SELF-MADE-ATC
With the “SELF-MADE-ATC” project, DLR researchers address key challenges in maintaining and customizing automatic speech recognition and understanding systems (ASRU) in the air traffic control domain. Building on a successful technology transfer to the industry, SELF-MADE-ATC introduces an automated process to maintain and customize such ASRU systems.
The new SELF-MADE-ATC system eliminates labour-intensive manual processes when adapting these systems to new airports or applications. This reduces the costs and risks associated with system customisation and maintenance. These costs can otherwise quickly exceed the usual licence fees by a factor of 3 to 7, when using ASRU technology.
A positive side effect of SELF-MADE-ATC is the availability of a vendor independent speech recognizer as part of the adaptation and maintenance pipeline, to streamline the speech-to-text process and the conversion into concepts. The goal is to ensure consistently high recognition accuracy at both the word and command levels, even within dynamic airspace environments due to, for example, monthly airspace adjustments.
Automatic speech recognition and understanding needs a high amount of training data, i.e. recordings from voice communication between aircraft pilots and air traffic controllers. Voice recordings are personal data and are, therefore, sensitive data, the processing of which may give rise to legal problems. SELF-MADE-ATC enables that the model training is performed directly in the environment of the end-users, i.e. sensitive voice recordings can remain within air traffic control centres.
Interested in participating or in testing SELF-MADE-ATC? Then get in touch with the project team!
Key data
Project | SELF-MADE-ATC |
Participants | DLR Institute for Flight Guidance (Coordinator) DLR Transfer Management DLR Economic Cooperation |
Duration | 2025 – 2027 |
Funding | Institutional Funding |
Website |
