Signal alignment for cross-datasets in P300 brain-computer interfaces.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Machine Learning

Background:

• Transfer learning is crucial in Brain-Computer Interface (BCI) research.
• Existing subject-to-subject transfer learning is limited, with few studies on cross-dataset or cross-paradigm transfer.
• P300 event-related potential (ERP) signals are key for BCI applications.

Purpose of the Study:

• To propose a novel signal alignment (SA) method for P300 ERP signals.
• To enable intuitive, computationally inexpensive, and effective cross-dataset transfer learning.
• To facilitate paradigm-to-paradigm transfer in BCI.

Main Methods:

• Developed a linear signal alignment (SA) technique.
• SA utilizes P300 latency, amplitude scale, and reverse factor for signal transformation.
• Evaluated SA on four diverse datasets: two P300 Speller BCIs, one face-stimuli P300 Speller, and one auditory oddball paradigm.

Main Results:

• The SA approach improved the average precision (AP) score from 25.5% to 35.8%.
• An average of 36.0% of subjects showed performance improvement using SA.
• The P300 Speller dataset with face stimuli demonstrated higher comparability across different datasets.

Conclusions:

• Proposed a simple and intuitive ERP signal alignment method.
• Demonstrated the feasibility of cross-dataset transfer learning, even between different paradigms.
• SA enhances the generalizability and applicability of BCI models across diverse datasets.