Hearing-aid Speech Quality Index Research Articles

A Non-Intrusive speech quality evaluation algorithm for hearing aids via an auxiliary training task

With the increasing severity of hearing problems caused by an aging population, it has become more urgent to study speech quality evaluation methods for hearing aids to automatically judge the effect of parameter adjustments to solve the problem of lack of audiologists. A new evaluation algorithm, which is based on the multi-task learning strategy that combines the main task of non-intrusive speech quality evaluation with the auxiliary task of quality classification, is proposed. The main task network is composed of a bi-directional long short-term memory (BiLSTM) network combined with the self-attention mechanism, and the auxiliary task network is formed by fully connected layers based on batch normalization. The proposed algorithm firstly uses a convolutional neural network (CNN) to extract frame-level features. Then, the features are inputted into the main task network and the auxiliary task network to extract utterance-level features, respectively. Finally, the objective scores and quality ratings are obtained by the different mappers. Evaluation experiments of the Hearing Aid Speech Quality Index (HASQI) show that compared with others algorithms, the proposed algorithm effectively improves the speech quality evaluation. And it also shows strong robustness and good generalization performances under different distorted conditions.

Acoustic feedback control in hearing aids with frequency warping

Acoustic Feedback Control continues to be a challenging problem due to the emerging form factors in advanced hearing aids and hearables, especially for severe and profound losses. In order to mitigate the effects of feedback in hearing aids, we present a novel use of frequency warping called “Freping,” for manipulating signals in the frequency domain. Freping is a non-linear transformation which breaks the similarity between the input and acoustic feedback signals of a hearing aid to prevent feedback. In this work, we implemented Freping in realtime for feedback management, as well as novel techniques to preserve the naturalness of sound. Our prior work has shown that Freping improves Added Stable Gain (ASG) in the presence of feedback while maintaining the same Hearing Aid Speech Quality Index (HASQI). We present new results for feedback management with Freping by performing “on-ear” measurements on an acoustic KEMAR manikin using standard hearing aid verification systems. We conducted feedback tests to measure the ASG with various configurations of Freping for various hearing loss levels using the NAL-NL2 prescriptions on a realtime open-source speech processing system for hearing aids. We also compare Freping feedback performance with that of commercial hearing aids.

An objective comparison of speech intelligibility and quality of hearing aid of different companies

Abstract Purpose: This study aimed to verify the intelligibility and quality of hearing aid-processed speech across companies through objective measures. The output of hearing aids programmed to different prescriptive formulae (Company fit, NAL NL 1, NAL NL 2, and DSLv5) across companies (Company I, Company II, Company III, Company IV, and Company V) were recorded and stored in the sound level meter connected to the manikin. Materials and Methods: Offline analysis of speech quality of the recorded speech was done using the Hearing Aid Speech Quality Index (HASQI), the intelligibilty was measured using the Hearing Aid Speech Perception Index (HASPI) and Speech Intelligibility Index (SII), and Long-Term Average Speech Spectrum (LTASS) was used to measure the spectral composition of speech. Results: When the companies are compared, the hearing aid of Company V showed significantly higher SII values for all prescriptive formulae indicative of better speech intelligibility provided by Company V hearing aids. In terms of speech quality, all the hearing aids showed similar HASQI values. Conclusions: The reasons for each of these findings are highlighted and recommendations for best practices on the use of objective measures for verifying the hearing aid output in clinical setups are suggested.

Robust intelligibility and quality evaluation of combined temporal and spectral processing for hearing impaired

• Combined temporal-spectral processing objectively evaluated in terms of HASQI and HASPI. • Six types of audiograms based on different slope and frequency losses are selected. • -5, -3, 0, 3, 5, 10 dB SNRs noisy environments in cafe, traffic, station and train are considered. • Proposed enhanced algorithm shows markable improvements in quality and intelligibility. Hearing-impaired people face numerous challenges with speech perception in the presence of interfering background noise. To suppress interfering background noise, the common approach widely used is speech enhancement. Inspired by the improved results of combined temporal and spectral processing in speech enhancement, this research study proposes temporal enhancement combined with two different spectral enhancement methods, with a novel approach of soft masking using priori and posterior signal to noise ratio uncertainty. The present study investigates quality and intelligibility objective evaluations, namely, hearing aid speech quality index and hearing aid speech perception index, of spectral and a combination of temporal-spectral speech enhancement methods for typical pattern of hearing loss characterized by six audiograms. For evaluation, clean speech files from the NOIZEUS database are mixed with four local noises, namely, cafeteria, traffic, station, and train at -5, -3, 0,3,5, and 10 dB SNRs. These local noises are quite common, which are encountered by people in their day-to-day lives. In most of the testing conditions, the new combined temporal spectral enhancement shows improved results in comparison with the purely spectral processing methods.

Comparison of laboratory and clinical computational metrics of acoustic measures of hearing-aid processed speech

Computational auditory metrics are used to characterize hearing-aid fittings in the laboratory and the clinic. The purpose of this study is to compare the speech intelligibility index (SII) as a clinical metric to the laboratory based SII and the hearing-aid speech perception index (HASPI) and hearing-aid speech quality index (HASQI) laboratory metrics. These comparisons are drawn from a comprehensive dataset of hearing-aid fittings for 120 hearing-aid recipients from a hospital-based audiology clinic. Hearing-aid devices are drawn from multiple manufacturers and multiple technology levels, and a total of nine hearing-aid devices are included. Acoustic recordings are made with the hearing-aids positioned on the KEMAR manikin and include multiple hearing-aid settings for each patient (manufacturer’s recommended fit, first fit by the audiologist, and the final fitting selected by the patient). Metric values are computed for each of the settings, and these metric comparisons highlight the advantages and challenges of each metric in characterizing hearing-aid signal processing. In addition, the results provide insight regarding hearing-aid fittings from the manufacturer’s fitting versus clinician/patient driven fittings. [Work supported by NIH R01 DC012289, NIH NRSA T32DC012280, and GN Hearing OCG6790B.]

Feasibility of a hearing-aid speech quality index designed for noisy measurement conditions

A goal in fitting hearing aids is to find settings that improve listener judgments of the amplified speech quality. Objective metrics that predict speech quality could assist in this fitting procedure and, thus, become useful tools in the clinic. The Hearing-Aid Speech Quality Index version 2 (HASQI v2) is effective in predicting speech quality ratings for speech processed through a hearing aid. However, HASQI v2 requires a signal-to-noise ratio (SNR) of approximately 40 dB to provide reliable measurements of low levels of nonlinear distortion, and this favorable an SNR is not always found in a clinical examination room or medical office where hearing-aid measurements are performed. Two modifications to HASQI v2 are considered in this presentation to improve the measurement accuracy in noisy situations. The first is applying dynamic-range compression to the test signal to enhance low-intensity speech sounds that could be masked by background noise, and the second is using a multilayer neural network to match noisy measurements to the corresponding measurements made in a quiet laboratory. The benefits of these modifications to HASQI are evaluated using a hearing-aid simulation, and verification is provided using commercial hearing-aid measurements made in an audiology clinic.

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.

SPQER: Speech Quality Evaluation Using Word Recognition for VoIP Communication in Lossy and Mobile Networks

In this paper, we introduce SPQER (pronounced speaker), a novel approach to evaluate the quality of experience for real-time Voice over IP (VoIP) communication in mobile and lossy networks. Traditional speech quality metrics, e.g., Perceptual Evaluation of Speech Quality (PESQ) or the Hearing-Aid Speech Quality Index (HASQI), directly compare frequencies and amplitudes to calculate the received signal distortions. SPQER instead uses machine learning classification to evaluate the percentage of recognizable words in conjunction with a time-based decay function to penalize delay and cross-talking. So instead of evaluating noise, SPQER directly answers the question: What percentage of words is the recipient able to understand? We presented a sensitivity analysis, which is based on testbed experiments for different packet loss rates and simulated delays, to asses the impact of challenging link conditions. A final correlation analysis to a short user study shows that SPQER can better evaluate the amount of understandable words than PESQ and HASQI, while still giving a more precise indication about the voice quality than the Word Error Rate (WER) metric.

Quantifying the Range of Signal Modification in Clinically Fit Hearing Aids.

Hearing aids provide various signal processing techniques with a range of parameters to improve the listening experience for a hearing-impaired individual. In previous studies, we reported significant differences in signal modification for mild versus strong signal processing in commercially available hearing aids. In this study, the authors extend this work to clinically prescribed hearing aid fittings based on best-practice guidelines. The goals of this project are to determine the range of cumulative signal modification in clinically fit hearing aids across manufacturers and technology levels and the effects of listening conditions including signal to noise ratio (SNR) and presentation level on these signal modifications. We identified a subset of hearing aids that were representative of a typical clinical setting. Deidentified hearing aid fitting data were obtained from three audiology clinics for adult hearing aid users with sensorineural hearing loss for a range of hearing sensitivities. Matching laboratory hearing aids were programmed with the deidentified fitting data. Output from these hearing aids was recorded at four SNRs and three presentation levels. The resulting signal modification was quantified using the cepstral correlation component of the Hearing Aid Speech Quality Index which measures the speech envelope changes in the context of a model of the listener's hearing loss. These metric values represent the hearing aid processed signal as it is heard by the hearing aid user. Audiometric information was used to determine the nature of any possible association with the distribution of signal modification in these clinically fit hearing aids. In general, signal modification increased as SNR decreased and presentation level increased. Differences across manufacturers were significant such that the effect of presentation level varied differently at each SNR, for each manufacturer. This result suggests that there may be variations across manufacturers in processing various listening conditions. There was no significant effect of technology level. There was a small effect of pure-tone average on signal modification for one manufacturer, but no effect of audiogram slope. Finally, there was a broad range of measured signal modification for a given hearing loss, for the same manufacturer and listening condition. The signal modification values in this study are representative of commonly fit hearing aids in clinics today. The results of this study provide insights into how the range of signal modifications obtained in real clinical fittings compares with a previous study. Future studies will focus on the behavioral implications of signal modifications in clinically fit hearing aids.

On mitigating acoustic feedback in hearing aids with frequency warping by all-pass networks

Acoustic feedback due to the acoustic coupling between the microphones and loudspeakers at high gains can cause the hearing aid system to become unstable. This instability results in brief “howling” artifacts as magnitude and phase conditions fulfill the Nyquist stability criterion (NSC). We present a novel approach that uses all-pass networks to perform nonlinear spectral mapping that we call “freping,” a portmanteau for frequency warping. In this contribution, we focus on spectral manipulations on top of adaptive feedback cancellation (AFC) to break NSC for improved feedback reduction. A real-time, multichannel realization is presented, which has individual control of the warping degree in each frequency band. Freping helps mitigate the NSC and thus leads to improved feedback control, while distortions due to freping are fairly benign based on informal subjective assessments. Our current findings indicate that improvements in terms of the perceptual evaluation of speech quality (PESQ) and the hearing-aid speech quality index (HASQI) can be achieved with freping for a basic AFC (PESQ: 2.56 to 3.52 and HASQI: 0.65 to 0.78) at medium amplification; and an advanced AFC (PESQ: 2.75 to 3.17 and HASQI: 0.66 to 0.73) at high amplification. Additional results will be included in the final presentation.

On Mitigating Acoustic Feedback in Hearing Aids with Frequency Warping by All-Pass Networks.

Acoustic feedback control continues to be a challenging problem due to the emerging form factors in advanced hearing aids (HAs) and hearables. In this paper, we present a novel use of well-known all-pass filters in a network to perform frequency warping that we call "freping." Freping helps in breaking the Nyquist stability criterion and improves adaptive feedback cancellation (AFC). Based on informal subjective assessments, distortions due to freping are fairly benign. While common objective metrics like the perceptual evaluation of speech quality (PESQ) and the hearing-aid speech quality index (HASQI) may not adequately capture distortions due to freping and acoustic feedback artifacts from a perceptual perspective, they are still instructive in assessing the proposed method. We demonstrate quality improvements with freping for a basic AFC (PESQ: 2.56 to 3.52 and HASQI: 0.65 to 0.78) at a gain setting of 20; and an advanced AFC (PESQ: 2.75 to 3.17 and HASQI: 0.66 to 0.73) for a gain of 30. From our investigations, freping provides larger improvement for basic AFC, but still improves overall system performance for many AFC approaches.

Learning-Based Reference-Free Speech Quality Measures for Hearing Aid Applications

Objective measures of speech quality are highly desirable in benchmarking and monitoring the performance of hearing aids (HAs). Existing HA speech quality indices such as the hearing aid speech quality index (HASQI) are intrusive in that they require a properly time-aligned and frequency-shaped reference signal to predict the quality of HA output. Two new reference-free HA speech quality indices are proposed in this paper, based on a model that amalgamates perceptual linear prediction (PLP), hearing loss (HL) modeling, and machine learning concepts. For the first index, HL-modified PLP coefficients and their statistics were used as the feature set, which was subsequently mapped to the predicted quality scores using support vector regression (SVR). For the second index, HL-impacted gammatone auditory filterbank energies and their second-order statistics constituted the feature set, which was again mapped using SVR. Two databases involving HA recordings were collected and utilized for the evaluation of the robustness and generalizability of the two indices. Experimental results showed that the index based on the gammatone filterbank energies not only correlated well with HA quality ratings by hearing impaired listeners, but also exhibited robust performance across different test conditions and was comparable to the full-reference HASQI performance.

Using Objective Metrics to Measure Hearing Aid Performance.

The performance of hearing aids is generally characterized by a small set of standardized measurements. The primary goals of these procedures are to measure basic aspects of the hearing aid performance and to ascertain that the device is operating properly. A more general need exists for objective metrics that can predict hearing aid outcomes. Such metrics must consider the interaction of all the signal processing operating in the hearing aid and must do so while also accounting for the hearing loss for which the hearing aid has been prescribed. This article represents a first step in determining the clinical applicability of the hearing aid speech perception index (HASPI) intelligibility and hearing aid speech quality index (HASQI) speech quality metrics. The goals of this article are to demonstrate the feasibility of applying these metrics to commercial hearing aids and to illustrate the anticipated range of measured values and identify implementation concerns that may not be present for conventional measurements. This article uses the HASPI intelligibility and HASQI speech quality metrics to measure the performance of commercial hearing aids. These metrics measure several aspects of the processed signal, including envelope fidelity, modifications of the temporal fine structure, and changes in the long-term frequency response, all in the context of an auditory model that reproduces the salient aspects of the peripheral hearing loss. The metrics are used to measure the performance of basic and premium hearing aids from three different manufacturers. Test conditions include the environmental factors of signal to noise ratio and presentation level, and the fitting configurations were varied to provide different degrees of processing from linear to aggressive nonlinear processing for two different audiograms. The results show that the metrics are capable of measuring statistically significant differences across devices and processing settings. HASPI and HASQI measure both audibility and nonlinear distortion in the devices, and conditions are identified where predicted intelligibility is high but predicted speech quality is substantially reduced. The external signal properties of signal to noise ratio and presentation level are both statistically significant. Hearing loss is significant for HASPI but not for HASQI, and degree of processing is significant for both metrics. A quadratic model for manufacturer showed large effect sizes for HASPI and HASQI, but basic versus premium hearing aid model is not significant. The results presented in this article represent a first step in applying the HASPI and HASQI metrics to commercial hearing aids. Modern hearing aids often use several different processing strategies operating simultaneously. The proposed metrics provide a way to predict the total effect of this processing, including algorithm interactions that may be missed by conventional measurement procedures. The measurements in this article show significant differences between manufacturers, processing settings, and adjustment for different hearing losses. No significant differences were found between basic and premium hearing aid models. Further research will be needed to determine the clinical relevance of these measurements and to provide target values appropriate for successful fittings.

Electroacoustic and behavioral evaluation of an open source audio processing platform

Hearing-aid (HA) research in academia is limited by the lack of wearable, reconfigurable, and reprogrammable audio processing platforms. We are developing such a platform, called the Tympan, which includes a Teensy 3.6 processor board that leverages the Arduino development environment while providing powerful computational capabilities. Custom-designed electronics are used for audio control, power management, and wireless communication. User-friendly software can be used to test the relative benefits of amplification variants. Users can also implement new algorithms and modify parameters of algorithms. This study evaluated an eight-channel HA implemented on the Tympan, relative to a commercially available HA. Gain was prescribed using NAL-NL1. Eighteen participants with hearing loss were tested. Electroacoustic tests included ANSI 3.22-2009 standard clinical measurements, the HA speech perception index, and the HA speech quality index. Behavioral tests included recognition of CASPA words in quiet and AzBio sentences in noise. The Tympan HA performed similar to the commercially-available HA on all tasks. Efforts are ongoing to miniaturize the Tympan hardware, increase flexibility of the software, and add features such as feedback management. The collaborative development and open sharing of algorithms facilitated by the Tympan will lead to advances in HA research and in other audio signal processing fields.

Speech quality evaluation of a sparse coding shrinkage noise reduction algorithm with normal hearing and hearing impaired listeners

Although there are numerous papers describing single-channel noise reduction strategies to improve speech perception in a noisy environment, few studies have comprehensively evaluated the effects of noise reduction algorithms on speech quality for hearing impaired (HI). A model-based sparse coding shrinkage (SCS) algorithm has been developed, and has shown previously (Sang et al., 2014) that it is as competitive as a state-of-the-art Wiener filter approach in speech intelligibility. Here, the analysis is extended to include subjective quality ratings and a method called Interpolated Paired Comparison Rating (IPCR) is adopted to quantitatively link the benefit of speech intelligibility and speech quality.The subjective quality tests are performed through IPCR to efficiently quantify noise reduction effects on speech quality. Objective measures including frequency-weighted segmental signal-to-noise ratio (fwsegSNR), perceptual evaluation of speech quality (PESQ) and hearing aid speech quality index (HASQI) are adopted to predict the noise reduction effects.Results show little difference in speech quality between the SCS and the Wiener filter algorithm but a difference in quality rating between the HI and NH listeners. HI listeners generally gave better quality ratings of noise reduction algorithms than NH listeners. However, SCS reduced the noise more efficiently at the cost of higher distortions that were detected by NH but not by the HI.SCS is a promising candidate for noise reduction algorithms for HI. In general, care needs to be taken when adopting algorithms that were originally developed for NH participants into hearing aid applications. An algorithm that is evaluated negatively with NH might still bring benefits for HI participants.

Quality of older voices processed by hearing aids: Acoustic factors explaining inter-talker differences

Hearing aid signal processing algorithms are often evaluated with professional recordings of voices. However, hearing aid users often listen to older talkers who may have poorer voices than younger talkers. The purpose of this study was to quantify the extent to which the acoustic and perceptual consequences of hearing aid digital signal processing algorithms differ for talkers that vary in their vocal characteristics. There were six older talkers (3 males and 3 females) selected from a larger database of 79 talkers; their voices had good, moderate, or poor quality based on perceptual data from younger and older listeners. The voices were presented in quiet and in the presence of babble noise (10 dB SNR). The voices were processed with varying amounts of frequency compression, wide dynamic range compression, and noise suppression (spectral subtraction). There were interactions between signal processing and voice characteristics as measured using the Hearing Aid Speech Quality Index (HASQI) and listener ra...

Evaluating the Generalization of the Hearing Aid Speech Quality Index (HASQI)

Many developers of audio signal processing strategies rely on objective measures of quality for initial evaluations of algorithms. As such, objective measures should be robust, and they should be able to predict quality accurately regardless of the dataset or testing conditions. Kates and Arehart have developed the Hearing Aid Speech Quality Index (HASQI) to predict the effects of noise, nonlinear distortion, and linear filtering on speech quality for both normal-hearing and hearing-impaired listeners, and they report very high performance with their training and testing datasets [Kates, J. and Arehart, K., Audio Eng. Soc., 58(5), 363-381 (2010)]. In order to investigate the generalizability of HASQI, we test its ability to predict normal-hearing listeners' subjective quality ratings of a dataset on which it was not trained. This dataset is designed specifically to contain a wide range of distortions introduced by real-world noises which have been processed by some of the most common noise suppression algorithms in hearing aids. We show that HASQI achieves prediction performance comparable to the Perceptual Evaluation of Speech Quality (PESQ), the standard for objective measures of quality, as well as some of the other measures in the literature. Furthermore, we identify areas of weakness and show that training can improve quantitative prediction.

Effects of noise, nonlinear processing, and linear filtering on perceived music quality

Objective: The purpose of this study was to determine the relative impact of different forms of hearing aid signal processing on quality ratings of music. Design: Music quality was assessed using a rating scale for three types of music: orchestral classical music, jazz instrumental, and a female vocalist. The music stimuli were subjected to a wide range of simulated hearing aid processing conditions including, (1) noise and nonlinear processing, (2) linear filtering, and (3) combinations of noise, nonlinear, and linear filtering. Study sample: Quality ratings were measured in a group of 19 listeners with normal hearing and a group of 15 listeners with sensorineural hearing impairment. Results: Quality ratings in both groups were generally comparable, were reliable across test sessions, were impacted more by noise and nonlinear signal processing than by linear filtering, and were significantly affected by the genre of music. Conclusions: The average quality ratings for music were reasonably well predicted by the hearing aid speech quality index (HASQI), but additional work is needed to optimize the index to the wide range of music genres and processing conditions included in this study.SumarioObjetivo: El propósito de este estudio fue determinar el impacto relativo de diferentes formas de procesamiento de señal de los auxiliares auditivos en cuanto a la calificación cualitativa de la música. Diseño: la calidad de la música se evaluó utilizando una escala de calificación para tres tipos de música: clásica orquestral, jazz instrumental y mujer vocalista. Los estímulos musicales estuvieron sujetos a un amplio número de simulaciones de procesamiento de los auxiliares auditivos, incluyendo (1) ruido y procesamiento no-linear, (2) filtración linear y (3) combinaciones de ruido, no-linear y filtración linear. Mestra: Se obtuvieron los índices de calidad en un grupo de 19 sujetos normoyentes y otro grupo de 15 sujetos con hipoacusia sensorineural. Resultados: Las calificaciones cualitativas en ambos grupos fueron generalmente comparables, confiables en las diferentes sesiones, fueron más alteradas por el ruido y por el procesamiento no linear de la señal que por la filtración linear y fueron afectadas significativamente por el género de la música. Conclusiones: El índice de calidad del lenguaje de los auxiliares auditivos (HASQI), pronosticó razonablemente bien el promedio de calificación de la calidad de la música, pero todavía es necesario continuar la investigación para optimizar el índice para que éste pueda cubrir una gama más amplia de géneros y condiciones de procesamiento incluidas en este estudio.