In the context of the MEVAC project, Vocapia in partnership with LIMSI, will study solutions to monitor vocal exchanges in the cockpit of fighter planes. The goal is to develop automatic monitoring that can be used both under operational conditions as well as on simulators in order to detect critical situations during missions and for post-mission debriefing. Speech recognition systems for in-flight use currently target the recognition of a limited number of spoken commands. The ambition of the MEVAC project is to process all cockpit communications, be it between the pilot and co-pilot or with ground control, taking into account the specific constraints and conditions of fighter planes. Flighter plane cockpit conversations are very challenging for automatic systems accumulating the challenges posed by different variations simultaneously, including: noisy communication channels, distortions, voice affected by stress, speech spoken with different accents, including non-native, multiple languages. The communications are also quite difficult for untrained humans to understand. Advanced machine learning techniques will be explored, building upon existing methods and maintaining a close interaction with users in order to propose solutions that address their real needs.

About Vocapia Research