layout | title |
---|---|
page |
Improving speech-in-noise for hearing devices (Clarity-2025) |
<font color="#882885" size="2pt">
<p><big><i><b>22nd August 2025, Rotterdam, The Netherlands</b></i></big></p>
</font>
We are now pleased to announce the 6th Clarity Workshop for Improving Speech-in-Noise for Hearing Devices, to be collocated with Interspeech 2025, Rotterdam, The Netherlands. This will be an in-person Clarity event, following the format of the 4th Clarity workshop that was a satellite of Interspeech 2023.
Understanding speech in the presence of background noise remains one of the most significant challenges for hearing-impaired listeners. Social noise levels can drastically reduce speech intelligibility, leading to isolation and reduced quality of life. With over 360 million people worldwide affected by disabling hearing impairment, this issue is exacerbated by an aging population. While amplification devices restore audibility, they often fail to address the complexities of noisy environments.
Recent advances in hearing aid devices and machine learning have opened the door to groundbreaking solutions. For example, advancements in low-power digital signal processors (DSPs) and edge AI hardware have equipped hearing aids with the ability to run simplified versions of deep learning models locally. Further, low energy wireless communication between hearing aids and mobile devices enables computationally demanding tasks to be offloaded, overcoming traditional power constraints. Techniques from statistical speech modeling and acoustic scene analysis can now be harnessed for hearing aid processing.
The Clarity Workshop provides a platform for collaboration between speech technologists, hearing researchers, and hearing aid developers. By fostering dialogue, we aim to inspire innovative approaches to addressing speech-in-noise challenges in real-world settings.
We welcome any work related to speech-in-noise processing for hearing devices, including (but not limited to) the following areas:
- Binaural technology for speech enhancement and source separation
- Multi-microphone processing techniques
- Real-time and low-latency approaches to speech enhancement
- Statistical model-driven approaches to hearing aid processing
- Machine learning for diverse target listeners
- Machine learning models of hearing impairment
- DNN-based speech enhancement under model size and low power constraints
- Audio quality and intelligibility assessment for hearing aid and cochlear implant users
- Efficient integration of psychoacoustic testing in machine learning
Clarity-2025 will host the 3rd Clarity Prediction Challenge, focused on developing innovative intrusive and non-intrusive methods for predicting hearing-aid speech intelligibility. The Challenge launches on 1st March. If you’re interested in participating, please sign up for our Google group to receive further updates.
<div class="board-list">
<ul>
<li> Michael Akeroyd<br /><a href="#" class="text-muted">University of Nottingham, UK</a></li>
<li> Jon Barker<br /> <a href="#" class="text-muted">University of Sheffield, UK </a></li>
<li> Fei Chen<br /> <a href="#" class="text-muted">Southern University of Science and Tech., China</a></li>
<li> Trevor Cox<br /> <a href="#" class="text-muted">University of Salford, UK </a></li>
<li> John Culling<br /> <a href="#" class="text-muted">University of Cardiff, UK</a></li>
<li> Jordi de Vries <br /> <a href="#" class="text-muted">TU Ghent, The Netherlands</a></li>
<li> Simone Graetzer<br /> <a href="#" class="text-muted">University of Salford, UK </a></li>
<li> Graham Naylor<br /> <a href="#" class="text-muted">University of Nottingham, UK</a></li>
<li> <a href="organisers.html"> More info </a> </li>
</ul>
</div>
<h2> Keynote Speakers </h2>
<div class="board-list">
<ul>
<h2> Invited Talks </h2>
<div class="board-list">
<ul>
</div>
<h2> Registration </h2>
<div class="board-list">
<ul>
<li> Registration opens in April 2025.</li>
<!-- <li> Paper submission deadline 2nd June</li> -->
<!-- <li> <a href="dates.html">More info</a> </li> -->
</ul>
</div>
</div>