Detecting English Speech in the Air Traffic Control Voice Communication

TL;DR

This paper presents a robust, lightweight English language detection system for air traffic control speech, significantly outperforming previous acoustic-based methods in both in-domain and out-of-domain scenarios.

Contribution

The proposed Bayesian subspace multinomial model-based ELD system is novel in using ASR-derived embeddings, achieving superior accuracy and robustness over existing x-vector based systems.

Findings

Achieved 0.0439 EER in in-domain detection.

Reduced EER by 50% compared to x-vector systems.

Reduced EER by 33% in out-of-domain detection.

Abstract

We launched a community platform for collecting the ATC speech world-wide in the ATCO2 project. Filtering out unseen non-English speech is one of the main components in the data processing pipeline. The proposed English Language Detection (ELD) system is based on the embeddings from Bayesian subspace multinomial model. It is trained on the word confusion network from an ASR system. It is robust, easy to train, and light weighted. We achieved 0.0439 equal-error-rate (EER), a 50% relative reduction as compared to the state-of-the-art acoustic ELD system based on x-vectors, in the in-domain scenario. Further, we achieved an EER of 0.1352, a 33% relative reduction as compared to the acoustic ELD, in the unseen language (out-of-domain) condition. We plan to publish the evaluation dataset from the ATCO2 project.