When people board an airplane, they enter an environment where the passenger experience has been carefully engineered. But for millions of travelers who are Deaf or Hard of Hearing (DHH), one vital part of that experience — spoken communication — is missing. Meal service options, turbulence notices, gate changes — all are moments when hearing passengers are informed, and others are isolated.
What’s happening: Boeing has demonstrated a speech-to-text transcription (STT) concept that displays captions of cabin announcements in nearly real time.
- Driven by artificial intelligence (AI) and designed to function offline, the developing technology could bridge a long-standing accessibility gap and make air travel more inclusive.
Why it matters: “With each accessibility advancement, our global team is building toward a more inclusive travel experience for everyone,” said Boeing engineer Ashley Badger.
What they're saying: “The real problem wasn’t technology,” said Bill Harkness, Boeing accessibility engineering leader. “It was access. It was dignity.”
Harkness, who is Deaf, has long advocated for better communication tools for passengers who, like him, rely on visual information.
“This was our chance to change that,” he said. “To build a system that doesn’t assume one way of listening fits all.”
Above the noise: The STT concept emerged as Boeing engineers and research scientists began to collaborate.
Jinri Kim, a machine learning researcher at the Boeing Korea Engineering & Technology Center (BKETC), recalls how a simple idea gained momentum.
“Developing the prototype, we had to select a model that could understand context — what’s a crew announcement and what isn’t,” Kim said.
- Aircraft cabins present serious technical challenges, including persistent engine noise, overlapping audio sources and limited Wi-Fi connectivity.
- To meet those constraints, the BKETC AI team evaluated multiple open-source architectures and adapted a model to work in noisy, offline conditions.
- The system processes incoming audio in two-second chunks, allowing near-real-time captions to appear on local devices or embedded displays.
