This study aimed to assess the emotional experiences of stroke patients and normal people using electroencephalogram (EEG) signals in happiness and sadness. The brain behaviors under both emotional states in the EEG signal were analyzed through signal processing methods. In this study, the EEG signals of normal control (NC) and stroke patients with left brain damage (LBD) and right brain damage (RBD) were analyzed through Hjorth parameters. The extracted Hjorth parameters showed significant differences between happiness and sadness in alpha, beta, and gamma frequency bands, which implied the two emotions exhibiting different brain behavior in different EEG bands. The topographic mapping of the Hjorth parameters presented different activation patterns in the subject groups, and the higher frontal activation can be observed from the NC group for both emotions. Also, the Hjorth Mobility and Complexity parameters were lower in LBD and RBD in the frontal regions of the alpha band. The significant difference channels between the emotions were analyzed by statistical analysis using ANOVA. Moreover, the features of each subject group were used for emotion classification by the application of machine learning-based algorithm. The KNN classification results achieved an average accuracy of 92.35% for NC, 90.84% for LBD, and 95.59% for RBD in classifying happiness and sadness. The emotion classification showed that the emotional dominance frequency bands were the beta and gamma bands. However, the alpha band activity showed left frontal lateralization in the NC group, while right frontal lateralization in the LBD and RBD groups suggested different brain activation of the stroke groups and the controls during happiness and sadness, which reflected the emotional impairment in stroke groups.
