A novel heterogeneous transfer learning method based on data stitching for the sequential coding brain computer...

Area of Science:

• Neuroscience
• Machine Learning
• Biomedical Engineering

Background:

• Brain-computer interfaces (BCI) require substantial electroencephalography (EEG) data for training classification models.
• High-dimensional BCI operations create significant data acquisition burdens.
• Existing methods struggle with efficient EEG data collection for complex BCI tasks.

Purpose of the Study:

• To develop a novel heterogeneous transfer learning method to alleviate the EEG data acquisition burden in BCI.
• To improve the efficiency and performance of BCI systems through advanced transfer learning techniques.

Main Methods:

• Proposed the multi-band data stitching with label alignment and tangent space mapping (MDSLATSM) algorithm for sequential coding paradigms.
• Utilized multi-band filtering and data stitching to create artificial signals bridging source and target domains.
• Employed label alignment to minimize distribution differences between domains and Riemannian manifold tangent space mapping for feature extraction.

Main Results:

• Achieved an average classification accuracy of 78.28% for cross-label heterogeneous transfer learning.
• Demonstrated superior performance in cross-subject transfer learning with an average accuracy of 64.01% compared to existing methods.
• Validated the effectiveness of MDSLATSM on a dataset including EEG signals from 16 subjects.

Conclusions:

• The novel heterogeneous transfer learning method, combining multiple techniques, offers superior performance for BCI applications.
• This approach significantly reduces the data acquisition burden, promoting the practical use of BCI systems.
• MDSLATSM represents a significant advancement in transfer learning for EEG-based BCI.