Shin Youngseo, Master's Student at Chosun University, Ranks 8th Worldwide in AI Acoustic Recognition Technology International Competition

Ranked 3rd Among Domestic Universities... High Evaluation in Music Classification and Location Detection Fields

[Asia Economy Honam Reporting Headquarters Reporter Yoon Jamin] Chosun University announced on the 13th that Shin Youngseo, a master's student in the Department of Computer Engineering (advisor Jeon Chanjun), achieved 8th place worldwide in an international competition for acoustic recognition technology using artificial intelligence (AI). This ranks 3rd among domestic universities.

The DCASE, a global competition for AI-based acoustic event and scene recognition technology, is the only sound technology-related competition hosted by the world's largest Institute of Electrical and Electronics Engineers (IEEE) AASP, with participation from leading global organizations such as Google, Intel, Amazon, IBM, Samsung, and LG.

The field in which master's student Shin Youngseo participated, Task 3, is "Sound source localization and direction estimation evaluated in real spatial acoustic scenes."

This field involves listening to sounds recorded with various types of sound sources and identifying the classification and location of the sound sources.

It requires accurately determining the classification and location of multiple sound sources, including both fixed and moving sources, and the evaluation is conducted on real-world sound sources rather than synthesized ones, making generalization performance crucial.

This recognition and detection technology can also be applied as "risk avoidance and alert technology" for groups such as the elderly or hearing-impaired who have difficulty hearing sounds and recognizing situations.

Additionally, it is expected to be utilized in various fields such as monitoring and surveillance that quickly track accident locations based on information like gunshots or screams, and robotics.

Chosun University designed and applied ▲ a "multi-generator technique" to overcome the limitations of data with class imbalance ▲ the ACCDOA format and transformer technique that encode all information related to sound source detection and direction estimation into a single output, achieving excellent results.

Master's student Shin Youngseo said, "I am pleased to achieve meaningful results in a major international competition," and added, "Deep learning-based acoustic recognition technology is continuously advancing, and I will study harder to achieve even better results next time."

This content was produced with the assistance of AI translation services.