Most search engine technologies based on drones use cameras which may fail when people are under rubbles or at night or in dark places. The new development put forward by Tokyo Institute of Technology, Kumamoto and Waseda University under a JST contract, uses search drones based on sound detection to get around such problems. The system consists of the open-source Robot’s Ear” software HARK (abbreviation for Honda Research Institute Japan Audition for Robots with Kyoto University), which can detect sound sources even under the drones’ noises, combined with a 3D sound source position estimation and map display technology as an interface for operators, and an all weather-type microphone array installed in the drone consisting of 16 microphones. The team has also developed a simultaneous speech recognition technology that allows identification of people who talk at the same time. It is named after Prince Shotoku Taishi (574 – 622) who is said to have been able to listen to 11 speakers simultaneously. The software is already used in a „moderator robot“ for quiz shows when multiple persons respond at the same time.”
JST news release, December 7, 2017