EEG-Based 3D Input Convolutional Neural Network for Stress Detection During Arithmetic Task

Recent studies show that more than three-quarters of adults have an overload of induced stress that directly and immediately impacts an individual's body feel, movement, and function. A multi-channel EEG signal-based 3D input convolution neural network (CNN) is proposed for detecting the change...

Full description

Saved in:
Bibliographic Details
Published inTENCON ... IEEE Region Ten Conference pp. 559 - 562
Main Authors Vijay, Rahul Kumar, Sharma, Aditi, Sharma, Khushi, Garg, Mahak, Hooda, Monika
Format Conference Proceeding
LanguageEnglish
Published IEEE 01.12.2024
Subjects
Arithmetic
Arithmetic task
Classification
Convolutional Neural Network
Convolutional neural networks
Electroencephalogram
Electroencephalography
Entropy
Human factors
IEEE Regions
Induced-Stress
Monitoring
Neural networks
Recording
Three-dimensional displays
Online AccessGet full text

Cover

More Information
Summary:Recent studies show that more than three-quarters of adults have an overload of induced stress that directly and immediately impacts an individual's body feel, movement, and function. A multi-channel EEG signal-based 3D input convolution neural network (CNN) is proposed for detecting the changes in mental stress during an activity. This CNN-based binary classifier detects the stress level during an arithmetic task and relaxation state according to the 32-channel EEG recordings. The experimental work is conducted on a publicly available dataset: SAM 40 (released in 2022) to monitor the induced stress during different tasks based on 40 subjects' EEG recordings with three trials per task. The 2D locations of 32 channels are mapped from 1D EEG recordings into interpolated 3D images with Differential Entropy (DE) in feature engineering. This 3D EEG image-driven CNN outperforms and can classify induced stress during an arithmetic task and relaxed state with an accuracy of 96.18%.
ISSN:2159-3450
DOI:10.1109/TENCON61640.2024.10903036