PP419 - How Strong Do You Prefer Your DNN? Effect of Coupling on Preferred DNN Strength

Abstract:
The incorporation of AI into hearing devices is a growing trend in the field as a novel noise reduction system. This study investigated the effect of acoustic coupling on the preferred deep neural network (DNN) strength for a commercially available device. Behavioral performance was gathered from older adults with moderate-to-severe hearing configurations using the DNN and three vent options for their acoustic couplings. Findings will reveal how acoustic coupling affects preferred DNN strength. Results may inform clinical decision-making for hearing device programming.

Summary:
The role of AI in hearing aids is quickly becoming a trending subject within the field. Denoising deep neural networks (DNNs) are the latest method for addressing challenging listening environments. To date, all behavioral measures of speech intelligibility benefit with the novel noise reduction system of the commercially available device used in this study have been collected using an occluded coupling. This gives an indication of the effect of the DNN in the absence of direct sound, but it is not reflective of how products are fit in the field. Data with a variety of venting options will provide a way to better understand performance under typical fitting conditions.

Behavioral performance was gathered from older adults with moderate-to-severe hearing configurations. Participants were tested on speech in noise performance (SIN) across three vent options for their acoustic couplings. Experiment 1 was executed at a fixed moderate DNN strength. In experiment 2 participants self-adjusted the novel noise reduction feature via an app to find their preferred DNN strength setting. SNR for the SIN task was determined by an SRT-50 using CNC word lists. The AZ-Bio corpus was used for SIN testing. All test materials will be presented in the sound field via twelve loudspeakers at a distance of 1 m. Target speech will be played from 0° azimuth; multitalker noise played from all 12 speakers.

Outcomes yet to be revealed, but researchers hypothesize the participants will have higher performance scores with increasingly occluding venting in experiment one, and preference for the strength of the DNN will correlate with the amount of venting with more occluding venting rendering lower preferred strength settings. The results of this investigation will strengthen our understanding of the combined effects of venting and DNN strength on behavioral performance. It will also provide insight into what DNN strength patients may prefer.

Understanding potential consequences of acoustic coupling in relation to this feature will lead to more informed device fittings in the field. The clinical takeaway for this poster includes clinical guidance for programming DNN strength as it relates to a patients acoustic coupling.