ATCOSIM: Air Traffic Control Simulation Speech Corpus

The ATCOSIM Air Traffic Control Simulation Speech corpus is a speech database of air traffic control (ATC) operator speech, provided by Graz University of Technology (TUG) and Eurocontrol Experimental Centre (EEC). It consists of ten hours of speech data, which were recorded during ATC real-time simulations using a close-talk headset microphone. The utterances are in English language and pronounced by ten non-native speakers. The database includes orthographic transcriptions and additional information on speakers and recording sessions.

Getting Started

A brief introduction about the ATCOSIM corpus can be found in the LREC 2008 paper and poster.
You can browse the entire corpus without having to download the full database. 

BROWSE

 

A good starting point is the DOC directory where you can find the readme file as well as the full manual of the corpus. You can also dive right into the corpus (~3MB) to see the transcriptions and listen to audio files. Various tables are located in the HTMLdata directory. The audio recordings themselves are located in the WAVdata directory, sorted by speakers and recording session. The TXTdata directory contains, among other things, the full database (without audio) in the CSV format and a wordlist.
 

Obtaining the Corpus

The corpus is available on-line and provided free of charge. The entire corpus including the recordings and all meta data has a size of approximately 2.5 gigabyte and is available in digital form as electronic ISO disk image.

The disk image is based on an ISO9660 Level 2 file system with Rock Ridge extensions. ISO9660 is the standard cross-platform interchange format for CDs and some DVDs, and is understood by virtually all operating systems. Also the Joliet extensions to ISO9660 and a UDF file system are present on the disk image.

You can browse through the corpus before downloading the full disk image.

DOWNLOAD

(Size: ~2.5 GB)

You will possibly encounter a warning that the certificate of the website is invalid, which you can safely ignore.
If you should receive the image from any other source, you can verify its integrity via this MD5 checksum.
If the download should fail, please contact Konrad Hofbauer to obtain a copy of the ATCOSIM corpus on DVD.

Support

A small MATLAB script demonstrates how to the access the database and search for and select certain types of utterances.
It requires MATLAB 7 (R14) and SoX. You need to adapt the script to your needs! No error checking is performed.
In case of problems or questions concerning ATCOSIM, please contact Konrad Hofbauer.

Contact: 
SPSC Cross Reference
Research Area: 
Acronym: 
ATCOSIM
Speech
Language: 
English
Native: 
Non-native
Gender: 
Female
Male
Content Type: 
Conversational Speech
Segmentation Level: 
Utterance
Transcription Format: 
orthographic plain-text 7-bit ASCII encoding
Laryngograph: 
No
Number of speakers: 
10
Number of utterances: 
10078
Sampling Frequency: 
32kHz
Multi-channel: 
No
Condition: 
Clean
Type: 
Database