Artificial intelligence based BCI using SSVEP signals with single channel EEG.

Area of Science:

• Neuroscience and Biomedical Engineering
• Signal Processing and Machine Learning

Background:

• Brain-Computer Interfaces (BCIs) facilitate direct brain-device communication.
• Steady-State Visual-Evoked Potentials (SSVEPs) offer rapid communication and minimal calibration for BCIs.
• Existing SSVEP classification methods struggle with accuracy using limited EEG channels in real-world scenarios.

Purpose of the Study:

• To enhance SSVEP signal classification accuracy in BCIs using machine learning.
• To extract dominant frequency features from SSVEP data for improved classification.
• To reduce the number of required EEG channels for practical BCI applications.

Main Methods:

• Utilized the Benchmark Dataset from Tsinghua BCI Lab with 64 EEG channels.
• Applied wavelet decomposition (db4) to extract frequency features (7.8-15.6 Hz) from the Oz channel.
• Classified extracted features using Decision Tree (DT), Linear Discriminant Analysis (LDA), and Support Vector Machine (SVM) models.

Main Results:

• Achieved high classification accuracies: 95.8% for DT and 96.7% for LDA and SVM.
• Demonstrated significant performance improvement compared to existing SSVEP classification techniques.
• Validated the effectiveness of the proposed feature extraction and classification approach.

Conclusions:

• SSVEP classification using machine learning significantly improves accuracy and efficiency in BCIs.
• Wavelet decomposition and machine learning provide a robust method for SSVEP-based BCIs.
• The developed method shows strong potential for assistive technologies and diverse BCI applications.