Electroencephalogram Based Emotion Recognition Using Variational Model Decomposition and Long Short-Term Memory

An Emotion Recognition (ER) is one of the significant research topics for its universal applications in the area of human-computer interactions (HCI). The human emotional features are utilized for recognition of various emotions and also recently numerous researches are developed on ER. Recently, Machine Learning (ML) and Deep Learning (DL) approaches are developed based on Electroencephalogram (EEG) signals for the ER, however, the ML-based approaches have time complexity. In this research, the hybrid method of Variational Model Decomposition (VMD) and Long Short-Term Memory (LSTM) is proposed for EEG-based ER and classification. Initially, this research collects the publicly available DEAP as well as SEED dataset for estimating the model performance. The Z-score normalization technique is used for normalizing the input data in the pre-processing step. Then, the pre-processed data is extracted by using Mel-Frequency Cepstral Coefficient (MFCC) technique. Finally, classification is done by using VMD-LSTM. The experimental results shows that the proposed method achieves better results and its accuracy of 98.45% and 98.56% for arousal and valence. The effectiveness of the proposed method is compared with the existing methods such as Linear Formulation of Differential Entropy (LF- <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">DfE</tex> ) and Generative Adversarial Network (GAN).