Emotion recognition from EEG signals is a pivotal aspect of human–computer interaction and affective computing. This research introduces a sophisticated system designed for real-time emotion identification, leveraging EEG data from the DEAP dataset. Employing a meticulous methodology, our approach ensures stability and reliability through a hybrid data augmentation strategy, z-score normalization, and targeted attention to pertinent frequency bands during preprocessing. The feature selection process is enhanced through the application of Genetic Algorithm, while the combination of Empirical Mode Decomposition and Wavelet Transform optimizes feature extraction. In the classification phase, cutting-edge neural network architectures, including Long Short-Term Memory (LSTM), Recurrent Neural Network (RNN), and Convolutional Neural Network (CNN), are employed. The debut of a novel vote classification technique, coupled with refined hyperparameter adjustments, contributes to heightened accuracy of 89.9% for CNN with Wavelet Transform and Genetic Algorithm, 89.9% for LSTM with Wavelet Transform and Genetic Algorithm and 88.9% for RNN with Wavelet Transform and Genetic Algorithm. The culmination of these methodologies results in an ensemble model, incorporating Genetic Algorithm, achieving an impressive accuracy rate of 90.5%. In conclusion, our recommended strategy not only demonstrates the potential to accurately identify a diverse range of emotions in real time but also emphasizes precision and adaptability through the integration of sophisticated methodologies and systematic optimization. This research stands as a comprehensive and effective contribution to the field of emotion recognition (Emotions—Surprised, Confused, Angry, Happy, Sad) with implications for diverse applications in real-world scenarios.