Deep Denoising for Hearing Aid Applications

Hearing aids are the primary method to manage hearing loss. However, there are limited recommendations for when and how to set advanced hearing aid features. The purpose of this study is to describe how hearing aid features are utilized in clinically fit devices and to evaluate the relationship between the fitted hearing aid feature and the Quick Speech-in-Noise Test (QuickSIN). Data from two laboratories were evaluated retrospectively, resulting in 107 bilateral hearing aid participants who obtained their hearing aids at clinics in their communities. Ages ranged from 60 to 93 years. Degree of speech-in-noise difficulty was evaluated using the QuickSIN (mild, moderate, or severe). Settings for directionality, digital noise reduction (DNR), and hearing assistive technology (HAT) use were documented. Directionality was categorized as omnidirectional, fixed (full-time directional), or adaptive (adjusts automatically based on noise source). DNR was recorded as either on or off. HAT use was recorded as either yes or no. QuickSIN scores ranged from -1.5 to 25 dB SNR loss (M = 7). A moderate correlation was determined for QuickSIN scores and pure-tone averages. Adaptive directionality was used most often, most participants had DNR turned on, and HAT use was low. The biggest contributions to the Chi-square test for directionality and degrees of speech-in-noise difficulty together were fixed/severe, fixed/moderate, and adaptive/mild. In this clinical sample, there was limited HAT use and advanced features are not set in a way that is consistent with speech-in-noise abilities. It is likely that patients fit with noise management that is not suited to their listening abilities are experiencing increased difficulties in challenging listening environments that could potentially be mitigated with alternative management. Evidence-based research on prefitting measures of speech in noise to help inform patient-centered clinical decisions is needed.

There is no doubt that the hearing healthcare profession is at the dawn of an exciting new era in the advancement of hearing aid technology. Within the last few years there has been a major shift from dispensing linear hearing aid technology into a new era of so many unique hearing aid circuit choices. Today there are so many choices, from improved and more sophisticated linear circuits, i.e., class D amplifiers, to almost completely adjustable nonlinear analog and most recently fully digital circuits! Keeping current with today's hearing aid technology challenges even the most dedicated hearing healthcare professional. In fact, the idea of keeping current is what best summarizes the primary focus of this manuscript. At Frye Electronics, we are continually striving to find better, easier and more helpful methods of evaluating the electroacoustic performance of a hearing aid. An example of this commitment is the experimental “digital speech” software approach for evaluating the performance of the new digital signal processing hearing aids that we demonstrated at this year's American Academy of Audiology convention. We believe that the new Frye/SPL fitting test will help dispensers keep up with today's new technology by providing hearing healthcare professionals with an important next step that is both a practical and easy-to-use fitting approach for utilization with all types of hearing aids.

Speech intelligibility is usually assessed using subjective tests, but several objective measures of speech intelligibility have also been proposed. This study compares intelligibility of IEEE sentences in collocated speech-shaped noise, recorded through a low-cost amplifier and three current-issue NHS hearing aids with single-channel noise reduction settings switched on and off. Results from twenty-one normal hearing listeners indicates that single-channel noise reduction algorithms can significantly improve intelligibility of noisy speech (p < 0.05), but no differences can be seen between different hearing aid models currently available on the NHS. The low-cost, off-the shelf hearing amplifier performed significantly worse than the NHS prescribed devices. Three objective speech intelligibility measures were applied to the recordings. None of the objective metrics was found to accurately predict the outcomes seen in this study for all hearing aid conditions, though general trends can be predicted (r = 0.83 overall, but for any discrete signal-to-noise ratio across all conditions, maximum r = 0.64). Machine-learning approaches were used to refit parameters from the best performing of these metrics. Although performance was improved, problems in predicting particular conditions remain, suggesting that objective measures do not encompass all relevant factors. [Action on Hearing Loss (England/Wales Registered Charity Number 207720).]

One of the most frequent complaints of hearing-impaired individuals is difficulty understanding speech in noisy environments. In order to improve listening comforts in background noise, digital noise reduction (DNR) has been incorporated into the hearing aid. However, each hearing aid manufacturer has their own proprietary algorithm for their DNR. Therefore, this study aimed to evaluate the efficacy of the DNR system in commercial hearing aids using the Malay speech-in-noise test. Electroacoustic testing has been performed on six hearing aids that have similar technology but different brands using newly developed Malay speech-in-noise test signals in the condition of DNR on and off. The total root-mean-square (RMS) gain reduction for each HA was calculated. The results show that the type of noise, the signal-to-noise ratio (SNR) and the gender of the speaker have a significant effect on the amount of gain reduction shown by the DNR of the tested hearing aids (p<0.05). In conclusion, clinicians need to measure and understand methods to verify the functionality of DNR before fitting a patient's hearing aid to ensure its effectiveness in real life.

Background and Aim: Background noise as a serious challenge mainly affects the speech perception in people with hearing loss. One of the methods used to control noise is digital noise reduction (DNR) technique. The present study aimed to investigate the effect of DNR program (activated and deactivated) on the frequency gain of basic and advanced behind-the-ear (BTE) Oticon hearing aids, using different DNR strategies to reduce background noise. Methods: Two behind-the-ear Oticon hearing aids (Opn1 S105 and GetP) were used in this study. The Affinity 2.0 test box was first used to measure their DNR (off/on) gains using the national acoustic laboratories-non linear2 (NAL-NL2) and desired sensation level multi-stage [input/output] (DSLm[I/O]) formulas at sound pressure levels of 45, 65 and 85 dB SPL at a frequency range of 250-8000 Hz for three hearing loss (HL) patterns using the international speech test signal and broad band noise. Results: There was a significant difference in DNR performance between the Opn1 S105 and GetP models for all three HL patterns at 45 and 65 dB SPL and most frequencies. Conclusion: The DNR performance of advanced and basic hearing aids is different for different HLs at 45 and 65 dB SPL and most frequencies. The performance of advanced hearing aids is significant using the DSLm[I/O] formula at most frequencies. Keywords: Digital noise reduction; international speech test signal; frequency response; broad band noise

Binaural noise reduction, with applications for instance in hearing aids, has been a very significant challenge. This task relates to the optimal utilization of the available microphone signals for the estimation of the ambient noise characteristics and for the optimal filtering algorithm to separate the desired speech from the noise. The additional requirements of low computational complexity and low latency further complicate the design. A particular challenge results from the desired reconstruction of binaural speech input with spatial cue preservation. The latter essentially diminishes the utility of multiple-input/single-output filter-and-sum techniques such as beamforming. In this paper, we propose a comprehensive and effective signal processing configuration with which most of the aforementioned criteria can be met suitably. This relates especially to the requirement of efficient online adaptive processing for noise estimation and optimal filtering while preserving the binaural cues. Regarding noise estimation, we consider three different architectures: interaural (ITF), cross-relation (CR), and principal-component (PCA) target blocking. An objective comparison with two other noise PSD estimation algorithms demonstrates the superiority of the blocking-based noise estimators, especially the CR-based and ITF-based blocking architectures. Moreover, we present a new noise reduction filter based on minimum mean-square error (MMSE), which belongs to the class of common gain filters, hence being rigorous in terms of spatial cue preservation but also efficient and competitive for the acoustic noise reduction task. A formal real-time subjective listening test procedure is also developed in this paper. The proposed listening test enables a real-time assessment of the proposed computationally efficient noise reduction algorithms in a realistic acoustic environment, e.g., considering time-varying room impulse responses and the Lombard effect. The listening test outcome reveals that the signals processed by the blocking-based algorithms are significantly preferred over the noisy signal in terms of instantaneous noise attenuation. Furthermore, the listening test data analysis confirms the conclusions drawn based on the objective evaluation.

Current standardized methods for the measurement of hearing‐aid performance (Specification of Hearing Aid Characteristics, S3.22‐1976, American National Standards Institute, New York) call for the measurement of two special characteristics of compression hearing aids, namely the dynamic and input‐output characteristics, at 2 kHz. A more thorough characterization of compression hearing aids requires measurements at more than one volume‐control setting to determine if the compression is controlled by the input or output signal. Also useful are random‐noise transfer functions and transmissibility measurements to determine the response to multifrequency signals, input‐output characteristics at several frequencies to determine the compression behavior in various frequency bands, and correlation coefficients to examine the influence of nonlinearities and internal noise in the hearing aid upon its performance. Measurements of these parameters are shown at two volume‐control settings for several commercially available hearing aids and their general applicability is discussed. [Work supported by the Veterans Administration.]

A directional hearing aid comprises a front microphone ( 11 ) and a back microphone ( 12 ), a delay processor ( 13, 14 ) for processing, according to a control parameter, the respective microphone signals, and means ( 8 a ) for adjusting the control parameter in order to minimize the output signal from the delay processor. The control parameter may be adjusted to change smoothly the function mode of a hearing aid between omnidirectional mode, a directional mode and a directional mode with a pair of null directions, symmetrical about the 180° direction. The directional controller may be implemented in a multichannel version. The invention provides a hearing aid, a method of controlling a hearing aid, a noise reduction system and a method of reducing noise in a hearing aid.

We completed a registered double-blind randomized control trial to compare acclimatization to two hearing aid fitting algorithms by experienced pediatric hearing aid users with mild to moderate hearing loss. We hypothesized that extended use (up to 13 months) of an adaptive algorithm with integrated directionality and noise reduction, OpenSound Navigator (OSN), would result in improved performance on auditory, cognitive, academic, and caregiver- or self-report measures compared with a control, omnidirectional algorithm (OMNI). Forty children aged 6 to 13 years with mild to moderate/severe symmetric sensorineural hearing loss completed this study. They were all experienced hearing aid users and were recruited through the Cincinnati Children's Hospital Medical Center Division of Audiology. The children were divided into 20 pairs based on similarity of age (within 1 year) and hearing loss (level and configuration). Individuals from each pair were randomly assigned to either an OSN (experimental) or OMNI (control) fitting algorithm group. Each child completed an audiology evaluation, hearing aid fitting using physically identical Oticon OPN hearing aids, follow-up audiological appointment, and 2 research visits up to 13 months apart. Research visit outcome measures covered speech perception (in quiet and in noise), novel grammar and word learning, cognition, academic ability, and caregiver report of listening behaviors. Analysis of outcome differences between visits, groups, ages, conditions and their interactions used linear mixed models. Between 22 and 39 children provided useable data for each task. Children using the experimental (OSN) algorithm did not show any significant performance differences on the outcome measures compared with those using the control (OMNI) algorithm. Overall performance of all children in the study increased across the duration of the trial on word repetition in noise, sentence repetition in quiet, and caregivers' assessment of hearing ability. There was a significant negative relationship between age at first hearing aid use, final Reading and Mathematical ability, and caregiver rated speech hearing. A significant positive relationship was found between daily hearing aid use and study-long change in performance on the Flanker test of inhibitory control and attention. Logged daily use of hearing aids related to caregiver rated spatial hearing. All results controlled for age at testing/evaluation and false discovery rate. Use of the experimental (OSN) algorithm neither enhanced nor reduced performance on auditory, cognitive, academic or caregiver report measures compared with the control (OMNI) algorithm. However, prolonged hearing aid use led to benefits in hearing, academic skills, attention, and caregiver evaluation.

In this study, two types of hearing aids were used. Both aids had the same frequency characteristics for frontal sound, but one employed an omnidirectional microphone and the other a directional microphone. The frequency characteristics of both hearing aids were measured for five azimuths on KEMAR and in situ in 12 normal-hearing subjects. For these subjects we also determined the speech reception threshold (SRT) with background noise in two rooms with different reverberation times. The direction of the speech stimuli was always frontal; the direction of the noise was varied. Additionally, directional hearing was measured with short noise bursts from eight loudspeakers surrounding the subject. In the less reverberant room, sounds coming from behind were less amplified by the hearing aid with the directional microphone than by the one with the omnidirectional microphone. In this room the monaural SRT values were largely determined by the level of the background noise. For the directional hearing aids there was an extra binaural advantage which depended on the direction of the background noise. Only for low-frequency noise bursts was directional hearing better with directional hearing aids. In the more reverberant room, no distinct differences between the frequency characteristics of the two hearing aid types were measured. However, a systematic difference between monaural SRT values measured through the two hearing aids was found. This difference was independent of noise azimuth. In conclusion, hearing aid(s) with a directional microphone showed no disadvantages and clear advantages under specific conditions.

Background and Aim: The goal of the present study was to objectively quantify the changes in acoustic parameters of the speech signal by comparing the unprocessed and hearing aid-processed speech stimuli. Materials and Methods: Eight hearing aids from two manufacturing companies were programmed for four different conditions: (a) noise reduction “OFF” and directionality “OFF;” (b) noise reduction “ON” and directionality “OFF;” (c) noise reduction “OFF” and directionality “ON;” and (d) noise reduction “ON” and directionality “ON.” One consonant-vowel-consonant-vowel word from the low-frequency, midfrequency, and high-frequency range each was used as stimuli. A sound level meter was used to record the speech stimuli output from the Knowles Electronics Manikin for Acoustic Research routed through the hearing aid (processed) and without the hearing aid (unprocessed). The recorded stimuli were further acoustically analyzed using PRAAT software. Results: The mean pitch, jitter, shimmer, and noise-to-harmonic ratio values differed between the unprocessed and hearing aid-processed speech stimuli, especially for the low-frequency and midfrequency stimuli. Whereas, for the high-frequency stimulus, PRAAT software failed to provide any value for the above-mentioned acoustic features for hearing aid-processed stimulus. Discussions: The processed speech had perturbations and more noise irrespective of the frequency of the speech stimulus. Hence, the outcome of the current study highlights the necessity to assess hearing aid-processed acoustic stimuli to check for the hearing aid’s performance.

Wide dynamic range compression (WDRC) and noise reduction both play important roles in hearing aids. WDRC provides level-dependent amplification so that the level of sound produced by the hearing aid falls between the hearing threshold and the highest comfortable level of the listener, while noise reduction reduces ambient noise with the goal of improving intelligibility and listening comfort and reducing effort. In most current hearing aids, noise reduction and WDRC are implemented sequentially, but this may lead to distortion of the amplitude modulation patterns of both the speech and the noise. This paper describes a deep learning method, called Neural-WDRC, for implementing both noise reduction and WDRC, employing a two-stage low-complexity network. The network initially estimates the noise alone and the speech alone. Fast-acting compression is applied to the estimated speech and slow-acting compression to the estimated noise, but with a controllable residual noise level to help the user to perceive natural environmental sounds. Neural-WDRC is frame-based, and the output of the current frame is determined only by the current and preceding frames. Neural-WDRC was compared with conventional slow- and fast-acting compression and with signal-to-noise ratio (SNR)-aware compression using objective measures and listening tests based on normal-hearing participants listening to signals processed to simulate the effects of hearing loss and hearing-impaired participants. The objective measures demonstrated that Neural-WDRC effectively reduced negative interactions of speech and noise in highly non-stationary noise scenarios. The listening tests showed that Neural-WDRC was preferred over the other compression methods for speech in non-stationary noises.

It is highly desirable that speech enhancement algorithms can achieve good performance while keeping low latency for many applications, such as digital hearing aids, mobile phones, acoustically transparent hearing devices, and public address systems. To improve the performance of traditional low-latency speech enhancement algorithms, a deep filter-bank equalizer (FBE) framework was proposed that integrated a deep learning-based subband noise reduction network with a deep learning-based shortened digital filter mapping network. In the first network, a deep learning model was trained with a controllable small frame shift to satisfy the low-latency demand, i.e., no greater than 4 ms, so as to obtain (complex) subband gains that could be regarded as an adaptive digital filter in each frame. In the second network, to reduce the latency, this adaptive digital filter was implicitly shortened by a deep learning-based framework and was then applied to noisy speech to reconstruct the enhanced speech without the overlap-add method. Experimental results on the WSJ0-SI84 corpus indicated that the proposed DeepFBE with only 4-ms latency achieved much better performance than traditional low-latency speech enhancement algorithms across several objective metrics. Listening test results further confirmed that our approach achieved higher speech quality than other methods.

Binaural hearing aids (HAs) can potentially perform advanced noise reduction algorithms, leading to an improvement over monaural/bilateral HAs. Due to the limited transmission capacities between the HAs and given knowledge of the complete joint noisy signal statistics, the optimal rate-constrained beamforming strategy is known from the literature. However, as these joint statistics are unknown in practice, sub-optimal strategies have been presented. In this paper, we present a unified framework to study the performance of these existing optimal and sub-optimal rate-constrained beamforming methods for binaural HAs. Moreover, we propose to use an asymmetric sequential coding scheme to estimate the joint statistics between the microphones in the two HAs. We show that under certain assumptions, this leads to sub-optimal performance in one HA but allows to obtain the truly optimal performance in the second HA. Based on the mean square error distortion measure, we evaluate the performance improvement between monaural beamforming (no communication) and the proposed scheme, as well as the optimal and the existing sub-optimal strategies in terms of the information bit-rate. The results show that the proposed method outperforms existing practical approaches in most scenarios, especially at middle rates and high rates, without having the prior knowledge of the joint statistics.

Hearing aids have become an indispensable part of the lives of some hearing-impaired people. Traditional hearing aids will be adjusted according to the personal hearing curve and allowing patients to avoid noise-induced harm. However, there is no sound classification or intelligent noise reduction, which cannot meet the higher demand for hearing aids. This paper designed a hearing aid based on a two-level neural network, and the Urbansound8K data set was used to train the neural network. It can simulate the human auditory attention mechanism and intelligently control the output volume. At the same time, the noise reduction model is used to perform corresponding noise reduction processing on different speech streams. Experimental results show that the hearing aid can differentially amplify various sounds in different scenes. The noise part of the sound heard by the user will be suppressed to a certain extent, which can improve the comfort of long-term wearing.