INTERNATIONAL JOURNAL OF SCIENTIFIC & TECHNOLOGY RESEARCH VOLUME 9, ISSUE 02, FEBRUARY 2020 ISSN 2277-8616 1827 IJSTR©2020 www.ijstr.org Emotion Recognition And Classification Using Eeg: A Review Nandini K. Bhandari, Manish Jain Abstract: Emotions result in physical and physiological changes which affect human intelligence and the world around us. Emotions which indicates inner feelings of a person is represented by EEG as a direct brain response to a stimuli. EEG-based emotion recognition is widely used in affect computing to improve communication between machines and human. In this paper we provide a comprehensive overview of methods proposed for emotion recognition using EEG published in last ten years. Our analysis is focused on feature extraction, selection and classification of EEG for emotion recognition. This survey will be a mile stone for researchers in enhancing the development of emotion recognition using EEG. Index Terms: DEAP, CNN, EEG, Electroencephalograph, EMD, emotion recognition, neural network, SVM. ——————————  —————————— 1 INTRODUCTION Emotions play a vital role in our daily life because they affect human cognition, perception, interaction, decision making ability along with human intelligence [1]. However, they were ignored by human computer interaction (HCI) systems till last decade. The HCI systems along with digital media, find potential applications in biomedical engineering, neuroscience, neuromarketing and other alternate areas of life, which are mainly affected by emotions. Hence, with increasing demand of HCI, automatic human emotion recognition is gaining the attention of researchers. The emotion recognition can be done with the help of text, speech, gesture movements and facial expressions [2] but electroencephalogram (EEG) gives better outcome as it directly measures true feelings. EEG is non-invasive and have high temporal resolution [3]. A rapid development in new wearable, handy, low cost wireless headsets measuring EEG and classification of EEG signals without trained professionals has enormously increased its use in other areas like, sleep management, e-learning, video games, cyber world, healing etc. This literature survey has covered recent methods used in EEG based emotion recognition, which will be helpful to researchers working in this field. The remaining paper is organized as follows. Section II describes emotions, characteristics of EEG signals and basic steps used in emotion recognition. Section III describes about database used in most of the papers. Section IV deals about preprocessing methods used on raw EEG signals. Section V contains information about various processes used in feature extraction and classification. Section VI discusses about the various aspects related to review and section VII gives conclusions extracted from this survey. 2 EMOTIONS AND EEG An emotion is complex physiological state which involves a person's experience, a physiological response and behavioral change. A person’s inner emotional state may become apparent by subjective experiences (how the person feels), internal/ inward expressions (physiological signals), and external/ outward expressions (audio/visual signals) [4]. These are temporary signals, having short duration and intensity variation. According to Paul Ekman and Friesen [5], there are six universal emotions, independent of various cultures in the world. They are happiness, fear, anger, sadness, disgust and surprise. Plutchik has considered eight emotions: anger, fear, sad, disgust, surprise, curious, acceptance and joy [6]. These emotions are highly complex in nature, varying from person to person. This complexity makes emotion recognition a challenging task. The studies of emotion recognition use different types of emotions techniques for classification: a. Discrete emotions: Happiness, fear, anger, sadness, disgust and surprise. Researchers may take single emotion or opposite emotions for detection. One may use four emotions namely happy, sad, fear and anger. b. Two emotions: Positive and negative. c. Valance arousal model: Valance means from very positive to very negative, arousal means sleepy to exited and dominance gives strength of emotion [7]. EEG (Electroencephalogram) The human cortex is divided into frontal (F), temporal (T), central (C), parietal (P) and occipital (O) lobes. EEG signal is the voltage fluctuation obtained by ionic current flow with synaptic connections of neuron. In an adult, EEG signal measured from scalp is a sinusoidal signal of range 10- 100μV. Useful information from brain is divided in five frequency bands namely, delta (0-4Hz), theta (4-8Hz), alpha (8-13Hz), beta (13-30Hz) and gamma (30-70Hz) [8]. Delta waves are obtained during deep sleep. Theta waves are associated with subconscious mind activities like sleeping and dreaming. Alpha waves occur during relaxed state and are more prominent in parietal and occipital lobe. Beta waves occur during focused mental activity. Gamma waves occur during hyper brain activity [9]. International 10/20 system [10] as shown in figure 1, is used for placing electrodes on skull to get EEG signals. The numbers 10 and 20 suggests, distance between neighboring electrodes (10% or 20% of total front-back or right-left distance of skull). ———————————————— • Nandini K. Bhandari is currently pursuing Ph. D Electrical and Electronics Engineering from Mandsaur University, Mandsaur. India. PH-+919922921252. E-mail: nandiniboob@gmail.com • Professor (Dr.) Manish Jain, Associate Professor, MCGER, MIAENG,Electrical and Electronics Engineering from Mandsaur University, Mandsaur. India,. E-mail: manish.jain@meu.edu.in