Sound classification using deep learning for hard of hearing and deaf people

Abstract

Our paper mainly focuses on developing an audio classification for people, who cannot hear properly, using Convolutional Neural Network (CNN) and Recurrent Neural Network (RNN). One of the many prevalent complaints from hearing aid users is excessive background noise. Hearing aids with background noise classification algorithms can modify the response based on the noisy environment. Speech, azan, and ambient noises are all examples of significant audio signals. Whenever a human hears a sound, they can easily identify the sound, however it’s not the same for computers, and we have to feed the algorithm data-sets in order to make it distinguish between different sounds[1]. Hence, we came up with the idea to build a system for people who have problems to hear. We have successfully managed to achieve a total of 98.67%, and 97.01% accuracy after training the data on our CNN and RNN model and testing it respectively.