Merged LSTM Model for emotion classification using EEG signals

The applicability of contemporary deep learning techniques have seen considerable improvements in the field of biomedical signal analysis. Emotion analysis using EEG signals is one such problem that has been studied and worked upon extensively in recent times. In this paper we have proposed a novel methodology to classify emotions using signal processing techniques such as wavelet transform and statistical measures for feature extraction and dimensionality reduction followed by developing state of the art neural architecture for the classification task. A merged LSTM model has been proposed for binary classification of emotions. The model's applicability and accuracy has been validated using DEAP dataset which is the benchmark dataset for emotion recognition.