Emotions play a major role in the day-to-day life of a human being, whether it is related to the process of making a decision or interacting with other people. Emotion recognition is all about knowing how a person feels at a particular point in time. Brain signal correlates with emotional state. Encephalography (EEG) signals have been used to measure brain signals. There are different types available to measure the emotional state of the subject such as text, facial expression, audio, video, and physiological signal. In the proposed system EEG signal is considered for analysis of the emotion of the subject. Emotion recognition is a topic that comes under the intersection of psychology, human-computer interaction, cognitive science, and neuroscience. At present, the classification of emotions based on Brain-Computer Interface (BCI) systems is an emerging topic in the emotion recognition research area. A major challenge faced in BCI systems is to give an accurate result in consideration of the Signal to Noise (SNR) ratio. In the real world, utilization of this technology is restricted for implementation in real applications since it is hard to classify and generalize features without proper technique as well as the accuracy of classification is lower than expected. To overcome the problems faced in this research field, deep learning algorithms have been implemented in the past few years in the BCI field. The main problem in traditional machine learning algorithms was that the features are not classified accurately. This problem has been overcome by the use of deep learning algorithms as they help in obtaining accurate features from the brain signals without selecting them manually. This in return gives a well-defined accuracy when applied to a problem along with a considerable size of the training set. Among various neural networks in the deep learning field, Recurrent Neural Networks (RNN) have gained importance in the past few years in the field of emotional analysis as they are used in situations that make use of sequential data as they are trained to recognize patterns across time. The main disadvantage which was being observed while using RNN algorithms is that when they learn through backpropagation through the hidden layers it gets affected by the vanishing gradient problem. To overcome this, Bidirectional Long Short-Term Memory (Bi-LSTM) was being utilized. The use of Bi-LSTM has given accurate results when applied to various problems involving sequential prediction, as it helps to overcome the vanishing gradient problem which occurs in the recurrent neural networks. Therefore, the Bi-LSTM algorithm will help provide efficient and accurate output in the BCI system. [ABSTRACT FROM AUTHOR]