By Haoshuai Zhou & Linkai Li, Orka Labs Inc.
We can now predict speech intelligibility before a patient’s first hearing aid trial. This tool provides an objective forecast to help clinicians optimize initial settings and manage patient expectations from day one.
The Universal Fitting Challenge
Every hearing care professional knows the scene: a patient returns from their first real-world test of new hearing aids, looking frustrated. “It was louder,” they say, “but I still couldn’t follow the conversation.”
This moment highlights the central challenge in our field. Making sounds audible is a technical achievement; making speech understandable is a human one. We bridge this gap through careful fine-tuning and verification—a process that, while effective, often relies on iterative cycles of trial and error that cost time and test patience.
The Promise of a First-Fit Forecast
Imagine moving beyond the time-consuming, back-and-forth cycle of reactive adjustments. Instead, you could start with a proactive forecast, enabling better expectation management and a more confident beginning for your patients. In short, you start closer to success.
This is no longer a hypothetical question. By analyzing the signal from a hearing aid in a noisy environment and the individual’s hearing profile, we can generate a reliable estimate of speech intelligibility. The goal is a closer first fit, so clinicians and users begin closer to success.
What a First-Fit Forecast Looks Like
This forecast is an intelligibility predictor. In practice, it works by analyzing two key pieces of information: the audio signal actually produced by the hearing aid in a noisy environment, and the individual patient’s audiogram. The tool then generates an easy-to-interpret output, such as a “high, medium, or low” expected comprehension score for that specific situation.
It is not a black box replacement for clinical judgment, but a data-driven aid designed to be consulted at the fitting desk.
Why a Smarter Starting Point Changes Everything
The traditional fitting loop is costly. A more predictive approach delivers immediate value across the entire ecosystem:
- For Clinicians: It transforms the first fitting from a best guess into a strategic conversation. You can set realistic expectations and collaboratively choose a starting point based on predicted performance in relevant listening scenarios.
- For Patients: It demystifies the process. Instead of an abstract promise of “better hearing,” they see a concrete, data-informed starting point, increasing confidence and engagement from the very beginning.
- For Manufacturers: It accelerates innovation by providing a rapid, objective metric for evaluating new algorithm designs long before lengthy human trials.
Grounded in Independent Evidence
The feasibility of this approach was recently put to the test in the 3rd Clarity Prediction Challenge (CPC3), an open benchmark managed by the UKRI-funded Clarity Project. The challenge tasked participants with predicting speech intelligibility for hearing-aid-processed speech in noise, comparing a wide range of algorithmic approaches.
The winning entry in this independent challenge demonstrated remarkably low prediction error, providing a strong proof point that an approach leveraging the final sound signal and the patient’s audiogram is not only viable but powerful. (CPC3 results can be viewed here)
A Partner for Clinical Wisdom
Think of this tool as a sophisticated dashboard, not an autopilot. It provides a clear readout of the road ahead—offering data on predicted performance. But the steering, the navigation, and the final decisions in complex traffic always remain in your hands.
It is designed to make you a more confident driver, not to take the wheel.
When the prediction aligns with real-ear measures and patient feedback, it provides powerful confirmation. When it diverges, the clinician’s expertise and the patient’s report must always guide the final decision.
From Benchmark to Clinical Practice
In the CPC3 benchmark, systems were ranked by the statistical root-mean-square error (RMSE) between predicted and actual scores. For everyday clinical use, this raw score can be translated into an intuitive, banded output—such as high, medium, or low expected comprehension—giving the clinician an immediate, actionable insight.
Conclusion: A More Confident Beginning
The question is no longer if we can predict speech understanding, but how this new capability will enhance our practice. By integrating an intelligibility forecast into the first fit, we shift the paradigm from fixing problems to preventing them. This means fewer “louder but…” moments, and more patients achieving the “I understand” breakthrough they seek—right from the start.
About the Authors
Haoshuai Zhou is Head of Algorithms at Orka Labs Inc, leading the development of predictive models for hearing technology. LinkedIn
Linkai Li is Co-Founder and Head of Product at Orka Labs Inc, driving user-centered innovation in hearing care. LinkedIn
