Hearing-impaired Speakers Research Articles

Acoustics of long vowels in Arabic-speaking children with hearing impairments

The present study investigates formant frequencies and vowel space area in Jordanian hearing-impaired individuals who received cochlear implants. To achieve this task, they were compared with their hearing aids and normal-hearing peers. The study involved 60 participants aged between 7 and 11 years, consisting of 20 cochlear implant (CI) users, 20 hearing aid (HA) users, and 20 normal-hearing (NH) children. All participants were native speakers of Jordanian Arabic. The study measured the fundamental frequency (F0), the first, second, and third formant frequencies (F1–F3) of the five long vowels: /i:/, /e:/, /a:/, /o:/, and /u:/ using Praat software (version 6.1.05). One-way repeated measure (ANOVA) and post hoc Tukey tests were used to compare the differences between the three groups. The results revealed significant differences between the three groups in the degree of vowel height and backness for all the long vowels. The vowel space in HA users was found to be smaller than in CI users. HA users had difficulty varying their tongue position in both the high–low and the front–back dimensions, while CI users had difficulty varying their tongue position only in the front–back dimension. The study also examined errors in long vowels and diphthongs. The findings indicated that after cochlear implantation, the formants of the vowels produced by hearing-impaired speakers shifted closer to those produced by their NH peers, and CI users had a greater vowel space compared to HA users. This research could potentially contribute to the development of effective interventions for hearing-impaired individuals and could provide insights for the design of speech therapy programs that aim to improve vowel production abilities.

Robust Hearing-Impaired Speaker Recognition from Speech using Deep Learning Networks in Native Language

Several research works in speaker recognition have grown recently due to its tremendous applications in security, criminal investigations and in other major fields. Identification of a speaker is represented by the way they speak, and not on the spoken words. Hence the identification of hearing-impaired speakers from their speech is a challenging task since their speech is highly distorted. In this paper, a new task has been introduced in recognizing Hearing Impaired (HI) speakers using speech as a biometric in native language Tamil. Though their speech is very hard to get recognized even by their parents and teachers, our proposed system accurately identifies them by adapting enhancement of their speeches. Due to the huge variety in their utterances, instead of applying the spectrogram of raw speech, Mel Frequency Cepstral Coefficient features are derived from speech and it is applied as spectrogram to Convolutional Neural Network (CNN), which is not necessary for ordinary speakers. In the proposed system of recognizing HI speakers, is used as a modelling technique to assess the performance of the system and this deep learning network provides 80% accuracy and the system is less complex. Auto Associative Neural Network (AANN) is used as a modelling technique and performance of AANN is only 9% accurate and it is found that CNN performs better than AANN for recognizing HI speakers. Hence this system is very much useful for the biometric system and other security related applications for hearing impaired speakers.

Hearing impaired speech recognition: Stockwell features and models

The development of speech recognition system for recognising the speeches of a reasonable person in various languages usually is in fashion and does not involve challenges to be faced by the researchers. For the past ten years, advances are taking place in analysing and recognising the speeches of the hearing impaired because of the deployment of sophisticated processing methods to study the characteristics of speech production. These technological advancements not only pave the way for developing algorithms for recognising the speeches of healthy persons, also recognising the utterances of the hearing impaired. The study on analysing the oral communication skills of hearing-impaired children has received the attention of the researchers, speech pathologists and audiologists to develop assistive tool/system because inadequacy of such skills dramatically affects the social, educational and career opportunities available to them at large. This paper mainly emphasises the need for the development of a more challenging speaker independent speech recognition system for hearing impaired so that the system can respond to the speech uttered by any HI. In this work, Modified Group Delay Features and Stockwell transform cepstral features are used at the front end and vector quantisation (VQ) and Multivariate Hidden Markov Models (MHMM) at the back end for recognising the speeches uttered by any person with hearing disability. Performance of the system is compared for the three modelling techniques VQ, Fuzzy C Means (FCM) clustering and MHMM for the recognition of isolated digits in Tamil. Recognition accuracy is 89.25% and 79.5% for speaker dependent and independent speech recognition system for the hearing impaired. Performance of the system reveals that this system may be deployed to understand the speeches uttered by any hearing impaired speaker, improve the social status of the people with hearing impairment and mitigate the social stigma in leading a normal life.

A strategic approach to recognize the speech of the children with hearing impairment: different sets of features and models

The automatic speech recognition system is developed and tested for recognizing the speeches of a normal person in various languages. This paper mainly emphasizes the need for the development of a more challenging speaker independent speech recognition system for hearing impaired to recognize the speeches uttered by any Hearing Impaired (HI) speaker. In this work, Gamma tone energy features with filters spaced an equivalent rectangular bandwidth (ERB), MEL & BARK scale, and MFPLPC features are used at the front end and vector quantization (VQ) & multivariate hidden Markov models (MHMM) at the back end for recognizing the speeches uttered by any hearing impaired speaker. Performance of the system is compared for the three modeling techniques VQ, FCM (Fuzzy C means) clustering and MHMM for the recognition of isolated digits and simple continuous sentences in Tamil. Recognition accuracy (RA) is 81.5% with speeches of eight speakers considered for training and speeches of the remaining two speakers considered for testing for speaker independent isolated digit recognition system. Accuracy is found to be 91% and 87.5% for considering 90% of the data for training and 10% for testing for speaker independent isolated digit and continuous speech recognition systems respectively. Accuracy can be further enhanced by having an extensive database for creating models/templates. Receiver operating characteristics (ROC) drawn between True Positive Rate and False Positive Rate is used to assess the performance of the system for HI. This system can be utilized to understand the speech uttered by any hearing impaired speaker and the system facilitates the provision of necessary assistance to them. It ultimately improves the social status of the hearing impaired people and their confidence level will be enhanced.

Use of formant centralization ratio for vowel impairment detection in normal hearing and different degrees of hearing impairment

Purpose: Hearing-impaired (HI) speakers show changes in vowel production and formant frequencies, as well as more cases of overlapping between vowels and more restricted formant space, than hearing speakers. This study was intended to explore whether the use of different acoustic parameters (Formant Centralization Ratio (FCR), Vowel Space Area (VSA), F2i/F2u ratio (second formant of/i,u/)) was suitable or not for characterizing impairments in the articulation of vowels in the speech of HL speakers. In fact, correlated acoustic parameters are used to determine the limits of tongue movements in vowel production in different severity degrees of hearing impairment.Methods: Speech recordings of 40 speakers with HL and 40 healthy controls were acoustically analyzed. The vowels (/a/,/i/,/u/) were extracted from the word context and, then, the first and second formants were calculated. The same vowel-formant elements were used to construct the FCR, expressed as (F2u + F2a + F1i + F1u)/(F2i + F1a), the F2i/F2u ratio, and the vowel space area (VSA), expressed as ABS((F1i*(F2a–F2u)+F1a*(F2u–F2i)+F1u*(F2i–F2a))/2).Results: The FCR differentiated HL groups from the control group and the discrimination was not gender-sensitive. All parameters were found to be strongly correlated with each other.Conclusions: The findings of this study showed that FCR was a more sensitive acoustic parameter than F2i/F2u ratio and VSA to distinguish speech of the HL groups from that of the normal group. Thus, FCR is considered to be applicable as an early objective measure of impaired vowel articulation in HL speakers.

Nasalance scores of Egyptian hearing impaired children

Context Speech of prelingual hearing impaired (HI) speakers is commonly characterized by the presence of resonance abnormalities which is related with nasality. Aim To investigate and compare nasalance scores of Egyptian HI children including cochlear implant (CI) and hearing aid (HA) users with normal hearing (NH) children. Settings and design It is an observational case–control study. Patients and methods The study consisted of group I: 41 HA children, group II: 24 CI children, and the control group: 31 age-matched and sex-matched children with NH. Nasometer II 6200 was used for evaluation of their nasalance scores. Statistical analysis used IBM SPSS, version 20, was used. Kruskal–Wallis and one-way analysis of variance tests were performed to compare the mean differences between the nasalance scores of the three studied groups. While Mann–Whitney and post-hoc tests were used for pairwise comparison. A correlation analysis was computed between the nasalance scores and many parameters using Spearman's and Pearson's coefficients. Results For oral sentence, both CI and HA children showed higher nasalance values in comparison with NH children. However, lower nasalance scores were observed for the nasal sentence. CI experience and enrollment into language therapy were found to have an effect on nasalance scores. Conclusion Resonance quality of the HI children is still at risk despite the fact that a significant number of them demonstrate normal resonance.

The special nature of Australian phonologies: Why auditory constraints on the sound systems of human languages are not universal

The phonological systems of human languages are constrained by what are often assumed to be universal properties of human auditory perception. However, the atypical phonologies found in many hearing-impaired speakers indicate that such constraints also operate at an individual level. The phonology of a child with chronic otitis media, for example, may lack a voicing distinction or sometimes have no fricatives. So, if a large group of speakers in a speech community operates with an atypical auditory system over many generations, then the phonology of the language(s) spoken by such a community might also over time be influenced by the particular properties of that common auditory system. Over half of the Australian Aboriginal population develop chronic otitis media with effusion in infancy and 50–70% of Aboriginal children have a significant hearing loss at both ends of the frequency range. Most Australian languages have phonologies which are atypical in world terms, having no voicing distinction and no fricatives or affricates, but an unusually large number of places of articulation. Acoustically, the sound systems of Australian languages appear to be a very good match for the hearing profiles of large numbers of their speakers. This paper reviews the evidence for a connection.

Recognition of emotional prosody in Mandarin: Evidence from a synthetic speech paradigm

Pitch-dominant information is reduced in the spectrally-impoverished signal transmitted by cochlear implants (CIs), leading to potential difficulties in perceiving voice emotion. However, this evidence comes from non-tonal languages such as English, in which pitch information is not required for lexical meaning. In order to better understand how hearing impaired (HI) speakers of a tone language with cochlear implants (CIs) process emotional prosody, an experiment was conducted with healthy normal hearing (NH) Mandarin-speaking adults listening to synthetic stimuli designed to resemble CI input. Listeners heard short sentences from a read-speech database produced by professional actors. Stimuli were selected to express four emotions (“angry,” “happy,” “sad,” and “neutral”), under four conditions which varied the lexical tones of Mandarin. Listeners heard natural speech and three noise-vocoded speech conditions (4-, 8-, and 16-spectral channels) and made a four-alternative, forced-choice decision about the basic emotion underlying each sentence. Preliminary results indicate more accurate emotional prosody recognition for natural speech than for synthesized speech, with greater accuracy for higher channel stimuli than lower channel stimuli. The findings also suggest NH Mandarin-speaking listeners show lower overall vocal emotional prosody accuracy compared with previous studies of non-tonal languages (e.g., English).

The Effect of Cochlear Implantation on Nasalance of Speech in Postlingually Hearing-Impaired Adults

Hypernasality is considered a prevalent speech abnormality that could significantly contribute to the unintelligibility of the hearing-impaired speakers. The aim of this study was to evaluate the effect of cochlear implantation and the duration of hearing loss on nasalance of speech of a postlingually impaired group of Saudi adult patients. Retrospective study. This study included 25 postlingually hearing-impaired patients who underwent cochlear implantation and 25 age-matched control subjects. Patients were divided into three groups according to the duration of hearing loss. The nasometric data of the hearing-impaired group were compared with the control group. Also, the preoperative values were compared with the postoperative values 6, 12, and 24 months after surgery. Significant differences were demonstrated between the preimplantation nasalance scores of the three subgroups and between the patients and control groups. There were statistically significant differences demonstrated between the pre- and the postimplantation nasalance values for the three groups of patients. Cochlear implantation appears to have significant effects on improving the nasalance of the speech of postlingually hearing-impaired adult patients. However, the degree of improvement might vary according to the duration of hearing loss the patients had preimplantation.

Vowel Formant Values in Hearing and Hearing-Impaired Children: A Discriminant Analysis

Hearing-impaired speakers show changes in vowel production and formant pitch and variability, as well as more cases of overlapping between vowels and more restricted formant space, than hearing speakers; consequently their speech is less intelligible. The purposes of this paper were to determine the differences in vowel formant values between 32 hearing speakers, 14 severely hearing-impaired speakers, and 25 profoundly hearing-impaired speakers, and to investigate the influence of perceptual constraints on the contrastiveness of spoken vowels in speakers with hearing loss, as these underline the importance of good phonation, articulation, and resonance in speech production. Several differences in formant values were confirmed with Anova-Welch tests, except for the F1 of open /e/ and /a/, the most open and loud vowels in Slovene. In addition, discriminant analysis showed real differences in vowel production (97.1% classification success), thus making it possible to differentiate between groups of hearing and hearing-impaired speakers based on this one function of vowel production. The hypothesis was confirmed that vowel production in hearing-impaired individuals is different from that in hearing individuals.The results suggest that speech rehabilitation should consider specific tasks and training for vowel production, including neuromuscular oral control and self-monitoring, to improve speech intelligibility. The results are also useful for those who are in contact with people with hearing impairment, such as teachers, audiologists, speech and language pathologists, nurses, and researchers in speech and hearing sciences.

Observations on Hearing Aid Users' Strategies for Controlling the Level of Their Own Voice

Evidence suggests that hearing-aid users have difficulties with own-voice level control, most likely because their auditory feedback is affected by hearing-aid amplification. The purpose of this study was to investigate how changes to auditory feedback affect the voice level of hearing-aid users. A correlational study was set up to investigate the relation between voice level and hearing-aid amplification. Seven hearing-impaired speakers participated. All were experienced hearing-aid users. The speakers projected their voice to a passive listener across different speaker-listener distances and with different prescriptions of gain in an experimental hearing aid. For each combination of conditions, produced voice level and self-perceived voice level was measured. These data were subjected to an analysis of variance assuming a mixture of random and fixed effects. In addition, all speakers took part in interviews. Three speakers reacted to the changes in auditory feedback in agreement with previous experiments with normal-hearing speakers: they compensated by changing produced voice level. In contrast, the voice levels in the other four speakers were largely unaffected by the changes to auditory feedback. A secondary observation was that while all speakers increased their voice level with distance, the two subgroups produced different growth rates of vocal level versus distance. It is hypothesized that the speakers in the former subgroup relied on auditory feedback for solving the experimental task, whereas the latter subgroup had developed an own-voice level-control strategy based on proprioceptory feedback, possibly because they have lost faith in their auditory feedback mechanism, which indeed is changed by both hearing loss and hearing-aid amplification. Comparison to "target" voice levels suggests that proprioceptory feedback is less effective than auditory feedback for achieving adequate level-distance growth rate.

‘Non‐vocalization’: A phonological error process in the speech of severely and profoundly hearing impaired adults, from the point of view of the theory of phonology as human behaviour

‘Non‐vocalization’ (N‐V) is a newly described phonological error process in hearing impaired speakers. In N‐V the hearing impaired person actually articulates the phoneme but without producing a voice. The result is an error process looking as if it is produced but sounding as if it is omitted. N‐V was discovered by video recording the speech of two groups, profoundly and severely hearing impaired adults in four elicitation tasks of varying difficulty, and analysing 2065 phonological error processes (substitutions, omissions, and N‐V) according to 24 criteria resulting in 49,560 data points. Results, which are discussed in view of the theory ‘Phonology as Human Behaviour’ (PHB), indicate that: (a) The more communicative the error process was; the more effort was made for its production and the more frequent its distribution; (b) The easier the elicitation task was, the more frequent the use of communicative error processes; c) The more difficult the elicitation task was, the more frequent the use of the relatively less communicative and easier to produce error processes; and d) The process of N‐V functioned like a communicative error process for the group of profoundly hearing impaired adults.

Motor control of speaking rate and oral diadochokinesis in hearing-impaired Farsi speakers

Although speech motor control has been studied intensively in normal-hearing and hearing-impaired (HI) speakers in America and Europe, essentially no research has been performed using Persian-speaking participants. A total of 46 prelingual hearing-impaired 15–18-year-old males and 15 normally hearing control participants from Iran participated in the study. Three speaking performance measures, oral diadochokinesis (DDK), speaking rate (words per minute), and intelligibility ratings, were obtained for the two groups and compared to previously published research for English-speaking participants. The DDK results in general showed that the normal-hearing group produced the fastest syllable rates, and the profoundly HI group produced the slowest. Similar results were obtained for speaking rates. Speech intelligibility was highest in the normal-hearing group and lowest in the profoundly HI group. Correlation analysis between DDK and speaking rates showed that for HI group only, a slow speaking rate corresponded to slow DDK rates. It is shown that generally there are significant differences in measures of speech motor control in normal-hearing and hearing-impaired participants. These results concord with those from other language groups.

An acoustic and perceptual study of initial stops produced by profoundly hearing impaired adolescents

This study investigated the role of Voice Onset Time (VOT) as perceptual cue to the aspiration contrast of Cantonese initial stops produced by adolescent profoundly hearing impaired speakers. Speakers with normal hearing signalled the aspiration contrast through VOT differences. Hearing impaired speakers produced initial stops with no significant VOT differences between aspirated and unaspirated stops; the accuracy of perception were above chance level for unaspirated stops but were below chance level for aspirated stops. Discriminant analysis conducted on data from a perceptual task showed that listeners with normal hearing relied on VOT cue for aspiration contrast of initial stops produced by control speakers. By contrast, stops produced by the hearing impaired speakers were not distinguished by listeners on the basis of VOT in that listeners were able to rely on VOT as a perceptual cue only to unaspirated stops, but not to aspirated stops.

Acoustic and Perceptual Study of Cantonese Tones Produced by Profoundly Hearing-Impaired Adolescents

This study investigated five acoustic parameters of fundamental frequency (f0), one related to average f0 level and four related to change of f0 (f0 change) over four consecutive quarters of the vocalic segment, for their roles as acoustic correlates to the perception of Cantonese lexical tones produced by profoundly hearing-impaired speakers. Speakers were 20 Cantonese adolescents, 10 of whom had normal-hearing (mean age = 13;05); the other 10 were profoundly hearing-impaired (mean age = 13;05). The hearing-impaired speakers were selected on the basis of the criterion of being prelingually deaf; they showed pure-tone average thresholds of 90 dB HL or more in the better ear at 0.5, 1.0, and 2.0 kHz. Twenty-four consonant-vowel Cantonese words made up four sets of words that minimally contrasted in the six contrastive Cantonese tones. The words were read by the speakers once in random order. Listeners were 10 final-year speech therapy students who were asked to identify the Cantonese tones. The f0 of each syllable (tone) was measured at five consecutive positions of the vocalic segment (initial, 25%, 50%, 75%, final). Discriminant analysis was used to determine how average f0 and f0 change accounted for both the production of intended tones and the perception of correctly identified tones. Average f0 and f0 change over the second half of the vocalic segment were reliable acoustic and perceptual correlates of tones produced by control speakers. By contrast, for hearing-impaired speakers, production of intended tones were not reliably distinguished by average f0 or f0 change. Furthermore, listeners mainly relied on f0 change over the second half of the vocalic segment for the distinction between tone 25 and level tones (55, 33, and 22) produced by hearing-impaired speakers. Average f0 and f0 change over the second half of the vocalic segment are important cues for accurate identification of Cantonese tones produced by speakers with normal-hearing. By contrast, these cues are less reliable for perceiving tones produced by profoundly hearing-impaired speakers.

An Acoustic and Perceptual Study of Final Stops Produced by Profoundly Hearing Impaired Adolescents

This study investigated formant frequencies for their role as acoustic and perceptual correlates to the place of articulation of Cantonese final stops produced by profoundly hearing impaired speakers. Speakers were 10 Cantonese adolescents (mean age=13;5 [years;months]) who were profoundly hearing impaired (HI). Control speakers were 10 adolescents (mean age=13;5) with normal hearing. Stimuli were Cantonese words that were minimally contrastive in place of final stops (/p, t, k/). Listeners were 10 final-year speech therapy students. The frequencies of F1, F2, and F3 were measured at the middle, 40 ms before the end, and at the end of the vocalic segments. Control speakers distinguished place contrasts through formant frequency differences at the end positions of the vowels. HI speakers produced final stops with missing formant transitions and neutralized vowels preceding final stops. Listeners relied on F2 transition cues for stops produced by control speakers, whereas F1 and F3 transition cues were used for stops produced by HI speakers. Formant frequencies of final stops produced by HI speakers showed reduced place distinction. When listeners identified the place of final stops produced by HI speakers, they relied on formant frequency cues that were different from those used for stops produced by control speakers.

Articulatory and acoustic measurements of vowels in hearing impaired speakers following treatment

The purpose of this study was to examine the relationships between ultrasound tongue shapes, electropalatography (EPG) contacts, formant data, and the perceptual data in the vowels of three hard of hearing adolescents. This pilot study examines the pre- and post-therapy speech of the participants in a 6 week group therapy programme with ultrasound and EPG technologies. Before treatment, participants showed particular difficulty with high vowels and the tense-lax distinction. Recordings were made of 2 males and 1 female with ultrasound and EPG. Three types of measurements were made; formant measurements, EPG tongue palate contacts, and perceptual judgements by experienced speech-language pathology researchers. Initial analysis values showed a change in the direction of EPG contact patterns, and perceptual categories. F1 values decreased and F2 values increased, moving in the direction of typical formant frequencies found in hearing people. Preliminary results of this study support a lingual-articulatory approach to treatment.

The effects of simultaneous communication on production and perception of speech.

This article reviews experiments completed over the past decade at the National Technical Institute for the Deaf and the State University of New York at Geneseo concerning speech produced during simultaneous communication (SC) and synthesizes the empirical evidence concerning the acoustical and perceptual characteristics of speech in SC. Comparisons are drawn between SC and other modes of rate-altered speech that have been used successfully to enhance communication effectiveness. Of particular importance are conclusions regarding the appropriateness of speech produced during SC for communication between hearing and hearing-impaired speakers and listeners and the appropriateness of SC use by parents and teachers for speech development of children with hearing impairment. This program of systematic basic research adds value to the discussion about the use of SC by focusing on the specific implications of empirical results regarding speech production and perception.

Intrinsic fundamental frequency effects in hearing impaired speakers

The source of the well-known intrinsic fundamental frequency (IF0) effect of vowel height has been controversial for decades. Previous work has found the average IF0 effect cross-linguistically to be 15.3 Hz [Whalen and Levitt, J. Phonetics (1995)]. The present study investigates IF0 for four hearing-impaired speakers. Based on previous observations that profoundly hearing impaired speakers vary voice pitch less than normal hearing speakers [Osberger and McGarr, Speech Lang. (1982)], our participants were expected to show a reduced IF0 effect. However, results show an average IF0 effect of 22 Hz, with a markedly wide range across speakers, from −4 to 48 Hz, with three of the four participants showing an above average-sized effect. Further, results of measures taken following speech articulation intervention using visual feedback [Bernhardt et al., Clin. Linguist Phonet. (2003)] show a decrease in IF0 for the speakers with an over-sized effect, and an increase in the speaker with an under-sized effect, despite that neither IF0 nor pitch in general were included in treatment. Results of this study support a lingual-articulatory origin for IF0, as well as suggesting that normal-hearing speakers may use auditory feedback to mediate what would otherwise be a larger effect. [Research supported by NSERC.]

Development of an Argentine Spanish hearing impaired speakers database

Argentine Spanish hearing impaired speech has not been systematically registered or studied. A database with standardized methods for data collection and labeling is presented here. Speech of 600 speakers, adults, and children, with different types of hearing losses is recorded at 44100 Hz and 16 bits. The corpus consists of word lists, grouped by tonality and frequency of use, and five short sentences, containing all the phonemes and allophones. Semi-spontaneous speech, consisting of a description of two sets of simple pictures, is also recorded. Data are acoustically analyzed and labeled at the segmental—graphemic, phonetic and acoustic tiers—and suprasegmental levels, using a labeling method that incorporates psychoacoustic information into a combination of categorical and continuous descriptions of fundamental frequency. Syntactic labels are added to semi-spontaneous speech data. Labeled files are loaded into a relational SQL database to allow different types of queries. A complementary database provides information about personal and medical data, and results of hearing tests. These databases contribute to the description of acoustic-phonetic features of deaf speech, a comparison between speech perception and production, and new approaches for the adjustment of auditory prostheses. [Work supported by FONCYT, CONICET and Mutualidad Argentina de HipoacB̄4usicos.]