Vol.1, No.1, 1-3 (2013) Occupational Diseases and Environmental Medicine
http://dx.doi.org/10.4236/odem.2013.11001
The complexity of occupational stress
electroencephalogram
Honger Tian1, Lili Cao1, Jun Wang2*, Tian Xu1, Yongguo Zhan1, Ling Liu1
1Key Laboratory of Environment Medicine and Engineering, Ministry of Education, School of Public Health, Southeast University,
Nanjing, China
2Image Processing and Image Communications Key Lab, College of Geo & Bio Information, Nanjing University of Posts & Tele-
comm, Nanjing, China; *Corresponding Author:
Received 18 September 2013; revised 20 October 2013; accepted 4 November 2013
Copyright © 2013 Honger Tian et al. This is an open access article distributed under the Creative Commons Attribution License,
which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. In accor-
dance of the Creative Commons Attribution License all Copyrights © 2013 are reserved for SCIRP and the owner of the intellectual
property Honger Tian et al. All Copyright © 2013 are guarded by law and by SCIRP as a guardian.
ABSTRACT
It is an important method for using electroen-
cephalogram (EEG) to detect and diagnose oc-
cupational Stress in clinical practice. In this pa-
per, the complexity analysis method based on
Jensen-Shannon Divergence was used to cal-
culate the complexity of occupational stress
electroencephalogram from students and nurses.
The study found that the complexity of nurses’
EEG was higher than that of students’ EEG. The
result can be used to assisted clinical diagnosis.
Keywords: Occupational Stress;
Electroencephalogram; Student s;
Nurses; Jensen-Shannon Divergence
1. INTRODUCTION
Since Jensen-Shannon Divergence (JSD) [1] (which
was used to measure the difference between the prob-
ability distribution of random variables) was proposed in
1991, it was widely applied to the symbol sequence
analysis and characterization [2], such as pattern recog-
nition [3], DNA sequence segmentation. JSD is the result
of symmetrizing and smoothing the Kullback-Leibler
Divergence (KLD). The non-negativity, symmetry, con-
tinuity [4,5] and boundness features of JSD have been
widely used in the analysis of time series. Electroen-
cephalogram (EEG) can also be seen as a time series, so
we consider using JSD complexity based analysis me-
thod to achieve recognition and detection of EEG. In this
paper, 12 graduate students from Southeast University
and 12 nurses from a third-grade class-A hospital
were chosen for comparative analysis.
Nurses need to withstand pressure from work, patient
and family. The occupational stress can cause changes in
psychology, physiology and behavior, such as anxiety,
irritability, depression, chronic fatigue syndrome, sleep
disorders, immune system suppression, cardiovascular
system disease, aggression, and bad habits [6-9]. Gradu-
ate students in universities relatively have lower occupa-
tional stress than the nurses. So their EEG should have a
different dynamic complexity. The Jensen-Shannon Di-
vergence was used to quantitatively analyze time series.
As a statistic parameter in information statistics, Jensen-
Shannon Divergence described the complexity between
different signals and the entropy value increased with
increased complexity.
2. JENSEN-SHANNON DIVERGENCE
(JSD)
Proposed that 1, 2 were two probability distribu-
tion of discrete random variables X. KULLBACK de-
fined direct difference I as:
p p
 

1
12 1
2
,log
xX
px
Ipppx px

 (1)
It can be seen from Equation (1) that differences I has
non-negativity and incremental feature, but it does not
have symmetry. In order to meet this point, the form of a
symmetric difference—J difference was proposed:



  


12
12 21
1
12
2
,
,,
log
xX
Jpp
Ipp Ipp
px
px pxpx




(2)
Distance change in the of two probability distributions
can meet the metric nature. The distance change was

H. E. Tian et al. / Occupational Diseases and Environmental Medicine 1 (2013) 1-3
2
defined as follows:
 
121 2
,
xX
Vpppxp x


 (3)
For the relationship between the direct differences I
and distance changes V, it was found the minimum value
of the differences I based on the V:
    

12112 212
,max,, ,IppLVpp LVpp (4)
where,







 
112
12 12
12
12 12
,
2,2,
log, 0,2
2,2,
LVpp
Vpp VppVpp
Vpp Vpp


 
(5)
 




 
24
12 12
212
6
12
12
,,
,236
,,0, 2
288
Vpp Vpp
LVpp
Vpp Vpp


(6)
But there are not a general representation of the max-
imum value for direct differences I and differences J
based on V. So there exists some limitations when meas-
uring the difference between the probability distribution.
For generally descripting the maximum limit of the
difference, papers [1,9] gave the amended definition of I
and J:
 
 
1
12 1
12
11 2
,log
11
22
11
,22
xX
px
Kpppx px px
Ip pp








(7)
The corresponding symmetrical form difference was
as follows:

1212 21
,,LppKppKp p

,


(8)
It can be seen from Equation (8) that, the difference
between K and L was not only meeting the non-negativ-
ity but also having limitation and semi-bounded. That is:
 
 
1212 11
1212 11
,,, ,
,,, ,
K
ppKpp Kpp
LppLpp Lpp
 
  (9)
  
12
121 2
,2 2
pp
LppHHpHp




 (10)
Now setting 1

and 2

were the weight of two
probability distributions, and meeting
121 2
,0, 1

 , Jensen-Shannon Divergence
(JSD) was defined as:
 


1211221122
,
J
SppH ppHpHp

   
(11)
where
 
1
ln
N
j
j
j
H
Ppp

 is Shannon Entropy.
For more than two but a limited number of probability
distributions , the corresponding weight
were
12
,,,
n
pp p
..., n12 , ,

 respectively, JSD was defined as:

12
11
,,,
nn
niii
ii

i
J
SpppHpHp








 (12)
3. DATA ANALYSIS
The EEG data we used were taken from 12 students
and 12 nurses. The sampling frequency was 200 Hz. We
used JSD to analyze the complexity of the EEG data of
12 students and 12 nurses and the corresponding results
were shown in Table 1.
According to Table 1, we can plot the complexity
measures of three ECG signals and twelve were shown in
Figure 1.
It can be seen from the Figure 1 that, the Jensen-
Shannon Divergence value of students is less than that of
nurses. The T test value of the two groups was equal to
4.414 which confidence probability is 0.001. It indicated
that high level of occupational stress had a higher Jensen-
Shannon Divergence value than low level, and EEG
should be more complex. So students and nurses can be
statistically distinguished.
4. CONCLUSION
In this paper, the complexity analysis method based on
Jensen-Shannon Divergence was used to calculate the
complexity of occupational stress electroencephalogram
Table 1. Complexity measures of twelve ECG signals (2000
points).
Subjects Student Nurse
1 0.2163 0.2215
2 0.2175 0.2787
3 0.0599 0.2813
4 0.2092 0.2403
5 0.1929 0.2659
6 0.0694 0.2793
7 0.2171 0.2727
8 0.2358 0.2809
9 0.0692 0.2714
10 0.2145 0.2707
11 0.2055 0.2798
12 0.0621 0.2743
mean ± STD 0.1641 ± 0.0738 0.2681 ± 0.0184

H. E. Tian et al. / Occupational Diseases and Environmental Medicine 1 (2013) 1-3 3
[2] Lamberti, P .W. and Majtey, A.P. (2003) Non-logarithmic
Jensen-Shannon divergence. Physica A: Statistical Me-
chanics and Its Applications, 329, 81-90.
http://dx.doi.org/10.1016/S0378-4371(03)00566-1
Student Nurse
0
0.1
0.2
0.3
0.4
0.5
Complexity
Student
Nurse
[3] Hung, W.L. and Yang, M.S. (2008) On the J-divergence
of intuitionistic fuzzy sets with its application to pattern
recognition. Information Sciences, 178, 1641-1650.
http://dx.doi.org/10.1016/j.ins.2007.11.006
[4] Rosso, O.A., Larrondo, H.A., Martin, M.T. and Fuentes,
M.A. (2007) Distinguishing Noise from Chaos. Physical
review letters, 99, Article ID: 154102.
http://dx.doi.org/10.1103/PhysRevLett.99.154102
[5] Martin, M.T., Plastino, A. and Rosso, O.A. (2006) Gen-
eralized statistical complexity measures: Geometrical and
analytical properties. Physica A: Statistical Mechanics
and Its Applications, 369, 439-462.
http://dx.doi.org/10.1016/j.physa.2005.11.053
Figure 1. Dynamic range of two kind signals’
complexity (N = 2000) (mean ± STD). ○ (Student),
* (Nurse).
from students and nurses. The study found that the com-
plexity of nurses’ EEG was higher than that of students’
EEG. The result can be used to assisted clinical diagno-
sis.
[6] Bagaajav, A., Myagmarjav, S., Nanjid, K., Otgon, S. and
Chae, Y.M. (2011) Burnout and job stress among Mon-
golian doctors and nurses. Industrial Health, 49, 582-588.
http://dx.doi.org/10.2486/indhealth.MS1256
[7] Kikuchi, Y., Nakaya, M., Ikeda, M., Narita, K., Takeda,
M. and Nishi, M. (2010) Effort-reward imbalance and
depressive state in nurses. Occupational Medicine, 60,
231-233. http://dx.doi.org/10.1093/occmed/kqp167
5. ACKNOWLEDGEMENTS
This work was supported by the National Natural Science Founda-
tion of China (Grant Nos. 61271082, 61201029, 61102094), the Natural
Science Foundation of Jiangsu Province (Grant Nos. BK2011759,
BK2011565), the Social development Foundation-science and technol-
ogy support projects of Jiangsu province (Grant Nos. BE2011777), and
Foundation of Nanjing University of Posts and Telecommunications
(JG03212JX02, JG03210JX19, 2011XSG11).
[8] V
oltmer, E., Wingenfeld, K., Spahn, C., Driessen, M. and
Schulz, M. (2012) Work-related behaviour and ex-
perience patterns of nurses in different professional stages
and settings compared to physicians in Germany. Inter-
national Journal of Mental Health Nursing, 22, 180-189.
http://dx.doi.org/10.1111/j.1447-0349.2012.00855.x
[9] Kikuchi, Y., Nakaya, M., Ikeda, M., Takeda, M. and Nishi,
M. (2013) Job stress and temperaments in female nurses.
Occupational Medicine, 63, 123-128.
http://dx.doi.org/10.1093/occmed/kqs212
REFERENCES
[1] Lin, J.H. (1991) Divergence measure based on the
Shannon entropy. IEEE Transactions on Information
Theory, 37, 145-151. http://dx.doi.org/10.1109/18.61115
OPEN ACC ESS