Improvement In Speech Quality Research Articles

Comparison the Voice Onset Time (VOT) of Postlingual Cochlear Implant Users and Normal Peers in the CAPE_V Sentences as Continues Speech Task

BackgroundAuditory perception plays a crucial role in speech and language development, emphasizing concerns about hearing loss. While cochlear implantation nearly tackles challenges associated with post-lingual hearing loss in adults, the importance of "auditory feedback" and acoustic assessment becomes crucial for evaluating speech disorders and devising effective treatments. This study aims to address the gap in assessing Voice Onset Time (VOT) as an indicator of nuanced variations in the speech of cochlear implant users during a continuous speech task. MethodRecordings of Persian CAPE-V sentences were obtained from 25 cochlear implant (CI) users and 25 healthy speakers, with a mean age of 33.2 years (SD=11.5, range=18-55 years). Ten words, incorporating both voiced and voiceless consonants, were selected from the CAPE-V sentences. VOT measurements for the specified stop consonants at the initial syllables of these chosen words were computed using PRAAT. A comparative analysis between the two age- and gender-matched groups was conducted using an independent sample t-test. Subsequently, ANCOVA was employed to examine the influence of demographic factors on VOT values among CI participants. ResultsUnvoiced consonant /p/ in /po/, /pɑ/, /pe/, and /pa/ syllables had higher VOT values in the healthy group, while the voiced consonant /d/ in /da/ and /di/ syllables demonstrated higher VOT values in the CI group. (P<0.05) Apart from /po/ and /di/ syllables, no significant impacts of demographic factors on VOT values were observed (P≥0.8). ConclusionDespite the improvement in speech quality after cochlear implantation, subtle differences persist. The motor theory, which underscores the impact of auditory inputs on temporal coordination, highlights the role of VOT in speech discrimination. Various linguistic factors affect VOT, including articulation position, vowel context, and raised vowels. While CI enhances syllable distinction, challenges in articulation for adults suggest a need for targeted training in rehabilitation programs, ultimately enhancing the quality of life for CI users.

ADVANCING GASIFICATION-COMBINED UP AND DOWN DRAFT GASIFIER-BASED TREATMENT OF TEXTILE WASTE: ASSESSING FEASIBILITY, ENVIRONMENTAL IMPACTS AND ENERGY RECOVERY POTENTIAL

Speech enhancement, a critical component in various applications including voice assistants and telecommunication systems, aims to improve the quality and intelligibility of spoken content. This paper introduces a novel approach leveraging Natural Language Processing (NLP) techniques for speech enhancement. By harnessing the power of NLP models, we propose a methodology that not only considers the acoustic features but also incorporates linguistic context in the enhancement process. The study encompasses data collection, preprocessing, and the application of advanced NLP algorithms to extract relevant linguistic information. The experimental results demonstrate a significant improvement in speech quality, as evidenced by a substantial increase in Signal-to-Noise Ratio (SNR) and Mean Opinion Score (MOS) compared to conventional methods. Furthermore, a comparative analysis with existing techniques showcases the superiority of our approach. This research not only contributes to the advancement of speech enhancement technology but also underscores the pivotal role of NLP in this domain. The findings pave the way for future research directions, suggesting potential applications in areas such as voice recognition systems and real-time communication platforms. Keywords:Acoustic Features, Linguistic Context, Mean Opinion Score (MOS), Natural Language Processing, NLP Algorithms, Signal-to-Noise Ratio (SNR), Speech Enhancement, Speech Quality, Telecommunication Systems, Voice Assistants.

Speech enhancement for helicopter headsets: Simulation and implementation on an FPGA platform

Clear and effective communication is crucial for the safe operation of aircraft. In helicopters, high levels of noise are generated by the engine, gears, and aerodynamics, which negatively impact speech intelligibility. To address this issue, modern aircraft headsets utilize active noise control (ANC) to reduce noise levels for both the crew and passengers. However, the speech signals captured by these headsets often contain high levels of background noise, thereby hindering internal and external flight communication. This paper introduces a dual microphone dual stage speech enhancement algorithm that combines basic spectral subtraction with a Wiener Filter, enhanced by the a priori and a posteriori signal-to-noise ratio. Audio data from within a helicopter cabin were recorded during a test flight. In a series of simulations, the Wiener Filter implementation is compared to other algorithms based only on spectral subtraction methods. The results are evaluated using established performance measures for speech quality. The Wiener Filter implementation results in the highest speech quality and is, therefore, implemented on an FPGA-platform for validation in a laboratory experiment. The simulations and measurements demonstrate significant improvements in speech quality and, consequently, enhance speech intelligibility using the proposed method.

American Cochlear Implant Alliance Task Force: Recommendations for Determining Cochlear Implant Candidacy in Adults.

The indications for cochlear implantation have expanded over time due to evidence demonstrating identification and implantation of appropriate cochlear implant (CI) candidates lead to significant improvements in speech recognition and quality of life (QoL). However, clinical practice is variable, with some providers using outdated criteria and others exceeding current labeled indications. As a results, only a fraction of those persons who could benefit from CI technology receive it. This document summarizes the current evidence for determining appropriate referrals for adults with bilateral hearing loss into CI centers for formal evaluation by stressing the importance of treating each ear individually and a "revised 60/60 rule". By mirroring contemporary clinical practice and available evidence, these recommendations will also provide a standardized testing protocol for CI candidates using a team-based approach that prioritizes individualized patient care. This manuscript was developed by the Adult Cochlear Implantation Candidacy Task Force of the American Cochlear Implant Alliance using review of the existing literature and clinical consensus. LEVEL OF EVIDENCE: N/A Laryngoscope, 134:S1-S14, 2024.

A real-time voice cloning system with multiple algorithms for speech quality improvement.

With the development of computer technology, speech synthesis techniques are becoming increasingly sophisticated. Speech cloning can be performed as a subtask of speech synthesis technology by using deep learning techniques to extract acoustic information from human voices and combine it with text to output a natural human voice. However, traditional speech cloning technology still has certain limitations; excessively large text inputs cannot be adequately processed, and the synthesized audio may include noise artifacts like breaks and unclear phrases. In this study, we add a text determination module to a synthesizer module to process words the model has not included. The original model uses fuzzy pronunciation for such words, which is not only meaningless but also affects the entire sentence. Thus, we improve the model by splitting the letters and pronouncing them separately. Finally, we also improved the preprocessing and waveform conversion modules of the synthesizer. We replace the pre-net module of the synthesizer and use an upgraded noise reduction algorithm combined with the SV2TTS framework to achieve a system with superior speech synthesis performance. Here, we focus on improving the performance of the synthesizer module to achieve higher-quality speech synthesis audio output.

CST: Complex Sparse Transformer for Low-SNR Speech Enhancement

Speech enhancement tasks for audio with a low SNR are challenging. Existing speech enhancement methods are mainly designed for high SNR audio, and they usually use RNNs to model audio sequence features, which causes the model to be unable to learn long-distance dependencies, thus limiting its performance in low-SNR speech enhancement tasks. We design a complex transformer module with sparse attention to overcome this problem. Different from the traditional transformer model, this model is extended to effectively model complex domain sequences, using the sparse attention mask balance model's attention to long-distance and nearby relations, introducing the pre-layer positional embedding module to enhance the model's perception of position information, adding the channel attention module to enable the model to dynamically adjust the weight distribution between channels according to the input audio. The experimental results show that, in the low-SNR speech enhancement tests, our models have noticeable performance improvements in speech quality and intelligibility, respectively.

Association of Cognitive Impairment Screening Scores With Improvements in Speech Recognition and Quality of Life After Cochlear Implantation

Many cochlear implant centers screen patients for cognitive impairment as part of the evaluation process, but the utility of these scores in predicting cochlear implant outcomes is unknown. To determine whether there is an association between cognitive impairment screening scores and cochlear implant outcomes. Retrospective case series of adult cochlear implant recipients who underwent preoperative cognitive impairment screening with the Montreal Cognitive Assessment (MoCA) from 2018 to 2020 with 1-year follow-up at a single tertiary cochlear implant center. Data analysis was performed on data from January 2018 through December 2021. Cochlear implantation. Preoperative MoCA scores and mean (SD) improvement (aided preoperative to 12-month postoperative) in Consonant-Nucleus-Consonant phonemes (CNCp) and words (CNCw), AzBio sentences in quiet (AzBio Quiet), and Cochlear Implant Quality of Life-35 (CIQOL-35) Profile domain and global scores. A total of 52 patients were included, 27 (52%) of whom were male and 46 (88%) were White; mean (SD) age at implantation was 68.2 (13.3) years. Twenty-three (44%) had MoCA scores suggesting mild and 1 (2%) had scores suggesting moderate cognitive impairment. None had been previously diagnosed with cognitive impairment. There were small to medium effects of the association between 12-month postoperative improvement in speech recognition measures and screening positive or not for cognitive impairment (CNCw mean [SD]: 48.4 [21.9] vs 38.5 [26.6] [d = -0.43 (95% CI, -1.02 to 0.16)]; AzBio Quiet mean [SD]: 47.5 [34.3] vs 44.7 [33.1] [d = -0.08 (95% CI, -0.64 to 0.47)]). Similarly, small to large effects of the associations between 12-month postoperative change in CIQOL-35 scores and screening positive or not for cognitive impairment were found (global: d = 0.32 [95% CI, -0.59 to 1.23]; communication: d = 0.62 [95% CI, -0.31 to 1.54]; emotional: d = 0.26 [95% CI, -0.66 to 1.16]; entertainment: d = -0.005 [95% CI, -0.91 to 0.9]; environmental: d = -0.92 [95% CI, -1.86 to 0.46]; listening effort: d = -0.79 [95% CI, -1.65 to 0.22]; social: d = -0.51 [95% CI, -1.43 to 0.42]). In this case series, screening scores were not associated with the degree of improvement of speech recognition or patient-reported outcome measures after cochlear implantation. Given the prevalence of screening positive for cognitive impairment before cochlear implantation, preoperative screening can be useful for early identification of potential cognitive decline. These findings support that screening scores may have a limited role in preoperative counseling of outcomes and should not be used to limit candidacy.

A Novel Jointly Optimized Cooperative DAE-DNN Approach Based on a New Multi-Target Step-Wise Learning for Speech Enhancement

In this paper, we present a new supervised speech enhancement approach based on a cooperative structure of deep autoencoders (DAEs) as generative models and deep neural networks (DNN). The DAE is used as a nonlinear alternative to nonnegative matrix factorization (NMF) for the extraction of harmonic structures and encoded features of the noise, clean and noisy signals, and a DNN is deployed as a nonlinear mapper. We introduce a deep network imitating NMF in a non-linear manner to overcome the problems of a simple linear model, such as performance degradation in non-stationary environments. Compared to combinatorial NMF and DNN methods, we do all of the decomposition, enhancement, and reconstruction processes in a nonlinear framework via a suitable cooperative structure of encoder, DNN, and decoders, and jointly optimize them. We also propose a supervised hierarchical multi-target training approach, performed in two steps, such that the DNN not only predicts the low-level encoded features as primary targets but it also predicts the high-level actual spectral signals as secondary targets. The first step acts as a pretraining for the second step which improves the learning strategy. Moreover, to exploit a more discriminative model for noise reduction, a DNN-based noise classification and fusion strategy (NCF) is also proposed. The experiments on TIMIT dataset reveal that the proposed methods outperform the previous approaches and achieve an average perceptual evaluation of speech quality (PESQ) improvement of up to about 0.3 for speech enhancement.

Noise Reduction Combining a General Microphone and a Throat Microphone.

In this study, we propose a method to reduce noise from speech obtained from a general microphone using the information of a throat microphone. A throat microphone records a sound by detecting the vibration of the skin surface near the throat directly. Therefore, throat microphones are less prone to noise than ordinary microphones. However, as the acoustic characteristics of the throat microphone differ from those of ordinary microphones, its sound quality degrades. To solve this problem, this study aims to improve the speech quality while suppressing the noise of a general microphone by using the information recorded by a throat microphone as reference information to extract the speech signal in general microphones. In this paper, the framework of the proposed method is formulated, and several experiments are conducted to evaluate the noise suppression and speech quality improvement effects of the proposed method.

E2E-V2SResNet: Deep residual convolutional neural networks for end-to-end video driven speech synthesis

Speechreading which infers spoken message from a visually detected articulated facial trend is a challenging task. In this paper, we propose an end-to-end ResNet (E2E-ResNet) model for synthesizing speech signals from the silent video of a speaking individual. The model is the convolutional encoder-decoder framework which captures the frames of video and encodes into a latent space of visual features. The outputs of the decoder are spectrograms which are converted into waveforms corresponding to a speech articulated in the input video. The speech waveforms are then fed to a waveform critic used to decide the real or synthesized speech. The experiments show that the proposed E2E-V2SResNet model is apt to synthesize speech with realism and intelligibility/quality for GRID database. To further demonstrate the potentials of the proposed model, we also conduct experiments on the TCD-TIMIT database. We examine the synthesized speech in unseen speakers using three objective metrics use to measure the intelligibility, quality, and word error rate (WER) of the synthesized speech. We show that E2E-V2SResNet model outscores the competing approaches in most metrics on the GRID and TCD-TIMIT databases. By comparing with the baseline, the proposed model achieved 3.077% improvement in speech quality and 2.593% improvement in speech intelligibility.

Enhancement of speech in noise using multi-channel, time-varying gains derived from the temporal envelope

The envelope of speech in noise has been used as a control signal to amplitude-modulate corrupted speech in order to reduce noise and potentially enhance speech quality and intelligibility. In this study, separate speech and noise time histories were mixed at different SNRs to produce time-aligned speech in noise with a range of SNRs. The temporal envelope of the control signal was manipulated by selecting mixtures with the same, or different, SNRs as the signal being controlled. Signals from 200 Hz to 6 kHz were processed into sixteen, parallel contiguous subbands with bandwidths approximating 1.5 times that of auditory filters. The subband envelopes were formed from the absolute value of signals and low pass filtered at 16 Hz. Eleven subjects aged 29 ± 8 years (mean and range) with normal hearing underwent the Modified Rhyme Test to assess speech intelligibility. Speech quality was assessed objectively by PESQ. When the SNRs of the controlling signal and the speech in noise being controlled were the same, no improvement in speech intelligibility was obtained. When the SNR of the controlling signal exceeded that of the signal being controlled, statistically significant increases in mean word scores of up to 36% could be obtained, hence providing an estimate for the maximum improvement in speech intelligibility achievable by reducing noise in the control signal. There were also small improvements in speech quality.

Auditory-cognitive training for adult cochlear implant recipients: a study protocol for a randomised controlled trial

BackgroundThere is an urgent need to develop new therapies to improve cognitive function in adults following cochlear implant surgery. This study aims to determine if completing at-home computer-based brain training activities improve memory and thinking skills in adults following their first cochlear implant.MethodsThis study will be conducted as a single-blind, head-to-head, randomised controlled trial (RCT). It will determine whether auditory training combined with adaptive computerised cognitive training will elicit greater improvement in cognition, sound and speech perception, mood, and quality of life outcomes in adult cochlear implant recipients, when compared to auditory training combined with non-adaptive (i.e. placebo) computerised cognitive training. Participants 18 years or older who meet the clinical criteria for a cochlear implant will be recruited into the study.ResultsThe results of this trial will clarify whether the auditory training combined with cognitive training will improve cognition, sound and speech perception, mood, and quality of life outcomes in adult cochlear implant recipients.DiscussionWe anticipate that our findings will have implications for clinical practice in the treatment of adult cochlear implant recipients.Trial registrationAustralian New Zealand Clinical Trials Registry ACTRN12619000609156. Registered on April 23 2019.

A new double backward distributive weighted adaptive filtering approach for speech quality improvement

A new double backward distributive weighted adaptive filtering approach for speech quality improvement

Speech Quality Improvement of Commercial Flat Screen TV-Sets

This work deals with the improvement of perceived speech quality of flat screen TV-sets on the various broadcasting platforms of the Italian radio and TV broadcasting company Rai. It is well known that the reduced thickness of flat TVs implied a degradation of the audio quality due to the adoption of miniaturised and cheap loudspeakers compared to the ones in the former Cathode Ray Tube televisions. The research activity gave rise to a Transfer Function (TF), which modifies in real-time the frequency spectrum of the audio signal from the TV station before the transmission to the broadcasting tower. In this way, the final users receive the processed audio signal without the need of additional hardware. A Digital Audio Optimizer dynamically equalizes the sound level, boosting the audio signal towards a flat frequency response without any increase in the loudness levels. Given that the majority of the broadcasted audio signals have speech or singing contents, the TF boosts the speech level in the frequency range that is more important for speech intelligibility, i.e., between 1 kHz and 4 kHz. Subjective evaluations of the proposed TF have been carried out in a laboratory in compliance with the standard ITU-R BS.1116. Three different commercial TV-set models, 30 audio excerpts from video tracks divided into the three genres Speech, Sing and Music, and Sport, and 72 subjects aged between 21 and 53 years, were part of the study. Overall, the perceived improvement in the audio quality compared to the non-processed signal was 25.3% on average among the three TV-set models and the three genres. In order to estimate the perceived improvement directly from the audio signal, regression analyses have been performed, which allow the estimation of subjective outcomes from objective indexes based on intensity features and frequency content of the signal, with standard errors between 10% and 14%.

Speech Enhancement for Hearing Impaired Based on Bandpass Filters and a Compound Deep Denoising Autoencoder

Deep neural networks have been applied for speech enhancements efficiently. However, for large variations of speech patterns and noisy environments, an individual neural network with a fixed number of hidden layers causes strong interference, which can lead to a slow learning process, poor generalisation in an unknown signal-to-noise ratio in new inputs, and some residual noise in the enhanced output. In this paper, we present a new approach for the hearing impaired based on combining two stages: (1) a set of bandpass filters that split up the signal into eight separate bands each performing a frequency analysis of the speech signal; (2) multiple deep denoising autoencoder networks, with each working for a small specific enhancement task and learning to handle a subset of the whole training set. To evaluate the performance of the approach, the hearing-aid speech perception index, the hearing aid sound quality index, and the perceptual evaluation of speech quality were used. Improvements in speech quality and intelligibility were evaluated using seven subjects of sensorineural hearing loss audiogram. We compared the performance of the proposed approach with individual denoising autoencoder networks with three and five hidden layers. The experimental results showed that the proposed approach yielded higher quality and was more intelligible compared with three and five layers.

Low-complexity artificial noise suppression methods for deep learning-based speech enhancement algorithms

Deep learning-based speech enhancement algorithms have shown their powerful ability in removing both stationary and non-stationary noise components from noisy speech observations. But they often introduce artificial residual noise, especially when the training target does not contain the phase information, e.g., ideal ratio mask, or the clean speech magnitude and its variations. It is well-known that once the power of the residual noise components exceeds the noise masking threshold of the human auditory system, the perceptual speech quality may degrade. One intuitive way is to further suppress the residual noise components by a postprocessing scheme. However, the highly non-stationary nature of this kind of residual noise makes the noise power spectral density (PSD) estimation a challenging problem. To solve this problem, the paper proposes three strategies to estimate the noise PSD frame by frame, and then the residual noise can be removed effectively by applying a gain function based on the decision-directed approach. The objective measurement results show that the proposed postfiltering strategies outperform the conventional postfilter in terms of segmental signal-to-noise ratio (SNR) as well as speech quality improvement. Moreover, the AB subjective listening test shows that the preference percentages of the proposed strategies are over 60%.

Performance analysis of various training targets for improving speech quality and intelligibility

Denoising a single-channel speech (recorded using one microphone) remains an open problem in many speech-related applications. Recently, supervised deep learning methods are used to denoise the speech signal. This work uses Deep Neural Network (DNN) to learn the Time–Frequency (T-F) mask of the clean speech from its noisy speech features. In general, Ideal Binary Mask (IBM) is used as the binary mask training target to improve speech intelligibility, and Ideal Ratio Mask (IRM) is used as a non-binary mask training target to improve speech quality. Still, it may not necessarily be the best T-F mask to analyze the performance of improvement in speech quality/intelligibility. However, an appropriate training target remains to be unclear for supervised deep learning methods. In this work, a non-binary novel soft T-F mask named Optimum Soft Mask (OSM) is proposed, analyzed and compared with different T-F mask types used for single-channel speech denoising methods. In addition, the target T-F mask is compared with the existing state of art approaches to show a clear performance advantage of supervised deep learning models. The performance of the binary and non-binary training targets of DNN is evaluated under different Signal-to-Noise-Ratio’s and noise conditions ti improve speech quality and intelligibility. The experimental results reveal that the binary mask IBM shows significant improvement in speech intelligibility; the non-binary mask IRM shows a substantial improvement in speech quality. At the same time, the proposed novel soft T-F mask shows notable improvement in both quality and intelligibility under various test conditions.

Occlusal and Functional Improvement of Patients With Midfacial Hypoplasia Corrected Using Intraoral Tooth-Borne Mid-Maxillary Osteodistractors at Teaching Hospital-Karapitiya, Sri Lanka

Mid-maxillary osteodistraction (MMOD) is a novel technique. This study seeks to assess the occlusal and functional improvement of patients with midfacial hypoplasia secondary to cleft lip and palate. This was done as a retrospective study at the Regional Cleft Centre of the Teaching Hospital-Karapitiya (THK). A total of 81 patient records were accessed on those who underwent MMOD with a tooth-borne intraoral device from January 1, 2007, to December 31, 2016, at THK. Records of the patients included pre- and post-treatment lateral cephalograms, study models, speech assessments, and photographs. An advancement ranging from 5 to 12 mm (mean, 8.14 mm) was achieved in all patients; 80% (n = 63) and 87.65% (n = 71) showed improvement in consonant production and quality of speech. These were statistically significant (P < .05). With regard to the Peer Assessment Rating score assessing the occlusal improvement, 97.6% showed an improvement whereas only 2 patients showed no improvement. There was a significant improvement in sella (S), nasion (N), and point A (SNA) angle; point A, the nasion, and point B (ANB) angle; and the maxillary length over the treatment period (P < .05). The change in the ANB angle and maxillary length significantly differed between genders with males having a comparatively higher improvement. The profile improvements were noteworthy on the profile photographs. This relatively novel approach to maxillary advancement by maxillary distraction using a combined surgical and modified orthodontic technique has proven to be successful. This protocol and the technique have improved function, occlusion, and the aesthetics.

Hearing aid use and cognition in older adults: Can we delay decline or even improve cognitive function?

AbstractBackgroundHearing loss is associated with accelerated cognitive decline in older adults, and has been identified as a modifiable risk factor for dementia. Although hearing aids are a successful treatment for hearing loss, it is unknown whether hearing aid use can delay the onset of cognitive decline or improve cognitive function in older adults with hearing loss. This prospective longitudinal study is investigating the effect of hearing aid use on cognitive decline and addresses many of the methodological limitations of previous studies.MethodPatients are recruited from the University of Melbourne Academic Hearing Aids Audiology Clinic and assessed before and every 18 months after hearing aid fitting. When numbers are sufficient, results will be compared with those of a control group of approximately 400 participants of a cohort study of ageing in older adults. Assessments include cognitive function (visually presented computerised assessment battery), hearing, speech perception, quality of life, activity, diet, loneliness and isolation, anxiety, depression, medical health and genetic risk.ResultAt baseline, multiple linear regression for 98 participants (62‐82 years), controlling for age, gender, cardiovascular conditions, education, and working/retired, with mean better ear PTA 31dB showed hearing loss and age predicted significantly poorer executive function, while tertiary education predicted significantly better executive function and visual learning. At 18‐month follow up, results for a subset of participants showed significant improvements in speech perception, listening disability and quality of life. Group mean scores across the cognitive assessment battery showed no significant decline and executive function was significantly improved. There was either clinically significant improvement or stability for executive function for over 97% of the subset, and for females for working memory, visual learning and visual attention, despite significantly greater hearing loss over this period.ConclusionDespite a small initial sample size, hearing loss, age and education significantly predicted baseline cognitive function. At 18‐month follow up, the majority of participants had either remained stable or demonstrated a meaningful clinical improvement in cognitive function, with significant group mean improvements for executive function and working memory. Treatment of hearing loss with hearing aids may delay cognitive decline. These are initial results, and further follow up is required.

Improving Speech Quality for Hearing Aid Applications Based on Wiener Filter and Composite of Deep Denoising Autoencoders

In hearing aid devices, speech enhancement techniques are a critical component to enable users with hearing loss to attain improved speech quality under noisy conditions. Recently, the deep denoising autoencoder (DDAE) was adopted successfully for recovering the desired speech from noisy observations. However, a single DDAE cannot extract contextual information sufficiently due to the poor generalization in an unknown signal-to-noise ratio (SNR), the local minima, and the fact that the enhanced output shows some residual noise and some level of discontinuity. In this paper, we propose a hybrid approach for hearing aid applications based on two stages: (1) the Wiener filter, which attenuates the noise component and generates a clean speech signal; (2) a composite of three DDAEs with different window lengths, each of which is specialized for a specific enhancement task. Two typical high-frequency hearing loss audiograms were used to test the performance of the approach: Audiogram 1 = (0, 0, 0, 60, 80, 90) and Audiogram 2 = (0, 15, 30, 60, 80, 85). The hearing-aid speech perception index, the hearing-aid speech quality index, and the perceptual evaluation of speech quality were used to evaluate the performance. The experimental results show that the proposed method achieved significantly better results compared with the Wiener filter or a single deep denoising autoencoder alone.