With the rapid development of EEG signal acquisition devices, EEGnowadays is more often used in real life, and EEG-based emotion recog-nition has become a popular research topic. This paper focuses on how toapply EEG signal to emotion recognition during watching videos. Energyspectrum is a traditional feature used in this task, whereas in this paperwe employ diferential entropy feature which can describe emotion relatedEEG signal better. Meanwhile, feature selection method is applied to fndkey frequency bands and key brain area for emotion recognition, whichprovides evidence for the neurophysiology to explore the machanism ofemotion. Linear dynamic system is employed to smooth the feature se-quence, in order to increases the accuracy of the recognition. WhetherEEG-based emotion recognition can be promoted in the real life is largelydependent on the stability of the model. In this paper, it is shown that themodelweproposedisstablewhentestingthemodelinarelativelylongdu-ration. The average accuracy of the given recognition algorithm achieves85%on the18experiments of6subjects. |