Emotion Recognition Based on EEG Brain Rhythm Sequencing Technique

This work proposes a technique that analyzes electroencephalography (EEG) using brain rhythms ( $\delta $ , $\theta $ , $\alpha $ , $\beta $ , and $\gamma $ ) presented in a sequential format and applies it for emotion recognition. Although brain rhythms are regarded as reliable parameters in EEG-based emotion recognition, to achieve high accuracy by considering fewer optimal multichannel rhythmic features (MCRFs) has not been addressed in detail. Thus, the rhythm sequence for each channel is generated by choosing the strongest brain rhythm having the maximum instantaneous power for every 200-ms time bin. A ${k}$ -nearest neighbor ( ${k}$ -NN) classifier is employed for evaluating the rhythmic features extracted from different sequences, and the experimental validation was performed on three well-known emotional databases (DEAP, MAHNOB, and SEED). The results showed that approximately 30% of MCRFs for as high as 87%–92%, achieving high classification accuracies with a small number of data. Further investigation revealed that the frontal and parietal regions are active during the emotional process, as consistent as earlier studies. Therefore, the proposed technique demonstrates its availability and reliability for emotion recognition. It also provides a novel solution to find optimal channel-specific rhythmic features in EEG signal analysis.