Jv
Jan-Willem van Doorn
info
Please Note
<p>This page displays the records of the person named above and is not linked to a unique person identifier. This record may need to be merged to a profile.</p>
2 records found
1
Whisper-ATC
Open Models for Air Traffic Control Automatic Speech Recognition with Accuracy
Current advancements in machine learning have provided new architectures, such as encoder-decoder transformers, for automatic speech recognition. For generic speech recognition, very high accuracies are already achievable. However, in air traffic control, automatic speech recognition models traditionally rely on domain-specific models constructed from limited training data. This study introduces this newly developed transformer model for air traffic control and provides a set of fully open automatic speech recognition models with high accuracies. This paper demonstrates how a large-scale, weakly supervised automatic speech recognition model, Whisper, is fine-tuned with various air traffic control datasets to improve model performance. We also evaluated the performance of different sizes of Whisper models. In the end, it was possible to achieve word error rates of 13.5% on the ATCO2 dataset and 1.17% on the ATCOSIM dataset with a random split (or 3.88% with speaker split). The study also reveals that finetuning with region-specific data can enhance performance by up to 60% in real-world scenarios. Finally, we have open-sourced the code base and the models for future research.
...
Current advancements in machine learning have provided new architectures, such as encoder-decoder transformers, for automatic speech recognition. For generic speech recognition, very high accuracies are already achievable. However, in air traffic control, automatic speech recognition models traditionally rely on domain-specific models constructed from limited training data. This study introduces this newly developed transformer model for air traffic control and provides a set of fully open automatic speech recognition models with high accuracies. This paper demonstrates how a large-scale, weakly supervised automatic speech recognition model, Whisper, is fine-tuned with various air traffic control datasets to improve model performance. We also evaluated the performance of different sizes of Whisper models. In the end, it was possible to achieve word error rates of 13.5% on the ATCO2 dataset and 1.17% on the ATCOSIM dataset with a random split (or 3.88% with speaker split). The study also reveals that finetuning with region-specific data can enhance performance by up to 60% in real-world scenarios. Finally, we have open-sourced the code base and the models for future research.
Conference paper
(2021)
-
Junzi Sun, Emmanuel Sunil, Ralph Koerse, Stijn van Selling, Jan-Willem van Doorn, Thomas Brinkman
The METeo Sensors in the Sky (METSIS) project, funded by SESAR’s Engage knowledge transfer network, investigated the use of drones as an aerial wind sensor network for U-space applications. The concept aims to provide accurate, lowcost and hyperlocal wind nowcasts for drones using data collected by drones themselves and the Meteo-Particle Model (MPM) for wind field reconstruction. In this paper, we describe the METSIS concept and a proof-of-concept experiment that was performed using four drones to determine the feasibility and accuracy of the concept at low altitudes. For the experiment, ultrasonic anemometers were mounted to each drone to measure local winds. The calibration of the wind sensors was tested using the NLR Anechoic Wind Tunnel. Subsequently, flight-tests were performed at the NLR Drone Center to evaluate the effect of obstacles, drone motion, measurement density, and measurement errors on concept accuracy. Wind fields estimated during the flight-tests were published to the AirHub Drone Operations Center (DOC) system to demonstrate the communication of this data to U-space end-users in real-time. The results indicated that the METSIS concept is a promising solution for the wind nowcast component of the U-space weather information service. Further research is planned to improve the accuracy and sclability of the METSIS concept.
...
The METeo Sensors in the Sky (METSIS) project, funded by SESAR’s Engage knowledge transfer network, investigated the use of drones as an aerial wind sensor network for U-space applications. The concept aims to provide accurate, lowcost and hyperlocal wind nowcasts for drones using data collected by drones themselves and the Meteo-Particle Model (MPM) for wind field reconstruction. In this paper, we describe the METSIS concept and a proof-of-concept experiment that was performed using four drones to determine the feasibility and accuracy of the concept at low altitudes. For the experiment, ultrasonic anemometers were mounted to each drone to measure local winds. The calibration of the wind sensors was tested using the NLR Anechoic Wind Tunnel. Subsequently, flight-tests were performed at the NLR Drone Center to evaluate the effect of obstacles, drone motion, measurement density, and measurement errors on concept accuracy. Wind fields estimated during the flight-tests were published to the AirHub Drone Operations Center (DOC) system to demonstrate the communication of this data to U-space end-users in real-time. The results indicated that the METSIS concept is a promising solution for the wind nowcast component of the U-space weather information service. Further research is planned to improve the accuracy and sclability of the METSIS concept.