How Laboratory Experiments Can Be Exploited forMonitoring Stress in the Wild: A Bridge BetweenLaboratory and Daily Life

Relieving the burden of intensive labeling for stress monitoring in the wild by using semi-supervised learning

Stress, a natural process affecting individuals' wellbeing, has a profound impact on overall quality of life. Researchers from diverse fields employ various technologies and methodologies to investigate it and alleviate the negative effects of this phenomenon. Wearable devices, such as smart bands, capture physiological data, including heart rate variability, motions, and electrodermal activity, enabling stress level monitoring through machine learning models. However, labeling data for model accuracy assessment poses a significant challenge in stress-related research due to incomplete or inaccurate labels provided by individuals in their daily lives. To address this labeling predicament, our study proposes implementing Semi-Supervised Learning (SSL) models. Through comparisons with deep learning-based supervised models and clustering-based unsupervised models, we evaluate the performance of our SSL models. Our experiments show that our SSL models achieve 77% accuracy with a classifier trained on an augmented dataset prepared using the label propagation (LP) algorithm. Additionally, our deep autoencoder network achieves 76% accuracy. These results highlight the superiority of SSL models over unsupervised learning techniques and their comparable performance to supervised learning models, even with limited labeled data. By relieving the burden of labeling in daily life stress recognition, our study advances stress-related research, recognizing stress as a natural process rather than a disease. This facilitates the development of more efficient and accurate stress monitoring methods in the wild.

A New Algorithm for Estimating a Noiseless, Evenly Sampled, Heart Rate Modulating Signal

Heart rate variability (HRV) is commonly intended as the variation in the heart rate (HR), and it is evaluated in the time and frequency domains with various well-known methods. In the present paper, the heart rate is considered as a time domain signal, at first as an abstract model in which the HR is the instantaneous frequency of an otherwise periodic signal, such as with an electrocardiogram (ECG). In this model, the ECG is assumed to be a frequency modulated signal, or carrier signal, where HRV or HRV(t) is the time-domain signal which is frequency modulating the carrier ECG signal around its average frequency. Hence, an algorithm able to frequency demodulate the ECG signal to extract the signal HRV(t) is described, with possibly enough time resolution to analyse fast time-domain variations in the instantaneous HR. After exhaustive testing of the method on simulated frequency modulated sinusoidal signals, the new procedure is eventually applied on actual ECG tracings for preliminary nonclinical testing. The purpose of the work is to use this algorithm as a tool and a more reliable method for the assessment of heart rate before any further clinical or physiological analysis.

Generalizable machine learning for stress monitoring from wearable devices: A systematic literature review

INTRODUCTION

Wearable sensors have shown promise as a non-intrusive method for collecting biomarkers that may correlate with levels of elevated stress. Stressors cause a variety of biological responses, and these physiological reactions can be measured using biomarkers including Heart Rate Variability (HRV), Electrodermal Activity (EDA) and Heart Rate (HR) that represent the stress response from the Hypothalamic-Pituitary-Adrenal (HPA) axis, the Autonomic Nervous System (ANS), and the immune system. While Cortisol response magnitude remains the gold standard indicator for stress assessment [1], recent advances in wearable technologies have resulted in the availability of a number of consumer devices capable of recording HRV, EDA and HR sensor biomarkers, amongst other signals. At the same time, researchers have been applying machine learning techniques to the recorded biomarkers in order to build models that may be able to predict elevated levels of stress.

OBJECTIVE

The aim of this review is to provide an overview of machine learning techniques utilized in prior research with a specific focus on model generalization when using these public datasets as training data. We also shed light on the challenges and opportunities that machine learning-enabled stress monitoring and detection face.

METHODS

This study reviewed published works contributing and/or using public datasets designed for detecting stress and their associated machine learning methods. The electronic databases of Google Scholar, Crossref, DOAJ and PubMed were searched for relevant articles and a total of 33 articles were identified and included in the final analysis. The reviewed works were synthesized into three categories of publicly available stress datasets, machine learning techniques applied using those, and future research directions. For the machine learning studies reviewed, we provide an analysis of their approach to results validation and model generalization. The quality assessment of the included studies was conducted in accordance with the IJMEDI checklist [2].

RESULTS

A number of public datasets were identified that are labeled for stress detection. These datasets were most commonly produced from sensor biomarker data recorded using the Empatica E4 device, a well-studied, medical-grade wrist-worn wearable that provides sensor biomarkers most notable to correlate with elevated levels of stress. Most of the reviewed datasets contain less than twenty-four hours of data, and the varied experimental conditions and labeling methodologies potentially limit their ability to generalize for unseen data. In addition, we discuss that previous works show shortcomings in areas such as their labeling protocols, lack of statistical power, validity of stress biomarkers, and model generalization ability.

CONCLUSION

Health tracking and monitoring using wearable devices is growing in popularity, while the generalization of existing machine learning models still requires further study, and research in this area will continue to provide improvements as newer and more substantial datasets become available.

Driver Drowsiness Detection: A Machine Learning Approach on Skin Conductance

The majority of car accidents worldwide are caused by drowsy drivers. Therefore, it is important to be able to detect when a driver is starting to feel drowsy in order to warn them before a serious accident occurs. Sometimes, drivers are not aware of their own drowsiness, but changes in their body signals can indicate that they are getting tired. Previous studies have used large and intrusive sensor systems that can be worn by the driver or placed in the vehicle to collect information about the driver's physical status from a variety of signals that are either physiological or vehicle-related. This study focuses on the use of a single wrist device that is comfortable for the driver to wear and appropriate signal processing to detect drowsiness by analyzing only the physiological skin conductance (SC) signal. To determine whether the driver is drowsy, the study tests three ensemble algorithms and finds that the Boosting algorithm is the most effective in detecting drowsiness with an accuracy of 89.4%. The results of this study show that it is possible to identify when a driver is drowsy using only signals from the skin on the wrist, and this encourages further research to develop a real-time warning system for early detection of drowsiness.

Cross Dataset Analysis for Generalizability of HRV-Based Stress Detection Models

Stress is an increasingly prevalent mental health condition across the world. In Europe, for example, stress is considered one of the most common health problems, and over USD 300 billion are spent on stress treatments annually. Therefore, monitoring, identification and prevention of stress are of the utmost importance. While most stress monitoring is carried out through self-reporting, there are now several studies on stress detection from physiological signals using Artificial Intelligence algorithms. However, the generalizability of these models is only rarely discussed. The main goal of this work is to provide a monitoring proof-of-concept tool exploring the generalization capabilities of Heart Rate Variability-based machine learning models. To this end, two Machine Learning models are used, Logistic Regression and Random Forest to analyze and classify stress in two datasets differing in terms of protocol, stressors and recording devices. First, the models are evaluated using leave-one-subject-out cross-validation with train and test samples from the same dataset. Next, a cross-dataset validation of the models is performed, that is, leave-one-subject-out models trained on a Multi-modal Dataset for Real-time, Continuous Stress Detection from Physiological Signals dataset and validated using the University of Waterloo stress dataset. While both logistic regression and random forest models achieve good classification results in the independent dataset analysis, the random forest model demonstrates better generalization capabilities with a stable F1 score of 61%. This indicates that the random forest can be used to generalize HRV-based stress detection models, which can lead to better analyses in the mental health and medical research field through training and integrating different models.

Examining Stress and Residual Symptoms in Remitted and Partially Remitted Depression Using a Wearable Electrodermal Activity Device: A Pilot Study

Consistent evidence suggests residual symptoms and stress are the most reliable predictors of relapse in remitted depression. Prevailing methodologies often do not enable continuous real-time sampling of stress. Thus, little is known about day-to-day interactions between residual symptoms and stress in remitted depression. In preparation for a full-scale trial, this study aimed to pilot a wrist-worn wearable electrodermal activity monitor: ADI (Analog Devices, Inc.) Study Watch for assessing interactions between physiological stress and residual depressive symptoms following depression remission. 13 individuals remitted from major depression completed baseline, daily diary, and post-daily diary assessments. Self-reported stress and residual symptoms were measured at baseline and post-daily diary. Diary assessments required participants to wear ADI's Study Watch during waking hours and complete self-report questionnaires every evening over one week. Sleep problems, fatigue, energy loss, and agitation were the most frequently reported residual symptoms. Average skin conductance responses (SCRs) were 16.09 per-hour, with an average of 11.30 hours of wear time per-day. Increased residual symptoms were associated with enhanced self-reported stress on the same day. Increased SCRs on one day predicted increased residual symptoms on the next day. This study showed a wearable electrodermal activity device can be recommended for examining stress as a predictor of remitted depression. This study also provides preliminary work on relationships between residual symptoms and stress in remitted depression. Importantly, significant findings from the small sample of this pilot are preliminary with an aim to follow up with a 3-week full-scale study to draw conclusions about psychological processes explored. Clinical and Translational Impact Statemen-ADI's wearable electrodermal activity device enables a continuous measure of physiological stress for identifying its interactions with residual depressive symptoms following remission. This novel procedure is promising for future studies.

Control, anxiety and test performance: Self-reported and physiological indicators of anxiety as mediators

BACKGROUND

This study investigated the role of different test anxiety components (affective, cognitive, motivational and physiological) as mediators between control and performance as proposed by Pekrun's control-value theory (CVT). While all components were assessed via self-report, the physiological component was additionally assessed via electrodermal activity (EDA).

AIMS

We examined the relative impact of the self-reported anxiety components and EDA in this mediating mechanism to identify the most relevant assessment(s) (i.e., self-reported anxiety components and/or EDA) for predicting test performance.

SAMPLE

The study comprised 50 eighth graders.

METHODS

Data were collected during a mathematics test comprising six task blocks. State self-reports of control and anxiety components along with test performance and other test emotions were collected block-wise (i.e., repeated assessments within students). EDA was continuously recorded.

RESULTS

Consistent with CVT, intra-individual mediation analysis with multiple mediators revealed that higher control predicted lower anxiety (i.e., all self-reported components). Unexpectedly, higher control was associated with increased EDA. Follow-up analyses taking other test emotions into account suggested this might reflect positive activation. Correlations between EDA and control and self-reported anxiety components differed depending on which test emotion was dominant in each situation. Regarding test performance, only the cognitive component was a significant mediator and thus seems to play a pivotal role in the relationship between control and performance.

CONCLUSIONS

Distinguishing between anxiety components and including unbiased physiological measures improve our understanding of the mechanisms behind the relationship between test anxiety and performance. Higher physiological arousal may be a sign of anxiety but can also be a sign of positive activation. When aiming to reduce negative effects of anxiety on performance, targeting the cognitive component seems crucial. Implications of these findings for educational and psychological practice are discussed.

Toward Emotion Recognition From Physiological Signals in the Wild: Approaching the Methodological Issues in Real-Life Data Collection

Emotion, mood, and stress recognition (EMSR) has been studied in laboratory settings for decades. In particular, physiological signals are widely used to detect and classify affective states in lab conditions. However, physiological reactions to emotional stimuli have been found to differ in laboratory and natural settings. Thanks to recent technological progress (e.g., in wearables) the creation of EMSR systems for a large number of consumers during their everyday activities is increasingly possible. Therefore, datasets created in the wild are needed to insure the validity and the exploitability of EMSR models for real-life applications. In this paper, we initially present common techniques used in laboratory settings to induce emotions for the purpose of physiological dataset creation. Next, advantages and challenges of data collection in the wild are discussed. To assess the applicability of existing datasets to real-life applications, we propose a set of categories to guide and compare at a glance different methodologies used by researchers to collect such data. For this purpose, we also introduce a visual tool called Graphical Assessment of Real-life Application-Focused Emotional Dataset (GARAFED). In the last part of the paper, we apply the proposed tool to compare existing physiological datasets for EMSR in the wild and to show possible improvements and future directions of research. We wish for this paper and GARAFED to be used as guidelines for researchers and developers who aim at collecting affect-related data for real-life EMSR-based applications.