Perceptual Effects of Adjusting Hearing-Aid Gain by Means of a Machine-Learning Approach Based on Individual User Preference

Abstract

This study investigated a method to adjust hearing-aid gain by use of a machine-learning algorithm that estimates the optimal setting of gain parameters based on user preference indicated in an iterative paired-comparison procedure. Twenty hearing-impaired participants completed this procedure for 12 different sound scenarios. During the adjustment procedure, their task was to indicate a preference based on one of three sound attributes: Basic Audio Quality, Listening Comfort, or Speech Clarity. In a double-blind comparison of recordings of the processed scenarios, and using the same attributes as criteria, the adjusted gain settings were subsequently compared with two prescribed settings of the same hearing aid (with and without activation of an automatic sound-classification system). The results showed that the adjustment method provided a general improvement of Basic Audio Quality, an improvement of Listening Comfort in a traffic-noise scenario but not in three scenarios with speech babble, and no significant improvement of Speech Clarity. A large variation in gain adjustments was observed across participants, both among those who did benefit and among those who did not benefit from the adjustment. There was no clear connection between the gain adjustments and the perceived benefit, which indicates that the preferred gain settings for a given sound scenario and a given listening intention are highly individual and difficult to predict.

Hearing Aid Self-Adjustment: Effects of Formal Speech-Perception Test and Noise

“…There are alternative strategies involving such things as paired comparison and machine learning. Some of these strategies extend to other factors such as the criteria used by listeners and the properties of the sound input during self-adjustment (e.g., Jensen et al, 2019). Empirical comparisons of methods for self-fitting and postfitting selfreadjustment are clearly needed.…”

“…The efficacy of user adjustment of amplification has been demonstrated using a variety of protocols that incorporate changes of frequency response (Boothroyd & Mackersie, 2017;Dreschler et al, 2008;Jensen et al, 2019;Keidser & Convery, 2018;Keidser et al, 2008;Mackersie et al, 2019;Nelson et al, 2018). Groupmean user-adjusted frequency responses are generally close to threshold-based prescribed targets (typically within 5 dB), but substantial individual differences have been observed (Jensen et al, 2019;Keidser & Convery, 2018;Mackersie et al, 2019;Nelson et al, 2018;Punch et al, 1994). In addition, Nelson et al (2018) reported that on average, speech recognition outcomes of self-adjusted responses were not significantly different from those using the current version of the National Acoustics Laboratories prescription for Non-Linear hearing aids (NAL-NL2; Keidser et al, 2011).…”

Play&Tune: User Feedback in the Development of a Serious Game for Optimizing Hearing Aid Orientation

“…Advances are being made in provisions of technologies to assist HA users with calibrating their own HA, for example via smartphones i.e. Pasta et al (2019) andSøgaard Jensen et al (2019), and indeed many modern HAs are sold along with proprietary smartphone apps which enable some extent of personalisation. These services enable HA users to change their HA calibrations whilst in a "live" situation, so as to correct their hearing as they go about different everyday activities.…”

“…Efforts have been made to produce smartphone technology to help HA users to train their HA through personalising their settings in different environments. Emerging techniques using smartphone apps connected to Bluetooth enabled HAs have been recently described (Aldaz, Puria, & Leifer, 2016;Pasta, Petersen, Jensen, & Larsen, 2019;Søgaard Jensen, Hau, Bagger Nielsen, Bundgaard Nielsen, & Vase Legarth, 2019), in which user-preference is incorporated into an ongoing calibration process. These studies reported positive findings relating to different aspects of the hearing experience for conditions in which HAs had been optimized by the user.…”

Application of Bayesian Active Learning to the Estimation of Auditory Filter Shapes Using the Notched-Noise Method

“…This approach worked well for determining the similarity between images (Houlsby et al, 2013) and has also been used in auditory applications. For example, GPs have been used to search for the optimal setting of a hearing aid (Jensen et al, 2019;Nielsen et al, 2014), and for determining audiograms (Cox & de Vries, 2015;Schlittenlacher et al, 2018a;Song et al, 2015), equal-loudness contours (Schlittenlacher & Moore, 2020), and psychometric functions (Song et al, 2017). Other BAL approaches, often using parametric models but also maximizing mutual information or something similar, have been used to determine equal-loudness contours (Shen et al, 2018) or the edge frequency of a dead region (Schlittenlacher et al, 2018b).…”