Arousal detection by eeg entropy and ERP

Abstract

Human emotions govern our quality of relations within the society. It can be expressed as a state of mind which predicts the response of a person whether positive or negative to a particular situation. It is a versatile phenomenon which controls the reactions and behavior of the person depending upon his mental state whether he is happy, angry, frustrated, sad, excited and scared. Emotions can be predicted from gestures, sound processing and facial expression can be faked but emotion recognition using EEG signals is very powerful method to know the internal state of mind accurately. In this dissertation, the aim of the study is to design an interactive and a smart two class system for emotion recognition based on Electroencephalogram (EEG) signals. In this study EEG signal is acquired on frontal electrodes such as F3, F4 and FZ from five subjects for classification of emotions into two classes namely HVHA and LVHA. The images provided by International Affective Picture System (IAPS) have been used for evoking emotions. The data has been acquired by placing a cap on the head of a subject as per 10-20 International system. The obtained EEG signal has been filtered in offline mode by using low pass IIR filter, high pass IIR filter and a notch filter. The low pass IIR filter is followed by high pass IIR filter. IIR filters have been used to bring the EEG signal in the frequency range of 0.5 to 40Hz. The notch filter has been used to remove the 50Hz power noise interference. The ERP potentials such as P100, N100 and the two latencies corresponding to these bio potentials have been extracted for every class of emotion from filtered EEG signals. The EEG signal is decomposed into five different frequency bands namely delta (0-4 Hz), theta (4–8 Hz), alpha (8–16 Hz), beta (16–32 Hz) and gamma (32- 64Hz) by using filtering technique. The entropy attribute from these five frequency bands has been extracted. The training and testing has been performed on the eleven combinations extracted from four attributes of ERP. The six combinations extracted from Entropy attribute is also used for classification. The classification has been performed using LIBSVM classifier with 3 fold cross validation and RBF kernel to classify emotions into two classes.