Citation

Isa SM. ICIC Express Letters 2022; 16(11): 1157-1167.

Copyright

(Copyright © 2022)

DOI

10.24507/icicel.16.11.1157

PMID

unavailable

Abstract

Electroencephalogram (EEG) signals have an important role in identifying real human feelings or emotions to develop a Brain-Computer Interface (BCI) system. Emotion recognition systems can be applied to medical care and entertainment to overcome mental illness to relieve negative emotions to reduce suicide attempts. This study proposes a deep learning-based EEG human emotion classification system and applies different feature extraction methods. In the proposed system, 2-second segments of EEG are decomposed using Discrete Wavelet Transform (DWT) and Fast Fourier Transform (FFT) to obtain several sub-signals and extract their features. A Convolutional Neural Network (CNN) and Deep Neural Network (DNN) are trained and validated using these features to classify two different levels of arousal and valence. The system was evaluated and achieved the best performance for arousal and valence with 96.84% and 97.18% average accuracy, respectively. The results show that the use of data augmentation and feature extraction methods plays an important role in deep learning-based EEG human emotion classification systems and provides excellent performance. In addition, determining the number of channels can affect the performance of the classification system. ICIC International © 2022.

Language: en

Keywords

Affective state; CNN; DEAP dataset; DNN; DWT; EEG; Emotion recognition; FFT