Noise Filtering Mobile Application for Speech Enhancement using a Redundant Convolutional Encoder-Decoder

초록

Hearing aids are the most commonly used devices for hearing impairment. It has been proven to be effective in helping patients hear better, especially in a quiet and controlled environment. However, while hearing aids help amplify sounds, it does not consider noise; thus it can worsen hearing impairment in the long run. To address this issue, a system capable of filtering sounds in real-time using Redundant Convolutional Encoder-Decoder (R-CED) is developed. R-CED is implemented by extracting redundant representations of noise at the encoder and produces speech as an output. The results of the experiments have shown the accuracy of the proposed system as it was tested in noisy environments. Furthermore, it has been proven that there is no significant difference between the system prototype and noise-canceling microphone.