EEG-Based Emotion Recognition with Combined Deep Neural Networks using Decomposed Feature Clustering Model

• Published in: 2019 13th International Conference on Open Source Systems and Technologies (ICOSST) pp. 1 - 6
• Main Authors: Asghar, Muhammad Adeel, Fawad, Khan, Muhammad Jamil, Amin, Yasar, Akram, Adeel
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.12.2019
• Subjects: Accuracy; Brain modeling; Computational efficiency; Computational modeling; Convolutional neural networks; DNN; Electroencephalography; Emotion recognition; Feature Decomposition; Feature extraction; Machine learning; Spectrogram
• DOI: 10.1109/ICOSST48232.2019.9043994

Much attention has been paid to the recognition of human emotions with the help of EEG signals based on machine learning technology. Recognizing emotions is a difficult task due to the non-linear nature of the EEG signal. This paper presents an advanced signal processing method that uses the depth function to extract features from all channels related to emotion. A decomposed feature clustering model is presented in this paper to decrease the computational cost of recognizing emotions and achieve better results. In the proposed method, we convert the signal into a two-dimensional wavelet spectrogram and calculate the characteristics of each subject. An EEG-based emotion classification model using a deep convolutional neural network (DNN) is presented on the SJTU SEED dataset. Combined feature model using AlexNet, VGGNet and ResNet-50 machine learning models are used for feature extraction. SVM and k-NN are used to classify data into positive/negative/neutral dimensions for SEED dataset. The results showed that models with images are more accurate than traditional models for emotion recognition. The proposed model achieves 91.3% accuracy in the SEED dataset, which is more accurate as compared to the other state-of-the-art human emotions recognition methods.