Related Experiment Video
Updated: Aug 9, 2025
06:04
Systematic Hearing Performance Evaluation Process for Adolescents with Cochlear Implantation at Early Ages
Published on: March 24, 2023
431
Restoring speech intelligibility for hearing aid users with deep learning
Peter Udo Diehl1,2, Yosef Singer3, Hannes Zilly3
1Audatic, Berlin, Friedrichstr. 210, 10117, Berlin, Germany. peter.u.diehl@gmail.com.
Scientific Reports
|February 15, 2023
Summary
A new deep learning algorithm significantly improves speech understanding for hearing aid users in noisy environments. This advanced noise suppression technology restores intelligibility to normal hearing levels, offering hope to millions worldwide.
Area of Science:
- Artificial Intelligence
- Audiology
- Signal Processing
Background:
- Millions worldwide experience disabling hearing loss, with hearing aids offering only partial compensation.
- Understanding speech in noisy environments remains a significant challenge for hearing aid users.
Purpose of the Study:
- To develop a deep learning algorithm for selective noise suppression to enhance speech intelligibility.
- To restore speech understanding in hearing aid users to the level of individuals with normal hearing.
Main Methods:
- A deep neural network was trained on a large dataset of noisy speech signals.
- Neural architecture search and a novel deep learning-based intelligibility metric were used for optimization.
- The algorithm operates on a single microphone, unlike traditional beamforming methods.
Main Results:
- The algorithm achieved state-of-the-art denoising performance across various noise types.
- Speech intelligibility was restored to the level of normal-hearing control subjects.
- The system demonstrated real-time processing capabilities on a laptop.
Conclusions:
- Deep learning-based noise suppression offers a promising solution for improving hearing aid effectiveness.
- This technology has the potential for large-scale deployment on hearing aid chips, significantly improving quality of life for the hearing impaired.
