Electroencephalogram emotion recognition method and system based on multiple tasks and attention mechanism

The invention discloses an electroencephalogram emotion recognition method and system based on multiple tasks and an attention mechanism, and the method comprises the steps: S1, collecting corresponding electroencephalogram signal data according to the position of an electroencephalogram channel related to emotion in an international standard 10-20 electrode placement system; s2, performing preprocessing such as band-pass filtering, power frequency notch filtering, downsampling, baseline correction, artifact removal and signal segmentation on the acquired electroencephalogram signals, and segmenting the signals into windows with the duration of 0.5 second; s3, power spectral density and differential entropy feature extraction is performed on the electroencephalogram data of each time slice, feature vectors are represented by two-dimensional brain maps to reserve spatial information, brain maps of all wavebands and features are stacked in a three-dimensional space, and each map serves as input of a model; and