Cross-Subject EEG-Based Emotion Recognition Through Neural Networks With Stratified Normalization

Fdez, Javier; Guttenberg, Nicholas; Witkowski, Olaf; Pasquali, Antoine

doi:10.3389/fnins.2021.626277

Cited by 43 publications

(13 citation statements)

References 36 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Similarly, the results show that the frontotemporal brain areas associated with the beta frequency show the greatest contribution to the performance of the models (see Figure 8). Finally, the model's performance proposed in this research reaches values comparable to other research (Fdez et al, 2021) to classify emotions into two categories (negative and positive).…”

Section: Discussionsupporting

confidence: 76%

See 1 more Smart Citation

Framework for the Classification of Emotions in People With Visual Disabilities Through Brain Signals

López-Hernández

González‐Carrasco

López-Cuadrado

et al. 2021

Front. Neuroinform.

View full text Add to dashboard Cite

Nowadays, the recognition of emotions in people with sensory disabilities still represents a challenge due to the difficulty of generalizing and modeling the set of brain signals. In recent years, the technology that has been used to study a person’s behavior and emotions based on brain signals is the brain–computer interface (BCI). Although previous works have already proposed the classification of emotions in people with sensory disabilities using machine learning techniques, a model of recognition of emotions in people with visual disabilities has not yet been evaluated. Consequently, in this work, the authors present a twofold framework focused on people with visual disabilities. Firstly, auditory stimuli have been used, and a component of acquisition and extraction of brain signals has been defined. Secondly, analysis techniques for the modeling of emotions have been developed, and machine learning models for the classification of emotions have been defined. Based on the results, the algorithm with the best performance in the validation is random forest (RF), with an accuracy of 85 and 88% in the classification for negative and positive emotions, respectively. According to the results, the framework is able to classify positive and negative emotions, but the experimentation performed also shows that the framework performance depends on the number of features in the dataset and the quality of the Electroencephalogram (EEG) signals is a determining factor.

show abstract

Section: Discussionsupporting

confidence: 76%

“…A new normalization method of features named stratified normalization is studied to classify emotions from EEG signals (Fdez et al, 2021). In this research, the SEED dataset is used, and the data on the effects of three independent variables (labeling method, normalization method, and feature extraction method) are recorded.…”

Section: Related Workmentioning

confidence: 99%

Framework for the Classification of Emotions in People With Visual Disabilities Through Brain Signals

López-Hernández

González‐Carrasco

López-Cuadrado

et al. 2021

Front. Neuroinform.

View full text Add to dashboard Cite

show abstract

“…The normalization lack leads to reduce the performance of the models. Min-Max is one of the well-known normalization methods which use in EEG based emotion detection [12]. It has a simple calculation mechanism that scales features between 0 and 1.…”

Section: Normalizationmentioning

confidence: 99%

“…As already mentioned above EEG signals reveal precise information more than audio signals and facial expressions in emotion detection. Most emotion recognition studies have been classified by using traditional ML approaches such as SVM [10], K nearest neighbors (KNN) [9], principle component analysis (PCA) [11], random forest (RF) [12] and linear discriminant analysis (LDA) [13]. However, these learning methods generally have presented low accuracy in contrast of deep learning (DL) methods.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

TheRobust EEG Based Emotion Recognition using Deep Neural Network

Khojasteh¹

2021

ijisae

View full text Add to dashboard Cite

This paper focuses on a novel Electroencephalography (EEG) based one dimensional convolution neural network (CNN) to classify emotional states. Differential entropy (DE) is considered as a feature extraction method after pre-processing phase. Besides, feature smoothing-linear dynamic system (LDS) and min-max normalization are used on the DE features before feeding into deep model. We design a one dimensional CNN model with six convolutions and fully connected blocks which gives outstanding performance in six combinations of SEED dataset. The model presented average accuracy of 98.55% and 95.91% in binary and single sessions respectively by using 10 fold cross validation. The proposed results fully demonstrate that our method achieves out of the best performance compare with other EEG based emotion recognition systems. Therefore, this model can be applied to other emotional datasets as a classifier and health care decision support system (DSS) as well.

show abstract

Domain Adaptation for Fear of Heights Classification in a VR Environment Based on EEG and ECG

Apicella,

Arpaia,

Barbato

et al. 2024

Inf Syst Front

View full text Add to dashboard Cite

Three levels of fear of heights were detected in subjects with different severities of acrophobia, based on the electroencephalographic (EEG) and electrocardiographic (ECG) signals. The study aims to demonstrate the feasibility of a data-fusion-based method for real-time assessment of the fear of heights intensity to integrate into adaptive Virtual Reality Exposure Therapy for acrophobia. The generalization performance of classification tasks on fear states is improved by exploiting both trait-based clustering and Domain Adaptation methods. Participants were gradually exposed to increasing height levels through a Virtual Reality (VR) scenario representing a canyon. The initial severity of fear of heights, the level of distress at each height, and the anxiety level before and after the exposure were assessed through the Acrophobia Questionnaire, the Subjective Unit of Distress, and the State and Trait Anxiety Inventory, respectively. The Simulator Sickness Questionnaire was administered to exclude possible motion sickness interference in the experiment. The EEG and ECG signals were acquired through a 32-channel headset and 1 Lead ECG derivation during the exposure to the eliciting VR scenario. Four classifiers (i.e. Support Vector Machines, Deep Neural Networks, Random Forests, and k-Nearest Neighbors) were adopted in the experimental environment. Preliminary tests were performed in a within-subject experiment, achieving the best classification accuracy of $$87.1 \% \pm 7.8 \%$$ 87.1 % ± 7.8 % with a Deep Neural Network. As the cross-subject approach is concerned, three strategies, namely Domain Adaptation (DA), data fusion (combining EEG with ECG), and participant clustering (based on the acrophobia severity), were evaluated. DA resulted in the most effective strategies by determining an improvement of more than 20 % in classification accuracy. Random Forest performed the best classification accuracy for the severe acrophobia cluster with a mean of $$63.6 \%$$ 63.6 % and a standard deviation of $$ 13.4 \%$$ 13.4 % over three classes by exploiting Stratified Normalization.

show abstract

Cross-Subject EEG-Based Emotion Recognition Through Neural Networks With Stratified Normalization

Cited by 43 publications

References 36 publications

Framework for the Classification of Emotions in People With Visual Disabilities Through Brain Signals

Framework for the Classification of Emotions in People With Visual Disabilities Through Brain Signals

TheRobust EEG Based Emotion Recognition using Deep Neural Network

Domain Adaptation for Fear of Heights Classification in a VR Environment Based on EEG and ECG

Contact Info

Product

Resources

About