Personal Stress-Level Clustering and Decision-Level Smoothing to Enhance the Performance of Ambulatory Stress Detection With Smartwatches

Abstract

Researchers strive hard to develop effective ways to detect and cope with enduring high-level daily stress as early as possible to prevent serious health consequences. Although research has traditionally been conducted in laboratory settings, a set of new studies have recently begun to be conducted in ecological environments with unobtrusive wearable devices. Since patterns of stress are ideographic, person-independent models have generally lower accuracies. On the contrary, person-specific models have higher accuracies but they require a long-term data collection period. In this study, we developed a hybrid approach of personal level stress clustering by using baseline stress self-reports to increase the success of person-independent models without requiring a substantial amount of personal data. We further added decision level smoothing to our unobtrusive smartwatch based stress level differentiation system to increase the performance by correcting false labels assigned by the machine learning algorithm. In order to test and evaluate our system, we collected physiological data from 32 participants of a summer school with wrist-worn unobtrusive wearable devices. This event is comprised of baseline, lecture, exam and recovery sessions. In the recovery session, a stress management method was applied to alleviate the stress of the participants. The perceived stress in the form of NASA-TLX questionnaires collected from the users as self-reports and physiological stress levels extracted using wearable sensors are examined separately. By using our system, we were able to differentiate the 3-levels of stress successfully. We further substantially increase our performance by personal stress level clustering and by applying high-level accuracy calculation and decision level smoothing methods. We also demonstrated the success of the stress reduction methods by analyzing physiological signals and self-reports.

A Review on Mental Stress Detection Using Wearable Sensors and Machine Learning Techniques

“…Some parameters like measuring stress using nasal ST and videos, wearable sensors, mobile phones, blink detection, typing behavior, human voice were also focused. From various machine learning classifiers used in previous papers, Random forest (RF) [23], [26], [38], SVM [20], [23], [26], [35]- [38], and decision trees [24], [35], [37]- [38], were found to be the most effective among all due to their better results as compared to others. Also, GSR, HRV, and ST features were most useful in stress prediction.…”

“…The physiological signals most commonly used in stress detection approaches are Heart Rate (HR) [17], [22]- [26], Heart Rate Variability (HRV), Skin Temperature (ST) [23]- [26], Skin Conductance (also called Galvanic Skin Response (GSR)) [17], [19]- [22], [24]- [26], [35], [37], Blood Pressure (BP) [23], [37], and Respiration Rate (RR) [18], [36], [55]. HRV is the beat-to-beat variability and has time-domain, frequency-domain, and non-linear domain indices for analysis.…”

“…Compromised accuracy due to inaccurate data and inputs, need a regular update of rules of a fuzzy logic control system, requires a lot of testing for validation and verification with hardware, completely dependent on human knowledge and skill, not always broadly accepted due to inaccurate results. [20], [23]- [24], [26], [34], [38], [45]- [50].…”

Modeling of Child Stress-State Identification Based on Biometric Information in Mobile Environment

“…is app presents the stress state of a child in real time and according to time zones. It serves as a medium for real-time monitoring, which enables users to apply it habitually and constantly [35,36].…”

Biosensing and Actuation—Platforms Coupling Body Input-Output Modalities for Affective Technologies

“…These, although not an exhaustive list, are to some extent physiological signals that have become standard for physiology tracking research—slowly crossing disciplines and making their way into affective health tracking, interaction design, and other domains of interest. Moreover, with objectives that range from out-of-the-lab psychophysiology tracking [ 57 , 58 , 59 ] to new perspectives in interaction design [ 43 , 49 , 60 ] our work has often addressed biosignals through other available biosignal research platforms beyond BITalino, such as biosignalsplux [ 61 ], Empatica E4 [ 62 ], Arduino accessories like the Grove GSR [ 63 ], or even commercial wearables such as the Samsung Gear S2 [ 64 ] among others.…”