<p>Emotion recognition from electroencephalogram (EEG) signals plays a crucial role in human-computer interaction, mental health monitoring, and cognitive neuroscience. In this study, we propose a novel approach for emotion recognition that integrates a chaotic local binary pattern (CLBP) for effective feature extraction, combined with the cuckoo search algorithm (CSA) for feature selection. The extracted features, which represent the non-linear dynamics of the EEG signals, are optimized through CSA to enhance the discriminative power of the emotional states. To classify these features, we employ the XGBoost classifier, a gradient-boosting model known for its high performance on categorical data and the ability to handle complex decision boundaries. The proposed model is validated using publicly available EEG datasets GAMEEMO and DREAMER, achieving superior accuracy and robustness in recognizing emotional states compared to conventional methods. The proposed method achieves the highest classification accuracy of 99.20% and 98.16% on the GAMEEMO and DREAMER databases, respectively, for subject-independent emotion recognition using the XGBoost classifier. Our findings highlight the potential of combining CLBP and CSA with machine learning for real-time cross-subject emotion recognition, offering promising applications in affective computing and neurofeedback systems.</p>