“Help me!” Sound Used to Locate the Distressed

- KRISS-Postech developed the technology for real-time location of sound sources -

- Real-time identification of the location and loudness of sound with 10 times higher accuracy -

A team of Korean research scientists* have developed AI technology that can accurately visualize the location and loudness of sound. This technology is 10 times more accurate than existing technology, consuming one-tenth of the time for calculation.

 

 * Principal research scientist Jiho Chang of the Acoustics, Ultrasound and Vibration Metrology Group of the Korea Research Institute of Standards and Science (KRISS, President Hyun-Min Park), Professor Seungchul Lee of the Department of Mechanical Engineering and Ph.d. candidate Soo Young Lee of the Pohang University of Science and Technology (Postech, President Mu-hwan Kim) (together, KRISS-Postech joint research team).

 

The KRISS-Postech joint research team developed AI technology to transform the location and loudness of sound into images so that such information can be visualized like a map for easy identification of the location.

 

▲ Technology to trace the sound location based on deep learning, developed by the KRISS-Postech joint research team

 

 

This technology can be used in everyday life to locate distressed people in the mountains. Unmanned aircraft technology for drones has recently been used widely in the areas of reconnaisance, transportation and rescue without human workers. However, the technology to locate sound sources with unmanned aerial vehicles has a low degree of precision and inevitably suffers from seriously poor performance depending on the noise in the surrounding environment.

 

The sound localizing technology developed by the joint research team provides ten times more accurate information than existing technology, and therefore, can be used in unfavorable conditions with the noise from drone propellers or other ambient noise. When associated with reconnaisance, transportation and rescue, the technology is expected to contribute to successful drone activities for various non-face-to-face missions.

 

 

▲ Sound source maps enabled by the existing technology (top),

Sound source maps suggested by the KRISS-Postech joint research team (bottom)

 

Capable of accurately localizing audible sound, the new technology is expected to be used in areas to keep people safe by detecting the location of gas or electricity leaks. In addition, it can also be used to accurately identify the location of the source of floor noise, which can enhance convenience of the general public.

 

Sound contains much information. However, such various information contained in sound was not utilized sufficiently because of the lack of sufficient resolution and accuracy. Even if the location of certain sound was found, it was difficult to forecast the loudness. When various sounds were mixed together, the location accuracy became significantly low. Another downside was that it took much time to increase accuracy.

 

The joint research team developed a deep learning algorithm to raise the accuracy by shortening the time consumed. Even in unfavorable conditions where a number of sounds are mixed, the new technology is able to precisely distinguish the location and loudness of one sound source from another.

 

▲ KRISS-Postech joint research team

[Professor Seungchul Lee of Postech (left), Ph.D. candidate Soo-Young Lee of Postech (center), Principal research scientist Jiho Chang of KRISS (right)]

 

 

The joint research team installed 56 speakers in the shape of a ball within the lab to accommodate various sound data. When several speakers produce certain sounds, the algorithm developed by the team traced the location and loudness to visualize the data like a map. This is 10 times more accurate than existing methods and can significantly reduce the calculation time by one-tenth.

 

“KRISS and Postech were able to obtain a remarkable outcome through interdisciplinary research based on their expertise in sound and AI, respectively,” said Dr. Jiho Chang. “We will continue research to bring innovation to the existing market by commercializing the new technology.”

 

The research outcome was published in May in Mechanical Systems and Signal Processing (IF: 6.471, JCR Top 3.4%), a world renowned mechanical engineering journal.