ArticlePDF Available
Network Biology, 2018, 8(1): 25-34
IAEES www.iaees.org
Article
EEG-metric based mental stress detection
Gaurav1, R. S. Anand1, Vinod Kumar2
1Department of Electrical Engineering, Indian Institute of Technology Roorkee, Roorkee 247667, India
2Jaypee University of Information Technology, Waknaghat 712100, India
E-mail: gsbmi.dee2014@iitr.ac.in, anandfee@iitr.ac.in, vinodfee@iitr.ac.in
Received 13 October 2017; Accepted 20 December 2017; Published 1 March 2018
Abstract
Mental stress level is a vital parameter affecting physical well-being, cognition, emotions, and professional
efficiency. With growing adversities in modern living standards, causing abnormal mental stress, it is
necessary to measure to cure it. Regular personal stress profile generated can be used as neurofeedback for the
clinical as well as personal assessment. This paper describes a method to detect mental stress level based on
physiological parameters. In this method, an electroencephalogram (EEG)-metric parameters based binary and
ternary stress classifier is developed. This is validated through probabilistic stress profiler of differential stress
inventory (a questionnaire based evaluation). Nine channel EEG is used to extract physiological signal.
EEG-metric based cognitive state and workload outputs are generated for 41 healthy volunteers (37 males and
4 females, age; 24±5 years). All subjects were guided to perform three simple tasks of closed eye, focusing
vision on a red dot on center of dark screen and focusing on a white screen. Central tendencies (mean, median
and mode) and standard deviation were extracted of EEG-metric (sleep onset, distraction, low engagement,
high engagement and cognitive states) as features. Either of the two or three classes of stress are evaluated
from probabilistic stress profiler of differential stress inventory and used as training output classes. A
supervisory training of multiple layer perceptron based binary support vector machine classifier was used to
detect stress class one by one. 40 subject’s samples were used for training and interchanging one-by one 41th
subjects stress class is determined from the designed classifier. Out of 41 subjects, stress level of 30 subjects is
correctly identified by binary classifier and stress level of 26 subjects is correctly identified by ternary
classifier, using multi-layer perceptron kernel based SVM.
Keywords EEG-metric; differential stress inventory questionnaire; SVM-MLP.
1 Introduction
Psychological stress is a phenomenon related to thoughts, emotion, physiological changes and everyday
social activities. High level of psychological stress can become cause of mental or physiological illness, or
other health related problems. It has been observed in recent researches that high level of mental stress
NetworkBiology
ISSN22208879
URL:http://www.iaees.org/publications/journals/nb/onlineversion.asp
RSS:http://www.iaees.org/publications/journals/nb/rss.xml
Email:networkbiology@iaees.org
EditorinChief:WenJunZhang
Publisher:InternationalAcademyofEcologyandEnvironmentalSciences
Network Biology, 2018, 8(1): 25-34
IAEES www.iaees.org
adversely affects cardiovascular, endocrine and immune system health, etc. and psychological health (Cohen
et al., 2007). Stress is surveyed to be an unavoidable aspect of college student’s life. A survey conducted by
Nightline Association reported about 65% of university students observe some kind of mental or
psychological stress in their daily academic and other activities (Zheng et al., 2016). Cognitive as well as
physical performance can also be affected due to daily stress conditions. High risk professions, such as
defense services, industrial process plants, vehicle, locomotives and flight operation and control are more
prone to get affected due to high stress condition and can be highly dangerous (Driskell et al., 2013).
Diagnose and measurement of stress level profiles; quantitatively and qualitatively with abnormality spotting
can be achieved by majorly three methods: task oriented, questionnaire based and physiological parameters
assessment based evaluations. Psychological questionnaire based methods are widely used to examine the
stress profile, but this is mostly only empirical to subject’s response and is prone to misrepresentation or
manipulation, and it can result to incorrect measurement. Also, evaluation requires an extensive training and
expertise.
Physiological and psychological parameters have been proposed to comparatively identify anxiety or
stress level (Sharma et al., 2012; Glenn et al., 2014; Hermens et al., 2014). Also, questionnaire based stress
evaluation methods have been developed, such as; Cohens Perceived Stress Scale (PSS), Stress Response
Inventory (SRI) and Hamilton Depression Rating Scale (HDRS) (Cohen et al., 1983; Koh et al., 2000;
Williams, 1988). Few of the methods to induce mental stress in lab settings are as mentioned (Skoluda et al.,
2015). Changes in autonomic nervous system responses can be induced by anxiety, this can be observed
through changes in physiological factors such as heart rate, blood pressure and respiratory rhythm (Jung et al.,
2013). In previous studies, power spectrum density of electroencephalogram (EEG) is observed to changes in
specific way with change in emotion or mental states (Alonso et al., 2015; Dogra et al., 2018; Zhang, 2018).
Different EEG patterns are seen to be changed with increasing level of stress (Hsieh et al., 2013). Stress
evaluation is also possible with examination of electrocardiogram (ECG) based factors, such as heart rate and
heart rate variability (Xu et al., 2015).
Cardiac and respiration activity was found to offer better stress assessment biomarkers than speech,
galvanic skin response or skin temperature when recorded with wearable biomedical measurement systems
(Seoane et al., 2014). For evaluation of anxiety in daily life scenarios, wearable systems which work on
physiological parameters have been developed (Wu et al., 2012). A non-invasive EEG sensor for chronic
stress evaluation was proposed to be worked in everyday social or professional environments (Hu et al.,
2015). Multimodal physiological parameters based mental fatigue prediction methods are developed
(Laurentet al., 2013). More studies are needed to validate their efficacy in case of mental stress evaluation.
Present method is an attempt to develop a method to measure overall psychological stress advent
through physiological signals and it’s parameters, visually EEG. EEG signal is an electrical impression of
bioelectric potential from brain, during regular stimulus and triggering of neuronal activity, due to neuronal
cell-dendrite current dipole dynamic change. Using EEG as a signature of regular brain neuronal activity,
discrete stress levels can be evaluated. In this method, metrics generated from B-Alert X10 based EEG
system and questionnaire based; differential stress inventory (DSI), are together used for discrete level stress
profile assessment. Features processed and generated through EEG-based metrics and discrete levels of stress
profile evaluated from DSI are incorporated on a support vector machine (SVM) based supervisory learning
system, which is further used to assess discrete class of stress profile of particular subject.
2 Experiment and Methods
2.1 EEG-based metrics for cognitive states
26
Network Biology, 2018, 8(1): 25-34
IAEES www.iaees.org
B-Alert X10® is a wireless hardware system for EEG and ECG acquisition with 9 channel EEG and single
channel ECG (B-Alert.com, 2011). B-Alert X-10 system follows 10-20 system of EEG electrodes placement.
The nine electrodes are fixed at Fz, F3 and F4 i.e. at center, left and right position of frontal region scalp; Cz,
C3 and C4 at center, left and right position of central region of scalp; Poz, P3 and P4 at center of
parieto-occipital, and left and right region of parietal region of scalp. ECG electrodes are placed in lead-I
region according to Einthoven triangle. Each EEG electrode to scalp impedance is checked before recording
and kept below 30K. The sampling frequency for acquisition is kept at 265 samples per second and 16 bit
resolution. Before acquisition, an alertness and memory profiler test (AMP) is conducted on each subject
which comprises of 3-choice vigilance task (3CVT), visual psychomotor vigilance Task (VPVT) and auditory
psychomotor vigilance task (APVT), this takes approximately 20 minutes. After AMP test, the acquisition
can be started to get 9 channel raw EEG signal and 1 channel raw ECG signal. Along with EEG and ECG
signal, EEG-based metrics including cognitive state and workload outputs is also produced each epoch of
second. These EEG-metrics are probability of sleep onset, distraction, low engagement, high engagement,
cognitive states (sleep Onset, distraction, low engagement, high engagement), and 2-class model of workload
(high workload, low workload) (Berka et al., 2007). The EEG-metrics are generated as discriminant function
analysis of spectral powers of EEG signals in different ranges of frequencies. EEG-metrics is generated from
every second as epoch by epoch outputs from ABM model. In this study, probability of sleep onset,
distraction, low engagement, high engagement, cognitive states were taken as features to train. For
acquisition and analysis, Acknowledge 4.2 software of Biopac™ is used. Acknowledge 4.2 is equipped with
filtering, mathematical and analysis tools.
2.2 Questionnaire based stress profiler evaluation
DSI is a questionnaire based psychological battery test provided by Vienna test system to analyze and
differentiate behavioral stress and distribute to the specific category of stress experiences (Rost et al., 1989).
The stress questionnaire consists of 52 questions of causes, 21 questions on symptoms, 30 questions on
coping and 20 questions on stabilization. The subject has to choose the amount of accordance of themselves,
with the condition mentioned in the questionnaire. According to the responses of the subject’s, raw score of
the evaluation for stress causes, symptoms, coping and stress stability is evaluated. Final evaluation result
Fig. 1 (a). A participant fixed with B-Alert X10 system on scalp, perform AMP test. (b) Flowchart of task operation, data
acquisition and stress classification.
(a) (b)
27
Network Biology, 2018, 8(1): 25-34
IAEES www.iaees.org
consists of probabilistic profile classification of the subject in describable classes as Type-I: Normal type,
Type-II: Overstressed, Type-III: Stress resistant, Type-IV: Low-stress-successful coping, Type-V: High
stress-successful coping, with values between 0 and 1, sum of all profiler is equal to 1. This profiler is used
for the training the classes of stress detection system.
2.3 Data acquisition, cognitive tasks and evaluation method
For stress classification study, B-Alert X10 system’s EEG and ECG electrodes were attached to the subject’s
head scalp and chest. Before starting physiological signal acquisition, 15 minutes of baseline, AMP test, is
guided to perform by the subject, which generates a learnt, ready to use, definition system for cognitive state
and workload outputs. Three tasks are to be performed by all subjects in physically rest condition sitting on
a chair. The tasks are, first is 5 minutes of eyes closed (EC), then 5 minutes of eyes open focusing vision on a
red color dot on a dark screen (DOT) and 5 minutes of eyes open with focus on a bright screen (EO). First
two tasks were performed in a dark environment and third one in a bright environment. During the task,
simultaneously the EEG, ECG, cognitive state and workload outputs are acquired and recorded. After the
completion of physiological parameters acquisition, a psychological questionnaire of differential stress
inventory is conducted on the subjects and stress probabilistic profiling is done. Profiling of 41 subjects (37
males and 4 females) was done with their consent. All the subjects are healthy students from IIT Roorkee
with clean medical records, no chronic diseases and age range of 24±5 years. In Fig.1 (a) is shown, a subject
is performing AMP task while EEG profile getting recorded.
For training of SVM system, each subject’s total 60 features are taken. From the 5 minutes of EC, DOT
and EC; five EEG-metrics; sleep onset, distraction, low engagement, high engagement and cognitive state
values were used for feature formation. Mean, sum of mean and variance, median and mode of each
EEG-based metric for each task was taken as feature. This way total 60 features were accounted for each
subjects for training purpose. For output classification, DSI outputs are used. Probabilistic profiler of DSI is
combined to form two classes. Profiling is created by adding Type-II: Overstressed and Type-V: High stress
to get ‘High Stress’. Sum of Type-I: Normal type, Type-III: Stress resistant and Type-IV: Low-stress is to get
‘Low Stress’. Similar way a three state stress profile was also created. These 60 features along with stress
class of 40 subjects is fed to a multilayer perceptron based SVM classifier. Once the learning is complete, the
60 features of 41st subject are fed to the classifier system to get stress class. Fig. 1 (b) depicts the
experimental flow of task performance, DSI evaluation, feature formation and supervised learning based
classifier development.
2.4 Linear support vector machine based binary classifier
For classification of the stress profile based on EEG-metrics, support vector machine (SVM) is used. SVM is
a supervised learning scheme for classification and regression (Vapnik et al., 1998; Zhao and Hasan, 2013).
The principle behind SVM algorithm is theory of structural risk minimization. The hyperplane function of
classification is decided based on minimizing of generalization error for decision boundaries. Also SVM is
resistant to over-training, and performance increases with generalization. Simplest form of SVM does binary
classification, in which few points in the data space is identified to construct a hyperplane which separates
two classes of points.
The training data x consists of n data samples each of m dimensions and belonging to class y, is
expressed as:
),,(),...,,(),...,,( 11 nnii yxyxyx
1,1, yx m
SVM projects data (xi , yi) into an infinite dimensional hyperplane (xi , yi) by using dedicated normalized
kernel function and defines its decision rule as sign(f(x)). The discriminant function f(x) creates the optimum
28
Network Biology, 2018, 8(1): 25-34
IAEES www.iaees.org
hyperplane decision boundary by using weight vector w* and bias b*.
** )()( bxwxf
The optimum values of w* and b* are estimated by solving following optimization problem where, C is
regularization parameter and i
s are slack variables allowing inseparable data.
;
2
1
min 1
2
,
n
ii
wCw
,1))(.( iii bxwy
0
i
This problem is solved using Lagrange optimization through dual formation, which finally yields
optimum value for weight vector w* and bias b*. Since SVM estimates an infinite dimensional optimum
hyper plane, usually it performs better than the other supervised learning algorithms while solving
classification problems even on higher dimensional input features. Other optimal decision boundaries, such as
polynomial, radial basis function based and multi-layer perceptron based kernels can also be used in place of
linear decision boundary for better classification results.
3 Results
Fig. 2 (a) shows 10s raw signal sample of one channel ECG and nine channel EEG, acquired through B-Alert
X10 system, Fig. 2 (b) shows five of the eight EEG-metric based cognitive states chosen for feature selection,
with respect time. As shown in Fig. 2 (b) EEG-metric are generated every second epoch by epoch. The
EEG-metrics lies between 0 and 1. Fig. 3 shows boxplot of all EEG-metrics features (mean and median) of all
samples of ‘High Stress’ and ‘Low Stress’ taken together. Fig. 3 (a) and (b) are box plot of mean and median
of metrics: high engagement, low engagement, distraction and drowsy metrics for the task of eyes closed.
Similarly, in Fig. 3 (c) and (d), features are for the task of open eyes with vision focused on red dot on dark
screen, and Fig. 3 (e) and (d), features are for task of open eyes with vision focused on white screen. Fig. 3 (g)
and (h) are Box plot of means and medians of cognitive state classification of less stress and high stress
group for three tasks of dot focus, eyes closed and eyes open.
Fig. 2 (a). Plot of raw ECG and EEG signal sample acquired from B-Alert X10, (b) EEG-metric acquired from B-Alert X10.
(a) (b)
29
Network Biology, 2018, 8(1): 25-34
IAEES www.iaees.org
Table 1 is the representation of maximum, minimum, median and mode of all the features, visually,
cognitive state classification, high engagement, low engagement, distraction and drowsiness for three task
conditions, for two stress classes (Less stress and High stress). Table 1(a) are observation for task of open
eyes focusing on a bright dot on dark screen, Table 1(b) are observation for task of closed eyes focusing in a
dark room, Table 1(c) are observation for task of open eyes focusing on a bright screen. It is observed that
central tendencies (mean and median) for high engagement and low engagement metrics are higher for high
stress class than less stress class, whereas distraction and drowsy metrics are lesser for high stress than low
stress. Cognitive state classification metric for DOT and EO tasks is seen at ‘Low Engagement’ for both less
and high stress class. Whereas for EC task, less stress class is seen in ‘Distraction’ state and ‘Low
Engagement’ for high stress class. Binary class SVM with various decision boundaries: linear, quadratic,
polynomial, radial-basis function (RBF) and multi-layer perceptron (MLP), were used to develop classifiers
using 60 features of each of 40 subjects on two and three class of stresses. Similarly, by interchanging
training and prediction matrix of all 41 subjects, the efficacy of the SVM based system is to be checked.
30
Network Biology, 2018, 8(1): 25-34
IAEES www.iaees.org
Fig. 3 Box plot showing contrast out of 41 subjects of either ‘Less stress’ or ‘High Stress’ EEG-metrics features of (high
engagement, low engagement, Distraction, Drowsy) for the tasks of (a) mean while closed eyes (b) median while closed eyes
(c) mean while focusing on a dot on dark screen, (d) median while focusing on a dot on dark screen (e) mean while open eyes
(f) median while open eyes, (g) mean of cognitive state classification of less stress and high stress group for three tasks of dot
focus, eyes closed and eyes open, (h) medians cognitive state classification.
(a)
(h) (g)
(f) (e)
(d) (c)
(b)
31
Network Biology, 2018, 8(1): 25-34
IAEES www.iaees.org
Altogether, for two states stress classifiers, out of 41 binary outputs, 23 were correct for linear and
quadratic, 22 for polynomial, 27 for RBF and 30 for MLP. So accuracy of 56.09% for linear and quadratic,
53.66% for polynomial, 65.85% for RBF and 73.1% for MLP based decision boundary was observed. This
evaluates multi-layer perceptron based SVM as the best classifier for two state stress classes.
For three states stress classifiers, out of 41 binary outputs, 16 were correct for linear and quadratic, 17
for polynomial, 23 for RBF and 26 for MLP. So accuracy of 39.02% for linear and quadratic, 41.46% for
polynomial, 56.10% for RBF and 63.41% for MLP based decision boundary was observed. This evaluates
multi-layer perceptron based SVM as the best classifier for both: two state as well as three stress classes.
Table 1 (a) Comparison of maximum, minimum, median and mode of EEG-metric features for two stress classes (Less stress and
High stress) while focusing on a dot on a screen.
Less Stress
Class
CC (Mean )*
CC (Median )
CC (Mode )
HE (Mean )
HE (Median )
HE (Mode )
LE (Mean )
LE (Median )
LE (Mode )
Dist (Mean )
Dist (Median )
Dist (Mode )
Dro (Mean )
Dro (Median )
Dro (Mode )
Minimum 0.36 0.3 0.1 0 0 0 0 0 0 0.01 0 0 0 0 0
Maximum 0.88 1 1 0.92 1 1 0.44 0.39 0.62 0.51 0.56 0.19 0.48 0.39 1
Median 0.61 0.6 0.9 0.38 0.28 0 0.2 0.09 0 0.26 0.03 0 0.02 0 0
Mode 0.61 0.6 0.9 0.29 0 0 0.27 0 0 0.01 0 0 0 0 0
High Stress
Class
CC (Mean )
CC (Median )
CC (Mode )
HE (Mean )
HE (Median )
HE (Mode )
LE (Mean )
LE (Median )
LE (Mode )
Dist (Mean )
Dist (Median )
Dist (Mode )
Dro (Mean )
Dro (Median )
Dro (Mode )
Minimum 0.45 0.1 0.1 0.05 0 0 0 0 0 0 0 0 0 0 0
Maximum 0.76 0.9 0.9 0.6 0.7 1 0.85 1 1 0.63 0.87 1 0.5 0.43 1
Median 0.605 0.6 0.75 0.395 0.29 0 0.355 0.24 0 0.16 0.015 0 0.005 0 0
Mode 0.58 0.6 0.9 0.55 0.01 0 0.36 0.24 0 0.06 0 0 0 0 0
Table 1 (b) Comparison of maximum, minimum, median and mode of EEG-metric features for two stress classes (Less stress
and High stress) while eyes closed.
Less Stress
Class
CC (Mean )
CC (Median )
CC (Mode )
HE (Mean )
HE (Median )
HE (Mode )
LE (Mean )
LE (Median )
LE (Mode )
Dist (Mean )
Dist (Median )
Dist (Mode )
Dro (Mean )
Dro (Median )
Dro (Mode )
Minimum 0.21 0.1 0.1 0 0 0 0.01 0 0 0.01 0 0 0 0 0
Maximum 0.8 0.9 0.9 0.72 0.88 0.15 0.5 0.52 0.82 1 1 1 0.53 0.61 1
Median 0.46 0.3 0.3 0.21 0.04 0 0.18 0.03 0 0.41 0.19 0 0.09 0 0
Mode 0.56 0.3 0.3 0.03 0 0 0.02 0 0 0.41 0 0 0 0 0
High Stress
Class
CC (Mean )
CC (Median )
CC (Mode )
HE (Mean )
HE (Median )
HE (Mode )
LE (Mean )
LE (Median )
LE (Mode )
Dist (Mean )
Dist (Median )
Dist (Mode )
Dro (Mean )
Dro (Median )
Dro (Mode )
Minimum 0.31 0.3 0 0.02 0 0 0 0 0 0.01 0 0 0 0 0
32
Network Biology, 2018, 8(1): 25-34
IAEES www.iaees.org
Maximum 0.83 0.9 0.9 0.8 1 1 0.93 1 1 0.85 1 1 0.2 0 0
Median 0.55 0.6 0.45 0.22 0.025 0 0.255 0.105 0 0.29 0.07 0 0.01 0 0
Mode 0.55 0.3 0.3 0.22 0 0 0.2 0 0 0.01 0 0 0.01 0 0
Table 1 (c) Comparison of maximum, minimum, median and mode of EEG-metric features for two stress classes (Less stress and
High stress) while looking at a bright screen.
Less Stress
Class
CC (Mean )
CC (Median )
CC (Mode )
HE (Mean )
HE (Median )
HE (Mode )
LE (Mean )
LE (Median )
LE (Mode )
Dist (Mean )
Dist (Median )
Dist (Mode )
Dro (Mean )
Dro (Median )
Dro (Mode )
Minimum 0.31 0.3 0.1 0 0 0 0.01 0 0 0.01 0 0 0 0 0
Maximum 0.82 0.9 0.9 0.77 1 0.71 0.57 0.63 0.25 0.98 1 1 0.4 0.08 1
Median 0.59 0.6 0.6 0.36 0.24 0 0.21 0.07 0 0.24 0.04 0 0.04 0 0
Mode 0.55 0.6 0.9 0.26 0 0 0.13 0 0 0.24 0 0 0 0 0
High Stress
Class
CC (Mean )
CC (Median )
CC (Mode )
HE (Mean )
HE (Median )
HE (Mode )
LE (Mean )
LE (Median )
LE (Mode )
Dist (Mean )
Dist (Median )
Dist (Mode )
Dro (Mean )
Dro (Median )
Dro (Mode )
Minimum 0.11 0.1 0.1 0.01 0 0 0.01 0 0 0 0 0 0 0 0
Maximum 0.84 0.9 0.9 0.85 0.99 1 0.87 1 1 0.84 1 1 0.97 1 1
Median 0.62 0.6 0.75 0.42 0.345 0 0.28 0.2 0 0.185 0.005 0 0.015 0 0
Mode 0.62 0.6 0.9 0.53 0 0 0.37 0 0 0.03 0 0 0 0 0
*CC: Cognitive classification state, HE: High engagement metric, LE: Low Engagement metric, Dist: Distraction metric and Dro: Drowsiness metric
4 Discussion
More development in the methods are possible, to detect multiple class and multiple level outputs and to get
detailed stress profiler on the basis of physiological signals. Also scopes for better feature choice for
classification and feature dimension reduction without affecting the classification mechanism can be further
achieved. With mental stress emulator at lab setup, a real time stress or anxiety level detector is also possible.
Other than questionnaire based stress profiler, other methods such as task based stress level can be used for
training purpose. Further, the number of EEG can be reduced and validity of method with lesser electrodes is
to be checked. With other possible EEG-metrics and reduced dimension, the predictor can be developed. The
duration of task conduction can be minimized and real-time stress level detection methods can be possible.
5 Conclusion
A simple SVM based classification method for discrete stress level detection was developed using
EEG-metrics and cognitive states as features for training and discrete classes generated from evaluation of
differential stress inventory questionnaire based stress profiler as training outputs. The experimental study on
41 subjects successfully detected 30 subject’s stress level correctly for binary classifier and 26 subject’s for
ternary classifier, which establishes the efficacy of the method. The method can be used for general purpose
stress level detection.
33
Network Biology, 2018, 8(1): 25-34
IAEES www.iaees.org
References
Alonso JF, Romero S, Ballester MR, Antonijoan RM, Mañanas MA. 2015. Stress assessment based on EEG
univariate features and functional connectivity measures. Physiological Measurement, 36(7): 1351-1365
Berka C, Levendowski DJ, Lumicao MN, Yau A, Davis G, Zivkovic VT, Olmstead RE, Tremoulet PD, Craven
PL. 2007. EEG correlates of task engagement and mental workload in vigilance, learning, and memory
tasks. Aviation, Space, and Environmental Medicine, 78(5): B231-B244
B-Alert.com. 2011. Advanced Brain Monitoring B-Alert x-10. http://www.b-alert.com/pdf/x10.pdf. Accessed
on July 1, 2017
Cohen S, Janicki-Deverts D, Miller GE. 2007. Psychological stress and disease. JAMA, 298(14): 1685-1687
Cohen S, Kamarck T, Mermelstein R. 1983. A global measure of perceived stress. Journal of health and Social
Behavior, 24(4): 385-396
Dogra J, Prashar N, Jain S, Sood M. 2018. Improved methods for analyzing MRI brain images. Network
Biology, 8(1)
Driskell JE, Salas E. 2013. Stress and Human Performance. Psychology Press, USA
Glenn T, Monteith S. 2014. New measures of mental state and behavior based on data collected from sensors,
smartphones, and the Internet. Current psychiatry reports, 16(12): 523
Hermens H, op den Akker H, Tabak M, Wijsman J, Vollenbroek M. 2014. Personalized coaching systems to
support healthy behavior in people with chronic conditions. Journal of electromyography and
kinesiology, 24(6): 815-826
Hsieh CS, Tai CC. 2013. A fatigue state evaluation system based on the band energy of electroencephalography
signals. Sensors and Materials, 25(9): 697-706
Hu B, Peng H, Zhao Q, Hu B, Majoe D, Zheng F, Moore P. 2015. Signal Quality assessment model for wearable
EEG sensor on prediction of mental stress. IEEE transactions on nanobioscience, 14(5): 553-561
Jung SJ, Chung WY. 2013. Non-intrusive healthcare system in global machine-to-machine networks. IEEE
Sensors Journal, 13(12): 4824-4830
Rost DH, Schermer FJ. 1989. Diagnostik des Leistungsangsterlebens. Diagnostica, 35(4): 287-314
Skoluda N, Strahler J, Schlotz W, Niederberger L, Marques S, Fischer S, Thoma MV, Spoerri C, Ehlert U, Nater
UM. 2015. Intra-individual psychological and physiological responses to acute laboratory stressors of
different intensity. Psychoneuroendocrinology, 51: 227-236
Sharma N, Gedeon T. 2012. Objective measures, sensors and computational techniques for stress recognition
and classification: A survey. Computer methods and programs in biomedicine, 108(3): 1287-1301
Vapnik VN, Vapnik V. 1998. Statistical Learning Theory (Vol. 1). Wiley, New York, USA
Williams JB. 1988. A structured interview guide for the Hamilton Depression Rating Scale. Archives of General
Psychiatry, 45(8): 742-747
Wu W, Gil Y, Lee J. 2012. Combination of wearable multi-biosensor platform and resonance frequency training
for stress management of the unemployed population. Sensors, 12(10): 13225-13248
Xu Q, Nwe TL, Guan C. 2015. Cluster-based analysis for personalized stress evaluation using physiological
signals. IEEE journal of biomedical and health informatics, 19(1): 275-281
Zhang WJ. 2018. Fundamentals of Network Biology. World Scientific, Singapore
Zhao Y, Hasan YA. 2013. Machine learning algorithms for predicting roadside fine particulate matter
concentration level in Hong Kong Central. Computational Ecology and Software, 3(3): 61-73
Zheng Y, Wong TC, Leung BH, Poon CC. 2016. Unobtrusive and multimodal wearable sensing to quantify
anxiety. IEEE Sensors Journal, 16(10): 3689-3696
34
... Subhani et al. [31] developed a machine learning framework involving EEG signal analysis of stressed participants based on frequency-domain features. Anand and Kumar [32] developed a method to detect low/average/high mental stress levels based on physiological parameters. In their approach, EEG-metric time-domain parameters, such as standard deviation, were used to classify stress. ...
... In a similar study, Subhani et al. [31] developed a machine learning framework involving EEG signal analysis and showed that the proposed framework produced 83.4% accuracy for multiple level identification. Anand and Kumar [32] showed the maximum accuracy of 73.1% in detecting stress levels based on EEG-metric time-domain parameters, such as standard deviation. Accordingly, in order to test the trained 3-SVM models quantitatively, 40% of samples that had not been used in the training process were used. ...
Article
Stress is one of the most significant health problems in the 21st century, and should be dealt with due to the costs of primary and secondary cares of stress-associated psychological and psychiatric problems. In this study, the brain network states exposed to stress were monitored based on electroencephalography (EEG) measures extracted by complex network analysis. To this regard, 23 healthy male participants aged 18–28 were exposed to a stress test. EEG data and salivary cortisol level were recorded for three different conditions including before, right after, and 20 min after exposure to stress. Then, synchronization likelihood (SL) was calculated for the set of EEG data to construct complex networks, which are scale reduced datasets acquired from multi-channel signals. These networks with weighted connectivity matrices were constructed based on original EEG data and also by using four different waves of the recorded signals including δ, θ, α, and β. In addition to these networks with weighted connectivity, networks with binary connectivity matrices were also derived using threshold T. For each constructed network, four measures including transitivity, modularity, characteristic path length, and global efficiency were calculated. To select the sensitive optimal features from the set of the calculated measures, compensation distance evaluation technique (CDET) was applied. Finally, multi-class support vector machine (SVM) was trained in order to classify the brain network states. The results of testing the SVM models showed that the features based on the original EEG, α and β waves have got better performances in monitoring the brain network states. © 2018 Nalecz Institute of Biocybernetics and Biomedical Engineering of the Polish Academy of Sciences
... In [29], Alyasseri et al. used subject-independent discrete wavelet transform (DWT) based statistical features along with entropy to classify five mental tasks for seven subjects using artificial neural network (ANN). In [30], EEG signals recorded from 41 subjects during three mental tasks have been classified using subject-independent statistical features and MLP kernel based SVM. An immune-feature weighted SVM has been proposed to classify five mental tasks for seven subjects with approximate entropy feature in [31]. ...
Article
Cognitive/mental task classification using single/limited channel(s) electroencephalogram (EEG) signals in real-time play an important role in designing portable brain-computer interface (BCI) and neurofeedback (NFB) systems. However, real-time recorded EEG signals are often contaminated with ocular artifacts (OAs) and muscle artifacts (MAs), which deteriorate the hand-crafted features extracted from EEG signal, resulting inadequate classification of cognitive tasks. Therefore, we investigate the use of deep learning techniques which do not require manual feature extraction or artifact suppression. In this paper, we propose a shallow one-dimensional convolutional neural network (1D-CNN) architecture for cognitive task classification. The robustness of the proposed architecture is evaluated using artifact-free and artifact-contaminated EEG signals taken from two publicly available databases (i.e, Keirn and Aunon (K) database and EEGMAT (E) database) and in-house (R) database recorded using single-channel device in performing not only cognitive/non-cognitive binary task classification but also cognitive/cognitive multi-tasks classification. Evaluation results demonstrate that the proposed architecture achieves the highest subject-independent classification accuracy of 99.70% and 100.00% for multi-class classification and pair-wise classification respectively in database K. Further, subject-independent classification accuracies of 99.00% and 98.00% are achieved in databases E and R respectively. Comparative performance analysis demonstrates that the proposed architecture outperforms existing approaches not only in terms of classification accuracy but also in robustness against artifacts.
... In [25], Alyasseri et al. used subjectindependent discrete wavelet transform (DWT) based statistical features along with entropy to classify five mental tasks for seven subjects using artificial neural network (ANN). In [26], EEG signals recorded from 41 subjects during three mental tasks have been classified using subject-independent statistical features and multi-layer perceptron (MLP) kernel based SVM. An immune-feature weighted SVM has been proposed to classify five mental tasks for seven subjects with approximate entropy feature in [27]. ...
Preprint
Mental task identification and classification using single/limited channel(s) electroencephalogram (EEG) signals in real-time play an important role in the design of portable brain-computer interface (BCI) and neurofeedback (NFB) systems. However, the real-time recorded EEG signals are often contaminated with noises such as ocular artifacts (OAs) and muscle artifacts (MAs), which deteriorate the hand-crafted features extracted from EEG signal, resulting inadequate identification and classification of mental tasks. Therefore, we investigate the use of recent deep learning techniques which do not require any manual feature extraction or artifact suppression step. In this paper, we propose a light-weight one-dimensional convolutional neural network (1D-CNN) architecture for mental task identification and classification. The robustness of the proposed architecture is evaluated using artifact-free and artifact-contaminated EEG signals taken from two publicly available databases (i.e, Keirn and Aunon ($K$) database and EEGMAT ($E$) database) and in-house ($R$) database recorded using single-channel neurosky mindwave mobile 2 (MWM2) EEG headset in performing not only mental/non-mental binary task classification but also different mental/mental multi-tasks classification. Evaluation results demonstrate that the proposed architecture achieves the highest subject-independent classification accuracy of $99.7\%$ and $100\%$ for multi-class classification and pair-wise mental tasks classification respectively in database $K$. Further, the proposed architecture achieves subject-independent classification accuracy of $99\%$ and $98\%$ in database $E$ and the recorded database $R$ respectively. Comparative performance analysis demonstrates that the proposed architecture outperforms existing approaches not only in terms of classification accuracy but also in robustness against artifacts.
... Gaurav et al. [17] have portrayed a strategy to distinguish mental stress level dependent on physiological parameters. ey had used SVM based on the binary classifier and classified stress in 2 levels and 3 levels, respectively. ...
Article
Full-text available
Stress is categorized as a condition of mental strain or pressure approaches because of upsetting or requesting conditions. There are various sources of stress initiation. Researchers consider human cerebrum as the primary wellspring of stress. To study how each individual encounters stress in different forms, researchers conduct surveys and monitor it. The paper presents the fusion of 5 algorithms to enhance the accuracy for detection of mental stress using EEG signals. The Whale Optimization Algorithm has been modified to select the optimal kernel in the SVM classifier for stress detection. An integrated set of algorithms (NLM, DCT, and MBPSO) has been used for preprocessing, feature extraction, and selection. The proposed algorithm has been tested on EEG signals collected from 14 subjects to identify the stress level. The proposed approach was validated using accuracy, sensitivity, specificity, and F1 score with values of 96.36%, 96.84%, 90.8%, and 97.96% and was found to be better than the existing ones. The algorithm may be useful to psychiatrists and health consultants for diagnosing the stress level.
... In significant visual stimulus experiments, synchronize between flickering light frequency induced as stimulus to eyes and EEG responses with coinciding frequencies has been found, the phenomenon called as photic driving [3]. Alpha, beta and theta correlates were found for absorbed meditative attention states and attentional stress factors [4], [5]. Because of the precise temporal resolution of electrophysiological recordings, the event-related potential (ERP) technique has proven particularly valuable for testing theories of perception and attention [6]. ...
Article
Various automated/semi-automated medical diagnosis systems based on human physiology have been gaining enormous popularity and importance in recent years. Physiological features exhibit several unique characteristics that contribute to reliability, accuracy and robustness of systems. There has also been significant research focusing on detection of conventional positive and negative emotions after presenting laboratory-based stimuli to participants. This paper presents a comprehensive survey on the following facets of mental stress detection systems: physiological data collection, role of machine learning in Emotion Detection systems and Stress Detection systems, various evaluation measures, challenges and applications. An overview of popular feature selection methods is also presented. An important contribution is the exploration of links between biological features of humans with their emotions and mental stress. The numerous research gaps in this field are highlighted which shall pave path for future research. © 2019 Nalecz Institute of Biocybernetics and Biomedical Engineering of the Polish Academy of Sciences
Book
Full-text available
Description As the first comprehensive title on network biology, this book covers a wide range of subjects including scientific fundamentals (graphs, networks, etc) of network biology, construction and analysis of biological networks, methods for identifying crucial nodes in biological networks, link prediction, flow analysis, network dynamics, evolution, simulation and control, ecological networks, social networks, molecular and cellular networks, network pharmacology and network toxicology, big data analytics, and more. Across 12 parts and 26 chapters, with Matlab codes provided for most models and algorithms, this self-contained title provides an in-depth and complete insight on network biology. It is a valuable read for high-level undergraduates and postgraduates in the areas of biology, ecology, environmental sciences, medical science, computational science, applied mathematics, and social science. Sample Chapter(s) 1. Fundamentals of Graph Theory Contents: Mathematical Fundamentals: Fundamentals of Graph Theory Graph Algorithms Fundamentals of Network Theory Other Fundamentals Crucial Nodes/Subnetworks/Modules, Network Types, and Structural Comparison: Identification of Crucial Nodes and Subnetworks/Modules Detection of Network Types Comparison of Network Structure Network Dynamics, Evolution, Simulation and Control: Network Dynamics Network Robustness and Sensitivity Analysis Network Control Network Evolution Cellular Automata Self-Organization Agent-based Modeling Flow Analysis: Flow/Flux Analysis Link and Node Prediction: Link Prediction: Sampling-based Methods Link Prediction: Structure- and Perturbation-based Methods Link Prediction: Node-Similarity-based Methods Node Prediction Network Construction: Construction of Biological Networks Pharmacological and Toxicological Networks: Network Pharmacology and Toxicology Ecological Networks: Food Webs Microscopic Networks: Molecular and Cellular Networks Social Networks: Social Network Analysis Software: Software for Network Analysis Big Data Analytics: Big Data Analytics for Network Biology Readership: Advanced undergraduates and graduate students and researchers in biology, ecology, pharmacology, applied mathematics, computational science, etc.
Article
Full-text available
This paper aims to develop an objective index for anxiety based on features derived from electroencephalogram (EEG) and photoplethysmogram (PPG) collected from wearable headset and glasses. The 20 subjects were asked to ride at his most comfortable speed in Task 1 and ride while imagining competing with another person in Task 2. A Competitive State Anxiety Inventory-2 questionnaire was conducted before each task to evaluate the anxiety level of each participant. Various features were extracted from EEG and PPG. The results of this paper showed that the mean value and average power of alpha band wavelet coefficients and that of beta band are highly correlated with the anxiety level ( r=-0.49 and-0.58, p<0.01 for alpha band, and r=-0.51 and-0.58, p<0.01 for beta band, respectively). Features extracted from partial autocorrelation of EEG showed moderate correlation with the anxiety level. Mean pulse rate also acts as a potential anxiety marker for individualized anxiety measurement. Using both EEG and PPG features, the classification accuracy of three-level anxiety by principle component analysis and k-nearest neighbors can achieve 62.5% across subjects. To conclude, wearable sensors have the potential to be used for assessing anxiety level objectively and unobtrusively to facilitate on-site sports performance enhancement and mental-stress-related studies.
Article
Full-text available
The biological response to stress originates in the brain but involves different biochemical and physiological effects. Many common clinical methods to assess stress are based on the presence of specific hormones and on features extracted from different signals, including electrocardiogram, blood pressure, skin temperature, or galvanic skin response. The aim of this paper was to assess stress using EEG-based variables obtained from univariate analysis and functional connectivity evaluation. Two different stressors, the Stroop test and sleep deprivation, were applied to 30 volunteers to find common EEG patterns related to stress effects. Results showed a decrease of the high alpha power (11 to 12 Hz), an increase in the high beta band (23 to 36 Hz, considered a busy brain indicator), and a decrease in the approximate entropy. Moreover, connectivity showed that the high beta coherence and the interhemispheric nonlinear couplings, measured by the cross mutual information function, increased significantly for both stressors, suggesting that useful stress indexes may be obtained from EEG-based features.
Article
Full-text available
Technology development in wearable sensors and biosignal processing has made it possible to detect human stress from the physiological features. However, the intersubject difference in stress responses presents a major challenge for reliable and accurate stress estimation. This research proposes a novel cluster-based analysis method to measure perceived stress using physiological signals, which accounts for the intersubject differences. The physiological data are collected when human subjects undergo a series of task-rest cycles, incurring varying levels of stress that is indicated by an index of the State Trait Anxiety Inventory. Next, a quantitative measurement of stress is developed by analyzing the physiological features in two steps: 1) a k -means clustering process to divide subjects into different categories (clusters), and 2) cluster-wise stress evaluation using the general regression neural network. Experimental results show a significant improvement in evaluation accuracy as compared to traditional methods without clustering. The proposed method is useful in developing intelligent, personalized products for human stress management.
Article
Image segmentation is a part of image processing for region or object extraction from the background area. Owing to the complex background, contrast of the infected portion, low intensity difference values, intricate inner body parts etc.; the problem of region extraction in segmentation is very challenging. Among various image segmentation techniques, thresholding is one of the simplest techniques, in which the region of interest is extracted from the background by comparing the pixel values with the threshold value. The threshold value is obtained from histogram of the image. The technique presented in the paper involves graph cut method in which the initial centroids are automatically selected by exploiting the symmetrical nature of the MRI images. The results obtained by the thresholding technique in this research work shows that any abnormality can be localized easily in horizontal divided MRI brain image rather than in vertical divided MRI image. Graph cut results show better segmentation than thresholding technique which is justified by PSNR and SSIM values
Article
Cranial nerve information can be used to correctly analyze fatigue states. Spectral analysis is the major method of identifying fatigue states. Various frequency bands can be distinguished by digital filters owing to their high accuracy and driftless features. The electroencephalography EEG signal is sent to a personal computer (PC) via a universal serial bus USB interface from a microcontroller and passed through digital filters within 200 taps, and thus, the spectrum of individual signals can be analyzed. This study has investigated the four EEG frequency bands, delta ()δ, theta (θ), alpha (α), and beta (β), using four algorithms to evaluate the fatigue state based on the EEG signals. We compared the four algorithms and determined the best one.
Article
Electroencephalogram (EEG) plays an important role in E-healthcare systems, especially in the mental healthcare area, where constant and unobtrusive monitoring is desirable. In the context of OPTIMI project, a novel, low cost, and light weight wearable EEG sensor has been designed and produced. In order to improve the performance and reliability of EEG sensors in real-life settings, we propose a method to evaluate the quality of EEG signals, based on which users can easily adjust the connection between electrodes and their skin. Our method helps to filter invalid EEG data from personal trials in both domestic and office settings. We then apply an algorithm based on Discrete Wavelet Transformation (DWT) and Adaptive Noise Cancellation (ANC) which has been designed to remove ocular artifacts (OA) from the EEG signal. DWT is applied to obtain a reconstructed OA signal as a reference while ANC, based on recursive least squares, is used to remove the OA from the original EEG data. The newly produced sensors were tested and deployed within the OPTIMI framework for chronic stress detection. EEG nonlinear dynamics features and frontal asymmetry of theta, alpha and beta bands have been selected as biological indicators for chronic stress, showing relative greater right anterior EEG data activity in stressful individuals. Evaluation results demonstrate that our EEG sensor and data processing algorithms have successfully addressed the requirements and challenges of a portable system for patient monitoring, as envisioned by the EU OPTIMI project.
Article
Objectives: The phenomenon of stress is understood as a multidimensional concept which can be captured by psychological and physiological measures. There are various laboratory stress protocols which enable stress to be investigated under controlled conditions. However, little is known about whether these protocols differ with regard to the induced psycho-physiological stress response pattern. Methods: In a within-subjects design, 20 healthy young men underwent four of the most common stress protocols (Stroop test [Stroop], cold pressor test [CPT], Trier Social Stress Test [TSST], and bicycle ergometer test [Ergometer]) and a no-stress control condition (rest) in a randomized order. For the multidimensional assessment of the stress response, perceived stress, endocrine and autonomic biomarkers (salivary cortisol, salivary alpha-amylase, and heart rate) were obtained during the experiments. Results: All stress protocols evoked increases in perceived stress levels, with the highest levels in the TSST, followed by Ergometer, Stroop, and CPT. The highest HPA axis response was found in the TSST, followed by Ergometer, CPT, and Stroop, whilst the highest autonomic response was found in the Ergometer, followed by TSST, Stroop, and CPT. Conclusions: These findings suggest that different stress protocols differentially stimulate various aspects of the stress response. Physically demanding stress protocols such as the Ergometer test appear to be particularly suitable for evoking autonomic stress responses, whereas uncontrollable and social-evaluative threatening stressors (such as the TSST) are most likely to elicit HPA axis stress responses. The results of this study may help researchers in deciding which stress protocol to use, depending on the individual research question.