StressNet: Hybrid model of LSTM and CNN for stress detection from electroencephalogram signal (EEG)

Everyday tasks can cause stress, which can lead to serious medical conditions, such as depression. EEG signal processing can assist medical professionals in managing the emotional state of patients. To detect stress states in EEG signals, we propose a new architecture, StressNet, which is a combinat...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Results in control and optimization Jg. 11; S. 100231
Hauptverfasser: Mane, Swaymprabha Alias Megha, Shinde, Arundhati
Format: Journal Article
Sprache:Englisch
Veröffentlicht: Elsevier B.V 01.06.2023
Elsevier
Schlagworte:
Accuracy
Azimuthal projection
Brain–computer interfaces (BCIs)
Convolutional neural network (CNN)
Deep learning
Electroencephalography (EEG)
Human stress detection
Long short-term memory (LSTM)
Stress
Deep learning
Human stress detection
Accuracy
Electroencephalography (EEG)
Brain–computer interfaces (BCIs)
Convolutional neural network (CNN)
Long short-term memory (LSTM)
Stress
Azimuthal projection
ISSN:2666-7207, 2666-7207
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:Everyday tasks can cause stress, which can lead to serious medical conditions, such as depression. EEG signal processing can assist medical professionals in managing the emotional state of patients. To detect stress states in EEG signals, we propose a new architecture, StressNet, which is a combination of a two-dimensional convolutional neural network (CNN) and a long short-term memory (LSTM) network. The EEG signals are first decomposed into alpha, beta, and theta signals, which are then used to generate azimuthal projection-based images. These images are fed into the 2D CNN for feature extraction, which are then passed to the LSTM for further processing. The LSTM’s ability to remember past states makes it particularly effective in modeling the temporal dynamics of EEG signals. Finally, the StressNet model classifies the features using fully connected layers into either stress or normal classes. We evaluate the performance of our model on the DEEP and SEED datasets and compare it to other methods. The results show that the proposed StressNet model outperforms the human stress detection accuracy, achieving an accuracy of 97.8%.
ISSN:2666-7207
2666-7207
DOI:10.1016/j.rico.2023.100231