Vowel Identification Scores Research Articles

Modeling Concurrent Vowel Scores Using the Time Delay Neural Network and Multitask Learning

The difference in fundamental frequency (F0) between vowels is a segregation cue for identifying concurrent vowels. Younger listeners with normal hearing (YNH) showed that the percent identification score of both vowels improved with increasing F0 difference and asymptoted at higher F0 difference. The current study developed a deep-neural-network model consisting of a time-delay neural network and multitask learning (TDNN-MTL) to predict the concurrent vowel scores. The input to the TDNN-MTL were temporal responses of concurrent vowels obtained using the auditory-nerve model. The TDNN-MTL was trained at a 3-Hz F0 difference until the score (80%) became closer to the YNH score (85%). The TDNN-MTL learned the formant coding and aided in the segregation of concurrent vowels into dominant and recessive vowels. The total weighted loss was tailored to mimic the dominant-recessive relationship of the YNH scores. The TDNN-MTL was tested against the other five F0 differences. The TDNN-MTL model scores of both vowels were successful in predicting the YNH scores. The chi-square tests revealed that the TDNN-MTL model scores correlated well with the YNH scores compared with other F0-segregation and multi-layer perceptron models. The TDNN-MTL also correctly predicted the one vowel identification scores and F0 benefit. These findings suggest that the trained TDNN-MTL accurately predicts the formant coding of the concurrent vowels across F0 differences, which aided in capturing the YNH scores. The TDNN-MTL can be extended to validate the behavioral studies on concurrent vowels obtained across acoustic (e.g., vowel duration and level) and auditory changes (e.g., aging and hearing loss).

Single-Channel Focused Thresholds Relate to Vowel Identification in Pediatric and Adult Cochlear Implant Listeners.

Speech recognition outcomes are highly variable among pediatric and adult cochlear implant (CI) listeners. Although there is some evidence that the quality of the electrode-neuron interface (ENI) contributes to this large variability in auditory perception, its relationship with speech outcomes is not well understood. Single-channel auditory detection thresholds measured in response to focused electrical fields (i.e., focused thresholds) are sensitive to properties of ENI quality, including electrode-neuron distance, intracochlear resistance, and neural health. In the present study, focused thresholds and speech perception abilities were assessed in 15 children and 21 adult CI listeners. Focused thresholds were measured for all active electrodes using a fast sweep procedure. Speech perception performance was evaluated by assessing listeners’ ability to identify vowels presented in /h-vowel-d/ context. Consistent with prior literature, focused thresholds were lower for children than for adults, but vowel identification did not differ significantly across age groups. Higher across-array average focused thresholds, which may indicate a relatively poor ENI quality, were associated with poorer vowel identification scores in both children and adults. Adult CI listeners with longer durations of deafness had higher focused thresholds. Findings from this study demonstrate that poor-quality ENIs may contribute to reduced speech outcomes for pediatric and adult CI listeners. Estimates of ENI quality (e.g., focused thresholds) may assist in developing customized programming interventions that serve to improve the transmission of spectral cues that are important in vowel identification.

Comparison of temporal fine structure sensitivity and concurrent vowel perception between children with and without reading disability

Background: Children with reading disabilities (RD) exhibit difficulty in perceiving speech in background noise due to poor auditory stream segregation. There is a dearth of literature on measures of temporal fine structure sensitivity (TFS) and concurrent vowel perception abilities to assess auditory stream segregation in children with reading disabilities. Hence the present study compared temporal fine structure sensitivity (TFS) and concurrent vowel perception abilities between children with and without reading deficits. Method: The present research consisted of a total number of 30 participants, 15 children with reading disabilities (RD) and fifteen typically developing (TD) children within the age range of 7-14 years and were designated as Group 1 and Group 2 respectively. Both groups were matched for age, grade, and classroom curricular instructions. The groups were evaluated for TFS and concurrent vowel perception abilities and the performance was compared using independent ‘t’ test and repeated measure ANOVA respectively. Results: Results revealed that the children with RD performed significantly (p < 0.001) poorer than TD children on both TFS and concurrent vowel identification task. On concurrent vowel identification tasks, there was no significant interaction found between reading ability and F0 difference suggesting that the trend was similar in both the groups. Conclusion: The study concludes that the children with RD show poor temporal fine structure sensitivity and concurrent vowel identification scores compared to age and grade matched TD children owing to poor auditory stream segregation in children with RD.

Comparison of temporal fine structure sensitivity and concurrent vowel perception between children with and without reading disability

Background: Children with reading disabilities (RD) exhibit difficulty in perceiving speech in background noise due to poor auditory stream segregation. There is a dearth of literature on measures of temporal fine structure sensitivity (TFS) and concurrent vowel perception abilities to assess auditory stream segregation in children with reading disabilities. Hence the present study compared temporal fine structure sensitivity (TFS) and concurrent vowel perception abilities between children with and without reading deficits. Method: The present research consisted of a total number of 30 participants, 15 children with reading disabilities (RD) and fifteen typically developing (TD) children within the age range of 7-14 years and were designated as Group 1 and Group 2 respectively. Both groups were matched for age, grade, and classroom curricular instructions. The groups were evaluated for TFS and concurrent vowel perception abilities and the performance was compared using independent ‘t’ test and repeated measure ANOVA respectively. Results: Results revealed that the children with RD performed significantly (p < 0.001) poorer than TD children on both TFS and concurrent vowel identification task. On concurrent vowel identification tasks, there was no significant interaction found between reading ability and F0 difference suggesting that the trend was similar in both the groups. Conclusion: The study concludes that the children with RD show poor temporal fine structure sensitivity and concurrent vowel identification scores compared to age and grade matched TD children owing to poor auditory stream segregation in children with RD.

Comparison of temporal fine structure sensitivity and concurrent vowel perception between children with and without reading disability.

Background: Children with reading deficits (RD) exhibit difficulty in perceiving speech in background noise due to poor auditory stream segregation. There is a dearth of literature on measures of temporal fine structure sensitivity (TFS) and concurrent vowel perception abilities to assess auditory stream segregation in children with reading deficits. Hence the present study compared temporal fine structure sensitivity (TFS) and concurrent vowel perception abilities between children with and without reading deficits. Method: The present research consisted of a total number of 30 participants, 15 children with reading deficits (RD) and fifteen typically developing (TD) children within the age range of 7-14 years and were designated as Group 1 and Group 2 respectively. Both groups were matched for age, grade, and classroom curricular instructions. The groups were evaluated for TFS and concurrent vowel perception abilities and the performance was compared using independent 't' test and repeated measure ANOVA respectively. Results: Results revealed that the children with RD performed significantly (p < 0.001) poorer than TD children on both TFS and concurrent vowel identification task. On concurrent vowel identification tasks, there was no significant interaction found between reading ability and F0 difference suggesting that the trend was similar in both the groups. Conclusion: The study concludes that the children with RD show poor temporal fine structure sensitivity and concurrent vowel identification scores compared to age and grade matched TD children owing to poor auditory stream segregation in children with RD.

Modeling concurrent vowel identification for shorter durations

Behavioral studies on concurrent vowels with shorter duration (50 ms) show that overall identification scores across fundamental frequency (F0) differences are reduced when compared to longer duration (200 ms). In this current study, we investigated the effect of shorter durations on concurrent vowel scores using the temporal responses of an auditory-nerve model (Zilany et al., 2014) with a modified version of the Meddis and Hewitt (1992) F0-guided vowel segregation algorithm. The F0 representations of 50 ms vowel pairs were reduced but could not fully account for reduced identification, suggesting a limited ability to avail an F0-guided vowel segregation cue. Thus, by limiting only this cue in the segregation algorithm for 200 ms, the model was qualitatively successful in capturing lower scores across F0 differences for 50 ms. This differential ability to utilize an F0-guided segregation cue might contribute to a variable F0 benefit across normal-hearing listeners. The model also predicted an improvement in identification scores of the second vowel of the pair, when the duration was increased from 50 ms to 200 ms. These model predictions suggest that the inability to utilize an F0-guided segregation cue at shorter durations can account for reduced concurrent vowel identification scores across F0 differences.

Effects of Age and Cochlear Implantation on Spectrally Cued Speech Categorization.

Purpose Weighting of acoustic cues for perceiving place-of-articulation speech contrasts was measured to determine the separate and interactive effects of age and use of cochlear implants (CIs). It has been found that adults with normal hearing (NH) show reliance on fine-grained spectral information (e.g., formants), whereas adults with CIs show reliance on broad spectral shape (e.g., spectral tilt). In question was whether children with NH and CIs would demonstrate the same patterns as adults, or show differences based on ongoing maturation of hearing and phonetic skills. Method Children and adults with NH and with CIs categorized a /b/-/d/ speech contrast based on two orthogonal spectral cues. Among CI users, phonetic cue weights were compared to vowel identification scores and Spectral-Temporally Modulated Ripple Test thresholds. Results NH children and adults both relied relatively more on the fine-grained formant cue and less on the broad spectral tilt cue compared to participants with CIs. However, early-implanted children with CIs better utilized the formant cue compared to adult CI users. Formant cue weights correlated with CI participants' vowel recognition and in children, also related to Spectral-Temporally Modulated Ripple Test thresholds. Adults and child CI users with very poor phonetic perception showed additive use of the two cues, whereas those with better and/or more mature cue usage showed a prioritized trading relationship, akin to NH listeners. Conclusions Age group and hearing modality can influence phonetic cue-weighting patterns. Results suggest that simple nonlexical categorization tests correlate with more general speech recognition skills of children and adults with CIs.

Consonant and Vowel Identification in Cochlear Implant Users Measured by Nonsense Words: A Systematic Review and Meta-Analysis.

The purpose of this systematic review and meta-analysis was to establish a baseline of the vowel and consonant identification scores in prelingually and postlingually deaf users of multichannel cochlear implants (CIs) tested with consonant-vowel-consonant and vowel-consonant-vowel nonsense syllables. Six electronic databases were searched for peer-reviewed articles reporting consonant and vowel identification scores in CI users measured by nonsense words. Relevant studies were independently assessed and screened by 2 reviewers. Consonant and vowel identification scores were presented in forest plots and compared between studies in a meta-analysis. Forty-seven articles with 50 studies, including 647 participants, thereof 581 postlingually deaf and 66 prelingually deaf, met the inclusion criteria of this study. The mean performance on vowel identification tasks for the postlingually deaf CI users was 76.8% (N = 5), which was higher than the mean performance for the prelingually deaf CI users (67.7%; N = 1). The mean performance on consonant identification tasks for the postlingually deaf CI users was higher (58.4%; N = 44) than for the prelingually deaf CI users (46.7%; N = 6). The most common consonant confusions were found between those with same manner of articulation (/k/ as /t/, /m/ as /n/, and /p/ as /t/). The mean performance on consonant identification tasks for the prelingually and postlingually deaf CI users was found. There were no statistically significant differences between the scores for prelingually and postlingually deaf CI users. The consonants that were incorrectly identified were typically confused with other consonants with the same acoustic properties, namely, voicing, duration, nasality, and silent gaps. A univariate metaregression model, although not statistically significant, indicated that duration of implant use in postlingually deaf adults predict a substantial portion of their consonant identification ability. As there is no ceiling effect, a nonsense syllable identification test may be a useful addition to the standard test battery in audiology clinics when assessing the speech perception of CI users.

Age-Related Performance on Vowel Identification and the Spectral-temporally Modulated Ripple Test in Children With Normal Hearing and With Cochlear Implants.

Children’s performance on psychoacoustic tasks improves with age, but inadequate auditory input may delay this maturation. Cochlear implant (CI) users receive a degraded auditory signal with reduced frequency resolution compared with normal, acoustic hearing; thus, immature auditory abilities may contribute to the variation among pediatric CI users’ speech recognition scores. This study investigated relationships between age-related variables, spectral resolution, and vowel identification scores in prelingually deafened, early-implanted children with CIs compared with normal hearing (NH) children. All participants performed vowel identification and the Spectral-temporally Modulated Ripple Test (SMRT). Vowel stimuli for NH children were vocoded to simulate the reduced spectral resolution of CI hearing. Age positively predicted NH children’s vocoded vowel identification scores, but time with the CI was a stronger predictor of vowel recognition and SMRT performance of children with CIs. For both groups, SMRT thresholds were related to vowel identification performance, analogous to previous findings in adults. Sequential information analysis of vowel feature perception indicated greater transmission of duration-related information compared with formant features in both groups of children. In addition, the amount of F2 information transmitted predicted SMRT thresholds in children with NH and with CIs. Comparisons between the two CIs of bilaterally implanted children revealed disparate task performance levels and information transmission values within the same child. These findings indicate that adequate auditory experience contributes to auditory perceptual abilities of pediatric CI users. Further, factors related to individual CIs may be more relevant to psychoacoustic task performance than are the overall capabilities of the child.

Vowel identification by cochlear implant users: Contributions of duration cues and dynamic spectral cues.

A recent study from our laboratory assessed vowel identification in cochlear implant (CI) users, using full /dVd/ syllables and partial (center- and edges-only) syllables with duration cues neutralized [Donaldson, Rogers, Cardenas, Russell, and Hanna (2013). J. Acoust. Soc. Am. 134, 3021-3028]. CI users' poorer performance for partial syllables as compared to full syllables, and for edges-only syllables as compared to center-only syllables, led to the hypotheses (1) that CI users may rely strongly on vowel duration cues; and (2) that CI users have more limited access to dynamic spectral cues than steady-state spectral cues. The present study tested those hypotheses. Ten CI users and ten young normal hearing (YNH) listeners heard full /dVd/ syllables and modified (center- and edges-only) syllables in which vowel duration cues were either preserved or eliminated. The presence of duration cues significantly improved vowel identification scores in four CI users, suggesting a strong reliance on duration cues. Duration effects were absent for the other CI users and the YNH listeners. On average, CI users and YNH listeners demonstrated similar performance for center-only stimuli and edges-only stimuli having the same total duration of vowel information. However, three CI users demonstrated significantly poorer performance for the edges-only stimuli, indicating apparent deficits of dynamic spectral processing.

Relative contributions of vowels and consonants in recognizing isolated Mandarin words

This study investigated the relative contributions of vowels and consonants in recognizing isolated Mandarin words. Normal-hearing native-Mandarin listeners were instructed to recognize isolated Mandarin words and identify consonants, vowels and tones with stimuli synthesized to contain different proportions of consonant or vowel segments, including five conditions of consonant-only, vowel-only, consonant or vowel plus consonant–vowel (C–V) transition, and C–V transition. The recognition score of the vowel-only Mandarin words was significantly higher than that of the consonant-only words; and word recognition scores had a higher correlation with vowel identification scores than consonant identification scores. Moreover, adding a small portion of C–V transition significantly improved the recognition score of the consonant-only Mandarin words. In the conditions of C–V transition and consonant plus C–V transition, the duration of preserved portion predicted modestly well the scores of isolated word recognition, and vowel, consonant and lexical tone identification in Mandarin. These findings suggest that there is a greater contribution of vowels than consonants to isolated word recognition in Mandarin, which is different from previous outcomes in English.

Effect of spectral amplitude elevation for vowel identification in simulated combined electric-and-acoustic hearing

Though the benefits of combined electric-and-acoustic stimulation (EAS) have been widely reported, its underlying mechanisms responsible are not well understood yet. This study assessed the effect of balanced spectral amplitudes between acoustic and electric portions in EAS to vowel identification via two experiments. Four vowels (i.e., /iy/, /eh/, /oo/ and /ah/) were synthesized using pure-tone harmonic components. In Experiment 1, the spectral amplitudes of natural vowels were elevated for either acoustic (below 600 Hz) or electric portions (above 600 Hz) from 5 dB to 30 dB. Vowel identification scores were collected from eight normal-hearing listeners. In Experiment 2, the synthesized vowel stimuli were processed by the signal processing condition simulating the electric-and-acoustic stimulation. The results of identifying synthesized vowels and synthesized EAS-processed vowels in the two experiments showed the similar patterns of declined identification rates as a function of (acoustic or electric) sp...

Vowels and intelligibility in dysarthric speech

Diane Kewley-Port’s work in vowel perception under challenging listening conditions and in the relation between vowel perception and production in second language learners has important implications for disordered speech. Vowel space area has been widely used as an index of articulatory working space in speakers with hypokinetic dysarthria related to Parkinson disease (PD), with the assumption that a larger vowel space is associated with higher speech intelligibility. Although many studies have reported acoustic measures of vowels in Parkinson disease, vowel identification and transcription tasks designed to relate changes in production with changes in perception are rarely performed. This study explores the effect of changes in vowel production by six talkers with PD speaking at habitual and loud levels of effort on listener perception. The relation among vowel acoustic measures (including vowel space area and measures of temporal and spectral distinctiveness), vowel identification scores, speech intelligibility ratings, and sentence transcription accuracy for speakers with dysarthria will be discussed.

Computational model predictions of cues for concurrent vowel identification.

Although differences in fundamental frequencies (F0s) between vowels are beneficial for their segregation and identification, listeners can still segregate and identify simultaneous vowels that have identical F0s, suggesting that additional cues are contributing, including formant frequency differences. The current perception and computational modeling study was designed to assess the contribution of F0 and formant difference cues for concurrent vowel identification. Younger adults with normal hearing listened to concurrent vowels over a wide range of levels (25-85 dB SPL) for conditions in which F0 was the same or different between vowel pairs. Vowel identification scores were poorer at the lowest and highest levels for each F0 condition, and F0 benefit was reduced at the lowest level as compared to higher levels. To understand the neural correlates underlying level-dependent changes in vowel identification, a computational auditory-nerve model was used to estimate formant and F0 difference cues under the same listening conditions. Template contrast and average localized synchronized rate predicted level-dependent changes in the strength of phase locking to F0s and formants of concurrent vowels, respectively. At lower levels, poorer F0 benefit may be attributed to poorer phase locking to both F0s, which resulted from lower firing rates of auditory-nerve fibers. At higher levels, poorer identification scores may relate to poorer phase locking to the second formant, due to synchrony capture by lower formants. These findings suggest that concurrent vowel identification may be partly influenced by level-dependent changes in phase locking of auditory-nerve fibers to F0s and formants of both vowels.

Role of slow temporal modulations in speech identification for cochlear implant users

Objective: This study aimed to assess whether the capacity of cochlear implant (CI) users to identify speech is determined by their capacity to perceive slow (< 20 Hz) temporal modulations. Design: This was achieved by studying the correlation between (1) phoneme identification in quiet and in a steady-state or fluctuating (8 Hz) noises, and (2) amplitude-modulation detection thresholds (MDTs) at 8 Hz (i.e. slow temporal modulations). Study sample: Twenty-one CI users, unilaterally implanted with the same device, were tested in free field with their everyday clinical processor. Results: Extensive variability across subjects was observed for both phoneme identification and MDTs. Vowel and consonant identification scores in quiet were significantly correlated with MDTs at 8 Hz (r = − 0.47 for consonants, r = − 0.44 for vowels; p < 0.05). When the masker was a steady-state noise, only consonant identification scores tended to correlate with MDTs at 8 Hz (r = − 0.4; p = 0.07). When the masker was a fluctuating noise, consonant and vowel identification scores were not significantly correlated with MDTs at 8 Hz. Conclusions: Sensitivity to slow amplitude modulations is correlated with vowel and consonant perception in CI users. However, reduced sensitivity to slow modulations does not entirely explain the limited capacity of CI recipients to understand speech in noise.

Multichannel compression hearing aids: Effect of channel bandwidth on consonant and vowel identification by hearing-impaired listeners

Aided consonant and vowel identification was measured in 13 listeners with high-frequency sloping hearing losses. To investigate the influence of compression-channel analysis bandwidth on identification performance independent of the number of channels, performance was compared for three 17-channel compression systems that differed only in terms of their channel bandwidths. One compressor had narrow channels, one had widely overlapping channels, and the third had level-dependent channels. Measurements were done in quiet, in speech-shaped noise, and in a three-talker background. The results showed no effect of channel bandwidth, neither on consonant nor on vowel identification scores. This suggests that channel bandwidth per se has little influence on speech intelligibility when individually prescribed, frequency-varying compressive gain is provided.

Spectral modulation transfer function and vowel identification.

Psychophysical and physiological studies have demonstrated selectivity for spectral modulation frequency in the auditory system, suggesting that auditory perception of complex spectra might be based on spectral modulation channels. However, an early study [Liu and Eddins (2008)] reported that vowel identification was significantly reduced by filtering in spectral modulation frequency domain, although there was no significant correlation between spectral modulation detection thresholds and vowel identification. The latter result may have been due to the small number of listeners (n = 5) in the previous study. Using a larger number of listeners, the present study measured spectral modulation transfer functions and vowel identification performance with and without spectral modulation filtering in 11 normal‐hearing listeners. Vowel identification performance was significantly correlated with spectral modulation detection thresholds at 0.5 cyc/oct (i.e., higher spectral modulation detection thresholds associated with poorer vowel identification scores). In addition, average spectral modulation detection thresholds were significantly correlated with the change in vowel identification associated with filtering in the spectral modulation domain. Both of these results highlight the relation between the perception of sinusoidal spectral modulation and vowel identification.

Relationship between perception of spectral ripple and speech recognition in cochlear implant and vocoder listeners

Spectral resolution has been reported to be closely related to vowel and consonant recognition in cochlear implant (CI) listeners. One measure of spectral resolution is spectral modulation threshold (SMT), which is defined as the smallest detectable spectral contrast in the spectral ripple stimulus. SMT may be determined by the activation pattern associated with electrical stimulation. In the present study, broad activation patterns were simulated using a multi-band vocoder to determine if similar impairments in speech understanding scores could be produced in normal-hearing listeners. Tokens were first decomposed into 15 logarithmically spaced bands and then re-synthesized by multiplying the envelope of each band by matched filtered noise. Various amounts of current spread were simulated by adjusting the drop-off of the noise spectrum away from the peak (40-5 dBoctave). The average SMT (0.25 and 0.5 cyclesoctave) increased from 6.3 to 22.5 dB, while average vowel identification scores dropped from 86% to 19% and consonant identification scores dropped from 93% to 59%. In each condition, the impairments in speech understanding were generally similar to those found in CI listeners with similar SMTs, suggesting that variability in spread of neural activation largely accounts for the variability in speech perception of CI listeners.

Coding of vowellike signals in cochlear implant listeners.

Neural-population interactions resulting from excitation overlap in multi-channel cochlear implants (CI) may cause blurring of the "internal" auditory representation of complex sounds such as vowels. In experiment I, confusion matrices for eight German steady-state vowellike signals were obtained from seven CI listeners. Identification performance ranged between 42% and 74% correct. On the basis of an information transmission analysis across all vowels, pairs of most and least frequently confused vowels were selected for each subject. In experiment II, vowel masking patterns (VMPs) were obtained using the previously selected vowels as maskers. The VMPs were found to resemble the "electrical" vowel spectra to a large extent, indicating a relatively weak effect of neural-population interactions. Correlation between vowel identification data and VMP spectral similarity, measured by means of several spectral distance metrics, showed that the CI listeners identified the vowels based on differences in the between-peak spectral information as well as the location of spectral peaks. The effect of nonlinear amplitude mapping of acoustic into "electrical" vowels, as performed in the implant processors, was evaluated separately and compared to the effect of neural-population interactions. Amplitude mapping was found to cause more blurring than neural-population interactions. Subjects exhibiting strong blurring effects yielded lower overall vowel identification scores.

Effect of pitch fluctuation of vowels on vowel perception under a noisy environment

Previous paper reported that natural vowel identification scores were higher under harmonic complex tones than under a white noise, in spite of the tone and the noise having the same flat spectral envelope [K. Ishizuka and K. Aikawa, J. Acoust. Soc. Am. 109, 2439 (2001)]. Natural vowels have pitch fluctuation. This paper reports a new finding that the pitch fluctuation has the effect of improving vowel identification under noisy environments. An experiment was conducted to examine the effect of pitch fluctuation of vowels on vowel perception under a harmonic complex tone at SNRs ranging from −10 to −15 dB. The Japanese female vowels were synthesized with the pitches fluctuated according to the 1/f characteristics [N. Aoki and T. Ifukube, J. Acoust. Soc. Am. 106, 423–433 (1999)]. Five pitch fluctuation bandwidth conditions for vowels were tested. The bandwidths ranged from 0 to 4.3% (F0 = 255.3 Hz). The harmonic complex tone was composed of the same F0 of the vowels and 21 components in the cosine phase. The subjects were forced to choose the vowel they heard for each stimulus. The mean vowel identification scores increased from 33.2% to 46.5%, proportional to the increase of the pitch fluctuation bandwidth.