Scientists train drones to track the sounds of people's screaming and hunt them down.It does sound like a dystopia of Terminator/Quiet Place, but it's not. These drones are not intended to kill people.Fraunhofer FKIE, Germany's Fraunhofer Institute for Communication, Information Processing and Ergonomics, has developed scream-seeking drones that can quickly locate people injured or trapped in emergencies. The researchers mount a variety of microphones on a drone to identify screams. They then use processing techniques like beamforming to pinpoint which microphone is closest to the sound, and then focus on that one.Macarena Varela, a researcher at Mashable, explained to Mashable via email that she and her colleagues WulfDieter Wirth (and Manfred Okum) had previously created a similar system in 2016. However, it was too large and heavy to mount on a drone.Varela stated that Dr Wirth convinced him that MEMS microphones could be used instead of traditional condenser microphones. This made the system smaller. "This was the beginning to our dream to be in a position to mount it on a drone to perform Search and Rescue (Search and Rescue).The research was presented to the Acoustical Society of America's 180th Meeting in June.Credit: Macarena VarelaVarela says that the researchers currently have 32 microphones in their array. However, Varela points out that they haven’t yet tried to determine how many it can use while still being efficient. They believe that more is better.Varela stated that MEMS microphones are small and inexpensive, so we plan to increase the number of microphones rather than reduce them.Researchers will be able to accurately calculate the angle of the sounds they hear, and also pick up audio from further away. This will allow the drone to pinpoint the exact location of the victim more precisely.Varela stated that beamforming techniques can be achieved by using an array of identical microphones, delivering synchronous data. Crows Nest is a particular array where microphones are placed randomly in a sphere. This array is omnidirectional and provides sound coverage in all directions.Varela continued, "All data from microphones are combined, after adding phases or delays to it in order to achieve maximum sensitivity for the selected direction and so forming a beam of sensitivity." The search for sound sources can then be made by scanning or changing the direction.SEE ALSO : Two swimmers were saved by a new drone in just 70 secondsResearchers use monopulse radar to determine the angle of the sound. This technique compares two simultaneous beams from slightly different directions to determine which signal is stronger for detecting the location of a target.Valera and her coworkers are currently testing and developing filtering techniques to reduce noise, such as that produced by the drone's propeller. They are also testing various detection methods, such as AI and neural networks, to pick up distress sounds. Fraunhofer FKIE's researchers use an audio database that includes "impulsive sounds" for both of these purposes. victims might produce are tapping, clapping, and screaming.Varela stated that "In previous lab tests, we were able detect impulsive sound, such as clapping with rotor noises present", "We are currently processing data from drone flying."Mashable received video from Varela illustrating their ongoing research.Drones can reach areas that are hard to reach and cover a lot more ground than dogs or humans. These efforts rely on humans monitoring the drone's camera. It is critical that victims of disasters are located quickly so technology that assists first responders is helpful.However, it may take some time to see the system in action. Varela states that it depends on the amount of time Varela and her coworkers can devote to it. The testing is ongoing and researchers don't have any specific deadlines. They believe that the system still has great potential.Varela stated that "a big part of our work is to transfer the methods from the large system to the smaller one." "But, we still face new challenges like the drone noise while flying. It is a matter for time, because we have the expertise within our team.
