripple tests, which can be administered at different presentation levels. However, it is not known how presentation level affects spectrotemporal modulation thresholds. In this work, we present behavioral data for normal-hearing adults which show that at higher ripple densities (2 and 4 ripples/oct), increasing presentation level led to worse discrimination thresholds. Results of a computational model suggested that the higher thresholds could be explained by a worsening of the spectrotemporal representation in the auditory nerve due to broadening of cochlear filters and neural activity saturation. Our results demonstrate the importance of taking presentation level into account when administering spectrotemporal modulation detection tests.

Development and validation of a spectro-temporal processing test for cochlear-implant listeners

Psychophysical tests of spectro-temporal resolution may aid the evaluation of methods for improving hearing by cochlear implant (CI) listeners. Here the STRIPES (Spectro-Temporal Ripple for Investigating Processor EffectivenesS) test is described and validated. Like speech, the test requires both spectral and temporal processing to perform well. Listeners discriminate between complexes of sine sweeps which increase or decrease in frequency; difficulty is controlled by changing the stimulus spectro-temporal density. Care was taken to minimize extraneous cues, forcing listeners to perform the task only on the direction of the sweeps. Vocoder simulations with normal hearing listeners showed that the STRIPES test was sensitive to the number of channels and temporal information fidelity. An evaluation with CI listeners compared a standard processing strategy with one having very wide filters, thereby spectrally blurring the stimulus. Psychometric functions were monotonic for both strategies and five of six participants performed better with the standard strategy. An adaptive procedure revealed significant differences, all in favour of the standard strategy, at the individual listener level for six of eight CI listeners. Subsequent measures validated a faster version of the test, and showed that STRIPES could be performed by recently implanted listeners having no experience of psychophysical testing.