An Improved Method for Identifying SSVEP Single-Channel Target Frequency for BCI Systems

Steady-state visual evoked potential (SSVEP) is a brain response to flickering visual stimuli in the occipital area. SSVEP detection methods usually utilize multichannel configurations, which can be intricate and uncomfortable. Conversely, single-channel configurations are simple and user-friendly, but attaining the necessary accuracy is difficult, requiring the development of a reliable and accurate single-channel technique. For this purpose, this study introduces a hybrid VMD-LGBM approach to single-channel SSVEP detection. The proposed technique utilizes VMD as a preprocessor that decomposes the SSVEP signal into harmonic, intrinsic mode functions (IMFs), and noisy IMFs. To separate harmonic IMFs from noisy IMFs, the crest factor of the cross-correlation function was introduced. Then the restored SSVEP signal is used to extract features using power spectral density analysis (Welch's approach), followed by classification using a Light Gradient Boosting Machine. Our proposed approach performs better than available techniques in terms of classification accuracy and information transfer rate. The proposed single-channel approach provides better efficiency and user-friendliness for practical applications.