May 17, 2025

“Distinguishing even the sound of hair breaking amidst noise”… Mounted on robots and drones for application [Heo Jin-seok’s Talk Talk Startup]

News│Focus

Revolutionizing Safety and Quality Inspection Through AI Acoustic Detection Technology — LOAS

Microscopic Cracks and Imbalances Ultimately Reveal Themselves Through Sound

Defects Once Detected by Human Hearing Are Now Identified and Tracked Through Microphones

Supplying Solutions to Major Corporations Pursuing “Zero Defect” Manufacturing — “Every Factory and Industrial Facility in the World Is Our Market”

The intelligent acoustic inspection system exhibited by LOAS at last year’s Smart Tech Korea exhibition. The demonstration showcased the real-time inspection process used to detect defects in products such as air conditioners and air purifiers. / Courtesy of LOAS

LOAS, a startup based in Songpa-gu, Seoul, possesses a unique technology: acoustic detection, tracking, and inspection technology capable of identifying microscopic sounds generated by defective products or industrial equipment. The technology operates effectively even in noisy environments. According to LOAS, while the technology determines defects through sound analysis, there are currently no other companies in Korea — or globally — with equivalent capabilities.

For every manufacturing company, preventing defective products from reaching customers is critically important. If products do not fail during warranty periods, companies can significantly reduce after-sales service costs and manpower requirements. Companies seeking to build premium brands must even predict and prevent potential failures before they occur.

“Even in this highly advanced era, most home appliance factories still rely on workers physically placing their ears close to products during final inspections to check for abnormal sounds. When air purifiers move along conveyor belts stretching dozens of meters, workers walk alongside them, hugging the products and concentrating on tiny sounds like ‘zzzt’ or ‘tick, tick’ to identify defective units. In some ways, it’s a bizarre scene.”

This was explained by Jaehyun Lee (39), whom we met on the 8th at the LOAS office in Songpa-gu, Seoul.

LOAS has already supplied its “SMART” intelligent acoustic inspection system to Korean home appliance manufacturers to detect defects through sound analysis. The company is now expanding the application of this technology to drones and robots for inspecting abnormalities in large-scale facilities such as semiconductor factories and petrochemical plants.

● AI Replacing Human Hearing

The defects detected by the AI acoustic inspection system are often impossible to identify visually. Many involve motors, bearings, fans, and other internal components operating inside products.

For example, air purifiers often rotate 360 degrees at the top to redirect airflow. If abnormal sounds occur during this rotation, the product is likely defective. Similar hidden defects are frequently found in the airflow direction control components of air conditioners. Refrigerators may also contain microscopic cracks in refrigerant pipelines invisible to the naked eye, which the system can detect.

The core strength of LOAS technology lies in accurately detecting, tracking, and inspecting sounds generated by specific products even within noisy environments. LOAS refers to its proprietary acoustic detection software engine as “AI Square.”

Lee explained, “We are the only company in the world capable of using AI to track and visualize the origin and path of sounds in real time within three-dimensional spaces, while automatically separating external noise and amplifying only subtle abnormal acoustic signals for analysis.” He added, “Many people imagined this technology, but we are proud to be the first to realize it.”

AI Square continues operating even when surrounding environmental noise suddenly changes or becomes significantly louder. Lee stated, “Developing the technology capable of accurately detecting abnormal sounds despite changing environments — while processing vast amounts of acoustic data quickly enough to deliver results within 0.8 seconds — was especially difficult.”

He further added, “Our acoustic detection technology is capable of instantly locating drones flying from distances of approximately 350 meters in real time.”

LOAS constructs SMART systems using computers equipped with AI Square, two circular microphone arrays containing dozens of microphones, and cameras used to indicate sound source locations. Since sound locations are overlaid directly onto camera footage, users can intuitively identify where issues are occurring.

This year, installations are actively being carried out across multiple domestic and international factories. Lee, who personally participates in installation projects, stated, “Unless there are special circumstances, defect inspection processes can be automated within approximately five days without significantly modifying production lines, even in noisy factory environments.”

By automating acoustic inspections previously dependent on human hearing, the system can detect extremely subtle sounds such as a strand of hair snapping, as well as ultrasonic and infrasonic frequencies beyond the human hearing range.

Lee emphasized, “Most importantly, unusual noise patterns related to defects — which were difficult to preserve through human hearing-based inspections — can now be stored and analyzed.” He added, “This ultimately leads to improved product quality, enabling companies to ship products so refined that they may never require returns to service centers during warranty periods.”

● Expanding Safety Inspection Through Factory Patrol Robots and Drones

LOAS integrated its AI acoustic detection technology into autonomous driving robots to create “Tfoi,” and into drones to create “Tfos.” These systems are used to monitor regions difficult for humans to access or too vast for manual inspection.

The autonomous robot Tfoi independently navigates factory interiors containing pipelines, fans, motors, bearings, and major industrial equipment while collecting all surrounding acoustic data in real time through multi-channel microphones and AI acoustic analysis systems.

Lee explained, “Last year, we successfully completed demonstration testing at Korea Western Power. The system detected microscopic gas leaks and helped prevent accidents. This year, the Korea Western Power robot system version will officially launch in June.”

The AI analyzes sound data collected during robot patrols, distinguishing between normal and abnormal sounds. When abnormal acoustic patterns are detected, alerts are immediately sent to monitoring systems.

The drone system Tfos primarily flies across large industrial plants, wind power facilities, and rooftop equipment areas at semiconductor factories — locations difficult for humans to access. The system was developed and supplied specifically for facility inspections at one of Korea’s leading semiconductor companies.

Lee stated, “The system is being used to detect abnormal sounds and chemical gas leaks occurring in rooftop pipelines and facilities difficult for humans to approach.” He added, “Flying approximately 10 meters away from target facilities along predefined autonomous routes helps reduce accident risks.”

● Frequent Factory Visits Led to the Startup — Three Years Devoted Entirely to Technology Development

Lee majored in physics at Hallym University before building his career as an application engineer at a German instrumentation company, designing optimized solutions based on customer needs.

During this period, he traveled extensively across industrial sites in the semiconductor, display, and automotive industries, handling vibration and acoustic measurement system design and on-site installation work. Through these experiences, he realized firsthand that abnormalities in industrial facilities are ultimately expressed through sound.

He began preparing for his startup in 2019. “I traveled across 12 European countries searching for the sensors we needed. After returning to Korea, I personally sought out experts in signal processing, acoustics, electronics, and software. After enduring more than three months of rejection, I finally gathered mentors and colleagues.”

Among those connections was Professor Baek Jong-ho of the Department of Software Convergence at Seoul Women's University.

Following the company’s founding in 2020, LOAS spent more than three years focused exclusively on technology development. It took 1 year and 8 months to create the initial prototype, another year to develop exception-handling algorithms capable of responding to real-time environmental noise, and seven additional months to create black-box technologies capable of amplifying and generating fault data.

The team sacrificed weekends and worked day and night. At times, they survived through personal credit loans due to severe financial difficulties. Even last year, the company continued facing hardships. Because no comparable technology existed, convincing investors proved extremely challenging.

Employees even took out loans themselves to complete the autonomous robot under development. Using those funds, the company completed its robot prototype and participated in the Smart Tech Korea exhibition.

That exhibition became the turning point that attracted investors and eventually led to KRW 4 billion in investment funding this year. Orders from major corporations that had been pending also began materializing in earnest this year.

Lee reflected, “It feels like we went from hell to heaven between last year and this year.” He added, “Now, I’m traveling to factories around the world to install our systems.”

LOAS is targeting KRW 5 billion in revenue this year.

Lee concluded, “Because the technology was so unique, persuasion was never easy. But once demonstration projects succeeded, everything started to move forward.” He added, “We plan to actively expand into manufacturing-heavy countries worldwide and become the company that defines the new standard for quality inspection and industrial safety diagnostics.”