BLeRIOT-A corpus: raw cockpit speech

A cockpit-oriented corpus was recorded, containing French and English pilot vocabulary from French-native speakers (ICAO English median level of 5). It is composed of two sub-corpora, recorded by BEA (civil) and by DGA Essais propulseurs (military). BEA Volunteers (9 males and 3 females; 24- to 64-year-old) with experience as pilots (private, airline, military; median number of flight hours: 475), or as air traffic controllers, were recorded while reading airplane take-off and checklists (Boeing, Airbus; total of 177 min) and non-sensitive transcripts from Airbus airplane CVRs (total of 47 min). Speakers wore a headset to hear their own voice as in a cockpit. The measurements were performed in a room with partial audio treatment on walls and an anti-sound-reflection screen (Reflexound, Plugger Sound) around a low-noise directional microphone (MKH 60 P48, Sennheiser), located at close range from the speaker, and plugged into a portable recorder (H4n Pro, Zoom; 32-bit float wave files at 44.1 kHz. DGA Volunteers (6 males, 22- to 56-year-old) with experience as pilots (private, flight engineer; median number of flight hours: 100), air traffic controllers or CVR analysts, were recorded while reading aircraft checklists (total of 21 min) as well as non-sensitive transcripts (total of 14 min). A portable recorder (H6, Zoom; 16-bit float wave files at 44.1 kHz) and its integrated stereo microphones were used, located at close range from the speaker. Only the left channels were kept.