A Convolutional Neural Network Feature Fusion Framework with Ensemble Learning for EEG-based Emotion Classification

Guo, Kailing; Han, Moonsik; Xie, Xiaona; Xu, Xiangmin

doi:10.1109/imbioc.2019.8777738

Cited by 8 publications

(5 citation statements)

References 9 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Different ordering of electrodes was investigated in the study [7] for emotional EEG classification with three connectivity features, PCC, PLV and TE; the obtained result showed that data-driven channel ordering is better than random ordering. Feature fusion approach was used by Guo et al with connectivity feature PCC and synchronization likelihood (SL) where ensemble of CNN was used for classification and higher ER accuracy obtained in Valence dimension than Arousal dimension [83]. Wang et al used PSI with CNN and showed that connectivity feature shows superior performance, compared with the input of raw EEG data [47].…”

Section: B Er Using Connectivity Featurementioning

confidence: 99%

A Comparative Study on Prominent Connectivity Features for Emotion Recognition From EEG

et al. 2023

View full text Add to dashboard Cite

Classifying distinct human emotions, the fundamental purpose of brain-computer interface research, is essential for providing instant personalized services and assistance to individuals. With such emerging applications for individuals, several techniques have been proposed recently to explore interactions between brain regions, such as correlation, synchronization, and dependence. Notably, functional and effective connectivity methods are applied to assess the relationships between different brain areas. The primary objective of this study is to compare the frequently used functional and effective connectivity methods to recognize emotion using Electroencephalogram (EEG) signals. This paper uses a benchmark emotional EEG dataset consisting of 32 channels of EEG signals collected from 32 subjects while they were watching 40 inspirational music videos. Specifically, correlation, phase synchronization, and mutual information are used to measure functional brain connectivity, and transfer entropy is used to acquire effective brain connectivity. After extracting the features, they are represented in a two-dimensional connectivity feature map (CFM). The CFMs are then used to classify emotions by a convolutional neural network model. The results of classified emotions are analyzed regarding compatible EEG bands, accuracy, and time. Notably, the Gamma band is found as the most compatible band. The comparative study has demonstrated that though the connectivity method named Pearson correlation coefficient requires less time, the normalized mutual information is the most accurate method with advantageous detecting capability of nonlinear dependencies.

show abstract

Section: B Er Using Connectivity Featurementioning

confidence: 99%

A Comparative Study on Prominent Connectivity Features for Emotion Recognition From EEG

et al. 2023

View full text Add to dashboard Cite

show abstract

“…The fusion of spatial-temporal features ensures that more EEG features are extracted, so many researchers focus on the spatial-temporal fusion strategies. Guo et al [15] fed the correlation coefficient matrix and synchronous likelihood matrix to the feature fusion framework of the emotional network and the inception network is adopted to fuse the latent features. Liu et al [16] proposed a 3D convolution attention neural network composed of spatial-temporal feature extraction module and channel attention weight learning module, and the internal spatial correlations of multi-channel EEG signals during continuous period time are extracted.…”

Section: Multiple Domain Feature Learningmentioning

confidence: 99%

“…The multi-layer perceptron (MLP) is applied to learn and extract temporal features, and spatial-temporal feature fusion by Bi-LSTM. Although the spatial-temporal fusion strategies [15]- [17] adopt abundant information to improve the performance, it is inadequate learning for temporal contexts which is also discriminative to the emotional states.…”

Section: Multiple Domain Feature Learningmentioning

confidence: 99%

“…For each 6S-long EEG segment, PSD features are extracted in five frequency bands: Delta-band (1-4 Hz), theta-band (4-8 Hz), alpha-band (8-12 Hz), beta-band (12-20 Hz), and gamma-band . The Cartesian coordinates of 32-channel are obtained by the EEGLAB toolbox [15] (As shown in Fig. 2), so that the power topography of the corresponding frequency bands can be obtained from the generated Cartesian coordinates.…”

Section: Eeg Preprocessing and Feature Extractionmentioning

confidence: 99%

See 1 more Smart Citation

Spatial-Temporal Feature Fusion Neural Network for EEG-Based Emotion Recognition

Wang

Zhang

et al. 2022

IEEE Trans. Instrum. Meas.

View full text Add to dashboard Cite

The spatial correlations and the temporal contexts are indispensable in Electroencephalogram (EEG)-based emotion recognition. However, the learning of complex spatial correlations among several channels is a challenging problem. Besides, the temporal contexts learning is beneficial to emphasize the critical EEG frames because the subjects only reach the prospective emotion during part of stimuli. Hence, we propose a novel Spatial-Temporal Information Learning Network (STILN) to extract the discriminative features by capturing the spatial correlations and temporal contexts. Specifically, the generated 2D power topographic maps capture the dependencies among electrodes, and they are fed to the CNN-based spatial feature extraction network. Furthermore, Convolutional Block Attention Module (CBAM) recalibrates the weights of power topographic maps to emphasize the crucial brain regions and frequency bands. Meanwhile, Batch Normalizations (BNs) and Instance Normalizations (INs) are appropriately combined to relieve the individual differences. In the temporal contexts learning, we adopt the Bidirectional Long Short-Term Memory Network (Bi-LSTM) network to capture the dependencies among the EEG frames. To validate the effectiveness of the proposed method, subject-independent experiments are conducted on the public DEAP dataset. The proposed method has achieved the outstanding performance, and the accuracies of arousal and valence classification have reached 0.6831 and 0.6752 respectively.

show abstract

“…Severe stenosis can trigger cardiac arrest (hypoxia) (Padmanabhan & Semmlow, 1994;Nikdelfaz & Jalili, 2018). Plaques come in three types: calcified, mixed, and non-calcified, with varying symptoms depending on type and extent (Guo, et al, 2019;Xiao, et al, 2020).…”

Section: Introductionmentioning

confidence: 99%

Enhancing Heart Disease Detection Using Political Deer Hunting Optimization-Based Deep Q-Network with High Accuracy and Sensitivity

Aher

2023

MEDIN

View full text Add to dashboard Cite

Heart disease is a clinical syndrome generally caused due to the impairment of structural and functional tissues. According to World Health Organization (WHO), heart disease has been recognized as the major reason for death occurring worldwide and the survey reported that 23% of death in the US is mainly because of heart-related diseases. Therefore, quick and precise detection of heart-related diseases is significant to enhance the health issues of heart disease-affected people and also reduces the mortality rate of the patients by protecting them from dangerous death by offering better medical services. Nowadays, deep learning techniques play a significant part in the domain of medical science and are also utilized to diagnose various diseases. Due to poor feature representation and some imbalance problems, predicting heart disease is very difficult using simple models. To encounter such problems, an effective strategy is employed in this research for detecting heart disease using the proposed Political Deer Hunting Optimization (PDHO) algorithm-based Deep Q-Network. However, the proposed PDHO algorithm is derived by the integration of the Political Optimizer (PO) and Deer Hunting Optimization (DHO) algorithms. Moreover, the proposed PDHO-based Deep Q-network for heart disease detection achieved a maximum accuracy of 93.4%, maximum sensitivity of 96.2%, and maximum specificity of 89.2%. This study leverages a diverse and extensive dataset from the Gene Expression Omnibus dataset [GSE98583] to advance our understanding of coronary artery disease.

show abstract

A Convolutional Neural Network Feature Fusion Framework with Ensemble Learning for EEG-based Emotion Classification

Cited by 8 publications

References 9 publications

A Comparative Study on Prominent Connectivity Features for Emotion Recognition From EEG

A Comparative Study on Prominent Connectivity Features for Emotion Recognition From EEG

Spatial-Temporal Feature Fusion Neural Network for EEG-Based Emotion Recognition

Enhancing Heart Disease Detection Using Political Deer Hunting Optimization-Based Deep Q-Network with High Accuracy and Sensitivity

Contact Info

Product

Resources

About