In contrast to traditional emotion classification using text, speech, and video data, a relatively new approach of using Electroencephalography (EEG) and brain wave patterns is discussed in this paper. EEG is already being used in the medical field to study and diagnose patients with sleep disorders, coma, encephalopathies, epilepsy, and brain tumors. The proposed approach is applied to the prerecorded EEG dataset (DEAP) to classify emotions. Emotions are categorized as Low (class 0) and High (class 1) for four parameters. In this paper, a new approach is carried out to extract the features from the EEG signals using empirical wavelet transform and various entropy computations to remove the dataset’s bias towards the majority class. A fusion of the most suitable classification techniques is evaluated by calculating the F1 score as a performance metric. The achieved F1 score is 78.925 for Valence, 70.04 for Arousal, 74.129 for Dominance, and 81.14 for Liking.