Computer games and movies evoke a wide range of emotions in youths, some of which may lead to psychological problems like depression, insomnia, suicidal tendencies, etc. Electroencephalogram (EEG) signals, reflecting brain activity, offer a reliable means for emotion recognition using frequency rhythms: \(\delta \) , \(\theta \) , \(\alpha \) , \(\beta \) , and \(\gamma \) . This study conducts channel-dependent and subject-independent analyses by decomposing EEG signals into various rhythms. It extracts joint time-frequency scattering (JTFS) features, known for their invariance, stability, and excellent representation of temporal and spectral information. Additionally, silhouette analysis validates the distinctiveness of the identified emotional clusters. JTFS features fed into KNN, SVM, Ensemble, and Neural Network classifiers demonstrate that KNN achieves the best performance–80.39% (Dataset I) and 99.3% (Dataset II). Among rhythms, \(\delta \) , \(\theta \) , and \(\alpha \) outperform \(\beta \) and \(\gamma \) , with the combined \(\delta \) + \(\theta \) rhythm displaying the highest accuracy.
