eduzhai > Applied Sciences > Engineering >

Automatic Speech Recognition Benchmark for Air-Traffic Communications

  • Save

... pages left unread,continue reading

Document pages: 5 pages

Abstract: Advances in Automatic Speech Recognition (ASR) over the last decade openednew areas of speech-based automation such as in Air-Traffic Control (ATC)environment. Currently, voice communication and data links communications arethe 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 methodmandatory for oceanic messages and limited for some domestic issues. ASRsystems on ATCo environments inherit increasing complexity due to accents fromnon-English speakers, cockpit noise, speaker-dependent biases, and smallin-domain ATC databases for training. Hereby, we introduce CleanSky EC-H2020ATCO2, a project that aims to develop an ASR-based platform to collect,organize and automatically pre-process ATCo speech-data from air space. Thispaper conveys an exploratory benchmark of several state-of-the-art ASR modelstrained on more than 170 hours of ATCo speech-data. We demonstrate that thecross-accent flaws due to speakers accents are minimized due to the amount ofdata, making the system feasible for ATC environments. The developed ASR systemachieves an averaged word error rate (WER) of 7.75 across four databases. Anadditional 35 relative improvement in WER is achieved on one test set whentraining a TDNNF system with byte-pair encoding.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×