A blockchain-based secure Internet of medical things framework for stress detection

Information Sciences, Journal Year: 2023, Volume and Issue: 628, P. 377 - 390

Published: Feb. 1, 2023

Language: Английский

---

AI-powered biometrics for Internet of Things security: A review and future vision

Journal of Information Security and Applications, Journal Year: 2024, Volume and Issue: 82, P. 103748 - 103748

Published: March 21, 2024

Language: Английский

---

Classification of Mental Stress from Wearable Physiological Sensors Using Image-Encoding-Based Deep Neural Network

Biosensors, Journal Year: 2022, Volume and Issue: 12(12), P. 1153 - 1153

Published: Dec. 9, 2022

The human body is designed to experience stress and react it, experiencing challenges causes our produce physical mental responses also helps adjust new situations. However, becomes a problem when it continues remain without period of relaxation or relief. When person has long-term stress, continued activation the response wear tear on body. Chronic results in cancer, cardiovascular disease, depression, diabetes, thus deeply detrimental health. Previous researchers have performed lot work regarding using mainly machine-learning-based approaches. most methods used raw, unprocessed data, which cause more errors thereby affect overall model performance. Moreover, corrupt data values are very common, especially for wearable sensor datasets, may lead poor performance this regard. This paper introduces deep-learning-based method detection by encoding time series raw into Gramian Angular Field images, promising accuracy while detecting levels an individual. experiment been conducted two standard benchmark namely WESAD (wearable detection) SWELL. During studies, testing accuracies 94.8% 99.39% achieved SWELL respectively. For dataset, chest taken experiment, including modalities such as three-axis acceleration (ACC), electrocardiogram (ECG), temperature (TEMP), respiration (RESP), etc.

Language: Английский

---

A Comparison of Personalized and Generalized Approaches to Emotion Recognition Using Consumer Wearable Devices: Machine Learning Study

JMIR AI, Journal Year: 2024, Volume and Issue: 3, P. e52171 - e52171

Published: March 23, 2024

Background There are a wide range of potential adverse health effects, ranging from headaches to cardiovascular disease, associated with long-term negative emotions and chronic stress. Because many indicators stress imperceptible observers, the early detection remains pressing medical need, as it can enable intervention. Physiological signals offer noninvasive method for monitoring affective states recorded by growing number commercially available wearables. Objective We aim study differences between personalized generalized machine learning models 3-class emotion classification (neutral, stress, amusement) using wearable biosignal data. Methods developed neural network problem data Wearable Stress Affect Detection (WESAD) set, multimodal set physiological 15 participants. compared results participant-exclusive generalized, participant-inclusive deep model. Results For problem, our model achieved an average accuracy 95.06% F1-score 91.71%; 66.95% 42.50%; 67.65% 43.05%. Conclusions Our emphasize need increased research in recognition given that they outperform certain contexts. also demonstrate viable achieve high performance.

Language: Английский

---

Multi-Class Stress Detection Through Heart Rate Variability: A Deep Neural Network Based Study

IEEE Access, Journal Year: 2023, Volume and Issue: 11, P. 57470 - 57480

Published: Jan. 1, 2023

Stress is a natural human reaction to demands or pressure, usually when perceived as harmful or/and toxic. When stress becomes constantly overwhelmed and prolonged, it increases the risk of mental health physiological uneasiness. Furthermore, chronic raises likelihood plagues such anxiety, depression, sleep disorder. Although measuring using parameters heart rate variability (HRV) common approach, how achieve ultra-high accuracy based on HRV measurements remains challenging task. not equivalent rate. While average value heartbeats per minute, represents variation time interval between successive heartbeats. The are related variance RR intervals which stand for R peaks. In this study, we investigate role features detection bio-markers develop machine learning-based model multi-class detection. More specifically, convolution neural network (CNN) developed detect stress, namely, no interruption pressure stress , both time- frequency-domain HRV. Validated through publicly available dataset, SWELL–KW, achieved score our has reached 99.9% ( xmlns:xlink="http://www.w3.org/1999/xlink">Precision = 1 xmlns:xlink="http://www.w3.org/1999/xlink">Recall $F1-$ xmlns:xlink="http://www.w3.org/1999/xlink">score xmlns:xlink="http://www.w3.org/1999/xlink">MCC 0.99 ), thus outperforming existing methods in literature. addition, study demonstrates effectiveness essential feature extraction technique, i.e., analysis variance.

Language: Английский

---

Detection of Stress from PPG and GSR Signals using AI Framework

Journal of The Institution of Engineers (India) Series B, Journal Year: 2025, Volume and Issue: unknown

Published: Jan. 3, 2025

Language: Английский

---

From Lab to Real-Life: A Three-Stage Validation of Wearable Technology for Stress Monitoring

MethodsX, Journal Year: 2025, Volume and Issue: 14, P. 103205 - 103205

Published: Feb. 5, 2025

Stress negatively impacts health, contributing to hypertension, cardiovascular diseases, and immune dysfunction. While conventional diagnostic methods, such as self-reported questionnaires basic physiological measurements, often lack the objectivity precision needed for effective stress management, wearable devices present a promising avenue early detection management. This study conducts three-stage validation of technology monitoring, transitioning from controlled experimental data real-life scenarios. Using WESAD dataset, binary five-class classification models were developed, achieving maximum accuracies 99.78 %±0.15 % 99.61 %±0.32 %, respectively. Electrocardiogram (ECG), Electrodermal Activity (EDA), Respiration (RESP) identified reliable biomarkers. Validation was extended SWEET representing data, confirm generalizability practical applicability. Furthermore, commercially available wearables supporting these modalities reviewed, providing recommendations optimal configurations in dynamic, real-world conditions. These findings demonstrate potential multimodal bridge gap between studies applications, advancing systems personalized management strategies.•Stress methods validated using (WESAD) (SWEET) datasets.•Commercial technologies offering insights into their applicability monitoring.

Language: Английский

---

Stress management with HRV following AI, semantic ontology, genetic algorithm and tree explainer

Scientific Reports, Journal Year: 2025, Volume and Issue: 15(1)

Published: Feb. 17, 2025

Heart Rate Variability (HRV) serves as a vital marker of stress levels, with lower HRV indicating higher stress. It measures the variation in time between heartbeats and offers insights into health. Artificial intelligence (AI) research aims to use data for accurate level classification, aiding early detection well-being approaches. This study's objective is create semantic model features knowledge graph develop an accurate, reliable, explainable, ethical AI predictive analysis. The SWELL-KW dataset, containing labeled conditions, examined. Various techniques like feature selection dimensionality reduction are explored improve classification accuracy while minimizing bias. Different machine learning (ML) algorithms, including traditional ensemble methods, employed analyzing both imbalanced balanced datasets. To address imbalances, various formats oversampling such SMOTE ADASYN experimented with. Additionally, Tree-Explainer, specifically SHAP, used interpret explain models' classifications. combination genetic algorithm-based using Random Forest Classifier yields effective results datasets, especially non-linear features. These optimized play crucial role developing management system within Semantic framework. Introducing domain ontology enhances representation acquisition. consistency reliability Ontology assessed Hermit reasoners, reasoning performance measure. significant indicator stress, offering its correlation mental well-being. While non-invasive, interpretation must integrate other assessments holistic understanding individual's response. Monitoring can help evaluate strategies interventions, individuals maintaining

Language: Английский

---

Deep Learning in Biometric Authentication: Challenges, Recent Advancements, and Future Trends

Journal of Advances in Information Technology, Journal Year: 2025, Volume and Issue: 16(4), P. 458 - 477

Published: Jan. 1, 2025

Language: Английский

---

A machine-learning Approach for Stress Detection Using Wearable Sensors in Free-living Environments

medRxiv (Cold Spring Harbor Laboratory), Journal Year: 2024, Volume and Issue: unknown

Published: April 30, 2024

Abstract Stress is a psychological condition due to the body’s response challenging situation. If person exposed prolonged periods and various forms of stress, their physical mental health can be negatively affected, leading chronic problems. It important detect stress in its initial stages prevent stress-related issues. Thus, there must alternative effective solutions for spontaneous monitoring. Wearable sensors are one most prominent solutions, given capacity collect data continuously real-time. sensors, among others, have been widely used bridge existing gaps monitoring thanks non-intrusive nature. Besides, they monitor vital signs, e.g., heart rate activity. Yet, works focused on acquired controlled settings. To this end, our study aims propose machine learning-based approach detecting onsets free-living environment using wearable sensors. The authors utilized SWEET dataset collected from 240 subjects via electrocardiography (ECG), skin temperature (ST), conductance (SC). In work, four learning models were tested set consisting subjects, namely K-Nearest Neighbors (KNN), Support vector classification (SVC), Decision Tree (DT), Random Forest (RF). These trained scenarios. Neighbor (KNN) model had highest accuracy 98%, while other also performed satisfactorily.

Language: Английский

---