CONIX Publication

Reinforcement Learning Framework for Augmenting Hearing Loss on the Open Speech Platform

Authors: Dhiman Sengupta, Rajesh Gupta, Harinath Garudadri


Parameter tuning for hearing aids is a challenging task. Hearing aid tuning requires extensive expertise due to the enormous search area and the subjectiveness of the user. Many other applications have shown reinforcement learning (RL) to be very adept at parameter tuning given a non-convex search area. Therefore, warranting the research community to explore applying RL to hearing aids and other hearable technologies. This work extends the open speech platform (OSP), an open-source hardware and software platform for the research and development of the next generation of hearable technologies, by adding an RL framework to allow for rapid exploration. We further discuss the motivations behind using an RL algorithm to fine-tune parameters in a hearing aid to augment the hearing of individual users. Even though the hearing aid is the primary application for this work, researchers can use the platform and framework for any hearable research, like augmenting our soundscape, which is how this work ties into CONIX.

Release Date: 10/20/2021
Uploaded File: View