Improving PSO-SVM for Fatigue Recognition

Excessive fatigue can cause harm in an individual's life. Therefore, the challenge is how to effectively detect fatigue. To this end, this paper uses an EEG multi-feature fusion method based on Linear Discriminant Analysis (LDA) dimensionality reduction, and uses the dimensionality reduction features as the input of Support Vector Machine (SVM) for classification. In order to improve the classification accuracy of SVM, an improved Particle Swarm Optimization (PSO) algorithm was proposed to optimize the SVM. The improved PSO-SVM algorithm is combined with other classification methods to classify the EEG signals of the subjects in the fatigued and awake states. The experimental results show that the improved PSO-SVM algorithm has achieved the best classification performance, the average accuracy rate reached 84.56%.