Segmentation-enhanced approach for emotion detection from EEG signals using the fuzzy C-mean and SVM.

The analysis of EEG signals for determining emotion is one of the most important topics in the field of artificial intelligence. It can be applied in a wide variety of areas, such as emotional health care and the man/machine interface. The purpose of the paper is the demonstration that emotions may be identified using EEG recordings in the hybrid approach based on the differentiated support vector machine (SVM) models with various types of kernel functions, as well as fuzzy C-means. The EEG signal of two subjects was recorded with the help of the Muse headband; the signal data was described as positive, neutral, or negative emotions. A Gauss kernel was the second-best outcome (95.78%), and a linear kernel was the best outcome (97.66%). Precision, recall, and F1-scores were used in establishing the performance of the SVM technique in emotion classification in conjunction with the fuzzy C-means classification approach. Besides covering the discussion on the importance of kernel choice in achieving good performance in SVM-based models, the analysis also showed that there was a potential to use EEG-based emotion detection. Moreover, one-way ANOVA statistical analysis has expressed that the linear kernel did perform significantly better as compared to other kernels (p < 0.05). To confirm that the proposed system would be rather robust, other deep learning models (CNN-LSTM hybrids) were designed and tested, the results of which proved that they had similar performance and at the same time less accurate results than the linear SVM. These results indicate the efficacy of SVM and the optimization of kernel parameters along with the integration of fuzzy logic in recognizing emotions based on EEG records.