Evaluating the Use of Feature Extraction and Windowing Using Neural Network in EEG-Based Emotion Recognition

Electroencephalogram (EEG) based emotion recognition has gained more attention in recent years due to the emergence of deep learning models, which enabled researchers to dive deeper into the pre-processing, the feature extraction methods, and choosing the right parameters for their model. This study focuses on the windows used to segment an EEG signal and the extracted features while using a deep neural network on the DEAP dataset. The extracted features consist of Deferential Entropy (DE), Power Spectral Density (PSD), Fast Fourier Transform (FFT), and Discrete Wavelet Transform (DWT). Testing the effect of those parameters on the accuracy of a combined neural network architecture. The model used is composed of a Convolutional Neural Network (CNN) and Long Short Term Memory (LSTM) network. The proposed method achieved the highest accuracy for a combination of FFT and a window size of 0.25 s.