Noise-band Vocoder Research Articles

Cochlear-implant simulated spectral degradation attenuates emotional responses to environmental sounds

Objective Cochlear implants (CI) provide users with a spectrally degraded acoustic signal that could impact their auditory emotional experiences. This study evaluated the effects of CI-simulated spectral degradation on emotional valence and arousal elicited by environmental sounds. Design Thirty emotionally evocative sounds were filtered through a noise-band vocoder. Participants rated the perceived valence and arousal elicited by each of the full-spectrum and vocoded stimuli. These ratings were compared across acoustic conditions (full-spectrum, vocoded) and as a function of stimulus type (unpleasant, neutral, pleasant). Study Sample Twenty-five young adults (age 19 to 34 years) with normal hearing. Results Emotional responses were less extreme for spectrally degraded (i.e., vocoded) sounds than for full-spectrum sounds. Specifically, spectrally degraded stimuli were perceived as more negative and less arousing than full-spectrum stimuli. Conclusion By meticulously replicating CI spectral degradation while controlling for variables that are confounded within CI users, these findings indicate that CI spectral degradation can compress the range of sound-induced emotion independent of hearing loss and other idiosyncratic device- or person-level variables. Future work will characterize emotional reactions to sound in CI users via objective, psychoacoustic, and subjective measures.

Spectral resolution and voice cue weighting in adults with normal hearing

Cochlear implant (CI) users use acoustic voice cues differently than normal hearing (NH) adults to identify a talker’s gender. Specifically, whereas NH listeners weight a talker’s fundamental frequency (f0) and resonance (operationalized as vocal tract length or VTL) equally, CI users rely almost exclusively on f0. CI users’ abnormal cue weighting may partially arise due to degraded auditory information delivered by the CI. We hypothesized that altering the amount of spectral information in the signal impacts voice cue weighting in NH listeners. Thirty NH adults performed a gender identification task. Auditory stimuli were monosyllabic words synthesized to have one of five f0 values and one of five VTL values. Synthesized voices were processed using a 16-, 8-, and 4-channel noise-band vocoder. Perceptual weights for each voice cue were estimated as the coefficients for f0 and VTL in regression models, with higher coefficients corresponding to stronger perceptual weightings. Listeners relied more on f0 in conditions with less spectral resolution than in conditions with greater spectral resolution. Cue weights for VTL did not change across conditions. Results suggest that removing frequency information from the auditory signal can modify the extent to which listeners use f0 to identify the gender of a talker.

Self-pacing ameliorates recall deficit when listening to vocoded discourse: a cochlear implant simulation.

In spite of its apparent ease, comprehension of spoken discourse represents a complex linguistic and cognitive operation. The difficulty of such an operation can increase when the speech is degraded, as is the case with cochlear implant users. However, the additional challenges imposed by degraded speech may be mitigated to some extent by the linguistic context and pace of presentation. An experiment is reported in which young adults with age-normal hearing recalled discourse passages heard with clear speech or with noise-band vocoding used to simulate the sound of speech produced by a cochlear implant. Passages were varied in inter-word predictability and presented either without interruption or in a self-pacing format that allowed the listener to control the rate at which the information was delivered. Results showed that discourse heard with clear speech was better recalled than discourse heard with vocoded speech, discourse with a higher average inter-word predictability was better recalled than discourse with a lower average inter-word predictability, and self-paced passages were recalled better than those heard without interruption. Of special interest was the semantic hierarchy effect: the tendency for listeners to show better recall for main ideas than mid-level information or detail from a passage as an index of listeners' ability to understand the meaning of a passage. The data revealed a significant effect of inter-word predictability, in that passages with lower predictability had an attenuated semantic hierarchy effect relative to higher-predictability passages. Results are discussed in terms of broadening cochlear implant outcome measures beyond current clinical measures that focus on single-word and sentence repetition.

Effects of spectral degradation on gated word recognition.

Although much is known about how normal-hearing listeners process spoken words under ideal listening conditions, little is known about how a degraded signal, such as speech transmitted via cochlear implants, affects the word recognition process. In this study, gated word recognition performance was measured with the goal of describing the time course of word identification by using a noise-band vocoder simulation. The results of this study demonstrate that spectral degradations can impact the temporal aspects of speech processing. These results also provide insights into the potential advantages of enhancing spectral resolution in the processing of spoken words.

Effects of spectral resolution and smearing on gated word recognition

In a word gating task, listeners are presented with increasing amounts of word-onset information (a series of increasingly longer temporal “gates”), and following each gate, they are asked to indicate what they think the target word is. Listeners with normal hearing (NH) typically recognize a target word well before they “hear” the entire word. Listeners with cochlear implants (CIs), however, need to hear almost the entire target word (require a greater amount of word-onset information) to recognize it. Here, we hypothesized that the poor spectral resolution, due to the limited number of channels and interactions between adjacent channels, may have a negative impact on gated word recognition performance. We manipulated spectral resolution by using: (1) a noise-band vocoder with a variable number of spectral channels; and (2) a vocoder with variable carrier filter slopes to simulate channel interaction. We determined the minimum amount of word-onset information required to recognize spoken words. Initial results suggest that the gated word recognition performance remained roughly unchanged as the amount of spectral degradation applied increased up to some extent, beyond which it deteriorated. With eight channels, which resemble the spectral resolution available to most CI users, even the most focused stimulation yielded poorer results.

Effects of Temporal Envelope Cutoff Frequency, Number of Channels, and Carrier Type on Brainstem Neural Representation of Pitch in Vocoded Speech.

The objective of this study was to determine if and how the subcortical neural representation of pitch cues in listeners with normal hearing is affected by systematic manipulation of vocoder parameters. This study assessed the effects of temporal envelope cutoff frequency (50 and 500 Hz), number of channels (1-32), and carrier type (sine-wave and noise-band) on brainstem neural representation of fundamental frequency (f o) in frequency-following responses (FFRs) to vocoded vowels of 15 young adult listeners with normal hearing. Results showed that FFR f o strength (quantified as absolute f o magnitude divided by noise floor [NF] magnitude) significantly improved with 500-Hz vs. 50-Hz temporal envelopes for all channel numbers and both carriers except the 1-channel noise-band vocoder. FFR f o strength with 500-Hz temporal envelopes significantly improved when the channel number increased from 1 to 2, but it either declined (sine-wave vocoders) or saturated (noise-band vocoders) when the channel number increased from 4 to 32. FFR f o strength with 50-Hz temporal envelopes was similarly small for both carriers with all channel numbers, except for a significant improvement with the 16-channel sine-wave vocoder. With 500-Hz temporal envelopes, FFR f o strength was significantly greater for sine-wave vocoders than for noise-band vocoders with channel numbers 1-8; no significant differences were seen with 16 and 32 channels. With 50-Hz temporal envelopes, the carrier effect was only observed with 16 channels. In contrast, there was no significant carrier effect for the absolute f o magnitude. Compared to sine-wave vocoders, noise-band vocoders had a higher NF and thus lower relative FFR f o strength. It is important to normalize the f o magnitude relative to the NF when analyzing the FFRs to vocoded speech. The physiological findings reported here may result from the availability of f o-related temporal periodicity and spectral sidelobes in vocoded signals and should be considered when selecting vocoder parameters and interpreting results in future physiological studies. In general, the dependence of brainstem neural phase-locking strength to f o on vocoder parameters may confound the comparison of pitch-related behavioral results across different vocoder designs.

Multidimensional Timbre Spaces of Cochlear Implant Vocoded and Non-vocoded Synthetic Female Singing Voices.

Many post-lingually deafened cochlear implant (CI) users report that they no longer enjoy listening to music, which could possibly contribute to a perceived reduction in quality of life. One aspect of music perception, vocal timbre perception, may be difficult for CI users because they may not be able to use the same timbral cues available to normal hearing listeners. Vocal tract resonance frequencies have been shown to provide perceptual cues to voice categories such as baritone, tenor, mezzo-soprano, and soprano, while changes in glottal source spectral slope are believed to be related to perception of vocal quality dimensions such as fluty vs. brassy. As a first step toward understanding vocal timbre perception in CI users, we employed an 8-channel noise-band vocoder to test how vocoding can alter the timbral perception of female synthetic sung vowels across pitches. Non-vocoded and vocoded stimuli were synthesized with vibrato using 3 excitation source spectral slopes and 3 vocal tract transfer functions (mezzo-soprano, intermediate, soprano) at the pitches C4, B4, and F5. Six multi-dimensional scaling experiments were conducted: C4 not vocoded, C4 vocoded, B4 not vocoded, B4 vocoded, F5 not vocoded, and F5 vocoded. At the pitch C4, for both non-vocoded and vocoded conditions, dimension 1 grouped stimuli according to voice category and was most strongly predicted by spectral centroid from 0 to 2 kHz. While dimension 2 grouped stimuli according to excitation source spectral slope, it was organized slightly differently and predicted by different acoustic parameters in the non-vocoded and vocoded conditions. For pitches B4 and F5 spectral centroid from 0 to 2 kHz most strongly predicted dimension 1. However, while dimension 1 separated all 3 voice categories in the vocoded condition, dimension 1 only separated the soprano stimuli from the intermediate and mezzo-soprano stimuli in the non-vocoded condition. While it is unclear how these results predict timbre perception in CI listeners, in general, these results suggest that perhaps some aspects of vocal timbre may remain.

Interactions between auditory quality, linguistic context, short-term memory, and speech recognition

The ability to recognize speech depends on several factors, including the quality of the auditory input, linguistic context, and the listener’s cognitive abilities. These factors correlate with speech recognition, but less consideration has been given to how these factors interact. Here, we present evidence from two studies indicating that interactions exist between quality, context, and short-term memory. In study one, we demonstrate that contextual expectations across stimulus sets can determine the relationship between auditory quality and short-term memory. Specifically, recall of lists of digits is not affected by noise-band vocoding, whereas vocoding impairs both item identification and recall for lists of single syllable words drawn from a large, untrained set. In study two, we demonstrate that correlations between digit list recall and PRESTO sentence recognition are strongest when auditory quality is poor, whereas correlations between digit and word list recall weaken with decreasing auditory quality. This finding suggests that auditory quality and semantic context moderate the relationship between memory and speech recognition. We conclude that incorporating experimental measures of auditory quality, short-term memory, and recognition of speech materials with different contexts will provide a clearer perspective on how these factors relate to speech recognition in listeners with hearing loss.The ability to recognize speech depends on several factors, including the quality of the auditory input, linguistic context, and the listener’s cognitive abilities. These factors correlate with speech recognition, but less consideration has been given to how these factors interact. Here, we present evidence from two studies indicating that interactions exist between quality, context, and short-term memory. In study one, we demonstrate that contextual expectations across stimulus sets can determine the relationship between auditory quality and short-term memory. Specifically, recall of lists of digits is not affected by noise-band vocoding, whereas vocoding impairs both item identification and recall for lists of single syllable words drawn from a large, untrained set. In study two, we demonstrate that correlations between digit list recall and PRESTO sentence recognition are strongest when auditory quality is poor, whereas correlations between digit and word list recall weaken with decreasing auditory qua...

Reducing Simulated Channel Interaction Reveals Differences in Phoneme Identification Between Children and Adults With Normal Hearing.

Channel interaction, the stimulation of overlapping populations of auditory neurons by distinct cochlear implant (CI) channels, likely limits the speech perception performance of CI users. This study examined the role of vocoder-simulated channel interaction in the ability of children with normal hearing (cNH) and adults with normal hearing (aNH) to recognize spectrally degraded speech. The primary aim was to determine the interaction between number of processing channels and degree of simulated channel interaction on phoneme identification performance as a function of age for cNH and to relate those findings to aNH and to CI users. Medial vowel and consonant identification of cNH (age 8-17 years) and young aNH were assessed under six (for children) or nine (for adults) different conditions of spectral degradation. Stimuli were processed using a noise-band vocoder with 8, 12, and 15 channels and synthesis filter slopes of 15 (aNH only), 30, and 60 dB/octave (all NH subjects). Steeper filter slopes (larger numbers) simulated less electrical current spread and, therefore, less channel interaction. Spectrally degraded performance of the NH listeners was also compared with the unprocessed phoneme identification of school-aged children and adults with CIs. Spectrally degraded phoneme identification improved as a function of age for cNH. For vowel recognition, cNH exhibited an interaction between the number of processing channels and vocoder filter slope, whereas aNH did not. Specifically, for cNH, increasing the number of processing channels only improved vowel identification in the steepest filter slope condition. Additionally, cNH were more sensitive to changes in filter slope. As the filter slopes increased, cNH continued to receive vowel identification benefit beyond where aNH performance plateaued or reached ceiling. For all NH participants, consonant identification improved with increasing filter slopes but was unaffected by the number of processing channels. Although cNH made more phoneme identification errors overall, their phoneme error patterns were similar to aNH. Furthermore, consonant identification of adults with CI was comparable to aNH listening to simulations with shallow filter slopes (15 dB/octave). Vowel identification of earlier-implanted pediatric ears was better than that of later-implanted ears and more comparable to cNH listening in conditions with steep filter slopes (60 dB/octave). Recognition of spectrally degraded phonemes improved when simulated channel interaction was reduced, particularly for children. cNH showed an interaction between number of processing channels and filter slope for vowel identification. The differences observed between cNH and aNH suggest that identification of spectrally degraded phonemes continues to improve through adolescence and that children may benefit from reduced channel interaction beyond where adult performance has plateaued. Comparison to CI users suggests that early implantation may facilitate development of better phoneme discrimination.

Multimodal and Spectral Degradation Effects on Speech and Emotion Recognition in Adult Listeners.

For cochlear implant (CI) users, degraded spectral input hampers the understanding of prosodic vocal emotion, especially in difficult listening conditions. Using a vocoder simulation of CI hearing, we examined the extent to which informative multimodal cues in a talker’s spoken expressions improve normal hearing (NH) adults’ speech and emotion perception under different levels of spectral degradation (two, three, four, and eight spectral bands). Participants repeated the words verbatim and identified emotions (among four alternative options: happy, sad, angry, and neutral) in meaningful sentences that are semantically congruent with the expression of the intended emotion. Sentences were presented in their natural speech form and in speech sampled through a noise-band vocoder in sound (auditory-only) and video (auditory–visual) recordings of a female talker. Visual information had a more pronounced benefit in enhancing speech recognition in the lower spectral band conditions. Spectral degradation, however, did not interfere with emotion recognition performance when dynamic visual cues in a talker’s expression are provided as participants scored at ceiling levels across all spectral band conditions. Our use of familiar sentences that contained congruent semantic and prosodic information have high ecological validity, which likely optimized listener performance under simulated CI hearing and may better predict CI users’ outcomes in everyday listening contexts.

Word segmentation from noise-band vocoded speech

ABSTRACTSpectral degradation reduces access to the acoustics of spoken language and compromises how learners break into its structure. We hypothesised that spectral degradation disrupts word segmentation, but that listeners can exploit other cues to restore detection of words. Normal-hearing adults were familiarised to artificial speech that was unprocessed or spectrally degraded by noise-band vocoding into 16 or 8 spectral channels. The monotonic speech stream was pause-free (Experiment 1), interspersed with isolated words (Experiment 2), or slowed by 33% (Experiment 3). Participants were tested on segmentation of familiar vs. novel syllable sequences and on recognition of individual syllables. As expected, vocoding hindered both word segmentation and syllable recognition. The addition of isolated words, but not slowed speech, improved segmentation. We conclude that syllable recognition is necessary but not sufficient for successful word segmentation, and that isolated words can facilitate listeners’ access to the structure of acoustically degraded speech.

Vowel and consonant confusions from spectrally manipulated stimuli designed to simulate poor cochlear implant electrode-neuron interfaces.

Suboptimal interfaces between cochlear implant (CI) electrodes and auditory neurons result in a loss or distortion of spectral information in specific frequency regions, which likely decreases CI users' speech identification performance. This study exploited speech acoustics to model regions of distorted CI frequency transmission to determine the perceptual consequences of suboptimal electrode-neuron interfaces. Normal hearing adults identified naturally spoken vowels and consonants after spectral information was manipulated through a noiseband vocoder: either (1) low-, middle-, or high-frequency regions of information were removed by zeroing the corresponding channel outputs, or (2) the same regions were distorted by splitting filter outputs to neighboring filters. These conditions simulated the detrimental effects of suboptimal CI electrode-neuron interfaces on spectral transmission. Vowel and consonant confusion patterns were analyzed with sequential information transmission, perceptual distance, and perceptual vowel space analyses. Results indicated that both types of spectral manipulation were equally destructive. Loss or distortion of frequency information produced similar effects on phoneme identification performance and confusion patterns. Consonant error patterns were consistently based on place of articulation. Vowel confusions showed that perceptions gravitated away from the degraded frequency region in a predictable manner, indicating that vowels can probe frequency-specific regions of spectral degradations.

Age-Related Differences in Listening Effort During Degraded Speech Recognition.

The purpose of the present study was to quantify age-related differences in executive control as it relates to dual-task performance, which is thought to represent listening effort, during degraded speech recognition. Twenty-five younger adults (YA; 18-24 years) and 21 older adults (OA; 56-82 years) completed a dual-task paradigm that consisted of a primary speech recognition task and a secondary visual monitoring task. Sentence material in the primary task was either unprocessed or spectrally degraded into 8, 6, or 4 spectral channels using noise-band vocoding. Performance on the visual monitoring task was assessed by the accuracy and reaction time of participants' responses. Performance on the primary and secondary task was quantified in isolation (i.e., single task) and during the dual-task paradigm. Participants also completed a standardized psychometric measure of executive control, including attention and inhibition. Statistical analyses were implemented to evaluate changes in listeners' performance on the primary and secondary tasks (1) per condition (unprocessed vs. vocoded conditions); (2) per task (single task vs. dual task); and (3) per group (YA vs. OA). Speech recognition declined with increasing spectral degradation for both YA and OA when they performed the task in isolation or concurrently with the visual monitoring task. OA were slower and less accurate than YA on the visual monitoring task when performed in isolation, which paralleled age-related differences in standardized scores of executive control. When compared with single-task performance, OA experienced greater declines in secondary-task accuracy, but not reaction time, than YA. Furthermore, results revealed that age-related differences in executive control significantly contributed to age-related differences on the visual monitoring task during the dual-task paradigm. OA experienced significantly greater declines in secondary-task accuracy during degraded speech recognition than YA. These findings are interpreted as suggesting that OA expended greater listening effort than YA, which may be partially attributed to age-related differences in executive control.

Loudness Contour Can Influence Mandarin Tone Recognition: Vocoder Simulation and Cochlear Implants.

Lexical tone recognition with current cochlear implants (CI) remains unsatisfactory due to significantly degraded pitch-related acoustic cues, which dominate the tone recognition by normal-hearing (NH) listeners. Several secondary cues (e.g., amplitude contour, duration, and spectral envelope) that influence tone recognition in NH listeners and CI users have been studied. This work proposes a loudness contour manipulation algorithm, namely Loudness-Tone (L-Tone), to investigate the effects of loudness contour on Mandarin tone recognition and the effectiveness of using loudness cue to enhance tone recognition for CI users. With L-Tone, the intensity of sound samples is multiplied by gain values determined by instantaneous fundamental frequencies (F0s) and pre-defined gain- F0 mapping functions. Perceptual experiments were conducted with a four-channel noise-band vocoder simulation in NH listeners and with CI users. The results suggested that 1) loudness contour is a useful secondary cue for Mandarin tone recognition, especially when pitch cues are significantly degraded; 2) L-Tone can be used to improve Mandarin tone recognition in both simulated and actual CI-hearing without significant negative effect on vowel and consonant recognition. L-Tone is a promising algorithm for incorporation into real-time CI processing and off-line CI rehabilitation training software.

A neural-based vocoder implementation for evaluating cochlear implant coding strategies

Most simulations of cochlear implant (CI) coding strategies rely on standard vocoders that are based on purely signal processing techniques. However, these models neither account for various biophysical phenomena, such as neural stochasticity and refractoriness, nor for effects of electrical stimulation, such as spectral smearing as a function of stimulus intensity. In this paper, a neural model that accounts for stochastic firing, parasitic spread of excitation across neuron populations, and neuronal refractoriness, was developed and augmented as a preprocessing stage for a standard 22-channel noise-band vocoder. This model was used to subjectively and objectively assess consonant discrimination in commercial and experimental coding strategies.Stimuli consisting of consonant-vowel (CV) and vowel-consonant-vowel (VCV) tokens were processed by either the Advanced Combination Encoder (ACE) or the Excitability Controlled Coding (ECC) strategies, and later resynthesized to audio using the aforementioned vocoder model. Baseline performance was measured using unprocessed versions of the speech tokens.Behavioural responses were collected from seven normal hearing (NH) volunteers, while EEG data were recorded from five NH participants. Psychophysical results indicate that while there may be a difference in consonant perception between the two tested coding strategies, mismatch negativity (MMN) waveforms do not show any marked trends in CV or VCV contrast discrimination.

Discriminability and Perceptual Saliency of Temporal and Spectral Cues for Final Fricative Consonant Voicing in Simulated Cochlear-Implant and Bimodal Hearing.

Multiple redundant acoustic cues can contribute to the perception of a single phonemic contrast. This study investigated the effect of spectral degradation on the discriminability and perceptual saliency of acoustic cues for identification of word-final fricative voicing in “loss” versus “laws”, and possible changes that occurred when low-frequency acoustic cues were restored. Three acoustic cues that contribute to the word-final /s/-/z/ contrast (first formant frequency [F1] offset, vowel–consonant duration ratio, and consonant voicing duration) were systematically varied in synthesized words. A discrimination task measured listeners’ ability to discriminate differences among stimuli within a single cue dimension. A categorization task examined the extent to which listeners make use of a given cue to label a syllable as “loss” versus “laws” when multiple cues are available. Normal-hearing listeners were presented with stimuli that were either unprocessed, processed with an eight-channel noise-band vocoder to approximate spectral degradation in cochlear implants, or low-pass filtered. Listeners were tested in four listening conditions: unprocessed, vocoder, low-pass, and a combined vocoder + low-pass condition that simulated bimodal hearing. Results showed a negative impact of spectral degradation on F1 cue discrimination and a trading relation between spectral and temporal cues in which listeners relied more heavily on the temporal cues for “loss-laws” identification when spectral cues were degraded. Furthermore, the addition of low-frequency fine-structure cues in simulated bimodal hearing increased the perceptual saliency of the F1 cue for “loss-laws” identification compared with vocoded speech. Findings suggest an interplay between the quality of sensory input and cue importance.

Information theoretic evaluation of a noiseband-based cochlear implant simulator

Noise-band vocoders are often used to simulate the signal processing algorithms used in cochlear implants (CIs), producing acoustic stimuli that may be presented to normal hearing (NH) subjects. Such evaluations may obviate the heterogeneity of CI user populations, achieving greater experimental control than when testing on CI subjects. However, it remains an open question whether advancements in algorithms developed on NH subjects using a simulator will necessarily improve performance in CI users. This study assessed the similarity in vowel identification of CI subjects and NH subjects using an 8-channel noise-band vocoder simulator configured to match input and output frequencies or to mimic output after a basalward shift of input frequencies. Under each stimulus condition, NH subjects performed the task both with and without feedback/training. Similarity of NH subjects to CI users was evaluated using correct identification rates and information theoretic approaches. Feedback/training produced higher rates of correct identification, as expected, but also resulted in error patterns that were closer to those of the CI users. Further evaluation remains necessary to determine how patterns of confusion at the token level are affected by the various parameters in CI simulators, providing insight into how a true CI simulation may be developed to facilitate more rapid prototyping and testing of novel CI signal processing and electrical stimulation strategies.

Perceptual learning of temporally interrupted spectrally degraded speech.

Normal-hearing (NH) listeners make use of context, speech redundancy and top-down linguistic processes to perceptually restore inaudible or masked portions of speech. Previous research has shown poorer perception and restoration of interrupted speech in CI users and NH listeners tested with acoustic simulations of CIs. Three hypotheses were investigated: (1) training with CI simulations of interrupted sentences can teach listeners to use the high-level restoration mechanisms more effectively, (2) phonemic restoration benefit, an increase in intelligibility of interrupted sentences once its silent gaps are filled with noise, can be induced with training, and (3) perceptual learning of interrupted sentences can be reflected in clinical speech audiometry. To test these hypotheses, NH listeners were trained using periodically interrupted sentences, also spectrally degraded with a noiseband vocoder as CI simulation. Feedback was presented by displaying the sentence text and playing back both the intact and the interrupted CI simulation of the sentence. Training induced no phonemic restoration benefit, and learning was not transferred to speech audiometry measured with words. However, a significant improvement was observed in overall intelligibility of interrupted spectrally degraded sentences, with or without filler noise, suggesting possibly better use of restoration mechanisms as a result of training.

Discriminability and perceptual saliency of acoustic cues for final consonant voicing in simulations of cochlear-implant and electric-acoustic stimulation

Multiple redundant acoustic cues can contribute to the perception of a single class of speech sounds. This study investigates the effect of spectral degradation on the discriminability and perceptual saliency of acoustic cues for phonetic identification of word-final fricative voicing in “loss” versus “laws,” and possible changes that occur when low-frequency acoustic cues are restored. Three acoustic cues that contribute to the word-final /s/-/z/ contrast (first formant frequency [F1] offset, vowel-consonant duration ratio, and consonant voicing duration) were systematically varied in modified natural speech syllables. The first experiment measured listeners' ability to discriminate differences among stimuli within a single cue dimension. The second experiment examined the extent to which listeners make use of a given cue to label a syllable as “loss” versus “laws” when multiple cues are available. Normal-hearing listeners were presented with stimuli that were either unprocessed, processed with 8-channel noise-band vocoder to approximate CI spectral degradation, or low-pass (LP) filtered to simulate low-frequency residual hearing. They were tested in four listening conditions: unprocessed, vocoder alone, LP alone, and vocoder + LP where low-frequency fine-structure cues could enhance F1 perception and voicing cues. The impact of listening condition on discriminability and weighting of different acoustic cues will be discussed.

Top–down restoration of speech in cochlear-implant users

In noisy listening conditions, intelligibility of degraded speech can be enhanced by top-down restoration. Cochlear implant (CI) users have difficulty understanding speech in noisy environments. This could partially be due to reduced top-down restoration of speech, which may be related to the changes that the electrical stimulation imposes on the bottom-up cues. We tested this hypothesis using the phonemic restoration (PhR) paradigm in which speech interrupted with periodic silent intervals is perceived illusorily continuous (continuity illusion or CoI) and becomes more intelligible (PhR benefit) when the interruptions are filled with noise bursts. Using meaningful sentences, both CoI and PhR benefit were measured in CI users, and compared with those of normal-hearing (NH) listeners presented with normal speech and 8-channel noise-band vocoded speech, acoustically simulating CIs. CI users showed different patterns in both PhR benefit and CoI, compared to NH results with or without the noise-band vocoding. However, they were able to use top-down restoration under certain test conditions. This observation supports the idea that changes in bottom-up cues can impose changes to the top–down processes needed to enhance intelligibility of degraded speech. The knowledge that CI users seem to be able to do restoration under the right circumstances could be exploited in patient rehabilitation and product development.