Design and Development of Speech Corpora for Air Traffic Control Training

Design and Development of Speech Corpora for Air Traffic Control Training Luboˇs ˇSm´ıdl, Jan ˇSvec, Daniel Tihelka, Jindˇrich Matouˇsek, Jan Romportl, Pavel Ircing Department of Cybernetics, University of West Bohemia Univerzitn´ı 8, 306 14 Plzeˇn, Czech Republic {smidl, honzas, dtihelka, jmatouse, rompi, ircing}@kky.zcu.cz Abstract The paper describes the process of creation of domain-specific speech corpora containing air traffic control (ATC) communication prompts. Since the ATC domain is highly specific both from the acoustic point-of-view (significant level of noise in the signal, non-native English accents of the speakers, non-standard pronunciation of some frequent words) and the lexical and syntactic perspective (prescribed structure of utterances, rather limited vocabulary), it is useful to collect and annotate data from this specific domain. Actually, the ultimate goal of the research effort of our team was to develop a voice dialogue system simulating the responses of the pilot that could be used for training aspiring air traffic controllers. In order to do so, we needed – among other modules – a domain-specific automatic speech recognition (ASR) and text-to-speech synthesis (TTS) engines. This paper concentrates on the details of the ASR and TTS corpora creation process but also overviews their usage in preparing practical applications and provides links to the distribution channel of the data. Keywords: speech corpus, speech recognition, text-to-speech, air traffic control communication 1. Introduction The air traffic control (ATC) constitutes a crucial segment of the whole air traffic industry – the air traffic controllers communicate with the pilots almost continuously in order to ensure the fluent and safe flow of the aerial traffic. The job of a controller is very demanding and requires – besides the specific personal prerequisites – an intensive training. This training is mainly focused on teaching and reinforcing the communication skills of the aspiring controller. The current state-of-the-art training procedure (at least in the Czech Republic) involves so-called pseudopilots. These are usually retired pilots that prepare training scenarios and consequently act as pilots of virtual plane (usually more than one at a time), communicate with the controller in training (trainee) and process the spoken prompts re- ceived from trainees to the form that can be entered into the software that simulates the plane movement on the radar screen. Two major drawbacks were identified in such a training set- ting. The first is rather obvious – the length of the controller training (approx. 2 years on average) and the relatively high salaries of the pseudopilots make the whole process very expensive. This was actually the first incentive that sparked the idea of developing an automatic training simu- lator based on the intelligent spoken dialogue system. Only after delving into the specifics of the ATC communication, we have realized that the scenario involving the pseudopi- lots in actually quite unrealistic in several aspects: First, In the real ATC scenarios, the controller will need to understand the English utterances pronounced mostly by non-native speakers (remember that we are talking about the air space situated in the Central Europe), sometimes with quite an exotic …

Content truncated to 500 words