Performance of Empirical Mode Decomposition for Frequency Identification in SSVEP Based BCI

Area of Science:

• Neuroscience
• Signal Processing
• Biomedical Engineering

Background:

• Steady State Visual Evoked Potentials (SSVEP) are crucial for brain-computer interfaces (BCIs).
• Accurate identification of SSVEP frequency components is vital for reliable BCI performance.
• Existing methods may face challenges in precisely detecting narrow band SSVEP frequencies.

Purpose of the Study:

• To propose and evaluate a novel Empirical Mode Decomposition based Conventional Correlation (EMDCC) method.
• To enhance the recognition accuracy of narrow band frequency components within SSVEP signals.
• To compare the performance of EMDCC against conventional correlation and Time-Weighting Canonical Correlation Analysis (TWCCA).

Main Methods:

• Empirical Mode Decomposition (EMD) was applied to decompose EEG signals.
• Conventional Correlation was utilized for frequency component identification.
• The proposed EMDCC method integrates EMD with conventional correlation.
• Performance was evaluated on a benchmark SSVEP dataset and an in-house dataset.

Main Results:

• The EMDCC method achieved a mean detection accuracy of 93.79% on the benchmark dataset, an improvement from 85.64% with conventional correlation.
• For the in-house dataset, EMDCC reached 82.5% accuracy, surpassing the conventional correlation's 67.5%.
• EMDCC outperformed TWCCA (91.04%) on the benchmark dataset, demonstrating superior detection accuracy.

Conclusions:

• The proposed EMDCC method significantly improves the detection accuracy of SSVEP frequency components.
• EMDCC offers a more robust approach for identifying narrow band frequencies compared to conventional methods.
• This advancement holds promise for enhancing the performance and reliability of SSVEP-based BCIs.