German researchers are developing drones equipped with microphones that will help search and rescue workers find and rescue survivors when every second counts.
The drones detect cries for help and signals from buried victims from the air and simultaneously pinpoint their location, according to Macarena Varela, a scientist at the Fraunhofer Institute for Communication, Information Processing and Ergonomics (FKIE) in Wachtberg, Germany.
The scientist, who presented her latest research at the 180th Meeting of the Acoustical Society of America, notes that drones “equipped with the latest technology can quickly scout a larger area than rescue workers or sniffer dogs. They can also reach places that are initially inaccessible to first responders and thus provide emergency managers with a good overview of the situation in the crisis area.”
Aerial photos are taken to document damage, and some drones are also equipped with a thermal imaging function to find buried victims.
In addition to visual identification, Varela’s research approach involves the use of acoustic sensors and clearly defined sound classification to locate people.
“We are developing a technology that can definitely save lives in the event of a disaster. In such situations, every minute counts,” says Varela. “For this approach to work, the sensors have to be able to distinguish between calls for help or other human signals and sounds that occur in nature. At the same time, the rotor noise of the drone has to be filtered out so that it does not drown out or change the calls for help picked up by the microphones.”
“If we consider the scenario of deployment to a disaster area, it is easy to imagine that other environmental noises, such as the noise of helicopters or emergency vehicles, have to be suppressed in order to positively detect the cries of people calling for help,” she explained.
Using artificial intelligence (AI), the prototype on which Macarena Varela and her colleague Wulf-Dieter Wirth are working with should learn not only how to suppress ambient noise, but also sound patterns such as shouting and clapping — in other words, sounds people make when they are trying to draw attention to themselves in emergency situations.
The AI is being trained using a database filled with “impulsive” human sounds, which the scientists are populating with their own self-recorded calls for help, the researchers noted.
Pinpointing the location of potential victims
To make the system capable not only of detecting sounds in an emergency, but more importantly of quickly locating people seeking help, the scientists are arranging the small, but very powerful microphones, in a specific shape: A so-called crow’s nest array (CNA) underneath the drone. In combination with modern signal processing techniques, known as beamforming, this made it possible to precisely track where the sound was coming from with accuracy, according to the scientists.
“So far, we have conducted highly successful tests in open fields,” Varela said. “Our goal is to identify the call for help, and determine the location of the potential victim in only a few seconds. Our system can already detect and localize impulsive sounds very precisely.”
But the project is not over yet: The next goal is to equip the drone with a microphone that has an even higher frequency in order to broaden the spectrum of the detailed signals and classify them with even greater precision. At the same time, the plan is to increase the distance from which the drone can detect the sounds.
For this purpose, the number of microphones will be increased. Currently, 32 microphones are used, while plans for the subsequent design envision twice that number, the researchers reported. This would also allow for a more precise calculation of the estimated angle of the sound, and even sounds from a greater distance could still be detected in real time.
Regarding actual use in disaster areas, Varela envisions the location data of victims being transmitted to emergency teams, who could then identify their precise positions on tablets.