Speech Intelligibility In Noisy Environments Research Articles

Effect of Digital Noise Reduction in Hearing Aids on Speech Intelligibility in Both Quiet and Noisy Environments.

Digital noise reduction (DNR) minimizes the effect of noise on speech signals by continuously monitoring frequency bands in the presence of noise. In the present study, we explored the effect of DNR technology on speech intelligibility in individuals using hearing aids (HAs) and investigated implications for daily use. Eighteen participants with bilateral moderate sensorineural hearing loss (aged 16-45 years) were included. Bilateral receiver-in-the-ear HAs were fitted in the participants. The adaptive and nonadaptive (with a signal-to-noise ratio (SNR) of +5 and -5 dB, respectively) Turkish matrix sentence test (TURMatrix) in noise and free-field hearing assessments, including hearing thresholds with hearing aids, speech recognition thresholds (SRT), and speech discrimination scores, were conducted in two different conditions: HA in the DNR-on and DNR-off conditions. No significant difference was observed between free-field hearing assessments with the HA in the DNR-off and DNR-on conditions (P > 0.05). Furthermore, the adaptive and nonadaptive TURMatrix revealed significant differences between the scores under the DNR-on and DNR-off conditions (P < 0.05). Nevertheless, under the DNR-on condition, there was no correlation between free-field hearing assessments with HA and TURMatrix results (P > 0.05). However, a significant correlation was observed between SRT scores with HA and TURMatrix scores (adaptive and nonadaptive, +5 and -5 dB SNR, respectively) under the DNR-off condition (P < 0.05). Our study findings suggest that DNR can improve speech intelligibility in noisy environments. Therefore, DNR can enhance an individual's auditory comfort by improving their capacity to grasp speech in background noise.

Effects of number of maxima and electrical dynamic range on speech-in-noise perception with an “n-of-m” cochlear-implant strategy

Cochlear implant (CI) recipients face great challenges in speech-in-noise recognition, partially due to the fact that only temporal envelopes from a limited number of bands are preserved in most CI signal processing strategies. In “n-of-m” strategies (e.g., the Advanced Combinational Encoder, ACE), the number of maxima (nmax) and electrical dynamic range (EDR) are two essential parameters that may affect the envelope representation and further influence speech perception. Speech recognition can be improved by optimizing parameter settings in CI programming. To investigate the effects of nmax and EDR on speech-in-noise perception, Mandarin speech reception thresholds (SRTs) in babble noise were measured in CI recipients using ACE. The nmax was set to 2, 4, 6, 8, and 16. The EDR was set to the base EDR (i.e., participants’ clinical EDR) and 50 % EDR (i.e., 50 %-compressed base EDR). Results showed that: 1) there was no significant interaction effect between nmax and EDR, 2) SRTs with nmax = 2, 4, and 16 were significantly higher (or worse) than those with moderate nmax (6–8), 3) narrower EDRs significantly lead to higher SRTs. Simulation experiments using a Gaussian-Enveloped Tones Vocoder in normal-hearing listeners were also conducted and provided both supportive and additional observations to the CI results. This study suggests that, in CI programming, nmax and EDR are two independent influencing factors. Large nmax (e.g., 16) is not recommended as it may harm speech intelligibility in noisy environments, and inaccurate EDR measurements should be avoided.

Audiovisual Speech Processing in Relationship to Phonological and Vocabulary Skills in First Graders.

It is generally accepted that adults use visual cues to improve speech intelligibility in noisy environments, but findings regarding visual speech benefit in children are mixed. We explored factors that contribute to audiovisual (AV) gain in young children's speech understanding. We examined whether there is an AV benefit to speech-in-noise recognition in children in first grade and if visual salience of phonemes influences their AV benefit. We explored if individual differences in AV speech enhancement could be explained by vocabulary knowledge, phonological awareness, or general psychophysical testing performance. Thirty-seven first graders completed online psychophysical experiments. We used an online single-interval, four-alternative forced-choice picture-pointing task with age-appropriate consonant-vowel-consonant words to measure auditory-only, visual-only, and AV word recognition in noise at -2 and -8 dB SNR. We obtained standard measures of vocabulary and phonological awareness and included a general psychophysical test to examine correlations with AV benefits. We observed a significant overall AV gain among children in first grade. This effect was mainly attributed to the benefit at -8 dB SNR, for visually distinct targets. Individual differences were not explained by any of the child variables. Boys showed lower auditory-only performances, leading to significantly larger AV gains. This study shows AV benefit, of distinctive visual cues, to word recognition in challenging noisy conditions in first graders. The cognitive and linguistic constraints of the task may have minimized the impact of individual differences of vocabulary and phonological awareness on AV benefit. The gender difference should be studied on a larger sample and age range.

Effects of Wearing Face Masks While Using Different Speaking Styles in Noise on Speech Intelligibility During the COVID-19 Pandemic.

The coronavirus pandemic has resulted in the recommended/required use of face masks in public. The use of a face mask compromises communication, especially in the presence of competing noise. It is crucial to measure the potential effects of wearing face masks on speech intelligibility in noisy environments where excessive background noise can create communication challenges. The effects of wearing transparent face masks and using clear speech to facilitate better verbal communication were evaluated in this study. We evaluated listener word identification scores in the following four conditions: (1) type of mask condition (i.e., no mask, transparent mask, and disposable face mask), (2) presentation mode (i.e., auditory only and audiovisual), (3) speaking style (i.e., conversational speech and clear speech), and (4) with two types of background noise (i.e., speech shaped noise and four-talker babble at −5 signal-to-noise ratio). Results indicate that in the presence of noise, listeners performed less well when the speaker wore a disposable face mask or a transparent mask compared to wearing no mask. Listeners correctly identified more words in the audiovisual presentation when listening to clear speech. Results indicate the combination of face masks and the presence of background noise negatively impact speech intelligibility for listeners. Transparent masks facilitate the ability to understand target sentences by providing visual information. Use of clear speech was shown to alleviate challenging communication situations including compensating for a lack of visual cues and reduced acoustic signals.

The effect of induction loop system use training on adults with hearing aids

The study aimed to investigate the effect of training on hearing aid users to benefit from induction loop systems. A five-question scale was developed to evaluate whether individuals using hearing aids could use the induction loop system effectively. In the first step, validity-reliability studies were performed with 264 individuals using hearing aids. In the second step, 30 individuals using hearing aids were given verbal and hands-on training on the induction loop system. Before and after training with hearing aids (noiseless, noise, noise + induction system active) in three different environments, questions on the scale were asked twice in total from the beginning to the end of the study. The significance of the differences between the values obtained as a result of the application of word lists in three different test settings was examined by repeating the measurements variance analysis. As a result of the post hoc analysis, P = 0.002 between test 3 (10.7, 1.53) and test 1 (11.7, 0.7) was calculated. There appears to be a statistically significant difference with the present situation (P < 0.01). The average scores of the scale between pre and posttraining applications as a preliminary and final test were analysed with a t-test. The final test average was statistically significantly greater than the preliminary test (P < 0.01). This study shows how important it is for hearing aid users to be informed and to receive the necessary training in order to gain the expected benefit from induction loop systems developed to improve speech intelligibility in noisy environments.

Short report on the effects of SARS-CoV-2 face protective equipment on verbal communication.

ObjectiveTo predict the impact of face personal protective equipment on verbal communication during the SARS-CoV-2 pandemic.DesignWe assessed the effect of common types and combinations of face personal protective equipment on speech intelligibility in quiet and in a simulated noisy environment.ResultsWearing face personal protective equipment impairs transmission of middle-to-high voice frequencies and affects speech intelligibility. Surgical masks are responsible for up to 23.3% loss of speech intelligibility in noisy environments. The effects are larger in the condition of advanced face personal protective equipment, accounting for up to 69.0% reduction of speech intelligibility.ConclusionThe use of face personal protective equipment causes significant verbal communication issues. Healthcare workers, school-aged children, and people affected by voice and hearing disorders may represent specific at-risk groups for impaired speech intelligibility.

Speech-in-Noise Test results of compensation claimants for noise induced hearing loss in Korean male workers: Words-in-Noise Test (WIN) and quick-Hearing-in-Noise Test (HINT).

BackgroundPure-tone audiometry is used as a gold standard for hearing measurement. However, since communication in the work environment occurs in noise, it might be difficult to evaluate the actual communication ability accurately based on pure-tone audiometry only. Therefore, the purpose of this study is to evaluate speech intelligibility in noisy environments by using Speech-in-Noise Tests and to check its relationship with pure-tone audiometry.MethodsFrom January 2017 to September 2018, for 362 workers who visited a university hospital for the purpose of compensating for noise-induced hearing loss, several tests were conducted: pure-tone audiometry, speech reception threshold, speech discrimination score, and Speech-in-Noise Tests (Words-in-Noise Test [WIN] and quick-Hearing-in-Noise Test [quick-HINT]). The subjects were classified into serviceable hearing group and non-serviceable hearing group based on 40 dB hearing level (HL) pure-tone average. In both groups, we conducted age-adjusted partial correlation analysis in order to find out the relationship between pure-tone threshold, speech reception threshold, speech discrimination score and WIN and quick-HINT respectively.ResultsIn non-serviceable hearing group, all results of partial correlation analysis were statistically significant. However, in serviceable hearing group, there were many results which showed little or no significant relationship between pure-tone threshold and Speech-in-Noise Tests (WIN and quick-HINT).ConclusionsThe relationship between Speech-in-Noise Tests and the pure-tone thresholds were different by the hearing impairment levels; in mild to moderate hearing loss workers, there was little or no relationship; in severe cases, the relationship was significant. It is not enough to predict the speech intelligibility of hearing-impaired persons, especially in mild to moderate level, with pure-tone audiometry only. Therefore, it would be recommended to conduct Speech-in-Noise Test.

Frequency-lowering processing to improve speech-in-noise intelligibility in patients with age-related hearing loss.

Modern hearing aids use various signal-processing strategies to improve speech intelligibility. In this manuscript, we studied the linear frequency transposition (LFT), a frequency-lowering algorithm, in patients with age-related hearing loss. Frequency-lowering algorithms transpose high-frequency sounds to a lower-frequency band. The study aimed to assess whether LFT could be used as a tool to improve speech intelligibility in patients with a better high-frequency preservation. The study population consisted of 77 patients (age average 74.8 ± 12.4) wearing hearing aids with an open or tulip coupling system with age-related hearing loss. The unaided air conduction pure tone average (PTA) at 500, 1000, 2000Hz was 43.5 ± 14.3dB; the unaided word recognition score (WRS) average was 53.7 ± 12.5%. We compared WRS in all patients with the hearing aid turned on, in "quiet" and using a "pink" and "babble" masking noise. Three hearing aid settings were tested in each acoustic conditions: no transposition (NT), high transposition (HT), and low transposition (LT). "High" and "low" refer to the "start frequency"; all sounds above the start frequency are transposed in a lower-frequency band. When the start frequency was suggested by the fitting software, we called the condition "high transposition"; when the start frequency was set at the lowest possible value provided by the fitting software, we called the condition "low transposition". The quality of the voice was also assessed asking the patient to give a score from 1 to 10, where 10 was the maximum listening comfort [quality of voice score (QVS)]. Collected data were compared for each condition (NT, HT, LT, in quiet, pink noise and babble noise) and no statistically significant differences were found in WRS and QVS (quiet WRS p = 0.07, pink noise WRS p = 0.18, babble noise WRS p = 0.11, QVS p = 0.91). We selected 33 patients with a better WRS in babble noise using transposition (high and low). In this group, the age was significantly lower than patients who did not use transposition (p = 0.01). Linear frequency transposition is not useful to improve speech-in-noise intelligibility in patients with age-related hearing loss. Despite that no statistically significant differences were found, younger people could get advantages from the LFT when babble noise disturbs the listening of speech. The use of FL algorithm as a way to improve speech intelligibility in noisy environments should be always considered.

Horizontal sound localisation and speech perception in Bonebridge-implanted single-sided deafness patients.

This study aimed to investigate the benefit of Bonebridge devices in patients with single-sided deafness. Five patients with single-sided deafness who were implanted with Bonebridge devices were recruited in a single-centre study. Participants' speech perception and horizontal sound localisation abilities were assessed at 6 and 12 months post-operatively. Speech intelligibility in noisy environments was measured in three different testing conditions (speech and noise presented from the front, speech and noise presented from the front and contralateral (normal ear) side separately, and speech presented from the ipsilateral (implanted Bonebridge) side and noise from the contralateral side). Sound localisation was evaluated in Bonebridge-aided and Bonebridge-unaided conditions at different stimuli levels (65, 70 and 75 dB SPL). All participants showed a better capacity for speech intelligibility in quiet environments with the Bonebridge device. The speech recognition threshold with the Bonebridge device was significantly decreased at both short- and long-term follow up in the speech presented from the ipsilateral (implanted Bonebridge) side and noise from the contralateral side condition (p < 0.05). Additionally, participants maintained similar levels of sound localisation between the Bonebridge-aided and unaided conditions (p > 0.05). However, the accuracy of localisation showed some improvement at 70 dB SPL and 75 dB SPL post-operatively. The Bonebridge device provides the benefit of improved speech perception performance in patients with single-sided deafness. Sound localisation abilities were neither improved nor worsened with Bonebridge implantation at the follow-up assessments.

Acoustic characteristics in relation to intelligibility reduction in noise for speakers with Parkinson’s disease

ABSTRACT Decreased speech intelligibility in noisy environments is frequently observed in speakers with Parkinson’s disease (PD). This study investigated which acoustic characteristics across the speech subsystems contributed to poor intelligibility in noise for speakers with PD. Speech samples were obtained from 13 speakers with PD and five healthy controls reading 56 sentences. Intelligibility analysis was conducted in quiet and noisy listening conditions. Seventy-two young listeners transcribed the recorded sentences in quiet and another 72 listeners transcribed in noise. The acoustic characteristics of the speakers with PD who experienced large intelligibility reduction from quiet to noise were compared to those with smaller intelligibility reduction in noise and healthy controls. The acoustic measures in the study included second formant transitions, cepstral and spectral measures of voice (cepstral peak prominence and low/high spectral ratio), pitch variation, and articulation rate to represent speech components across speech subsystems of articulation, phonation, and prosody. The results show that speakers with PD who had larger intelligibility reduction in noise exhibited decreased second formant transition, limited cepstral and spectral variations, and faster articulation rate. These findings suggest that the adverse effect of noise on speech intelligibility in PD is related to speech changes in the articulatory and phonatory systems.

A Personal Sound Amplification Product Compared to a Basic Hearing Aid for Speech Intelligibility in Adults with Mild-to-Moderate Sensorineural Hearing Loss.

Background and ObjectivesThis study aimed to compare functional hearing with the use of a personal sound amplification product (PSAP) or a basic hearing aid (HA) among sensorineural hearing impaired listeners.Subjects and MethodsNineteen participants with mild-to-moderate sensorineural hearing loss (SNHL) (26-55 dB HL; pure-tone average, 0.5-4 kHz) were prospectively included. No participants had prior experience with HAs or PSAPs. Audiograms, speech intelligibility in both quiet and noisy environments, speech quality, and preference were assessed in three different listening conditions: unaided, with the HA, and with the PSAP.ResultsThe use of PSAP was associated with significant improvement in pure-tone thresholds at 1, 2, and 4 kHz compared to the unaided condition (all p<0.01). In the quiet environment, speech intelligibility was significantly improved after wearing a PSAP compared to the unaided condition (p<0.001), and this improvement was better than the result obtained with the HA. The PSAP also demonstrated similar improvement in the most comfortable levels compared to those obtained with the HA (p<0.05). However, there was no significant improvement of speech intelligibility in a noisy environment when wearing the PSAP (p=0.160). There was no significant difference in the reported speech quality produced by either device or in participant preference for the PSAP or HA.ConclusionsThe current result suggests that PSAPs provide considerable benefits to speech intelligibility in a quiet environment and can be a good alternative to compensate for mild-to-moderate SNHL.

A Wearable, Extensible, Open-Source Platform for Hearing Healthcare Research.

Hearing loss is one of the most common conditions affecting older adults worldwide. Frequent complaints from the users of modern hearing aids include poor speech intelligibility in noisy environments and high cost, among other issues. However, the signal processing and audiological research needed to address these problems has long been hampered by proprietary development systems, underpowered embedded processors, and the difficulty of performing tests in real-world acoustical environments. To facilitate existing research in hearing healthcare and enable new investigations beyond what is currently possible, we have developed a modern, open-source hearing research platform, Open Speech Platform (OSP). This paper presents the system design of the complete OSP wearable platform, from hardware through firmware and software to user applications. The platform provides a complete suite of basic and advanced hearing aid features which can be adapted by researchers. It serves web apps directly from a hotspot on the wearable hardware, enabling users and researchers to control the system in real time. In addition, it can simultaneously acquire high-quality electroencephalography (EEG) or other electrophysiological signals closely synchronized to the audio. All of these features are provided in a wearable form factor with enough battery life for hours of operation in the field.

Evaluation of a near-end listening enhancement algorithm by combined speech intelligibility and listening effort measurements.

Previous studies showed that near-end listening enhancement (NELE) algorithms can significantly improve speech intelligibility in noisy environments. This study investigates the benefit of the NELE algorithm AdaptDRC in normal-hearing listeners at signal-to-noise ratios (SNRs) for which speech intelligibility is at ceiling, by evaluating listening effort for processed and unprocessed speech in the presence of speech-shaped and cafeteria noise. The results suggest that the NELE algorithm is able to reduce listening effort over a wide range of SNRs. Hence, listening effort seems to be applicable for evaluating NELE algorithms over a much wider SNR range than speech intelligibility.

A deep learning based segregation algorithm to increase speech intelligibility for hearing-impaired listeners in reverberant-noisy conditions.

Recently, deep learning based speech segregation has been shown to improve human speech intelligibility in noisy environments. However, one important factor not yet considered is room reverberation, which characterizes typical daily environments. The combination of reverberation and background noise can severely degrade speech intelligibility for hearing-impaired (HI) listeners. In the current study, a deep learning based time-frequency masking algorithm was proposed to address both room reverberation and background noise. Specifically, a deep neural network was trained to estimate the ideal ratio mask, where anechoic-clean speech was considered as the desired signal. Intelligibility testing was conducted under reverberant-noisy conditions with reverberation time T 60 = 0.6 s, plus speech-shaped noise or babble noise at various signal-to-noise ratios. The experiments demonstrated that substantial speech intelligibility improvements were obtained for HI listeners. The algorithm was also somewhat beneficial for normal-hearing (NH) listeners. In addition, sentence intelligibility scores for HI listeners with algorithm processing approached or matched those of young-adult NH listeners without processing. The current study represents a step toward deploying deep learning algorithms to help the speech understanding of HI listeners in everyday conditions.

New Method Boosts Speech Intelligibility in Noisy Environments

New Method Boosts Speech Intelligibility in Noisy Environments

Speech Noise Reduction Algorithm in Digital Hearing Aids Based on an Improved Sub-band SNR Estimation

To improve the speech intelligibility in noisy environments for persons with hearing impairments, a new method for reducing noise, based on improved sub-band signal-to-noise ratio (SNR) estimation, is proposed. First, the input signal is decomposed into several sub-band signals with an analysis filter bank. Then, under the assumption of a Gaussian model, maximum a posterior probability is applied to estimate the information embedded in adjacent frames in each sub-band, which is in the form of a joint probability density function, and the minimum of the noise spectrum is tracked to estimate the noise. Subsequently, the gain of each sub-band, which changes with the noise in the corresponding sub-band, is calculated with a linear proportional gain function. The obtained gains of the sub-bands are multiplied by the sub-band noisy signals to obtain the enhanced sub-band speech signals. Finally, all the sub-band signals are spliced to obtain the estimated speech signals. In this algorithm, the gains are calculated in the time domain, which avoids the process of the inverse Fourier transform and leads to a decrease in computational complexity. Compared with the traditional spectral subtraction and basic Wiener filtering method, the delay in this algorithm is reduced by 40.4 and 60.6%, respectively. It is also compared with the modulation depth integrated into hearing aids under an experimental simulation and a real scenario. The results indicate that the output SNR is improved by 1 dB under the software simulation and 3.1 dB in the real scenario when the input SNR is set as 10 dB. Compared with the simulation environment, the proposed algorithm only fell by 1.5% in the real scenario. Furthermore, the distance of the logarithmic spectrum and quality of speech perception are improved by 20.6 and 9.3%, respectively.

Our auditory results using the Vibrant Soundbridge on the long process of the incus: 20 years of data

ObjectiveAfter 20 years of experience with different types of middle ear implants, we analyzed our database about the Vibrant Soundbridge (VSB) to know the rate of complications, the effect on the residual hearing and the audiometric gain in our center. MethodThe study was retrospective and included all VSB implants bound to the long process of the incus in our tertiary medical center between january 1999 and february 2015. We observed the effect of surgery on residual hearing by comparing bone and air conduction thresholds before and after implantation. The functional results of the implant were quantified by measuring, at several post-operative intervals, the thresholds with the VSB in pure tone audiometry and speech audiometry, in quiet and in noise. Results53 VSB were implanted in 46 patients aged between 22 and 81 years old (average 53.9). 48 patients (90%) suffered from a sensorineural hearing loss, and 5 patients from a mixed hearing loss due to an otosclerosis (but only 3 of them have undergone stapedotomy). There were no major complications (e.g. facial palsy, dead ear or postoperative infection). The placement of the implant created an insignificant deterioration of the air conduction thresholds (5,6 dB HL) and bone conduction thresholds (2.2 dB HL) at 6 weeks post-implantation. The bone conduction thresholds increased by 4.7 dB HL 2.5 years after surgery in comparison with the preoperative results, which is also considered clinically insignificant. With the implant turned on, the pure tone audiometry thresholds in open field, in quiet, were significantly improved (gain of 13.9 dB on average on frequencies from 250 to 8000 Hz), particularly at frequencies of 1000, 2000 and 4000 Hz as the average gain on these frequencies amounted to 19.4 dB. The speech intelligibility in a cocktail party noise was also improved by 18.3% on average at 6, 52 and 104 weeks post-implantation. ConclusionThe Vibrant Soundbridge with the electromagnetic vibrator fixed to the long process of the incus is a safe active middle ear implant with no major complications; it has no significant impact on the residual hearing. The VSB is particularly suitable for patients who are unable to wear conventional hearing aids due to anatomical or infectious problems in the external ear canal, or in case of poor audiometric results with conventional hearing aids. The VSB brings significant hearing gain, as it is particularly efficient in frequencies for the speech range and higher, resulting especially in better speech intelligibility in noisy environments.

Speech intelligibility improvement in noisy environments based on energy correlation in frequency bands

A new speech processing algorithm is proposed to improve speech intelligibility in noisy environments without increasing speech energy. The method improves the near-end speech intelligibility by optimizing the frame-based spectral energy correlation between clean speech and noisy modified speech with a power constraint. This algorithm is developed based on short-time objective intelligibility (STOI) measure, which predicts speech intelligibility in background noise according to the correlation of clean and noisy speeches. The proposed method is compared with unprocessed speech and two baseline methods using two objective intelligibility measures and an intelligibility listening test under various noisy conditions. Results show large intelligibility improvements with the proposed method over the unprocessed noisy speech. In addition, compared with the baseline methods, the proposed algorithm provides the best intelligibility scores for all noisy conditions in the STOI measure and for low signal-to-noise ratios in the speech intelligibility index. The word recognition results also show that the proposed algorithm performs better than the unprocessed speech and the reference methods. An objective quality measure is applied to investigate the speech quality of the introduced method, which is proven to insignificantly affect speech quality.

Sound coding in the auditory nerve of gerbils

Gerbils possess a very specialized cochlea in which the low-frequency inner hair cells (IHCs) are contacted by auditory nerve fibers (ANFs) having a high spontaneous rate (SR), whereas high frequency IHCs are innervated by ANFs with a greater SR-based diversity. This specificity makes this animal a unique model to investigate, in the same cochlea, the functional role of different pools of ANFs. The distribution of the characteristic frequencies of fibers shows a clear bimodal shape (with a first mode around 1.5 kHz and a second around 12 kHz) and a notch in the histogram near 3.5 kHz. Whereas the mean thresholds did not significantly differ in the two frequency regions, the shape of the rate-intensity functions does vary significantly with the fiber characteristic frequency. Above 3.5 kHz, the sound-driven rate is greater and the slope of the rate-intensity function is steeper. Interestingly, high-SR fibers show a very good synchronized onset response in quiet (small first-spike latency jitter) but a weak response under noisy conditions. The low-SR fibers exhibit the opposite behavior, with poor onset synchronization in quiet but a robust response in noise. Finally, the greater vulnerability of low-SR fibers to various injuries including noise- and age-related hearing loss is discussed with regard to patients with poor speech intelligibility in noisy environments. Together, these results emphasize the need to perform relevant clinical tests to probe the distribution of ANFs in humans, and develop appropriate techniques of rehabilitation.This article is part of a Special Issue entitled <Annual Reviews 2016>.

Binary mask estimation for noise reduction based on instantaneous SNR estimation using Bayes risk minimisation

The binary mask approach has been researched to suppress noise and improve speech intelligibility in noisy environments. An algorithm that estimates the binary mask for noise-corrupted speech based on the instantaneous signal-to-noise ratio (SNR) estimation is proposed. The instantaneous SNR estimation is performed by minimising the Bayes risk with a weighted cost function. In the experiments, white noise was used for the training of the SNR estimator and the binary mask estimation was performed for babble, factory, speech-shaped noise. The experimental results show that the proposed method yields substantial improvements in terms of classification accuracy for the binary mask estimation.