Emotion analysis with EEG based deep learning

The brain is the basic organ that controls the regular functioning of the body of living things and the fulfillment of the tasks necessary for their survival. When considered from this point of view, it is obvious that the brain has a vital importance for human life and that any malfunctions in this organ will directly affect the body. Electroencephalography (EGG) is a tool that enables the detection of malfunctions in the brain and is a type of signal obtained as a result of electrical activity in the brain. The fact that EEG signals carry specific information about many neurological diseases has pushed researchers to examine these signals. One of the diseases that occur in the brain and that can be detected by using EEG signals is epilepsy, the incidence of which is increasing day by day and may result in death or permanent damage due to limited opportunities. The existence of the dangers arising from these impossibilities has led researchers to propose low-cost assistant systems. Within the scope of this study, we aimed to create an assistant approach in order to reduce the mistakes that the specialists can make due to environmental factors and to prevent the irreparable or unrecoverable mistakes that these mistakes may cause. In this direction, we aimed to analyze the data of healthy people and people with epilepsy by performing our study with two different data sets and to determine the states of the subjects as a result of these analyzes. Within the scope of this study, we aimed to create an assistant approach in order to reduce the mistakes that the specialists can make due to environmental factors and to prevent the irreparable or unrecoverable mistakes that these mistakes may cause. In this direction, we aimed to analyze the data of healthy people and people with epilepsy by performing our study with two different data sets and to determine the states of the subjects as a result of these analyzes. For this purpose, we classified The Hauz Khas dataset and Bonn Epilepsy dataset, which is frequently used in the literature, by applying the deep learning methods Resnet-50, Resnet-101, AlexNet, GoogleNet and VGG-19 approaches to the datasets. However, we tried to increase the classification accuracy and thus the acceptability of the method by applying the Short Time Fourier Transform (STFT) and Continuous Wavelet Transform (CWT) methods to the data in order to subject the data to the classification process with sharper limits rather than taking the data in a raw way. At the end of the study, as a result of using CWT as a feature extraction method, we were able to classify the preictal and ictal moments on The Hauz Khas dataset at 100% with all classifiers. In addition, we achieved 100% success rate with all classifiers by using the CWT feature extraction method in binary classifications of the Bonn dataset. Finally, we achieved a classification success rate of 99,33% in the classification of the AB-CD-E group.