Eyeblink Detection Algorithm Based on Joint Optimization of VME and Morphological Feature Extraction

Eyeblink detection is critical in areas such as electroencephalography (EEG) artifact removal and health monitoring. In this article, we propose a single-channel automatic eyeblink detection algorithm based on joint optimization of variational mode extraction (VME) algorithm and morphological feature extraction (MFE). First, we use the ${k}$ -means clustering algorithm and discrete Fourier transform (DFT) to automatically extract the center frequency of eyeblink signal. Simultaneously, we use singular spectrum analysis (SSA) to filter the EEG data. Then, eyeblink detection is performed based on VME and adaptive threshold extracted by MFE. Finally, the processing of eyeblink detection is globally optimized based on gray wolf optimization (GWO) to obtain the best combination of parameters in VME and MFE, and the performance of the algorithm is verified by experiments on semi-simulated dataset and collected real EEG database of nine subjects. Several traditional eyeblink detection methods are compared, and the results show that the joint optimization method obtains the best accuracy, sensitivity, and false negative rate of 97.63%, 92.64%, and 0.017, respectively.