Researchers at Embry-Riddle Aeronautical University have created an artificial intelligence (AI)( system designed to transcribe and translate aviation radio communications. The team said the technology addresses longstanding challenges that pilots and air traffic controllers face with unclear transmissions and aviation-specific phraseology.
To convert spoken radio transmissions into standardized text, the system combines automatic speech-recognition technology with natural language processing. The AI refines communications by standardizing terminology, formatting call signs and numbers, removing filler words, and identifying potential errors in pilot-controller exchanges.
Andrew Schneider, assistant professor in the College of Aviation who directs the university’s Speech and Language AI Lab, leads the project alongside Jianhua Liu, associate professor of electrical and computer engineering. Their research received $30,000 in grants from Embry-Riddle’s Boeing Center for Aviation and Aerospace Safety.
Initial testing using off-the-shelf speech recognition tools on recordings from 12 high-traffic U.S. airports revealed significant limitations, with an 80% word error rate. “Aviation English isn’t standard conversational grammar—it’s a condensed, highly specific phraseology spoken over a noisy radio where words get clipped and specialized jargon abounds,” Schneider explained.
Automatic Speech Recognition
Liu’s signal-processing expertise enabled customization of the automatic speech recognition system, reducing the word error rate to below 15%. The improved performance led to the system’s adoption in a NASA-funded project requiring the extraction of flight deck communications from high-background-noise audio.
Kristy Kiernan, associate director of the Boeing Center, noted the technology’s potential to analyze previously untapped qualitative data sources. “Large language models can open up whole new data sources that we can leverage to improve safety,” she said.
Future applications include real-time systems that could interface with aircraft systems to detect inconsistencies between verbal instructions and aircraft behavior, flag missed communications, or assist with checklist verification. Researchers envision the technology serving as an intelligent copilot to enhance situational awareness and prevent communication breakdowns.
The system could also provide immediate feedback to student pilots and help instructors identify specific communication training needs, potentially transforming how aviation communication skills are developed and maintained across the industry.