An Improved VMD and Wavelet Hybrid Denoising Model for Wearable SSVEP-BCI

The brain-computer interface (BCI) based on steady-state visual evoked potentials (SSVEP) has attracted considerable attention due to its non-invasiveness, low user training requirements, and efficient information transfer rate. To optimize the accuracy of SSVEP detection, we propose an innovative hybrid EEG denoising model combining variational mode decomposition (VMD) with wavelet packet transform(WPT). This model ingeniously integrates VMD decomposition and WPT denoising techniques, employing detrended fluctuation analysis (DFA) thresholding to deeply filter the noisy data collected from wearable devices. The filtered components are then reconstructed alongside the unprocessed components. Finally, three classification algorithms are used to validate the proposed method on a wearable SSVEP-BCI dataset. Our proposed algorithm achieves accuracies of 71.27% and 86.35% on dry and wet electrodes, respectively. Comparing the use of VMD combined with adaptive wavelet denoising and direct denoising with VMD, the classification accuracy of our method improved by 3.68% and 0.26% on dry electrodes, respectively, and by 3.28% and 0.66% on wet electrodes, respectively. The proposed approach demonstrates excellent performance and holds promising potential for application and generalization in the field of wearable EEG denoising.

@article{Xia2024,
  title     = {An Improved VMD and Wavelet Hybrid Denoising Model for Wearable SSVEP-BCI},
  journal   = {International Journal of Advanced Computer Science and Applications},
  volume    = {15},
  number    = {5},
  year      = {2024},
  publisher = {The Science and Information Organization},
  author    = {Yongquan Xia and Keyun Li and Duan Li and Jiaofen Nan and Ronglei Lu},
  doi       = {10.14569/IJACSA.2024.0150532},
  url       = {https://doi.org/10.14569/IJACSA.2024.0150532}
}