Optimizing Largest Lyapunov Exponent Utilizing an Intelligent Water Drop algorithm: A Brain Computer Interface Study

Abstract

Brain computer interface (BCI) based on imagery motor cortex is a critical issue for paralyzed patients. Several algorithms are developed to detect the imagery movement patterns based on electroencephalograph (EEG) signals. In this study, a chaotic feature named as Largest Lyapunov Exponent (LLE) is employed to diagnose the imagery movement patterns. To find the best LLE values, the input parameters for constructing a phase space lag are optimized using a Water Drop optimizer method. To evaluate the Water Drop algorithm for the LLE features, a task is designed to record EEG signals from 18 subjects. Results show that the traditional LLE and the optimized LLE accuracies are reach to 57.77% and 65.44% average accuracies, respectively. The Obtained paired T-test for the optimized LLE show that the number of subjects whose feature are significant are more than the traditional LLE method. It is concluded that the Water Drop optimizer algorithm improves values for reconstructing a phase space feature extraction than the traditional LLE method significantly.