EEG signals can be used as an accurate metric for human emotions. Using machine learning, and subsequently deep learning, we can accurately classify EEG signals and predict emotions. Several deep learning methodologies have been experimented with for this purpose. Thus, we have reviewed notable published research on the field and compared their results. EEG is a complex data format that holds further potential for analysis. Therefore, we have also compared and discussed the contribution of different feature extraction techniques in EEG data. We have tested traditional machine learning models (SVM, kNN, MLP and ELM) and deep learning models ViT and RNN. The results show deep learning models with attention mechanisms give the best accuracy in emotion recognition.