Emotion Recognition with Multi-Channel EEG Signals Using Visual Stimulus

Emotions are complex and may vary from person to person in a situation. The purpose of this study is to perform emotion analysis by using specific signal processing algorithms, to find the features and channels that are effective in the emotion recognition by using 60 visual stimuli with obtained EEG signals from the 32-channel EEG device that is branded Brain Products from 25 volunteers. The graphical user interface (GUI) has been designed to display visual stimuli at certain time intervals. Empirical mode decomposition (EMD) method has been applied to EEG signals and Intrinsic Mode Functions (IMFs) have been obtained. The most meaningful IMF has been investigated by using signal analysis methods such as Power Spectrum Density (PSD) using Periodogram and Welch's method. Feature extractions have been performed for filtered signal, IMF1, IMF2, IMF3 and average of first three IMFs. The classification has been done by the support vector machine (SVM). Different channels and features have been classified. It has been found that IMF1 and IMF2 are the most meaningful IMFs, Hjorth parameters have highest success rate as features and F4, T7, T8 EEG channels were the most effective channels with a success rate of 81.80%, 77.13%, 76% in emotions recognition.