Scientists Teach Drones to Track Human Cries for Help
If someone created a flying machine capable of tracking you and listening to your voice, it might seem creepy. But what if you ended up under a pile of rubble after a natural disaster and rescue services couldn't find you? Then a search drone doesn't seem like such a bad idea.
This concept is the focus of engineers at the German Fraunhofer FKIE institute, who created a prototype drone designed to find people by detecting human screams and identifying other signs of distress.
One of the lead engineers, Macarena Varela, demonstrated the progress of this work at the annual conference organized by the Acoustical Society of America in June 2021.
In a sci-fi horror film, it's easy to imagine drones searching for people. Varela argues that these devices are ideally suited for post-disaster scenarios such as earthquakes, hurricanes, and wildfires. They can hover over areas that are hard for rescuers to reach and determine where people are trapped.
“Drones can cover a large area in less time than rescuers or search dogs on the ground. If there is a collapsed building, it can alert and assist rescuers. It can get into places where they cannot reach on their own”
Unmanned aerial vehicles are typically used for search and rescue operations during natural disasters. Most often they take aerial photos of structural damage. Some drones have thermal imaging capabilities to scan body heat, while larger drones can deliver medicine and other supplies to people in isolated areas.
Researchers continue to find new ways to use an extra pair of eyes in the sky. Washington University is presenting drones that use smells to detect disaster survivors. Aerospace Corporation is working on drones that can visually identify dogs and report their location to rescue teams. The University of Zurich has developed a drone that changes shape in flight and fits into oddly shaped gaps.
Detecting people using aerial acoustics comes with certain difficulties. The hearing system will need to distinguish human screams from sounds commonly found in nature, such as wind and animal voices. It may also need to recognize patterns associated with foot stomping, clapping, and other ways people try to attract the attention of rescue teams.
FKIE engineers accounted for these situations when creating their conceptual drone. First, they needed a database of "impulsive" human noises to train the drone. They recorded their screams, knocking, and other sounds that could indicate people were in trouble.
They then analyzed each sound frequency to find common "signatures" and used them to train the artificial intelligence software. According to Varela, they also worked on filtering out noise produced by the drone's rotors and other environmental sounds.
Once the software part was complete, the engineers assembled a system of tiny digital microphones to capture "precise" angles from which sound originates. According to them, digital microphones, like those in smartphones and hearing aids, are better because they don't require bulky equipment like analog or traditional microphones.
The team placed a group of microphones under the drone and used signal processing methods that allowed it to track where human sounds were coming from. The system also increased the volume and clarity of speech.
So far, they have conducted successful "open field" experiments and found that the drone pinpoints a victim's location within seconds of picking up a sound. In a demonstration video, the drone was shown hovering above the ground before turning toward a researcher who was making noise and calling for help from the ground.
Engineers have already succeeded in detecting and locating impulsive sounds with very high accuracy despite drone noise. They are in the process of patenting their microphone-based sound tracking method. They would then like to add a higher-frequency microphone to the drone to capture more sound signals.
The idea is to pick up sound from hundreds of meters away. In the real world, a victim's location data could one day be sent over a wireless network to emergency services. After this development appeared, various organizations contacted the research team, interested in collaborating with them.
Primarily, the research institute experiments and builds prototypes. Sometimes it issues licenses for its inventions to companies focused on non-commercial products. The institute's head of business development said: "We are good at developing technologies and solving problems for which no solutions yet exist. But commercialization is not really our job."
