"Help Me" Finding Lost Person's Location in the Mountains by Sound

KRISS-POSTECH Develop Real-Time Sound Source Localization Technology
Real-Time Detection of Sound Position and Size, Over 10 Times More Accurate Than Before

[Asia Economy Reporter Kim Bong-su] It is difficult to locate the exact position of a sound even if one shouts deep in the mountains, making it challenging to rescue stranded individuals. However, a domestic research team has developed an artificial intelligence technology that can accurately visualize the location and magnitude of sound sources. It is expected to greatly aid in life-saving operations by being more than 10 times more accurate and faster than existing methods.

The Korea Research Institute of Standards and Science (KRISS) announced on the 17th that Jang Ji-ho, a principal researcher of the Acoustic Vibration Ultrasonic Standards Group, in collaboration with Professor Lee Seung-chul’s mechanical engineering research team at POSTECH, has developed an AI program that converts the location and magnitude of sounds into images, visualizing them like a map to easily identify their positions.

The research team developed a deep learning algorithm that shortens processing time while increasing accuracy. It can precisely distinguish the location and magnitude of individual sound sources even under adverse conditions where multiple sounds are mixed.

To implement various acoustic data, the team installed 56 speakers in a spherical arrangement in the laboratory. When specific sounds are emitted from multiple speakers, the developed algorithm tracks their location and magnitude and visualizes them like a map. This method is more than 10 times more accurate than previous ones and reduces computation time to one-tenth.

Jang Ji-ho, principal researcher at KRISS, said, “The multidisciplinary convergence research conducted by KRISS and POSTECH, each with expertise in acoustics and artificial intelligence respectively, has led to excellent results,” adding, “We plan to continue research to commercialize the technology and bring innovation to the existing market.”

In practical use, this technology can locate stranded individuals in the mountains by sound. Recently, unmanned aerial vehicle (UAV) technologies such as drones have been used around the clock in reconnaissance, transportation, and rescue without human intervention. However, sound source location tracking technology using UAVs has the drawback of low precision and severe performance degradation depending on surrounding noise environments.

The sound source location tracking technology developed by the joint research team provides information more than 10 times more accurate than existing methods, allowing it to be used even under adverse conditions with drone propeller noise or other background noises. Combining this technology with reconnaissance, transportation, and rescue missions in the future is expected to contribute to the success of various contactless drone operations.

This technology, which accurately identifies the location of audible sounds, is also expected to be actively utilized in public safety fields such as detecting the location of electrical leakage, gas leaks, and water leaks by sound. Furthermore, it is anticipated to be applicable in areas enhancing public convenience, such as accurately identifying the location of noise between floors.

The research results were published in May in the international journal of mechanical engineering, Mechanical Systems and Signal Processing.