The discovery of EEG signal in 1929 by the German psychiatrist Hans Berger changed the way we understand the structure and functioning of brain.The interaction between a machine and computer is increasing day by day and the need of the hour is to develop a Brain Computer Interface(BCI) which can help the humanity.EEG Emotion recognition system can be used to predict the emotions felt by disabled people.Studies have found that corticolimbic Theta electroencephalographic (EEG) oscillation is responsible for the emotions that one feels. Alpha, Theta, Beta and Delta sub-bands of EEG play a major role in brains emotion processing. The goal of this study is to identify emotions from an EEG data collected from a Single Channel EEG headset.13 subjects of varying age participated in the EEG experiment which were shown videos that helped in evoking three emotional states: neutral, calm and fear. After each video the participants were asked to rate the video on the basis of SAM Model and Valence-Arousal scale.The method used is based on Digital Signal Processing Techniques in order to remove arte-facts,clubbed with machine learning in order to design a system for predicting emotions using EEG signal.Stationary Wavelet Transform (SWT) with haar wavelet at level 6 decomposition with Garrote Thresholding is used to clean the signal and remove the noise. Higuchi Fractal Dimension is also calculated and added as one of the features and is found to have increase the classification accuracy due to its ability to identify the patterns from the data.Experimental results show that EEG based emotion classification can predict emotions with an average of 76% in case of pure EEG signal and 85% in case of EEG signals with Valence-Arousal scale using Recurrent Neural Networks. © 2019 Galgotias University.