EEG-Based Multiclass Workload Identification Using Feature Fusion and Selection

The effectiveness of workload identification is one the critical aspects in a monitoring instrument mental state. In this field, usually recognized as binary classes. There are scarce studies toward multiclass because challenge success much tough, even though more class added. Besides, most existing only utilized spectral power features from individual channels but ignoring abundant interchannel that represent interactions between brain regions. study, we representing intrachannel information and to classify multiple classes based on an electroencephalogram. We comprehensively compared each category contributing elucidated roles feature fusion selection for identification. results demonstrated combination (83.12% terms accuracy) enhanced classification performance with categories (i.e., band features, 75.90%, connection 81.72%, accuracy). With F-score selection, accuracy was further increased 83.47%. When graph metric were fused, reached 84.34%. Our study provided comprehensive comparisons methods played important role enhancement These could facilitate practical application