African vultures optimization algorithm based enhanced VMD with CWT and k-medoids for removal of ocular artifacts from EEG signals

Electroencephalogram (EEG) records the neuronal activities of the human brain, which helps in the detection and diagnosis of different neurological disorders. The recorded EEG signals are mostly contaminated by ocular artifacts (OAs). Variational mode decomposition (VMD) is a non-recursive signal decomposition technique that has been used to eliminate OAs. However, the performance of the VMD algorithm is affected by the selection of the number of modes (K) and penalty factor (α) parameters. Hence, an enhanced VMD is proposed in this research by optimizing the parameters of the VMD algorithm using the African vulture’s optimization algorithm with the combination of energy entropy and ensemble kurtosis as an objective function. First, an enhanced VMD decomposes an EEG signal into a defined number of band-limited intrinsic mode functions (BIMFs). Next, the BIMFs related to OAs are identified based on the power spectral density-based distance metric, and then the noisy BIMFs are added to estimate the OA signal. Then, the combination of continuous wavelet transform and k-medoids clustering techniques is used to extract only OA-contaminated intervals from the estimated OA signal. Finally, the extracted OA signal is subtracted from the contaminated EEG signal. The proposed method is validated with EEG signals from the Mendeley, Keirn EEG, and CAP sleep datasets. The comparative analysis shows the superiority of the proposed work over existing techniques. Further, the channel-wise and length-wise analysis reveals the efficacy of the proposed work.