Automated speech recognition in ATC environment

Air Traffic Management (ATM) in general, and Air Traffic Control (ATC) systems, in particular, are commonly limited by their capacity, as they are intended to balance demand in order to provide the best possible performance to the overall network. In the case of ATM, Air Traffic Controllers (ATCO) workload is identified as one of the main limiting factors to increase overall system capacity and properly match it to demand. ATCO workload can be measured by various methods, being one of them the identification of all the control events that he performs during his task. One key aspect in ATC is communication with pilots via voice (transmission and reception). Even though new technologies enable datalink communications, voice is still the main source of Ground-Air communications. Thereby, Automated Speech Recognition in ATM communications will lead to automated transcription of communications and, consequently -with a proper system at the back-endto automated detection of the control events that the controller has performed using voice, allowing calculation of estimated workload. This paper describes a prototype developed to perform this Automated Speech Recognition and controllers event detection, as well as the methodology used to reach it.