A technology has been developed that improves wireless image transmission efficiency by 45 times, and is expected to enable real-time visual tasks in various wireless channel environments.

Ulsan National Institute of Science and Technology (UNIST) announced on the 6th that a research team led by Professor Seonghwan Yoon of the Graduate School of Artificial Intelligence has developed AI technology that dramatically increases the efficiency of wireless image transmission.

This technology is expected to be of great help in fields such as autonomous driving, telemedicine, and the metaverse, which require real-time processing of large-scale image data.

The 'Task-Adaptive Semantic Communication' technology developed by the research team is a method that selects and transmits only the semantic information absolutely necessary for task performance, unlike the existing method of compressing and transmitting the entire image.

By analyzing the semantic structure of objects, layouts, and relationships within an image, it drastically reduces the amount of data by extracting and transmitting only the information necessary for the task's purpose.

For example, in a simple object classification task, only object information such as 'cat' and 'car' is transmitted, while in an image generation task, information on the arrangement and relationship of objects is also transmitted.

Especially “to peopleWe reduced unnecessary data transmission by applying a ‘semantic filtering’ algorithm that filters out information that is always true, such as “there is a league,” or redundant relationship expressions.

This technology has been proven to achieve transmission efficiency up to 45 times higher than conventional methods through simulation results, and to be capable of performing real-time visual tasks even in various wireless channel environments.

Professor Yoon Seong-hwan emphasized, “In the future, the core of communication will be not just sending accurately, but sending meaningfully,” and “This research is a signal that will change the landscape of intelligent wireless communication.”

“It has high potential for application in various fields, such as recognition systems for autonomous vehicles, remote surgery and diagnosis, and real-time rendering in the metaverse,” said first author Park Jeong-hoon.

The results of this study were published on October 20th in the IEEE Journal on Selected Areas in Communications (JSAC), the top journal in the IEEE communications field, and were conducted through numerous national research support projects including the Ministry of Science and ICT, the Institute for Information & Communications Technology Planning and Evaluation (IITP), the Ministry of Health and Welfare, and the National Research Foundation of Korea (NRF).