Speech Intelligibility Enhancement Research Articles

Influences of noise reduction on speech intelligibility, listening effort, and sound quality among adults with severe to profound hearing loss.

Noise reduction (NR) algorithms have been integrated into modern digital hearing aids to reduce noise annoyance and enhance speech intelligibility. This study aimed to evaluate the influences of a novel hearing aid NR algorithm on individuals with severe-to-profound hearing loss. Twenty-five participants with severe-to-profound bilateral sensorineural hearing loss underwent three tests (speech intelligibility, listening effort, and subjective sound quality in noise) to investigate the influences of NR. All three tests were performed under three NR strength levels (Off, Moderate, and Strong) for both speech in noise program (SpiN) and speech in loud noise program (SpiLN), comprising six different hearing aid conditions. NR activation significantly reduced listening effort. Subjective sound quality assessments also exhibited benefits of activated NR in terms of noise suppression, listening comfort, satisfaction, and speech clarity. Individuals with severe-to-profound hearing loss still experienced advantages from NR technology in both listening effort measure and subjective sound quality assessments. Importantly, these benefits did not adversely affect speech intelligibility.

Assessment and Analysis of Reverberation Time in Hyder-Burks Arena at Tennessee Technological University

This study addresses the issue of long reverberation times in large-scale arenas, specifically the HyderBurks Arena (HBA) at Tennessee Technological University. HBA primarily hosts cattle shows and agricultural events. The current multi-speaker sound system in the facility ensures audibility, but the long reverberation time accumulates noise that impairs communication and causes dissatisfaction. The objectives of this work conducted as a student project in an undergraduate noise control class include measuring the reverberation time, analyzing the data, researching noise reduction methods, and proposing feasible plans that consider cost, achievability, and ease of implementation. The current reverberation time measured in the arena is approximately 2.313 seconds, posing challenges for speechapplications. The aim was to reduce this to a range of 1.32 to 1.02 seconds. The study employed a Larson Davis Sound Advisor 831C sound level meter to measure the reverberation time at various points in the arena. Despite low-frequency analysis challenges, critical frequencies affecting speech intelligibility were identified. The study also highlights potential issues with the arena’s sound system and the need for a synchronized, centralized speaker system. The assessment revealed that areas near the concrete walls without sound absorption have the maximum T60, indicating the need for treatment. Recommendations include ceiling treatment to reduce the reverberation time and enhance speech intelligibility. The proposed plan, considering cost-effectiveness and financial feasibility, aims to improve the atmosphere and experience for HBA users.

Using deep learning to improve the intelligibility of a target speaker in noisy multi-talker environments for people with normal hearing and hearing loss.

Understanding speech in noisy environments is a challenging task, especially in communication situations with several competing speakers. Despite their ongoing improvement, assistive listening devices and speech processing approaches still do not perform well enough in noisy multi-talker environments, as they may fail to restore the intelligibility of a speaker of interest among competing sound sources. In this study, a quasi-causal deep learning algorithm was developed that can extract the voice of a target speaker, as indicated by a short enrollment utterance, from a mixture of multiple concurrent speakers in background noise. Objective evaluation with computational metrics demonstrated that the speaker-informed algorithm successfully extracts the target speaker from noisy multi-talker mixtures. This was achieved using a single algorithm that generalized to unseen speakers, different numbers of speakers and relative speaker levels, and different speech corpora. Double-blind sentence recognition tests on mixtures of one, two, and three speakers in restaurant noise were conducted with listeners with normal hearing and listeners with hearing loss. Results indicated significant intelligibility improvements with the speaker-informed algorithm of 17% and 31% for people without and with hearing loss, respectively. In conclusion, it was demonstrated that deep learning-based speaker extraction can enhance speech intelligibility in noisy multi-talker environments where uninformed speech enhancement methods fail.

Functional outcomes for speech-in-noise intelligibility of NAL-NL2 and DSL v.5 prescriptive fitting rules in hearing aid users.

The primary aim of this research study is to assess whether differences exist in the application of the NAL-NL2 and DSL v.5 prescription formulas in terms of speech-in-noise intelligibility. Data from 43 patients, were retrospectively evaluated and analyzed. Inclusion criteria were patients with bilateral conductive, sensorineural, or mixed hearing loss, already using hearing aids for at least 1year, and aged 18years or older. Patients were categorized into two groups based on the prescriptive method employed by the hearing aid: NAL-NL2 or DSL v.5. Pure tone audiometry, speech audiometry, free field pure tone and speech audiometry withthehearing aid, and Matrix sentence test were performed. The Abbreviated Profile of Hearing Aid Benefit (APHAB) questionnaire was used to assess the personal audiological benefit provided bythe hearing aid. No statistically significant differences were found comparing the free-field pure tone average (FF PTA) and the free-field Word Recognition Score (FF WRS). Comparing the Speech Reception Threshold (SRT) parameter of patients with NAL-NL2 vs DSL v.5, no statistically significant difference was found, thus highlighting a condition of comparability between the two prescription methods in terms of speech-in-noise intelligibility. Comparing the results of the APHAB questionnaire, no statistically significant differences were evident for all subscales and overall benefit. When conducting a comparison between male and female patients using the NAL-NL2 method, no differences were observed in SRT values, however, the APHAB questionnaire revealed a difference in the AV subscale score for the same subjects. Our analysis revealed no statistically significant differences in speech-in-noise intelligibility, as measured by the SRT values from the Matrix Sentence Test, when comparing the two prescriptive methods. This compelling result reinforces the notion that, functionally, both methods are comparably effective in enhancing speech intelligibility in real-world, noisy environments. However, it is crucial to underscore that the absence of differences does not diminish the importance of considering individual patient needs and preferences in the selection of a prescriptive method.

Evaluating the Effects of Building Characteristics of Worship Auditoriums on Speech Intelligibility in Makurdi Metropolis

To enjoy the comfort of the indoor environment of any enclosed space, it is important to have a reasonable measure of understanding of any spoken word in the room. In line with the above this study assessed the characteristics of worship auditoriums with regards to speech intelligibility in Makurdi metropolis with the view to identifying building characteristics that affect speech intelligibility worship auditoriums. The study was carried out through the employment of walkthrough survey and physical observation; a well structured checklist was employed for the study. Measurements of lengths, widths and heights were carried out, from which the room volume was obtained to have an insight of the space under consideration. The auditoriums of ten worship centres in Makurdi Metropolis were studied. The findings that showed all the auditoriums were rectangular in shape. It was found that there was use of inappropriate application of materials for surface finishes in the auditoriums. In terms of room volume, it was found that G recorded the highest volume with 17064.00m3 while E recorded the lowest volume of 756.00m3 . The mean volume of the auditoriums was calculated to be 517.64m3 , the Standard Deviation (SD) was 4733.44 and the Coefficient of Variation (CV) was found to be 81.36%. The study recommended that proper selection and application of surface finish materials be ensured during construction of worship centres. Also, that acoustic treatment is applied in auditoriums where defects are identified, with the aim of effectively enhancing speech intelligibility in worship auditoriums in Makurdi Metropolis.

The Effect of the Frequency and Energetic Content of Broadband Noise on the Lombard Effect and Speech Intelligibility

The Lombard effect is an unconscious reflex of speakers to increase vocal effort when disturbed by noise, aiming to enhance speech intelligibility. This study aims to evaluate the effect of noise with different energetic content and levels at various frequencies on the Lombard effect, communication disturbance, vocal comfort, and speech intelligibility. Twenty university students participated in the study, reading a six-sentence excerpt and performing an intelligibility test under 12 randomized noise conditions. These conditions included noises at low (20–500 Hz), medium (500–4000 Hz), and high frequencies (4000–20,000 Hz), at four levels (45 dB, 55 dB, 65 dB, 75 dB). After each condition, participants rated their perceived communication disturbance and vocal discomfort. The results indicated that noise with energetic content at medium frequencies produced the highest Lombard effect, produced the most detrimental effect on communication disturbance and vocal comfort, and caused the strongest decrease in speech intelligibility, whereas it was minimally affected by low- and high-frequency noise. In conclusion, this study highlights that medium-frequency noise has the greatest impact on vocal effort, communication disturbance, and vocal comfort, while low- and high-frequency noise has minimal effect on speech intelligibility.

The Effect of Noise on Initiation and Maintenance of Clear Speech and Associated Mental Demand.

The objectives of this study were to explore (a) the influence of different types of background noise and their informational content on the ability of speakers to initiate and maintain clear speech (CS), a widely utilized technique for enhancing speech intelligibility, and (b) the impact of background noise and CS usage on speakers' mental demand. Five adult females were asked to read sentences using both habitual and CS under four distinct noise conditions: quiet, multitalker (MT) noise, reversed multitalker (RevMT) noise, and speech-shaped (SS) noise. Following this, speakers rated their perceived level of mental demand for each speaking condition using the modified NASA Task Load Index scale. A two-part listening experiment with 48 listeners was conducted to evaluate the speakers' effectiveness in initiating and maintaining CS. Speakers initiated CS more successfully in noise than in quiet, with better performance observed in the presence of RevMT noise as compared to SS noise. Regarding the maintenance of CS, none of the speakers were successful in a quiet environment. Furthermore, the ability to maintain CS was most adversely affected in MT noise, followed by RevMT noise and SS noise. Our findings suggest that the effect of background noise on speech production is complex and multifaceted. The noise type affected speakers' ability to initiate and maintain CS as well as the mental demand associated with the speech task. The results underscore the importance of considering the characteristics of background noise and cognitive aspects of speech production when training and evaluating speakers' performance.

D2StarGAN: A Near-Far End Noise Adaptive StarGAN for Speech Intelligibility Enhancement

When using mobile communication, the voice output from the device is already relatively clear, but in a noisy environment, it is difficult for the listener to obtain the information expressed by the speaker with clarity. Consequently, speech intelligibility enhancement technology has emerged to help alleviate this problem. Speech intelligibility enhancement (IENH) is a technique that enhances speech intelligibility during the reception phase. Previous research has focused on IENH through normal versus different levels of Lombardic speech conversion, inspired by a well-known acoustic mechanism called the Lombard effect. However, these methods often lead to speech distortion and impair the overall speech quality. To address the speech quality degradation problem, we propose an improved (StarGAN)-based IENH framework by combining StarGAN networks with the dual discriminator idea to construct the conversion framework. This approach offers two main advantages: (1) Addition of a speech metric discriminator on top of StarGAN to optimize multiple intelligibility and quality-related metrics simultaneously; (2) a framework that is adaptive to different distal and proximal noise levels with different noise types. Experimental results from objective experiments and subjective preference tests show that our approach outperforms the baseline approach, and these enable IENH to be more widely used.

Improvement of speech by facial prosthesis in a midfacial defect with a gunshot injury: A case report.

Gunshot wounds in the maxillofacial region may cause serious defects that result in functional and esthetic disabilities. A 14-year-old male patient was referred to the prosthodontics clinic for provisional prosthetic restoration 5 months after an injury involving the midline midfacial region. An interim extraoral silicone maxillofacial prosthesis, also known as an epithesis, replacing the nose and upper lip, was fabricated to enhance appearance and speech intelligibility. The effect of the prosthesis insertion on speech outcome was assessed over 6 months. Blood pulse oxygen saturation levels and heart rate were monitored to assess ventilation following the prosthesis insertion. Results of the Assessment of Articulation Subtest revealed the patient produced the target speech sounds of 31 more words correctly at 6 months follow-up. Bilabial, labiodental, and postalveolar speech sounds in the phonetic inventory improved. Fabrication of an extraoral silicone maxillofacial prosthesis, modifying substructure design, and prosthesis usage over time contributed to the significant enhancement of speech intelligibility and did not significantly change the blood pulse oxygen saturation levels and heart rate. It led to an improvement in overall aesthetic outcomes and was a useful tool for the patient's social integration. This article is protected by copyright. All rights reserved.

A hybrid discriminant fuzzy DNN with enhanced modularity bat algorithm for speech recognition

In recent years, speech processing resides a major application in the domain of signal processing. Due to the audibility loss of some speech signals, people with hearing impairment have difficulty in understanding speech, which reintroduces a crucial role in speech recognition. Automatic Speech Recognition (ASR) development is a major challenge in research in the case of noise, domain, vocabulary size, and language and speaker variability. Speech recognition system design needs careful attention to challenges or issues like performance and database evaluation, feature extraction methods, speech representations and speech classes. In this paper, HDF-DNN model has been proposed with the hybridization of discriminant fuzzy function and deep neural network for speech recognition. Initially, the speech signals are pre-processed to eliminate the unwanted noise and the features are extracted using Mel Frequency Cepstral Coefficient (MFCC). A hybrid Deep Neural Network and Discriminant Fuzzy Logic is used for assisting hearing-impaired listeners with enhanced speech intelligibility. Both DNN and DF have some problems with parameters to address this problem, Enhanced Modularity function-based Bat Algorithm (EMBA) is used as a powerful optimization tool. The experimental results show that the proposed automatic speech recognition-based hybrid deep learning model is effectively-identifies speech recognition more than the MFCC-CNN, CSVM and Deep auto encoder techniques. The proposed method improves the overall accuracy of 8.31%, 9.71% and 10.25% better than, MFCC-CNN, CSVM and Deep auto encoder respectively.

Listening in complex acoustic scenarios: Ongoing advances on methodologies and approaches in relation to classrooms

The acoustic quality of the built environment where we spend our days strongly affects our social life and indirectly our health. Particularly, people with a hearing impairment (usually related to ageing), children with or without learning disorders and non-native speakers, require better room acoustics to enhance speech communication and be included in the challenging multicultural processes. In the specific framework of children in their everyday learning environments, i.e., in classrooms, enhanced speech intelligibility (SI) is necessary. Noise and reverberation in classrooms degrade SI, but the combined effect of reverberation, informational noise and position of target, listener and noise source on SI still needs insights. This work provides an overview on the recent research carried out in the field of SI under complex acoustic scenarios. Results from in-field measurements as well as from advanced predictive models will be presented and discussed. Particularly, measurements were integrated with listening tests based on the Matrix material, since it has been optimized for the Italian language and for many other foreign languages. Concerning the predictions’ results, the Binaural Speech Intelligibility Model (BSIM) implemented at the University of Oldenburg was used to validate an accurate method to predict the listeners’ ability at the classroom’s design stage.

Dual-task costs of clear and loud speech: Implications for talkers with Parkinson disease

Decades of research suggest that the speech production adjustments associated with adopting a clearer or louder speaking style enhance speech intelligibility, making these speech styles a common talker-oriented strategy in speech production interventions for talkers with dysarthria secondary to Parkinson disease (PD). Recent data suggest that explicit instructions to speak louder or clearer than usual may increase the processing load and effort associated with speech production. Empirical data will be presented that demonstrate these higher-effort speech styles may require greater attentional resources than habitual or conversational speech. A group of neurotypical talkers performed a sentence repetition task using Habitual, Loud, and Clear speech styles in isolation and while performing a concurrent visuomotor tracking task. Lip and jaw kinematics, speech intensity, and accuracy on the visuomotor task were collected to quantify performance on each task. Group-level data suggest that when faced with the competing task demands of the visuomotor task, neurotypical talkers prioritized the articulatory and phonatory adjustments associated with Clear and Loud speech at the cost of secondary task performance. Case examples from talkers with hypokinetic dysarthria resulting from PD will also be examined to frame and discuss clinical implications and considerations.

SNR-Based Inter-Component Phase Estimation Using Bi-Phase Prior Statistics for Single-Channel Speech Enhancement

The fundamental problem of phase-aware single-channel speech enhancement is the estimation of the harmonic phase of signal components from noisy observations. One approach to obtain an estimate of the harmonic phase is by smoothing the noisy harmonic phase in the time domain. Accurate identification of the smoothing regions is crucial to achieve the best enhancement results. Previous works have introduced a binary hypothesis framework for detecting these regions, which is formulated <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">individually</i> for each harmonic component based on von Mises statistics and local harmonic SNR levels. Later, smoothing frameworks have been proposed in the <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">inter-component phase domain</i> , where the inter-component phases follow the temporal constancy property for clean speech signals. However, these frameworks either do not formulate an SNR-dependent smoothing threshold or impose it empirically. The aim of this work is to formulate an SNR-dependent binary hypothesis framework for smoothing in the <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">inter-component phase domain</i> , using prior inter-component phase statistics. The framework then reconstructs the instantaneous phases of the enhanced speech signal components. This proposed framework results in superior speech intelligibility enhancement, reduces phase deviation, and does not introduce auditory buzziness artifacts compared to the previously developed framework that does not formulate an SNR-dependent smoothing threshold.

Acoustic Feedback Noise Cancellation in Hearing Aids Using Adaptive Filter

To enhance speech intelligibility for people with hearing loss, hearing aids will amplify speech using gains derived from evidence-based prescriptive methods, in addition to other advanced signal processing mechanisms. While the evidence supports the use of hearing aid signal processing for speech intelligibility, these signal processing adjustments can also be detrimental to hearing aid sound quality, with poor hearing aid sound quality cited as a barrier to device adoption. In general, an uncontrolled environment may contain degradation components like background noise, speech from other speakers etc. along with required speech components. In this paper, we implement adaptive filtering design for acoustic feedback noise cancellation in hearing aids. The adaptive filter architecture has been designed using normalized least mean square algorithm. By using adaptive filters both filter input coefficients are changeable during run-time and reduce noise in hearing aids. The proposed design is implemented in matlab and the simulations shows that the proposed architecture produces good quality of speech, accuracy, maintain stable steady state. The proposed design is validated with parameters like Noise Distortion, Perceptual Evaluation of Speech Quality, Signal to Noise Ratio, and Speech Distortion. The feedback canceller is implemented in MATLAB 9.4 simulink version release name of R2018a is used for validation with Echo Return Loss Enhancement (ERLE). The ERLE of the NMLS is reduced when the filter order is increases.Around 10% of the power spectrum density (PSD) is less when compared with existing designs.

Development of CNN-based robust dysarthric isolated digit recognition system by enhancing speech intelligibility

Development of CNN-based robust dysarthric isolated digit recognition system by enhancing speech intelligibility

A Novel Speech Intelligibility Enhancement Model based on Canonical Correlation and Deep Learning.

Current deep learning (DL) based approaches to speech intelligibility enhancement in noisy environments are often trained to minimise the feature distance between noise-free speech and enhanced speech signals. Despite improving the speech quality, such approaches do not deliver required levels of speech intelligibility in everyday noisy environments. Intelligibility-oriented (I-O) loss functions have recently been developed to train DL approaches for robust speech enhancement. Here, we formulate, for the first time, a novel canonical correlation based I-O loss function to more effectively train DL algorithms. Specifically, we present a canonical-correlation based short-time objective intelligibility (CC-STOI) cost function to train a fully convolutional neural network (FCN) model. We carry out comparative simulation experiments to show that our CC-STOI based speech enhancement framework outperforms state-of-the-art DL models trained with conventional distance-based and STOI-based loss functions, using objective and subjective evaluation measures for case of both unseen speakers and noises. Ongoing future work is evaluating the proposed approach for design of robust hearing-assistive technology.

Do cochlear-implanted children benefit from voice cue differences for understanding speech in speech masker?

Differences in voice cues of mean fundamental frequency (F0) and vocal-tract length (VTL) between a target and masker speech can help a listener to segregate the two, and enhance speech intelligibility. Postlingually deafened and implanted adult CI users have difficulties in perceiving F0 and VTL, and this possibly contributes to challenges in understanding speech in presence of a speech masker. Previous research showed that speech-on-speech understanding is not yet adult-like in normal-hearing (NH) children of school age, as this skill seems to develop over many years. Yet, like NH adults, NH children do seem to benefit from the voice cue differences in some situations. In the case of CI children, a number of factors could influence intelligibility of speech with speech maskers. Prelingually deafened and implanted CI children develop auditory skills primarily based on spectrotemporally degraded speech input, which could lead to different developmental patterns compared to NH children. On the other hand, the CI children would benefit from the optimal brain plasticity period for learning, which could lead to different perceptual patterns compared to CI adults. In this talk, I will present data from multiple projects with children and adults at our lab to investigate these potential expectations.

지하철 승강장에서 소음 및 신호안내음의 분포 특성

As the elderly frequently use subway stations, there is a need for the acoustical improvement of subway platforms from a barrier-free perspective. In this study, the distribution characteristics of noise and signal guide sounds on subway platforms are determined for different locations. The entering and leaving noise of the train and the signal guide sounds were measured for thirteen and fourteen points along the two subway platforms. As a result, the maximum difference (L<SUB>Aeq</SUB>) between the noise values for the train entering and leaving for the measurement points ranged from 5.0 dBA ~ 9.0 dBA and 9.2 dBA ~ 10.9 dBA, respectively. Additionally, the maximum difference (L<SUB>Aeq</SUB>) between the signal and guide sounds for the measurement points ranged from 6.2 dBA ~ 6.5 dBA and 4.7 dBA ~ 6.5 dBA, respectively. The maximum difference (L<SUB>Aeq</SUB>) in the signal to noise ratio (S/N ratio) between the noise and signal guide sounds for the measurement points ranged from 5.1 dBA ~ 14.6 dBA. It was also found that noise had more influence on the S/N ratio than the signal guide sound. These results suggest that there is a need for noise reduction in various positions of a subway platform to enhance speech intelligibility.

A Novel Method for Intelligibility Assessment of Nonlinearly Processed Speech in Spaces Characterized by Long Reverberation Times.

Objective assessment of speech intelligibility is a complex task that requires taking into account a number of factors such as different perception of each speech sub-bands by the human hearing sense or different physical properties of each frequency band of a speech signal. Currently, the state-of-the-art method used for assessing the quality of speech transmission is the speech transmission index (STI). It is a standardized way of objectively measuring the quality of, e.g., an acoustical adaptation of conference rooms or public address systems. The wide use of this measure and implementation of this method on numerous measurement devices make STI a popular choice when the speech-related quality of rooms has to be estimated. However, the STI measure has a significant drawback which excludes it from some particular use cases. For instance, if one would like to enhance speech intelligibility by employing a nonlinear digital processing algorithm, the STI method is not suitable to measure the impact of such an algorithm, as it requires that the measurement signal should not be altered in a nonlinear way. Consequently, if a nonlinear speech enhancing algorithm has to be tested, the STI—a standard way of estimating speech transmission cannot be used. In this work, we would like to propose a method based on the STI method but modified in such a way that it makes it possible to employ it for the estimation of the performance of the nonlinear speech intelligibility enhancement method. The proposed approach is based upon a broadband comparison of cumulated energy of the transmitted envelope modulation and the received modulation, so we called it broadband STI (bSTI). Its credibility with regard to signals altered by the environment or nonlinear speech changed by a DSP algorithm is checked by performing a comparative analysis of ten selected impulse responses for which a baseline value of STI was known.

Marathi Speech Intelligibility Enhancement Using I-AMS Based Neuro-Fuzzy Classifier Approach for Hearing Aid Users

Globally, 1.6 billion individuals suffered from hearing disability in 2019. According to the World Health Organization, by 2050, the number of people with hearing impairments will rise to 2.5 billion. Speech perception in noisy surroundings is a challenge for hearing aid users. This study aimed to design a novel methodology to improve the speech recognition ability of hearing aid users from various backgrounds. To improve speech enhancement, we propose a discrete cosine transform (DCT)-based improved amplitude-magnitude spectrogram (I-AMS) algorithm with a fuzzy classifier. First, the I-AMS approach disintegrates speech signals containing noise into time-frequency units and eliminates the noise present in the signal. Next, the time frequency units (t-f units), modulation frequency (f <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">m</sub> ), and centre frequency (f <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">c</sub> ) are extracted from the denoised signal. A neuro-fuzzy classifier was used to classify the background speech environment into three different classes. The proposed I-AMS algorithm was tested, achieved improvements in terms of sensitivity (+1.02%) and accuracy (+11.80%). Speech denoising revealed a 1.27% improvement in speech recognition performance.