Comparison of user interfaces for measuring the matrix sentence test on a smartphone.

OBJECTIVE: Smartphone-based self-testing could facilitate large-scale data collection and remote diagnostics. For this purpose, the matrix sentence test (MST) is an ideal candidate due to its repeatability and accuracy. In clinical practice, the MST requires professional audiological equipment and supervision, which is infeasible for smartphone-based self-testing. Therefore, it is crucial to investigate the feasibility of self-administering the MST on smartphones, including the development of an appropriate user interface for the small screen size. DESIGN: We compared the traditional closed matrix user interface (10 × 5 matrix) to three alternative, newly-developed interfaces (slide, type, wheel) regarding SRT consistency, user preference, and completion time. STUDY SAMPLE: We included 15 younger normal hearing and 14 older hearing-impaired participants in our study. RESULTS: The slide interface is most suitable for mobile implementation, providing consistent and fast SRTs and enabling all participants to perform the tasks effectively. While the traditional matrix interface works well for most participants, some participants experienced difficulties due to its small size on the screen. CONCLUSIONS: We propose the newly-introduced slide interface as a plausible alternative for smartphone screens. This might be more attractive for elderly patients that may exhibit more challenges with dexterity and vision than our test subjects employed here.

Conformities and gaps of clinical audiological data with the international classification of functioning disability and health core sets for hearing loss.

OBJECTIVE: The International Classification of Functioning Disability and Health (ICF) is a classification of health and health-related domains created by the World Health Organization and can be used as a standard to evaluate the health and disability of individuals. The ICF Core Set for Hearing Loss (CSHL) refers to the ICF categories found to be relative to Hearing Loss (HL) and the consequences of it on daily life. This study aimed to adapt the content of a database gathered in Hörzentrum Oldenburg gGmbH that included HL medical assessments and audiological data to the ICF. DESIGN: ICF linking rules were applied to these assessment methods including medical interviews, ear examinations, pure-tone audiometry, Adaptive Categorical Loudness Scaling, and speech intelligibility test. STUDY SAMPLE: 1316 subjects. RESULTS: In total, 44% of the brief and 18% of the comprehensive CSHL categories were addressed. The hearing functions were broadly evaluated. "Activities and Participation" and "Environmental Factors" were poorly examined (17% and 12% of the comprehensive CSHL categories, respectively). CONCLUSIONS: The HL correlation with day-to-day activities limitation, performance restriction, and environmental conditions were poorly addressed. This study showed the essence of incorporating these methodologies with approaches that assess the daily-life challenges caused by HL in rehabilitation.

Evaluation of adjustment behaviour in a semi-supervised self-adjustment fine-tuning procedure for hearing aids.

OBJECTIVE: This study investigated the adjustment behaviour of hearing aid (HA) users participating in a semi-supervised self-adjustment fine-tuning procedure for HAs. The aim was to link behaviour with the reproducibility and duration of the adjustments. DESIGN: Participants used a two-dimensional user interface to identify their HA gain preferences while listening to realistic sound scenes presented in a laboratory environment. The interface allowed participants to adjust amplitude (vertical axis) and spectral slope (horizontal axis) simultaneously. Participants were clustered according to their interaction with the user interface, and their search directions were analysed. STUDY SAMPLE: Twenty older experienced HA users were invited to participate in this study. RESULTS: We identified four different archetypes of adjustment behaviour (curious, cautious, semi-browsing, and full-on browsing) by analysing the trace points of all measurements for each participant. Furthermore, participants used predominantly horizontal or vertical paths when searching for their preference. Neither the archetype, nor the search directions, nor the participants' technology commitment was predictive of the reproducibility or the adjustment duration. CONCLUSIONS: The findings suggest that enforcement of a specific adjustment behaviour or search direction is not necessary to obtain fast, reliable self-adjustments. Furthermore, no strict requirements with respect to technology commitment are necessary.

Evaluation of a semi-supervised self-adjustment fine-tuning procedure for hearing aids.

OBJECTIVE: This study investigated the effects of different adjustment criteria and sound scenes on self-adjusted hearing-aid gain settings. Self-adjusted settings were evaluated for speech recognition in noise, perceived listening effort, and preference. DESIGN: This study evaluated a semi-supervised self-adjustment fine-tuning procedure that presents realistic everyday sound scenes in a laboratory environment, using a two-dimensional user interface, and enabling simultaneous changes in amplitude and spectral slope. While exploring the two-dimensional space of parameter settings, the hearing-aid users were instructed to optimise either listening comfort or speech understanding. STUDY SAMPLE: Twenty experienced hearing aid users (median age 69.5 years) were invited to participate in this study. RESULTS: Adjustment criterion and sound scenes had a significant effect on preferred gain settings. No differences in signal-to-noise ratios required for 50% speech intelligibility or in the perceived listening effort were observed between the adjusted settings of the two adjustment criteria. There was a preference for the self-adjusted settings over the prescriptive first fit. CONCLUSIONS: Listeners could reliably select their preferred gains to the two adjustment criteria and for different speech stimuli.

A model of speech recognition for hearing-impaired listeners based on deep learning.

Automatic speech recognition (ASR) has made major progress based on deep machine learning, which motivated the use of deep neural networks (DNNs) as perception models and specifically to predict human speech recognition (HSR). This study investigates if a modeling approach based on a DNN that serves as phoneme classifier [Spille, Ewert, Kollmeier, and Meyer (2018). Comput. Speech Lang. 48, 51-66] can predict HSR for subjects with different degrees of hearing loss when listening to speech embedded in different complex noises. The eight noise signals range from simple stationary noise to a single competing talker and are added to matrix sentences, which are presented to 20 hearing-impaired (HI) listeners (categorized into three groups with different types of age-related hearing loss) to measure their speech recognition threshold (SRT), i.e., the signal-to-noise ratio with 50% word recognition rate. These are compared to responses obtained from the ASR-based model using degraded feature representations that take into account the individual hearing loss of the participants captured by a pure-tone audiogram. Additionally, SRTs obtained from eight normal-hearing (NH) listeners are analyzed. For NH subjects and three groups of HI listeners, the average SRT prediction error is below 2 dB, which is lower than the errors of the baseline models.

Relation between hearing abilities and preferred playback settings for speech perception in complex listening conditions.

OBJECTIVE: This study investigated if individual preferences with respect to the trade-off between a good signal-to-noise ratio and a distortion-free speech target were stable across different masking conditions and if simple adjustment methods could be used to identify subjects as either "noise haters" or "distortions haters". DESIGN: In each masking condition, subjects could adjust the target speech level according to their preferences by employing (i) linear gain or gain at the cost of (ii) clipping distortions or (iii) compression distortions. The comparison of these processing conditions allowed investigating the preferred trade-off between distortions and noise disturbance. STUDY SAMPLE: Thirty subjects differing widely in hearing status (normal-hearing to moderately impaired) and age (23-85 years). RESULTS: High test-retest stability of individual preferences was found for all modification schemes. The preference adjustments suggested that subjects could be consistently categorised along a scale from "noise haters" to "distortion haters", and this preference trait remained stable through all maskers, spatial conditions, and types of distortions. CONCLUSIONS: Employing quick self-adjustment to collect listening preferences in complex listening conditions revealed a stable preference trait along the "noise vs. distortions" tolerance dimension. This could potentially help in fitting modern hearing aid algorithms to the individual user.

Inference of the distortion component of hearing impairment from speech recognition by predicting the effect of the attenuation component.

OBJECTIVE: A model-based determination of the average supra-threshold ("distortion") component of hearing impairment which limits the benefit of hearing aid amplification. DESIGN: Published speech recognition thresholds (SRTs) were predicted with the framework for auditory discrimination experiments (FADE), which simulates recognition processes, the speech intelligibility index (SII), which exploits frequency-dependent signal-to-noise ratios (SNR), and a modified SII with a hearing-loss-dependent band importance function (PAV). Their attenuation-component-based prediction errors were interpreted as estimates of the distortion component. STUDY SAMPLE: Unaided SRTs of 315 hearing-impaired ears measured with the German matrix sentence test in stationary noise. RESULTS: Overall, the models showed root-mean-square errors (RMSEs) of 7 dB, but for steeply sloping hearing loss FADE and PAV were more accurate (RMSE = 9 dB) than the SII (RMSE = 23 dB). Prediction errors of FADE and PAV increased linearly with the average hearing loss. The consideration of the distortion component estimate significantly improved the accuracy of FADE's and PAV's predictions. CONCLUSIONS: The supra-threshold distortion component-estimated by prediction errors of FADE and PAV-seems to increase with the average hearing loss. Accounting for a distortion component improves the model predictions and implies a need for effective compensation strategies for supra-threshold processing deficits with increasing audibility loss.

Development and evaluation of the Cantonese matrix sentence test.

OBJECTIVE: To develop the Cantonese matrix (YUEmatrix) test according to the international standard procedure and examine possible different outcomes in another tonal language. DESIGN: A 50-word Cantonese base-matrix was established. Word-specific speech recognition functions, speech recognition thresholds (SRT), and slopes were obtained. The speech material was homogenised in intelligibility by applying level corrections up to ± 3 dB. Subsequently, the YUEmatrix test was evaluated in five aspects: training effect, test-list equivalence, test-retest reliability, establishment of reference data for normal-hearing Cantonese-speakers, and comparison with the Cantonese-Hearing-In-Noise-Test. STUDY SAMPLE: Overall, 64 normal-hearing native Cantonese-speaking listeners. RESULTS: SRT measurements with adaptive procedures resulted in a reference SRT of -9.7 ± 0.7 dB SNR for open-set and -11.1 ± 1.2 dB SNR for the closed-set response format. Fixed SNR measurements suggested a test-specific speech intelligibility function slope of 15.5 ± 0.7%/dB. Seventeen 10-sentences base test lists were confirmed to be equivalent with respect to speech intelligibility. Training effect was not observed after two measurements of 20-sentences lists. CONCLUSIONS: The YUEmatrix yields comparable results to matrix tests in other languages including Mandarin. Level adjustments to homogenise sentences appear to be less effective for tonal languages than for most other languages developed so far.

Sensitivity and specificity of automatic audiological classification using expert-labelled audiological data and Common Audiological Functional Parameters.

OBJECTIVE: As a step towards the development of an audiological diagnostic supporting tool employing machine learning methods, this article aims at evaluating the classification performance of different audiological measures as well as Common Audiological Functional Parameters (CAFPAs). CAFPAs are designed to integrate different clinical databases and provide abstract representations of measures. DESIGN: Classification and evaluation of classification performance in terms of sensitivity and specificity are performed on a data set from a previous study, where statistical models of diagnostic cases were estimated from expert-labelled data. STUDY SAMPLE: The data set contains 287 cases. RESULTS: The classification performance in clinically relevant comparison sets of two competing categories was analysed for audiological measures and CAFPAs. It was found that for different audiological diagnostic questions a combination of measures using different weights of the parameters is useful. A set of four to six measures was already sufficient to achieve maximum classification performance which indicates that the measures contain redundant information. CONCLUSIONS: The current set of CAFPAs was confirmed to yield in most cases approximately the same classification performance as the respective optimum set of audiological measures. Overall, the concept of CAFPAs as compact, abstract representation of auditory deficiencies is confirmed.

Uni- and bilateral spectral loudness summation and binaural loudness summation with loudness matching and categorical loudness scaling.

OBJECTIVE: Current hearing aid prescription rules assume that spectral loudness summation decreases with hearing impairment and that binaural loudness summation is independent of hearing loss and signal bandwidth. Previous studies have shown that these assumptions might be incorrect. Spectral loudness summation was measured and compared for loudness scaling and loudness matching. DESIGN: In this study, the effect of bandwidth on binaural summation was investigated by comparing loudness perception of low-pass filtered, high-pass filtered, and broadband pink noise at 35 Categorical Units for both unilateral and bilateral presentation. STUDY SAMPLE: Sixteen hearing-impaired listeners. RESULTS: The results show that loudness differences between the three signals are different for bilateral presentation than for unilateral presentation. In specific, binaural loudness summation is larger for the low-pass filtered pink noise than for the high-pass filtered pink noise. Finally, individual variability in loudness perception near loudness discomfort level was found to be very large. CONCLUSIONS: Loudness matching is offered as a fast and reliable method to measure individual loudness perception. As discomfort with loud sounds is one of the major problems encountered by hearing aid users, measurement of individual loudness perception could improve hearing aid fitting substantially.

Machine learning for decoding listeners' attention from electroencephalography evoked by continuous speech.

Previous research has shown that it is possible to predict which speaker is attended in a multispeaker scene by analyzing a listener's electroencephalography (EEG) activity. In this study, existing linear models that learn the mapping from neural activity to an attended speech envelope are replaced by a non-linear neural network (NN). The proposed architecture takes into account the temporal context of the estimated envelope and is evaluated using EEG data obtained from 20 normal-hearing listeners who focused on one speaker in a two-speaker setting. The network is optimized with respect to the frequency range and the temporal segmentation of the EEG input, as well as the cost function used to estimate the model parameters. To identify the salient cues involved in auditory attention, a relevance algorithm is applied that highlights the electrode signals most important for attention decoding. In contrast to linear approaches, the NN profits from a wider EEG frequency range (1-32 Hz) and achieves a performance seven times higher than the linear baseline. Relevant EEG activations following the speech stimulus after 170 ms at physiologically plausible locations were found. This was not observed when the model was trained on the unattended speaker. Our findings therefore indicate that non-linear NNs can provide insight into physiological processes by analyzing EEG activity.

Influence of auditory attention on sentence recognition captured by the neural phase.

The aim of this study was to investigate whether attentional influences on speech recognition are reflected in the neural phase entrained by an external modulator. Sentences were presented in 7 Hz sinusoidally modulated noise while the neural response to that modulation frequency was monitored by electroencephalogram (EEG) recordings in 21 participants. We implemented a selective attention paradigm including three different attention conditions while keeping physical stimulus parameters constant. The participants' task was either to repeat the sentence as accurately as possible (speech recognition task), to count the number of decrements implemented in modulated noise (decrement detection task), or to do both (dual task), while the EEG was recorded. Behavioural analysis revealed reduced performance in the dual task condition for decrement detection, possibly reflecting limited cognitive resources. EEG analysis revealed no significant differences in power for the 7 Hz modulation frequency, but an attention-dependent phase difference between tasks. Further phase analysis revealed a significant difference 500 ms after sentence onset between trials with correct and incorrect responses for speech recognition, indicating that speech recognition performance and the neural phase are linked via selective attention mechanisms, at least shortly after sentence onset. However, the neural phase effects identified were small and await further investigation.

Prediction of individual speech recognition performance in complex listening conditions.

This study examined how well individual speech recognition thresholds in complex listening scenarios could be predicted by a current binaural speech intelligibility model. Model predictions were compared with experimental data measured for seven normal-hearing and 23 hearing-impaired listeners who differed widely in their degree of hearing loss, age, as well as performance in clinical speech tests. The experimental conditions included two masker types (multi-talker or two-talker maskers), and two spatial conditions (maskers co-located with the frontal target or symmetrically separated from the target). The results showed that interindividual variability could not be well predicted by a model including only individual audiograms. Predictions improved when an additional individual "proficiency factor" was derived from one of the experimental conditions or a standard speech test. Overall, the current model can predict individual performance relatively well (except in conditions high in informational masking), but the inclusion of age-related factors may lead to even further improvements.

Common Audiological Functional Parameters (CAFPAs) for single patient cases: deriving statistical models from an expert-labelled data set.

Objective: Statistical knowledge about many patients could be exploited using machine learning to provide supporting information to otolaryngologists and other hearing health care professionals, but needs to be made accessible. The Common Audiological Functional Parameters (CAFPAs) were recently introduced for the purpose of integrating data from different databases by providing an abstract representation of audiological measurements. This paper aims at collecting expert labels for a sample database and to determine statistical models from the labelled data set.Design: By an expert survey, CAFPAs as well as labels for audiological findings and treatment recommendations were collected for patients from the database of Hörzentrum Oldenburg.Study sample: A total of 287 single patient cases were assessed by twelve highly experienced audiological experts.Results: The labelled data set was used to derive probability density functions for categories given by the expert labels. The collected data set is suitable for estimating training distributions due to realistic variability contained in data for different, distinct categories. Suitable distribution functions were determined. The derived training distributions were compared regarding different audiological questions.Conclusions: The method-expert survey, sorting data into categories, and determining training distributions - could be extended to other data sets, which could then be integrated via the CAFPAs and used in a classification task.

Clinical validation of the Russian Matrix test - effect of hearing loss, age, and noise level.

OBJECTIVE: To validate the Russian matrix sentence test (RUMatrix) for the assessment of speech recognition in quiet and in noise in clinical praxis. The effect of hearing impairment, age, and masking-noise level on speech recognition was examined. DESIGN: All participants underwent pure tone audiometry, a monosyllabic speech test in quiet, and speech recognition measurements with RUMatrix in quiet (SRTQ) and in noise (SRTN). STUDY SAMPLE: One hundred and forty-two listeners divided into four groups: 1. Young normal-hearing listeners, 2. Older normal-hearing listeners, 3. Young hearing-impaired listeners, and 4. Older hearing-impaired listeners. RESULTS: Significant differences between groups of listeners were found in the SRTQ and SRTN. A strong correlation between hearing threshold and SRTQ (R2=0.88, p < 0.001) indicates a strong link between speech recognition in quiet and audibility. The pure-tone average explained less variance in SRTN (R2=0.67, p < 0.001), pointing out an additional influence of suprathreshold distortion. A high test sensitivity of 0.99 was found for SRTN and SRTQ. The monosyllabic test had a low sensitivity (0.21), indicating that the test is not suitable for separating normal-hearing and hearing-impaired listeners. CONCLUSIONS: RuMatrix is a reliable speech recognition assessment tool with a high sensitivity and validity for the main aspects of hearing impairment.

The Finnish simplified matrix sentence test for the assessment of speech intelligibility in the elderly.

Objective: A simplified version of the Finnish matrix sentence test (FMST) was developed to improve the reliability of hearing diagnostic for children and for patients with limited working memory capacity and/or vocabulary.Design: Study 1 evaluated the word matrix of the Finnish simplified matrix sentence test (FINSIMAT) to rule out systematic differences between the new FINSIMAT test lists, and to provide reference values for normal-hearing (NH) young adults (YA). In Study 2, the FINSIMAT and the FMST were evaluated in elderly listeners with mild-to-moderate hearing impairment (HI).Study sample: Twenty NH YAs participated in Study 1, and 16 elderly HI adults participated in Study 2.Results: For NH YAs, the reference speech reception threshold (SRT50) estimate and the slope for the FINSIMAT were -11.2 ± 1.0 dB signal-to-noise ratio (SNR) and 19.4 ± 1.9%/dB SNR. For the elderly HI listeners, the mean SRT50 estimates for the FINSIMAT and FMST were -4.1 and -3.6 dB SNR, respectively. The correlation between the FMST and FINSIMAT results was strong (r2 = 0.78, p < 0.001).Conclusion: The FINSIMAT showed comparable characteristics to the FMST and proved feasible for measurements in elderly HI listeners.

On the limitations of sound localization with hearing devices.

Limited abilities to localize sound sources and other reduced spatial hearing capabilities remain a largely unsolved issue in hearing devices like hearing aids or hear-through headphones. Hence, the impact of the microphone location, signal bandwidth, different equalization approaches, as well as processing delays in superposition with direct sound leaking through a vent was addressed in this study. A localization experiment was performed with normal-hearing subjects using individual binaural synthesis to separately assess the above-mentioned potential limiting issues for localization in the horizontal and vertical plane with linear hearing devices. To this end, listening through hearing devices was simulated utilizing transfer functions for six different microphone locations, measured both individually and on a dummy head. Results show that the microphone location is the governing factor for localization abilities with linear hearing devices, and non-optimal microphone locations have a disruptive influence on localization in the vertical domain, and an effect on lateral sound localization. Processing delays cause additional detrimental effects for lateral sound localization; and diffuse-field equalization to the open-ear response leads to better localization performance than free-field equalization. Stimuli derived from dummy head measurements are unsuited for evaluating individual localization abilities with a hearing device.

Common Audiological Functional Parameters (CAFPAs): statistical and compact representation of rehabilitative audiological classification based on expert knowledge.

OBJECTIVE: As a step towards objectifying audiological rehabilitation and providing comparability between different test batteries and clinics, the Common Audiological Functional Parameters (CAFPAs) were introduced as a common and abstract representation of audiological knowledge obtained from diagnostic tests. DESIGN: Relationships between CAFPAs as an intermediate representation between diagnostic tests and audiological findings, diagnoses and treatment recommendations (summarised as "diagnostic cases") were established by means of an expert survey. Expert knowledge was collected for 14 given categories covering different diagnostic cases. For each case, the experts were asked to indicate expected ranges of diagnostic test outcomes, as well as traffic light-encoded CAFPAs. STUDY SAMPLE: Eleven German experts in the field of audiological rehabilitation from Hanover and Oldenburg participated in the survey. RESULTS: Audiological findings or treatment recommendations could be distinguished by a statistical model derived from the experts' answers for CAFPAs as well as audiological tests. CONCLUSIONS: The CAFPAs serve as an abstract, comprehensive representation of audiological knowledge. If more detailed information on certain functional aspects of the auditory system is required, the CAFPAs indicate which information is missing. The statistical graphical representations for CAFPAs and audiological tests are suitable for audiological teaching material; they are universally applicable for real clinical databases.

Restoring Perceived Loudness for Listeners With Hearing Loss.

OBJECTIVES: Normalizing perceived loudness is an important rationale for gain adjustments in hearing aids. It has been demonstrated that gains required for restoring normal loudness perception for monaural narrowband signals can lead to higher-than-normal loudness in listeners with hearing loss, particularly for binaural broadband presentation. The present study presents a binaural bandwidth-adaptive dynamic compressor (BBDC) that can apply different gains for narrow- and broadband signals. It was hypothesized that normal perceived loudness for a broad variety of signals could be restored for listeners with mild to moderate high-frequency hearing loss by applying individual signal-dependent gain corrections. DESIGN: Gains to normalize perceived loudness for narrowband stimuli were assessed in 15 listeners with mild to moderate high-frequency hearing loss using categorical loudness scaling. Gains for narrowband loudness compensation were calculated and applied in a standard compressor. Aided loudness functions for signals with different bandwidths were assessed. The deviation from the average normal-hearing loudness functions was used for gain correction in the BBDC. Aided loudness functions for narrow- and broadband signals with BBDC were then assessed. Gains for a 65 dB SPL speech-shaped noise of BBDC were compared with gains based on National Acoustic Laboratories' nonlinear fitting procedure version 2 (NAL-NL2). The perceived loudness for 20 real signals was compared to the average normal-hearing rating. RESULTS: The suggested BBDC showed close-to-normal loudness functions for binaural narrow- and broadband signals for the listeners with hearing loss. Normal loudness ratings were observed for the real-world test signals. The proposed gain reduction method resulted on average in similar gains as prescribed by NAL-NL2. However, substantial gain variations compared to NAL-NL2 were observed in the data for individual listeners. Gain corrections after narrowband loudness compensation showed large interindividual differences for binaural broadband signals. Some listeners required no further gain reduction for broadband signals; for others, gains in decibels were more than halved for binaural broadband signals. CONCLUSION: The interindividual differences of the binaural broadband gain corrections indicate that relevant information for normalizing perceived loudness of binaural broadband signals cannot be inferred from monaural narrowband loudness functions. Over-amplification can be avoided if binaural broadband measurements are included in the fitting procedure. For listeners with a high binaural broadband gain correction factor, loudness compensation for narrowband and broadband stimuli cannot be achieved by compression algorithms that disregard the bandwidth of the input signals. The suggested BBDC includes individual binaural broadband corrections in a more appropriate way than threshold-based procedures.

Spectral directional cues captured by hearing device microphones in individual human ears.

Spatial hearing abilities with hearing devices ultimately depend on how well acoustic directional cues are captured by the microphone(s) of the device. A comprehensive objective evaluation of monaural spectral directional cues captured at 9 microphone locations integrated in 5 hearing device styles is presented, utilizing a recent database of head-related transfer functions (HRTFs) that includes data from 16 human and 3 artificial ear pairs. Differences between HRTFs to the eardrum and hearing device microphones were assessed by descriptive analyses and quantitative metrics, and compared to differences between individual ears. Directional information exploited for vertical sound localization was evaluated by means of computational models. Directional information at microphone locations inside the pinna is significantly biased and qualitatively poorer compared to locations in the ear canal; behind-the-ear microphones capture almost no directional cues. These errors are expected to impair vertical sound localization, even if the new cues would be optimally mapped to locations. Differences between HRTFs to the eardrum and hearing device microphones are qualitatively different from between-subject differences and can be described as a partial destruction rather than an alteration of relevant cues, although spectral difference metrics produce similar results. Dummy heads do not fully reflect the results with individual subjects.