Abstract
Overstress at work is a serious issue that affects both workers and companies. Stress impairs employees' well-being, reduces productivity, and raises absenteeism. Consequently, there may be an increase in organizational costs associated with subpar work output and sick leave. Additionally, high blood pressure brought on by stress and other factors may be harmful to heart health. A stress detector uses appropriate sensors, such as an electrocardiogram (ECG) or a galvanic skin response (GSR), to gather physiological signals to distinguish between a stressed and a normal person. The desired features that represent the stress level in working people are extracted from these signals through pre-processing. To categorize this extracted feature set, support vector machine (SVM) and k-nearest neighbour (KNN) are examined. According to the outcome, the feature vector with the best features has a significant impact on stress detection.
TopMultimodal sensor efficacy to detect stress of working people is experimentally discussed in (Widanti, N., Sumanto, B., Rosa, P., Miftahudin, and M.F.). This employs the sensor data from sensors such as pressure distribution, HR, Blood Volume Pulse (BVP) and Electrodermal activity (EDA). An eye tracker sensor is also used which systematically analyses the eye movements with the stressors like Stroop word test and information related to pickup tasks. The authors (Nakashima et. al., 2015) performed perceived stress detection by a set of non-invasive sensors which collects the physiological signals such as ECG (Xu et al., 2015), GSR, Electroencephalography (EEG), EMG, and Saturation of peripheral oxygen (SpO2). Continuous stress levels are estimated using the physio-logical sensor data such as GSR, EMG, HR, and Respiration in (N. Renee Baptiste). The stress detection is carried out effectively using Skin conductance level (SCL), HR, Facial EMG sensors (Tanev et. al., 2014) by creating ICT related Stressors. Stress detection is discussed in various literatures as it is a significant societal contribution that enhances the lifestyle of individuals. Sioni et al. (Sioni et. al., 2015) analysed stress using Respiration, Heart rate (HR), facial electromyography (EMG), Galvanic skin response (GSR) foot and GSR hand data with a conclusion that, features pertaining to respi-ration process are substantial in stress detection. Emerging research has brought to light significant advancements in healthcare technology across diverse sectors, encompassing areas like cognitive rehabilitation, cardiovascular disease management, stress detection, and the application of IoT devices (Kaushik et al., 2023). Notably, the synergistic integration of AI tools with IoT devices has been instrumental in enabling cognitive cardiac rehabilitation, facilitating personalized healthcare via brain-computer interfacing within home automation systems, and creating sophisticated cardiovascular disease classifiers for secure platforms. Additionally, ongoing research has explored innovative stress detection techniques, especially beneficial for cognitive rehabilitation during the challenges posed by the COVID-19 pandemic. This includes leveraging EEG-based smart advisor bots for stress monitoring during activities like gaming. Moreover, the convergence of IoT and AI technologies has extended to applications beyond healthcare, such as the development of solar-powered agriculture robots and advanced emotion detection mechanisms using generative adversarial networks, collectively illustrating the dynamic landscape of technological progress shaping the future of healthcare delivery. Widanti et al. (Widanti et. al., 2015) proposed a research to predict stress levels solely from Electrocardiogram (ECG). Features of ECG are analysed using GRNN model to measure the stress level. Heart rate variability (HRV) features and RR interval features are used to classify the stress level (Schmidt et. al., 2018). It is noticed that Support Vector Machine (SVM) was used as the classification algorithm predomi-nantly due to its generalization ability and sound mathematical background Various kernels were used to develop models using SVM and it is concluded in that a linear SVM on both ECG frequency features and HRV features performed best, outperforming other model choices (Incel et. al., 2013).
Key Terms in this Chapter
Stress: Stress is the response we give when we sense pressure or danger. It typically occurs when we feel helpless or uncontrollable in a circumstance. Stress can manifest itself in several ways. A member of a group, such as when a member of your family is going through a trying period like a bereavement or financial difficulties. An individual, for example, when you are finding it difficult to manage your many duties. a constituent of your community, for instance, if you are a member of a discriminated-against religious group A member of the public, for instance, amid calamities like the coronavirus pandemic or natural disasters.
Machine Learning: A subfield of computer science and artificial intelligence called machine learning (ML) is concerned with using data and algorithms to help AI mimic human learning processes and progressively become more accurate.
SVM (Support Vector Machines): Strong machine learning algorithms like Support Vector Machine (SVM) are utilized for tasks including regression, outlier identification, and linear or nonlinear classification. Text classification, picture classification, handwriting recognition, spam detection, face detection, gene expression analysis, and anomaly detection are just a few of the many applications for SVMs. Because SVMs can handle high-dimensional data and nonlinear relationships, they are versatile and effective in a wide range of applications.
ECG: An ECG, often known as an EKG, is a fast test to monitor heartbeat. It captures the electrical impulses produced by the heart. Test findings used to determine arrhythmias, or abnormal heartbeats, and heart attacks. Hospitals, operating rooms, medical offices, and ambulances all use ECG machines. Simple ECGs performed on some personal electronics, like smartwatches.
HRV: The modest variation in the intervals between your heartbeats is known as heart rate variability. These variations can signal present or potential health concerns, like as cardiac ailments and mental health disorders like anxiety and depression, even if they are invisible without the use of specialist sensors.
GSR: The GSR sensor measures the different levels of the skin carrying the electric current. An increase in skin sweat causes an increase in electrical current conductance. Therefore, following an occurrence, a higher level of skin conductivity may indicate either good or negative emotional arousal.