Enhancing noisy group conversations for people with hearing loss
CHiME-9 ECHI: A Machine Learning Challenge for Enhancing Conversations to Address Hearing Impairment
Computation and Language
Summary
Hearing conversations in noisy places like cafeterias is hard for people with hearing loss. The authors ran a challenge where teams created machine learning systems to make speech clearer and easier to understand from recordings made with special glasses and hearing aids. They found that the best systems improved how well people could understand and enjoy the conversations, even if the usual computer measures didn’t always show this improvement. This work helps develop better hearing support technology for real-life noisy settings.
What this means in practice
- •For hearing aid developers: Improve algorithms that enhance speech clarity in hearing aids for noisy group conversations.$Commercial implications: Makes possible hearing aids with better real-world performance by enhancing multi-speaker speech in noisy settings using machine learning.
- •For smart glasses manufacturers: Incorporate speech enhancement features using multi-microphone recordings to assist hearing-impaired users in social environments.
Authors
Robert Sutherland, Thomas Kuebert, Marko Lugger, Stefan Petrausch, Eline Borch Petersen, Juan Azcarreta Ortiz, Buye Xu, Stefan Goetze, Jon Barker
Abstract
This work presents the task and results of the CHiME-9 challenge for Enhancing Conversations to address Hearing Impairment. The challenge considers the scenario of four-party conversations in a noisy, cafeteria-style environment with interfering speech sources and sound effects. Participants are provided with audio recordings made with Meta Aria glasses and hearing aid microphones, and clean speech samples of the conversation participants. The task is to extract the speech of the conversation partners from the noisy multi-channel recordings with the goal of improving the intelligibility and quality of the speech, evaluated using objective metrics and subjective listening tests. This paper reviews submissions from seven teams and ranks them on a combination of subjective intelligibility and quality. Results show that while the objective metrics do not reflect listener performance, the top systems were able to make substantial improvements over the challenge baseline in both intelligibility and quality ratings.