Classification of Human Emotions from Electroencephalogram (EEG) Signal using Deep Neural Network

Al-Nafjan, Abeer; Hosny, Manar; Al-Wabil, Areej; Al-Ohali, Yousef

doi:10.14569/ijacsa.2017.080955

Cited by 85 publications

(68 citation statements)

References 22 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The comparison of our model with previous studies. research method average accuracy cross validation valence arousal Li et al [16] C-RNN 72.06% 74.12% 5-fold Al-Nafjan et al [17] PSD+DNN 82.00% 82.00% 10-fold Liu et al [18] Multimodal Deep Learning 85.20% 80.50% 10-fold our model ResNet+LFCC+KNN 90.39% 89.06% 10-fold low valence/high arousal low valence/low arousal high valence/high arousal high valence), the best accuracy of the proposed approach is 90.21%. And the performance with different classifiers is shown in Fig.…”

Section: Results For 10-fold Cross Validationmentioning

confidence: 99%

Multiple Feature Fusion for Automatic Emotion Recognition Using EEG Signals

Liu

Fang

et al. 2018

2018 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

View full text Add to dashboard Cite

show abstract

Section: Results For 10-fold Cross Validationmentioning

confidence: 99%

Multiple Feature Fusion for Automatic Emotion Recognition Using EEG Signals

Liu

Fang

et al. 2018

2018 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

View full text Add to dashboard Cite

show abstract

“…This is because, technically, including the channels in the model, which are not correlated with the emotion changes, does not help and on the contrary can adversely affect the accuracy. It is also known that the electrical relations between asymmetrical channels are determining the arousal and valence, hence the emotion [73,74]. Therefore, we chose four asymmetrical pairs of electrodes: AF1, F3, F4, F7, T7, AF2, F5, F8 and T8 from frontal and temporal lobes which are equally spread on the skull.…”

Section: Channel Selectionmentioning

confidence: 99%

Investigating the Use of Pretrained Convolutional Neural Network on Cross-Subject and Cross-Dataset EEG Emotion Recognition

Çimtay

Ekmekcioǧlu

2020

Sensors

196

108

View full text Add to dashboard Cite

The electroencephalogram (EEG) has great attraction in emotion recognition studies due to its resistance to deceptive actions of humans. This is one of the most significant advantages of brain signals in comparison to visual or speech signals in the emotion recognition context. A major challenge in EEG-based emotion recognition is that EEG recordings exhibit varying distributions for different people as well as for the same person at different time instances. This nonstationary nature of EEG limits the accuracy of it when subject independency is the priority. The aim of this study is to increase the subject-independent recognition accuracy by exploiting pretrained state-of-the-art Convolutional Neural Network (CNN) architectures. Unlike similar studies that extract spectral band power features from the EEG readings, raw EEG data is used in our study after applying windowing, pre-adjustments and normalization. Removing manual feature extraction from the training system overcomes the risk of eliminating hidden features in the raw data and helps leverage the deep neural network’s power in uncovering unknown features. To improve the classification accuracy further, a median filter is used to eliminate the false detections along a prediction interval of emotions. This method yields a mean cross-subject accuracy of 86.56% and 78.34% on the Shanghai Jiao Tong University Emotion EEG Dataset (SEED) for two and three emotion classes, respectively. It also yields a mean cross-subject accuracy of 72.81% on the Database for Emotion Analysis using Physiological Signals (DEAP) and 81.8% on the Loughborough University Multimodal Emotion Dataset (LUMED) for two emotion classes. Furthermore, the recognition model that has been trained using the SEED dataset was tested with the DEAP dataset, which yields a mean prediction accuracy of 58.1% across all subjects and emotion classes. Results show that in terms of classification accuracy, the proposed approach is superior to, or on par with, the reference subject-independent EEG emotion recognition studies identified in literature and has limited complexity due to the elimination of the need for feature extraction.

show abstract

“…We outline two different EEG preprocessing approaches (Section 2.1) and, in this context, we evaluate (Section 3) the discriminative capacity of various EEG features (Section 2.2), which were reported successful in previous related studies [23][24][25][28][29][30][31][32][33][34]. These EEG features are based on the following: Next, in Section 2.3 we outline the DEAP database, and in Section 2.4 the common experimental protocol used in all experiments.…”

Section: Methodsmentioning

confidence: 99%

“…In studies that use PSD as an EEG feature, traditionally all five-alpha, beta, gamma, delta and theta-frequency bands are considered [8,23,24]. In some cases, low frequency bands were omitted [25] or only specific bands (such as alpha and beta) were used [26,27]. The typical way of calculation for PSD-based features is through a short-time Discrete Fourier Transform (stDFT) (in fact, Fast Fourier Transform (FFT)) applied on non-overlapping frames of the segmented EEG signal [24,26].…”

Section: Introductionmentioning

confidence: 99%

Evaluation of Features in Detection of Dislike Responses to Audio–Visual Stimuli from EEG Signals

2020

View full text Add to dashboard Cite

There is a strong correlation between the like/dislike responses to audio–visual stimuli and the emotional arousal and valence reactions of a person. In the present work, our attention is focused on the automated detection of dislike responses based on EEG activity when music videos are used as audio–visual stimuli. Specifically, we investigate the discriminative capacity of the Logarithmic Energy (LogE), Linear Frequency Cepstral Coefficients (LFCC), Power Spectral Density (PSD) and Discrete Wavelet Transform (DWT)-based EEG features, computed with and without segmentation of the EEG signal, on the dislike detection task. We carried out a comparative evaluation with eighteen modifications of the above-mentioned EEG features that cover different frequency bands and use different energy decomposition methods and spectral resolutions. For that purpose, we made use of Naïve Bayes classifier (NB), Classification and regression trees (CART), k-Nearest Neighbors (kNN) classifier, and support vector machines (SVM) classifier with a radial basis function (RBF) kernel trained with the Sequential Minimal Optimization (SMO) method. The experimental evaluation was performed on the well-known and widely used DEAP dataset. A classification accuracy of up to 98.6% was observed for the best performing combination of pre-processing, EEG features and classifier. These results support that the automated detection of like/dislike reactions based on EEG activity is feasible in a personalized setup. This opens opportunities for the incorporation of such functionality in entertainment, healthcare and security applications.

show abstract

Classification of Human Emotions from Electroencephalogram (EEG) Signal using Deep Neural Network

Cited by 85 publications

References 22 publications

Multiple Feature Fusion for Automatic Emotion Recognition Using EEG Signals

Multiple Feature Fusion for Automatic Emotion Recognition Using EEG Signals

Investigating the Use of Pretrained Convolutional Neural Network on Cross-Subject and Cross-Dataset EEG Emotion Recognition

Evaluation of Features in Detection of Dislike Responses to Audio–Visual Stimuli from EEG Signals

Contact Info

Product

Resources

About