{"title":"Deep Learning Based Classification and Combined Transform Based Feature Extraction Approach for Mental Stress Prediction of Human Beings Using EEG","authors":"Shashibala Agarwal, Maria Jamal, Parmod Kumar","doi":"10.1002/ett.70155","DOIUrl":null,"url":null,"abstract":"<div>\n \n <p>Stress is a psychological condition in which a person feels overwhelmed with pressure. Early identification of psychological stress is critical for preventing illness progression and saving lives. Electroencephalography (EEG) is often used to collect psychological information such as brain rhythms in the form of electric waves. Traditional deep learning techniques face limitations like temporal dynamics and feature extraction issues. To address these shortcomings, a deep learning-based classification model was created, combining advanced transform-based feature extraction techniques to more effectively predict mental stress by using EEG signals. The process begins by utilizing physiological parameters extracted from the EEG Psychiatric Disorders Dataset. The raw EEG signals undergo pre-processing to enhance their quality, which includes smoothing, alignment, and addressing non-uniform sampling issues. The signals are then decomposed and their components extracted using the Adaptive Flexible Analytic Wavelet Transform (AFAWT). Short-Term Fourier Transform-Randon Transform (STFT-RT) approach is used to extract the key features of signals. Feature selection is optimized using the Young's Double Slit Experiment Optimizer (YDSE) to ensure only the most relevant features are chosen for classification. Finally, these selected features are fed into the Parallel Neural Networks with Extreme Efficiency (ParNeXt v1-DB) model, which utilizes a drop block mechanism to enhance model generalization and prevent overfitting, ensuring highly effective mental stress prediction. According to simulated research, the proposed approach demonstrated significant improvements over existing algorithms. In Dataset 1, the method achieved an accuracy of 97.8%, selectivity of 95.4%, while Dataset 2 recorded an accuracy of 96.3%, PPV of 93.8%. Thus, the proposed method is the most effective method for predicting human mental stress using EEG.</p>\n </div>","PeriodicalId":23282,"journal":{"name":"Transactions on Emerging Telecommunications Technologies","volume":"36 6","pages":""},"PeriodicalIF":2.5000,"publicationDate":"2025-05-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Transactions on Emerging Telecommunications Technologies","FirstCategoryId":"94","ListUrlMain":"https://onlinelibrary.wiley.com/doi/10.1002/ett.70155","RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q3","JCRName":"TELECOMMUNICATIONS","Score":null,"Total":0}
引用次数: 0
Abstract
Stress is a psychological condition in which a person feels overwhelmed with pressure. Early identification of psychological stress is critical for preventing illness progression and saving lives. Electroencephalography (EEG) is often used to collect psychological information such as brain rhythms in the form of electric waves. Traditional deep learning techniques face limitations like temporal dynamics and feature extraction issues. To address these shortcomings, a deep learning-based classification model was created, combining advanced transform-based feature extraction techniques to more effectively predict mental stress by using EEG signals. The process begins by utilizing physiological parameters extracted from the EEG Psychiatric Disorders Dataset. The raw EEG signals undergo pre-processing to enhance their quality, which includes smoothing, alignment, and addressing non-uniform sampling issues. The signals are then decomposed and their components extracted using the Adaptive Flexible Analytic Wavelet Transform (AFAWT). Short-Term Fourier Transform-Randon Transform (STFT-RT) approach is used to extract the key features of signals. Feature selection is optimized using the Young's Double Slit Experiment Optimizer (YDSE) to ensure only the most relevant features are chosen for classification. Finally, these selected features are fed into the Parallel Neural Networks with Extreme Efficiency (ParNeXt v1-DB) model, which utilizes a drop block mechanism to enhance model generalization and prevent overfitting, ensuring highly effective mental stress prediction. According to simulated research, the proposed approach demonstrated significant improvements over existing algorithms. In Dataset 1, the method achieved an accuracy of 97.8%, selectivity of 95.4%, while Dataset 2 recorded an accuracy of 96.3%, PPV of 93.8%. Thus, the proposed method is the most effective method for predicting human mental stress using EEG.
期刊介绍:
ransactions on Emerging Telecommunications Technologies (ETT), formerly known as European Transactions on Telecommunications (ETT), has the following aims:
- to attract cutting-edge publications from leading researchers and research groups around the world
- to become a highly cited source of timely research findings in emerging fields of telecommunications
- to limit revision and publication cycles to a few months and thus significantly increase attractiveness to publish
- to become the leading journal for publishing the latest developments in telecommunications