An Optimized Encoding BCI Framework: Implementing Massive Command With Minimal Calibration.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Computer Science

Background:

• Brain-computer interfaces (BCIs) are crucial for assistive technologies, but limited command sets and high calibration costs hinder practical use.
• Existing BCIs struggle to balance command quantity with user-friendly calibration procedures.

Purpose of the Study:

• To develop a functional optimization-based encoding framework for generating a large number of highly discriminable commands with reduced calibration.
• To enhance BCI control capacity and efficiency for real-world applications.

Main Methods:

• Implemented a functional optimization theory to improve command distinguishability by optimizing encoding functions.
• Utilized a few-shot training strategy and symbol reusability to minimize calibration data requirements.
• Employed a symbol-joint decoding approach to boost recognition accuracy.

Main Results:

• Developed an online BCI system with 1,008 commands, achieving state-of-the-art in command scale and calibration efficiency.
• Demonstrated a 13.94% increase in accuracy and a 26.12% increase in information transfer rate (ITR) compared to baseline methods.
• Achieved high accuracy (86.60%) and ITR (122.74 bits/min) with only 72 seconds of calibration across 15 subjects.

Conclusions:

• The functional optimization framework significantly enhances BCI performance, offering unprecedented command scale and calibration efficiency.
• The system's flexibility and compatibility across paradigms suggest broad applicability and improved practicality for large-scale BCIs.
• This approach overcomes key limitations in current BCIs, paving the way for more sophisticated and accessible brain-computer interfaces.