Assessing Cognitive Load Combining Features of Time,Frequency and Spatial Domain under Digital Graphical Interface

Evaluating operator’s cognitive load(CL)under digital graphical interface accurately can help to realize cognitive feedback mechanism and ultimately improve ergonomics.In order to further improve the robustness and generalization capability of the evaluation method,Att-BLSTM is applied to CL evaluation problems combining with EEG experiments.First,we train Multi-CNN to extract time and spatial domain features.Next,we train an Att-BLSTM to learn robust representations from raw EEG time series.Finally,a multi-feature fusion strategy is used to construct CL evaluation method.12 subjects were recruited,and EEG data were collected under two CL conditions.The average accuracy of our method on our dataset is 82%,which has a stronger EEG signal characterization capability than the traditional machine learning method,it can also extract the time domain characteristics of EEG more accurately and with stronger robustness than other deep learning methods.