Related Experiment Video
Updated: Jul 16, 2025
06:34
A Single-Channel and Non-Invasive Wearable Brain-Computer Interface for Industry and Healthcare
Published on: July 7, 2023
2.4K
A data expansion technique based on training and testing sample to boost the detection of SSVEPs for brain-computer
Xiaolin Xiao1,2,3, Lijie Wang1,4, Minpeng Xu1,2,3
1Academy of Medical Engineering and Translational Medicine, Tianjin University, Tianjin, People's Republic of China.
Journal of Neural Engineering
|September 8, 2023
Summary
Cyclic Shift Trials (CSTs) improve steady-state visual evoked potentials (SSVEPs) brain-computer interfaces (BCIs) by integrating unsupervised and supervised learning. This novel method enhances performance, especially with limited training data, reducing sample size dependency.
Area of Science:
- Neuroscience
- Biomedical Engineering
- Machine Learning
Background:
- Steady-state visual evoked potentials (SSVEPs)-based brain-computer interfaces (BCIs) offer high accuracy and speed.
- SSVEP-BCI performance degrades significantly with limited training data due to overfitting.
- Existing methods do not leverage unsupervised learning from test trials to improve supervised models.
Purpose of the Study:
- To introduce a novel method, Cyclic Shift Trials (CSTs), for SSVEP detection.
- To integrate unsupervised learning from test trials with supervised learning from training trials.
- To mitigate the overfitting effect in SSVEP-BCIs with limited training samples.
Main Methods:
- Proposed the Cyclic Shift Trials (CSTs) method for SSVEP detection.
- CST leverages the regularity and periodicity of SSVEPs to expand training samples.
- An online SSVEP-BCI system was developed and tested using CST with extended canonical correlation analysis and ensemble task-related component analysis.
Main Results:
- CST significantly enhanced signal-to-noise ratios of SSVEPs.
- The method improved system performance, particularly with few training samples and short stimulus durations.
- An online information transfer rate of 236.19 bits/min was achieved with only 36s calibration and one training sample per category.
Conclusions:
- CST effectively utilizes both supervised and unsupervised learning information.
- CST acts as a data expansion technique, enhancing SSVEP characteristics and reducing sample size dependency.
- CST offers a promising approach to improve SSVEP-BCI performance without increasing experimental burden.
