Publication Details

Automatic Speech Recognition Benchmark for Air-Traffic Communications

ZULUAGA-GOMEZ Juan, MOTLÍČEK Petr, ZHAN Qingran, VESELÝ Karel and BRAUN Rudolf. Automatic Speech Recognition Benchmark for Air-Traffic Communications. In: Proceedings of Interspeech 2020. Shanghai: International Speech Communication Association, 2020, pp. 2297-2301. ISSN 1990-9772. Available from: https://isca-speech.org/archive/Interspeech_2020/pdfs/2173.pdf
Czech title
Srovnávací test automatického rozpoznávání řeči pro hlasovou komunikací v leteckém provozu
Type
conference paper
Language
english
Authors
Zuluaga-Gomez Juan (IDIAP)
Motlíček Petr, Ing., Ph.D. (IDIAP)
Zhan Qingran (IDIAP)
Veselý Karel, Ing., Ph.D. (DCGM FIT BUT)
Braun Rudolf (IDIAP)
URL
Keywords

Speech Recognition, Air Traffic Control, Transfer Learning, Deep Neural Networks, Lattice-Free MMI

Abstract

Advances in Automatic Speech Recognition (ASR) over the last decade opened new areas of speech-based automation such as in Air-Traffic Control (ATC) environments. Currently, voice communication and data links communications are the only way of contact between pilots and Air-Traffic Controllers (ATCo), where the former is the most widely used and the latter is a non-spoken method mandatory for oceanic messages and limited for some domestic issues. ASR systems on ATCo environments inherit increasing complexity due to accents from non- English speakers, cockpit noise, speaker-dependent biases and small in-domain ATC databases for training. Hereby, we introduce CleanSky EC-H2020 ATCO2, a project that aims to develop an ASR-based platform to collect, organize and automatically pre-process ATCo speech-data from air space. This paper conveys an exploratory benchmark of several state-ofthe- art ASR models trained on more than 170 hours of ATCo speech-data. We demonstrate that the cross-accent flaws due to speakers accents are minimized due to the amount of data, making the system feasible for ATC environments. The developed ASR system achieves an averaged word error rate (WER) of 7.75% across four databases. An additional 35% relative improvement in WER is achieved on one test set when training a TDNNF system with byte-pair encoding.

Published
2020
Pages
2297-2301
Journal
Proceedings of Interspeech - on-line, vol. 2020, no. 10, ISSN 1990-9772
Proceedings
Proceedings of Interspeech 2020
Conference
Interspeech, Shanghai, CN
Publisher
International Speech Communication Association
Place
Shanghai, CN
DOI
UT WoS
000833594102086
EID Scopus
BibTeX
@INPROCEEDINGS{FITPUB12404,
   author = "Juan Zuluaga-Gomez and Petr Motl\'{i}\v{c}ek and Qingran Zhan and Karel Vesel\'{y} and Rudolf Braun",
   title = "Automatic Speech Recognition Benchmark for Air-Traffic Communications",
   pages = "2297--2301",
   booktitle = "Proceedings of Interspeech 2020",
   journal = "Proceedings of Interspeech - on-line",
   volume = 2020,
   number = 10,
   year = 2020,
   location = "Shanghai, CN",
   publisher = "International Speech Communication Association",
   ISSN = "1990-9772",
   doi = "10.21437/Interspeech.2020-2173",
   language = "english",
   url = "https://www.fit.vut.cz/research/publication/12404"
}
Back to top