Noise reduction of auditory steady-state response EEG signals based on WPT-VMD

To address the issue of noise in auditory steady-state response (ASSR) EEG signals while preserving meaningful signal information, this study proposes a novel noise reduction method based on the wavelet packet transform (WPT) and an improved variational mode decomposition (VMD). The enhanced VMD method integrates the Sparrow Search Algorithm (SSA) to optimize the decomposition parameters. Specifically, the proposed approach begins by decomposing the original EEG signals into seven levels using WPT, employing the db4 wavelet as the decomposition basis. Subsequently, SSA is applied to optimize VMD parameters, and the WPT-decomposed signals are further processed with VMD to extract intrinsic mode functions (IMFs). The correlation coefficients between each IMF and the WPT-decomposed signals are then calculated to identify and select the most relevant modal components. Finally, the retained IMFs are reconstructed to produce the denoised signals. Experimental results demonstrate that, compared to existing artifact removal techniques, the proposed method achieves a higher signal-to-noise ratio (SNR) improvement, reduced root mean square error (RMSE), and an enhanced power ratio.