Enhanced smart hearing aid using deep neural networks

Abstract

Being smart has become a feature of almost all the electronic devices that we use every day. Recently, deep neural network based speech enhancement has made a breakthrough to speech de-noising process; however, in some situations, noise could be more important than speech, like talking while a fire alarm is activated. Hearing impaired people are the ones that could be most hurt in those situations, because of their lower ability to hear, especially in noisy environment. Available hearing aids lack the smart feature that distinguishes between desired and undesired noise types, so their users have to rely on separate alerting systems to ensure their safety. The idea that takes our concern in this work is to develop a smart hearing aid that has the ability to detect important noise and make it audible. In this paper, we will present three noise of interest aware speech enhancement networks for hearing aids application using deep learning, so as to introduce the idea of smart hearing aid. Keywords: Deep learning, Dropout, Noise of interest awareness, Smart hearing aid, Speech enhancement