Summary:
UCLA researchers in the Department of Electrical and Computer Engineering have developed an artificial intelligence (AI) assistance framework for brain-machine interface (BMI) applications designed to enhance users' ability to complete tasks efficiently and swiftly.
Background:
Paralysis affects over 5 million individuals in the U.S., limiting their mobility and independence. The growing demand for therapeutics and technologies to restore mobility for those with physical impairments has spurred interest in brain-machine interfaces (BMIs). However, limited performance, high costs, and safety risks have hindered widespread adoption. Current strategies to enhance BMI performance, such as high-density electrode arrays, often face signal detection delays and limited adaptability. State-of-the-art BMIs require users to control motion trajectory, making even simple quality-of-life tasks inefficient and unnatural. There is an urgent need for innovative solutions to improve BMI performance, reduce user demands, and enable more natural, efficient control.
Innovation:
Professor Jonathan Kao and his research team have devised an innovative approach to boost BMI performance and significantly reduce user neural effort in BMI tasks. Their proposed invention presents a comprehensive framework to overcome the limitations of modern BMI applications. By combining computer vision and AI, the system creates a copilot that learns task structures and patterns, allowing users to focus less on minute details of motion, resulting in more efficient and natural control. The copilot predicts possible actions by learning neural information in trajectory motion and gauging the user's intentions. Tested in quality-of-life tasks like lifting a cup or controlling a computer cursor, the copilot framework can operate with various modalities, including intracortical spiking activity and electrocorticography, increasing versatility and applicability to a broader range of BMI users.
Potential Applications:
- BMI devices
- Wearable prostheses, exoskeletons, and other external device control for: amputees or individuals with paralysis, dangerous tasks (e.g., bomb defusal.), virtual, augmented, and mixed reality, gaming
- Robotics or teleoperation systems
Advantages:
- Effectively reduces the neural effort to perform tasks
- Completes BMI tasks with natural speeds and increased precision
- Compatible with various modalities
Development to Date:
Initial Conception: Sep/01/2017
Related Papers:
Olsen, Sebastian, Jianwei Zhang, Ken-Fu Liang, Michelle Lam, Usama Riaz, and Jonathan C. Kao. "An artificial intelligence that increases simulated brain–computer interface performance." Journal of Neural Engineering 18, no. 4 (2021): 046053.
Reference:
UCLA Case No. 2023-110
Lead Inventor:
Prof. Jonathan Kao.