Field Deployment of Personalized DSLv5 Hearing Aid Amplification by Bayesian Machine Learning
Aoxin Ni, Edward Lobariñas, Nasser Kehtarnavaz · 2024
This paper presents a real-time smartphone app that enables field deployment of a personalized DSLv5 amplification strategy based on a multi-band Bayesian machine learning algorithm. This implementation allows for the personalization of DSLv5 in real-world audio environments. The app includes a training and a testing session module. The training session allows reaching an optimum set of personalized gain values across a number of frequency bands. This is achieved by conducting paired audio comparisons by the user in a time-efficient manner. The testing session assesses comparisons between the personalized gain setting versus the standard DSLv5 prescription gain setting. The details of the steps taken to achieve this real-time implementation on smartphone platforms are presented. The results of a clinical experiment conducted on six participants with hearing loss show that the personalized settings on average are preferred over the standard settings by a factor of six times.