EEG-based analyzing human idiosyncrasies within unpredictable circumstances using 3D convolutional neural networks

Abstract

Automatic recognition of emotions is a very challenging of task. To detect emotions from EEG signals, a sophisticated learning algorithm is needed that can reflect high-level abstraction. Using electroencephalogram ( EEG) signals, in particular the learning of spatiotemporal characteristics, various methods have been used to improve the robustness of the emotion detection systems. The use of a model inspired by the Siamese Network in this approach, which deals with data by concatenation and 3D convolution. The Adam optimizer was used in this method for preparation. This model is a custom model which can deal with a stack for a single time instant, which gives better efficiency than 3D Convolutional Network. This type of model is usually used for Face Recognition, but use of such a model for this purpose opens a new horizon of opportunity for Automatic recognition of emotion. Highest accuracy of our project is around 70% has been achieved by Dense Net and lowest accuracy was 30%.