The Empirical Mode Decomposition-Decision Tree Method to Recognize the Steady-State Visual Evoked Potentials with...

Area of Science:

• Neuroscience
• Signal Processing
• Biomedical Engineering

Background:

• Steady-state visual evoked potentials (SSVEPs) are crucial for brain-computer interfaces.
• Empirical Mode Decomposition (EMD) analyzes SSVEPs by decomposing them into intrinsic mode functions (IMFs).
• Traditional EMD methods struggle with accuracy when the stimulation frequency range is wide.

Purpose of the Study:

• To develop an improved EMD-based method for SSVEP analysis.
• To enhance SSVEP recognition accuracy across a broad range of stimulation frequencies.
• To overcome limitations of existing EMD techniques in extended frequency ranges.

Main Methods:

• SSVEP signals were recorded from six subjects with stimulation frequencies from 6 to 16 Hz.
• EMD was employed to extract effective IMFs from the SSVEP signals.
• Features like peak spectrum frequency and normalized local energy were extracted from IMFs and their combinations.

Main Results:

• The proposed method achieved a total recognition accuracy of 79.75%.
• This accuracy surpasses EMD-Fast Fourier Transform (72.05%) and Canonical Correlation Analysis (77.31%).
• Individual IMFs and their combinations showed limited efficiency in wide frequency ranges.

Conclusions:

• The novel EMD approach significantly improves SSVEP recognition rates.
• The method offers a robust solution for SSVEP analysis in wide stimulation frequency ranges.
• Recognition rates improved by over 2.4% compared to CCA and 7.7% compared to EMD-FFT.