Ranked Third and Won Popularity Award Among 630 Teams
Implemented "Physical AI" That Selectively Listens Only to Necessary Sounds
Started from Local Industrial Site Ideas
Achieving Both Safety and Efficiency for Manufacturing Innovation
The research team led by Professor Jeong Imdu from the Department of Mechanical Engineering and Graduate School of Artificial Intelligence at UNIST (President Park Jongrae) has achieved outstanding results on the national stage by addressing the 'noise problem' in industrial sites using AI technology.
The UNIST 'SafeAngel' team won third place in the finals of the '2025 AI Champion Competition' hosted by the Ministry of Science and ICT, held at Dragon City Hotel in Yongsan, Seoul, on the 5th. The team was awarded the Minister's Award (AI Challenger Award).
A total of 630 teams from across the country participated, and after intense competition, only five teams received awards in the finals.
This competition was the first nationwide AI technology contest held this year. The field was narrowed down from 100 teams in the preliminaries, to 20, and finally to 8 teams in the final round. Among them, the UNIST SafeAngel team received high marks for practicality, technological completeness, and social impact.
The research team included Professor Jeong Imdu, along with researchers Kim Taekyung, Kim Kyunghwan, Kim Dohyun, Gong Byunghun, and Lee Yoonsu. In addition, researchers Seo Junyoung, Bang Jina, and Moon Youngmin from the Pohang Institute of Industrial Science and Technology (RIST) participated. Their project was titled 'On-Device Physical AI Technology for Industry-Specific Active Hearing Protection and Communication Devices.'
Professor Jeong Imdu (center front row) and the 'SafeAngel' team are cheering before the AI Champion finals. Photo by UNIST
Industrial sites are still filled with noise. According to statistics from the Ministry of Employment and Labor, 98.8% of occupational disease cases last year were noise-induced hearing loss. Most earplugs block out all sounds, making it impossible to hear alarms or colleagues' voices, which increases the risk of accidents.
The SafeAngel team directly tackled this issue by developing AI technology that selectively listens to only the necessary sounds instead of using earplugs. With physical AI, the system distinguishes and removes sounds such as high-frequency and impact noises that traditional noise-canceling cannot filter. This protects the ears while allowing warning signals and conversations to be heard clearly.
The core technology is the 'Physics Informed Neural Network' (PINN) method, which incorporates physical laws into the AI learning process. It analyzes ambient noise in real time and filters out only the dangerous frequency bands. In particularly noisy environments, the AI automatically adjusts the physical filter to block unnecessary loud noises. Additionally, the system can predict machine failures based on mechanical sounds and detect changes in workers' voice patterns to assess fatigue or potential accidents.
Professor Jeong Imdu stated, "We hope to help reduce chronic hearing loss and fatal accidents caused by communication failures in production sites through AI technology. It is important that AI not only protects workers by analyzing mechanical failure signals hidden within noise, but also helps improve production efficiency. With this idea, which originated in industrial sites, we aim to open up real possibilities for the advancement of physical AI."
In the final judging, the SafeAngel team also won the 'AI Impactor Award' (Popularity Award) by ranking first in the 'Best Technology' category through votes from a 200-member public jury, in addition to the expert evaluation.
With these achievements, the team secured both the 'AI Challenger Award' and the 'AI Impactor Award,' and will receive up to 500 million won in support for follow-up research.
© The Asia Business Daily(www.asiae.co.kr). All rights reserved.
