Automatic Speech Recognition for Air Traffic Control Communications

Abstract

A significant fraction of communications between air traffic controllers and pilots is through speech, via radio channels. Automatic transcription of air traffic control (ATC) communications has the potential to improve system safety, operational performance, and conformance monitoring, and to enhance air traffic controller training. We present an automatic speech recognition model tailored to the ATC domain that can transcribe ATC voice to text. The transcribed text is used to extract operational information such as call-sign and runway number. The models are based on recent improvements in machine learning techniques for speech recognition and natural language processing. We evaluate the performance of the model on diverse datasets.

Natural Language Processing (NLP) in Aviation Safety: Systematic Review of Research and Outlook into the Future

“…In terms of evaluation, no common method is observed from selected studies, potentially because of the diversity of applications. However, the F score was used to assess the accuracy of aircraft call-sign extraction and runway number extraction [8]. An important observation from this study [8] is that the F-scores decrease because of the errors from ASR, which reemphasized the importance of ASR accuracy when integrating ASR and information extraction in the same system.…”

“…However, the F score was used to assess the accuracy of aircraft call-sign extraction and runway number extraction [8]. An important observation from this study [8] is that the F-scores decrease because of the errors from ASR, which reemphasized the importance of ASR accuracy when integrating ASR and information extraction in the same system. Other evaluation methods adopted comparisons with different classifiers, such as support vector machine (SVM), k-nearest neighbors (k-NN), and random forest (RF) [16,45].…”

“…The selected articles show that the end-to-end speech recognition framework is a study trend for better ASR strategies over traditional HMM-based models, especially given the challenges of multilingual recognition, special use of language, and high speech rate. Among the five articles which include ASR for aviation as a primary research objective, three articles adopted an end-to-end framework [3,8,17]. Neural networks-based methods are also widely used in ASR.…”

“…The combined uses of LSTM, RNN, CNN, and DNN were explored in different studies; the bidirectional encoder representation from transformers (BERT) model was adopted by one study to check speech emotion [46]. In addition, name entity recognition, rule-based grammar, and Bayesian networks were also used [8,42]. In terms of evaluation, no common method is observed from selected studies, potentially because of the diversity of applications.…”

“…However, the analysis of such textual-based data usually requires significant human resource investment [7]. Considerable efforts in terms of time and money are needed for work, such as voice recognition [8], text mining and identification [9,10], topic categorization [11,12], and semantic reasoning [13]. Therefore, automatic and practical analytic approaches are needed to overcome these labor-intensive challenges.…”