Hierarchical Harris hawks optimization for epileptic seizure classification.

The intelligent recognition of electroencephalogram (EEG) signals is a valuable tool for epileptic seizure classification. Given that visual inspection of EEG signals is time-consuming, and that mutant signals dramatically increase the workload of neurologists, automatic epilepsy diagnosis systems are extremely helpful. However, the existing epilepsy diagnosis methods suffer from some shortcomings. For example, they tend to fall into local optima quickly because of their failure to fully consider the discriminative features of EEG signals. To tackle this problem, in this article, an enhanced automatic epilepsy diagnosis method is proposed using time-frequency analysis and improved Harris hawks optimization (IHHO) with a hierarchical mechanism. Specifically, the signal is decomposed into five rhythms using continuous wavelet transform, with the local and global features extracted using the local binary pattern and the gray level co-occurrence matrix. Discriminative features are then selected and further mapped to the final recognition results using both IHHO and the k-nearest neighbor classifier. To evaluate its performance, the proposed method was compared with a variety of classical meta-heuristic algorithms on 23 benchmark functions. Moreover, the proposed approach achieved more than 99.67% accuracy on the Bonn dataset and 99.06% accuracy on the CHB-MIT dataset, out-performing a multitude of state-of-the-art methods. Taken together, these results demonstrate the utility of our approach in the automatic diagnosis of epilepsy. Supportive datasets and source codes for this research are publicly available at https://github.com/sstudying/lzzhen, and latest updates for the HHO algorithm are provided at https://aliasgharheidari.com/HHO.html.