Understanding Tower Controller Communication for Support in Air Traffic Control Displays

Automatic speech recognition and understanding for air traffic control (ATC) communication has been deeply explored in the approach and en-route environment. Solution 97.2 is one of the first European attempts to analyze recognition rates and human performance of air traffic controllers (ATCos) in simulated tower and ground environments. Three validation exercises with 22 ATCos from four different European air navigation service providers were conducted in Germany, Norway, and Italy. The validated artificial intelligence-based prototypes of Assistant Based Speech Recognition systems (ABSR) supported ATCos fulfilling tasks in a ground and tower environment as well as multiple remote tower environment, respectively. Thus, in any relevant ATC display, (1) recognized callsigns of ATCo utterances have been highlighted, (2) fully recognized commands were shown, and (3) the ATCo was able to manually manipulate the ABSR output if needed or the output was automatically accepted by the ATC system otherwise. This paper evaluates callsign and command type recognition rates as well as ATCo performance. It compares the results for the three validation exercises. A callsign recognition rate of more than 98%, a command recognition rate of more than 91%, and a slight reduction in ATCo workload are achieve