Artificial intelligence in a communication system for air traffic controllers' emergency training

Abstract

In the last few years, there has been a lot of research into the use of machine learning for speech recognition applications. However, applications to develop and evaluate air traffic controllers' communication skills in emergency situations have not been addressed so far. In this study, we proposed a new automatic speech recognition system using two architectures: The first architecture uses convolutional neural networks and gave satisfactory results: 96% accuracy and 3% error rate on the training dataset. The second architecture uses recurrent neural networks and gave very good results in terms of sequence prediction: 99% accuracy and 𝑒 −7% error rate on the training dataset. Our intelligent communication system (ICS) is used to evaluate aeronautical phraseology and to calculate the response time of air traffic controllers during their emergency management. The study was conducted at International Civil Aviation Academy, with third-year air traffic control engineering students. The results of the trainees' performance prove the effectiveness of the system. The instructors also appreciated the instantaneous and objective feedback.