KAIST Professor Jeong Jae-seung Develops 'Robot Arm Controlled by Thought'

Cerebral Cortex Signal Analysis, Movement Prediction Brain Signal Decoding Technology
Operable Without Practice for Patients with Motor Disabilities Including Quadriplegia

[Asia Economy Reporter Kim Bong-su] A research team led by Professor Jae-Seung Jeong of KAIST, a renowned neuroscientist, has developed a system that decodes human brain signals to move a robotic arm in the desired direction by thought alone, without long-term training.

KAIST announced on the 24th that Professor Jeong's research team developed this brain-machine interface system. It is a decoding technology for imagined arm movement direction brain signals that predicts the intended arm movement of epilepsy patients by analyzing cerebral cortex signals observed when they imagine extending their arm. Since it does not require actual movement or complex motor imagery, patients with motor disabilities can naturally and easily control the robotic arm without long-term training.

Brain-machine interfaces are technologies that allow humans to control machines by thought alone. They have attracted significant attention as assistive technologies enabling patients with arm movement disabilities or amputations to control robotic arms and regain necessary arm functions in daily life.

To implement a brain-machine interface for robotic arm control, it is necessary to measure the electrical signals generated in the brain when a person moves their arm and decode these brain signals using various artificial intelligence analysis techniques such as machine learning to predict the intended movement from the brain signals.

However, patients with motor disabilities such as upper limb amputation find it difficult to move their arms physically, so an interface that can direct the robotic arm's direction by imagination alone is urgently needed. The brain signal decoding technology must predict the direction imagined by the user from imagined brain signals rather than actual movement signals. However, imagined brain signals have a significantly lower signal-to-noise ratio than actual movement brain signals, making it difficult to accurately predict the arm's direction?a long-standing challenge. To overcome this, previous studies attempted imagining other body movements with higher signal-to-noise ratios to move the arm, but the unnatural gap between the intended arm extension and the cognitive movement caused users to require long-term training, resulting in inconvenience.

Therefore, decoding technology that predicts the direction of arm extension during imagination has low accuracy and makes it difficult for patients to learn how to use it. This problem has been a long-standing challenge in the brain-machine interface field.

To solve this problem, the research team measured users' natural arm movement imagination using high spatial resolution cerebral cortex signals (electrocorticogram) and developed for the first time a decoding technology that calculates directional information of arm movements, which are difficult to measure directly, by applying a variational Bayesian least squares machine learning technique.

The team's arm movement imagination signal analysis technology is not limited to specific cerebral regions such as the motor cortex, allowing customized learning of imagined signals and cerebral region characteristics that may vary by user, thereby outputting optimal computational model parameter results.

The research team confirmed that cerebral cortex signal decoding can predict the direction of arm extension imagined by patients with over 80% accuracy.

Furthermore, by analyzing the computational model, the team revealed the spatiotemporal cerebral characteristics important for directional imagination and confirmed that the closer the cognitive process of imagination is to the actual arm extension process, the higher the direction prediction accuracy can be.

In February, the research team published preliminary research results on a high-accuracy robotic arm control brain-machine interface based on artificial intelligence and genetic algorithms in the world-renowned journal Applied Soft Computing. Building on that, this follow-up study simplified the computational algorithm, conducted robotic arm driving tests, and improved patients' imagination strategies to create a practical usage environment. They successfully tested actual robotic arm operation and confirmed that the robotic arm moved in the intended direction, demonstrating accurate reaching of targets by reading intentions in four directions.

The arm movement direction imagination brain signal analysis technology developed by the research team is expected to contribute to improving the accuracy and efficiency of brain-machine interfaces that control robotic arms for patients with motor disabilities, including those with quadriplegia.

Professor Jeong said, "The technology that analyzes individualized brain signals for each disabled person and allows robotic arm control without long-term training is an innovative result," adding, "This technology is expected to greatly contribute to the commercialization of robotic arms that can replace prosthetic arms in the future."

The research results were published in the September issue, Volume 19, Issue 5, of the international journal in the field of neuroengineering, Journal of Neural Engineering.

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