Personalization of Hearing Aid Compression by Human-in-the-Loop Deep Reinforcement Learning

Nasim Alamdari,Edward Lobarinas,Nasser Kehtarnavaz

doi:10.1109/access.2020.3035728

Nasim Alamdari, Edward Lobarinas + Show 1 more

Open Access

PDF Available

https://doi.org/10.1109/access.2020.3035728

Copy DOI

Export

Save

Cite

Journal: IEEE Access	Publication Date: Jan 1, 2020
Citations: 20	License type: CC BY 4.0

Affiliation: The University of Texas at Dallas

Abstract
Highlights/Summary
Full-Text PDF
Similar Papers

Abstract

Listen

Existing prescriptive compression strategies used in hearing aid fitting are designed based on gain averages from a group of users which may not be necessarily optimal for a specific user. Nearly half of hearing aid users prefer settings that differ from the commonly prescribed settings. This paper presents a human-in-the-loop deep reinforcement learning approach that personalizes hearing aid compression to achieve improved hearing perception. The developed approach is designed to learn a specific user's hearing preferences in order to optimize compression based on the user's feedbacks. Both simulation and subject testing results are reported. These results demonstrate the proof-of-concept of achieving personalized compression via human-in-the-loop deep reinforcement learning.

Highlights

In hearing impaired individuals, the relative intensity difference between barely audible and uncomfortably loud sound becomes smaller
To address the non-linearities of hearing perception, a human-in-loop interactive machine learning approach based on deep reinforcement learning (DRL) [22] is developed in this paper
Permutations are defined by a dictionary in which each action is mapped to a set of compression ratio adjustments in the frequency bands

Summary

INTRODUCTION

The relative intensity difference between barely audible and uncomfortably loud sound becomes smaller. It has been reported that up to half of individuals using fitted hearing aids preferred amplification or compression settings different than the prescription provided [4,5,6,7,8]. In [18,19,20], a machine learning approach for self-adjustment or self-tuning of compression was presented In these papers, a Gaussian regression model was used to achieve personalized compression by estimating its parameters from training data. To address the non-linearities of hearing perception, a human-in-loop interactive machine learning approach based on deep reinforcement learning (DRL) [22] is developed in this paper.

PERSONALIZED COMPRESSION APPROACH

Initializing the human-centered DRL system with the above compression ratios

14. End for

CONCLUSION