Cross-Subject fNIRS Signals Channel-Selection based on Multi-Objective NSGA-II Algorithm

In brain-computer interface (BCI) systems, finding an optimal channel set to decrease the cost of computation and portability of the signal acquisition system to achieve higher classification accuracy is vital. This study presents a multi-objective meta-heuristic algorithm to select optimal channels in the multi-channel fNIRS signals. We proposed non-dominated sorting multi-objective Genetic Algorithm (NSGA-II) to perform channel selection. We will find an optimal solution to the channel selection problem in brain-computer interface (BCI) systems to obtain the best channels in the multi-channel fNIRS signal dataset. Toward the classification of the fNIRS signals, a preprocessing task should be applied, followed by selecting the channels set and extracting related features of every trial. In the next step, in order to apply the feature selection method, the mRmR algorithm was applied. Classification accuracy with 10-fold cross-validation is then performed as an objective for the presented algorithm to select the best accuracy and best channel set. Finally, the results illustrate that the proposed selected method obtained an average of 27.25 best optimal channels per subject. Moreover, classification results are 67.9±11 % for the subjects. It was found that the LDA classification method resulted in the best performance compared to other methods.