The Optimal Number and Distribution of Channels in Mental Fatigue Classification Based on GA-SVM

Abstract

Mental fatigue is closely related to our daily life and work, a considerable number of studies have achieved good results in quantifying and predicting them. Although some studies have achieved a high accuracy by using only a single channel, and a few have explored the optimal solution for feature and channel selection. However, detailed research of optimally setting the electrodes position and determining the number channels are rarely seen. In this study, by designing a novel genetic operator and applying the GA-SVM model, we compared the maximum number of optimal channels and their distributions. The result suggests that the classification accuracy almost reaches its optimum (94.0±5.3 %) when the maximum number of channels reaches 5, and is not affected by the epoch length. The whole brain optimal channels topographic map analysis shows that the optimal channels are mainly distributed in the prefrontal, occipital and temporal lobes, while hardly any is located in the parietal lobe, which indicates that the mental fatigue induced by visual search task characterized similarly among different individuals and highly task-related.