Speech Enhancement and Denoising Audio for Hard-of-Hearing People in Universities

Technology for hearing aids is constantly evolving to make people with hearing problems feel more at home in the world. Every hearing aid model should be adjusted in advance to the individual's specific type of hearing loss. However, many people cannot afford such cutting-edge technology. In th...

Full description

Saved in:
Bibliographic Details
Main Authors: Kanev, Anton, Shapovalova, Veronika
Format: Conference Proceeding
Language:English
Subjects:
Auditory system
BLSTM
Fine Tuning
Generative adversarial networks
Hearing aids
HOH
Measurement
Neural Network
Noise
Noise reduction
Power engineering
Spectral Mask
Speech enhancement
Online Access:Request full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Technology for hearing aids is constantly evolving to make people with hearing problems feel more at home in the world. Every hearing aid model should be adjusted in advance to the individual's specific type of hearing loss. However, many people cannot afford such cutting-edge technology. In the Bauman Moscow State University there are groups with DHH persons. The assumption is that they might have troubles with some areas of studying. At first, the idea to create an app to process audio recordings of lectures appeared. There are many models for processing audio data. In this paper, it is an experiment to reduce specific noises that commonly occur in university (such as chalk on a blackboard, students' whisper in class etc.). Models are not pre-trained on such data, therefore, it is assumed the performance quality of existing models might be unsatisfying. Moreover, it is supposed that fine tuning existing models could increase audio quality in target cases.
ISSN:2831-7262
DOI:10.1109/REEPE60449.2024.10479921