Advanced automation support developed in Wave 1 of SESAR IR includes using of automatic speech recognition (ASR) to reduce the amount of manual data inputs by air-traffic controllers. Evaluation of controllers feedback has been subdued due to the limited recognition performance of the commercial of the shell ASR engines that were used, even in laboratory conditions. The reasons for the unsatisfactory conclusions include e.g. inability to distinguish controllers accents, deviations from standard phraseology and limited real-time recognition performance. Past exploratory research funded project MALORCA, however, has shown (on restricted use-cases) that satisfactory performance can be reached with novel datadriven machine learning approaches. Based on the results of MALORCA HAAWAII project aims to research and develop a reliable, error resilient and adaptable solution to automatically transcribe voice commands issued by both air-traffic controllers and pilots. The project will build on very large collection of data, organized with a minimum expert effort to develop a new set of models for complex environments of Icelandic en-route and London TMA. HAAWAII aims to perform proof-of-concept trials in challenging environments, i.e. to be directly connected with real-life data from ops room. As pilot read-back error detection is the main application, HAAWAII aims to significantly enhance the validity of the speech recognition models. The proposed work goes far beyond the work planned for the Wave 2 IR programme and will improve both safety and reduce controllers workload. The digitization of controller and pilot voice utterances can be used for a wide variety of safety and performance related benefits including, but not limiting to pre-fill entries into electronic flight strips and CPDLC messages. Another application demonstrated during proof-of-concept will be to objectively estimate controllers workload utilising digitized voice recordings of the complex London TMA.

Popis česky
Cílem projektu je výzkum a vývoj spolehlivého, proti chybám odolného a přizpůsobitelného řešení pro automatický přepis hlasových příkazů vyřčených jak pracovníky řízení letového provozu, tak piloty.

Klíčová slova
Artificial Intelligence , Machine Learning, Air-Traffic Control, Natural Language Processing, Automatic Speech Recognition, 

Klíčová slova česky
umělá inteligence, strojové učení, řízení letového provozu, zpracování přirozeného jazyka, automatické rozpoznávání řeči

H2020-SESAR-2019-2

angličtina

Smrž Pavel, doc. RNDr., Ph.D. - hlavní řešitel
Doležal Jan, Ing. - spoluřešitel
Dytrych Jaroslav, Ing., Ph.D. - spoluřešitel
Hradiš Michal, Ing., Ph.D. - spoluřešitel
Jírovec Martin, Ing. - spoluřešitel
Musil Martin, Ing., Ph.D. - spoluřešitel
Otrusina Lubomír, Ing. - spoluřešitel
Veselý Karel, Ing., Ph.D. - spoluřešitel

Ústav počítačové grafiky a multimédií
- odpovědné pracoviště (28.1.2020 - nezadáno)
Výzkumná skupina znalostních technologií
- interní (28.1.2020 - 30.11.2022)
Ústav počítačové grafiky a multimédií
- příjemce (28.1.2020 - 30.11.2022)

HELMKE, H.; ONDŘEJ, K.; SHETTY, S.; KLEINERT, M.; OHNEISER, O.; EHR, H.; ZULUAGA-GOMEZ, J.; SMRŽ, P. Readback Error Detection by Automatic Speech Recognition and Understanding - Results of HAAWAII project for Isavia's Enroute Airspace. SESAR Innovation Days 2022. Budapest: 2022. p. 1-9.
Detail

PRASAD, A.; ZULUAGA-GOMEZ, J.; MOTLÍČEK, P.; SARFJOO, S.; NIGMATULINA, I.; OHNEISER, O.; HELMKE, H. Grammar Based Speaker Role Identification for Air Traffic Control Speech Recognition. Proceedings of the 12th SESAR Innovation Days. Budapest: 2022. p. 1-9.
Detail

PRASAD, A.; ZULUAGA-GOMEZ, J.; MOTLÍČEK, P.; SARFJOO, S.; NIGMATULINA, I.; VESELÝ, K. Speech and Natural Language Processing Technologies for Pseudo-Pilot Simulator. Proceedings of the 12th SESAR Innovation Days. Budapest: 2022. p. 1-9.
Detail

ZULUAGA-GOMEZ, J.; SARFJOO, S.; PRASAD, A.; NIGMATULINA, I.; MOTLÍČEK, P.; ONDŘEJ, K.; OHNEISER, O.; HELMKE, H. BERTraffic: BERT-based Joint Speaker Role and Speaker Change Detection for Air Traffic Control Communications. In IEEE Spoken Language Technology Workshop, SLT 2022 - Proceedings. Doha: IEEE Signal Processing Society, 2023. p. 633-640. ISBN: 978-1-6654-7189-3.
Detail

NIGMATULINA, I.; ZULUAGA-GOMEZ, J.; PRASAD, A.; SARFJOO, S.; MOTLÍČEK, P. A Two-Step Approach to Leverage Contextual Data: Speech Recognition in Air-Traffic Communications. In ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings. Singapore: IEEE Signal Processing Society, 2022. p. 6282-6286. ISBN: 978-1-6654-0540-9.
Detail

Odpovědnost: Smrž Pavel, doc. RNDr., Ph.D.