blog — ATCO2: Automatic collection and processing of voice data from air-traffic communications

Air Traffic Control Conversations Collection – A legal introduction by ELDA

— by Petr Motlicek — last modified Sep 02, 2020 01:55 PM

ELRA, the European Language Resources Association, and its distribution agency, ELDA, have been funded in 1995 and have been a world-wide leading player in distributing Language Resources and providing other services to the speech language communities. In the course of the ATCO2 project, ELDA provides legal expertise for the collection of Air Traffic Control Conversations and Data Management in the project.

Is it possible to have a cross-accent speech recognizer that works for different airports?

— by Petr Motlicek — last modified Sep 02, 2020 01:56 PM

One of the main concerns in automatic speech recognition (ASR) for air-traffic communications is the presence of several non-English accents in the English communications between pilots and air traffic controllers (ATCO).

Bringing together what belongs together: Matching voice commands and radar data

— by Petr Motlicek — last modified Sep 02, 2020 01:57 PM

The whole ATCO² Project relies on two input streams. One input stream is represented by the voice command issued by air-traffic controler and the other input stream is provided by the automatic dependent surveillance-broadcast data. An independent collection of both streams makes a matching of the data inevitable. The following lines will give an insight in how the matching process is done.

Automatic speech recognition, how it works?

— by Petr Motlicek — last modified Aug 03, 2020 10:17 AM

[Updated, 3.4.2020]: ATCO2 project is closely aligned with the development of automatic speech recognition engines for Air-Traffic Controllers (ATCOs), particularly to automatically transcribe their communication with the pilots. This blogpost is giving some insight into the process of Automatic Speech Recognition, current trends, and some details on how it will be integrated in the ATCO2 project. We are describing Hybrid HMM-based speech recognizer, which is the current state-of-the-art speech recognition paradigm. The literature also suggests end-to-end systems. However, we did not consider using these, due to practical reasons. We use the toolkit Kaldi [1], both for training the baseline models, and processing the untranscribed data.

The annotation started

— by Petr Motlicek — last modified Sep 02, 2020 01:59 PM

[Updated 1. 2. 2020]: The work on ATCO2 project started during Christmas. We kicked-off one of the important phases, which is a data collection and transcription. Why do we need to start with this? Well, this will be explained in larger context below!

Kick-off meeting

— by admin — last modified Dec 03, 2019 03:56 PM

The kick-off meeting took place at Brno University of Technology on November 22, 2019.