METHOD AND APPARATUS FOR HUMAN STRESS DETECTION BASED ON ELECTROENCEPHALOGRAPHY SIGNALS (EEG) Emotions play an important role in our day-to-day life. They help in identifying a human state of mind. Information about the human state of mind can significantly help in human-machine interaction and brain-computer interface. Some of the existing researchers have used speech, text, gesture or facial expressions for emotion recognition. However, these factors vary across culture and nation. Because of which, it is difficult to detect emotions more accurately. Hence, present work considers EEG signals for emotion recognition which not only ignores external factors but also helps to detect real emotions arising directly from our brain. A benchmark DEAP (Dataset for Emotion Analysis using Physiological signals) dataset is used for emotion analysis. A novel feature extraction technique called Frequency Cepstral Coefficient (FCC) is proposed to extract features from DEAP dataset. FCC technique is compared with Kernel Density Estimation (KDE) on DEAP dataset for feature extraction. These extracted features are then classified into two emotional states - happy and sad using K-Nearest Neighbor (K-NN) classifier. The selection of most appropriate and reliable method of feature extraction greatly helps for accurate classification. Number of experiments was conducted to evaluate the efficiency of feature extraction techniques. The experimental results show that KDE gives 80% accuracy and FCC outperforms it by achieving 90% accuracy on DEAP.
