Cluster Comput (2018) 21:681–690

https://doi.org/10.1007/s10586-017-0977-2

Wearable sensor devices for early detection of Alzheimer disease

using dynamic time warping algorithm
R. Varatharajan1 · Gunasekaran Manogaran2 · M. K. Priyan2 ·
Revathi Sundarasekar3

Received: 7 April 2017 / Revised: 31 May 2017 / Accepted: 5 June 2017 / Published online: 22 June 2017
© Springer Science+Business Media, LLC 2017

Abstract Alzheimer disease is a significant problem in 1 Introduction

public health. Alzheimer disease causes severe problems
with thinking, memory and activities. Alzheimer disease Recently, a report from The Hindu states that nearly 50
affected more on the people who are in the age group of lakhs of people are affected by Alzheimer disease in India
80-year-90. The foot movement monitoring system is used [1–4]. Human motion recognition is used to detect the
to detect the early stage of Alzheimer disease. internets of Alzheimer disease in early stage. In order to monitor the
things (IoT) devices are used in this paper to monitor the human motion, various internet of things (IoT) devices are
patients’ foot movement in continuous manner. This paper developed recently [5–7]. IoT is applied in many applications
uses dynamic time warping (DTW) algorithm to compare to get better results [8–10]. Though, implementation of above
the various shapes of foot movements collected from the system consists of various challenges and issues [11]. For
wearable IoT devices. The foot movements of the normal example, IoT devices are usually communicated with other
individuals and people who are affected by Alzheimer disease devices with the help of wireless network. Hence, there is a
are compared with the help of middle level cross identifi- need to improve the communication system using an efficient
cation (MidCross) function. The identified cross levels are service oriented-architecture (SoA) [12]. In addition, there is
used to classify the gait signal for Alzheimer disease diagno- a need of standardization to develop an efficient and effec-
sis. Sensitivity and specificity are calculated to evaluate the tive IoT system and solve the gap between the customer and
DTW algorithm based classification model for Alzheimer service providers [13]. IoT devices generally connect with
disease. The classification results generated using the DTW cloud; hence, there is need of effective integration platform
is compared with the various classification algorithms such between cloud and IoT system [12,14].
as inertial navigation algorithm, K-nearest neighbor classi- IoT generates huge amount of data, hence, there is need
fier and support vector machines. The experimental results of advance scalable algorithms to process such kind of data
proved the effectiveness of the DTW method. [15]. IoT devices communicate with each other, hence, there
is a need to remove multicast/ broadcast flooding [16]. Sim-
Keywords Internets of things · Alzheimer disease · ilarly, when a device roaming from one place to another; it
Dynamic time warping · Middle level cross identification · is mandatory to reduce the transmission latency between the
Inertial navigation algorithm · K-nearest neighbor classifier · source to destination [17]. IoT devices generates enormous
Support vector machines amount of data, hence, there is need to have scalable data stor-
age platform in cloud [18–21]. Moreover, IoT devices should
have an appropriate security mechanism to protect and pre-
B R. Varatharajan vent from unauthorized access and data loss [22]. IoT devices
varathu21@yahoo.com should support various types of modern networking protocols
1
such as Mobile Internet Protocol version 6 (MIPv6), Inter-
Sri Ramanujar Engineering College, Chennai, India
net Control Message Protocol version 6 (ICMPv6) and so
2 VIT University, Vellore, India on [23]. Similarly, a modern IoT system should support star
3 Priyadarshini Engineering College, Vellore, India and mesh topology to provide an efficient network commu-

123

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

682 Cluster Comput (2018) 21:681–690

nication between the devices [24]. Hence, the modern IoT complicated image segmentation. Two level segmentation
devices must be identified at time and location, hence, the approach is followed in this approach namely background
mobile address of the devices should be enhanced. In addi- and target [43]. The detected retinal layer boundaries are
tion, efficient handoff scheme is to be identified to avoid effectively utilized to identify the Alzheimer disease and
jitter, delays, and interruptions in Internet of Vehicles (IoV) Glaucoma. The experimental results effectively segment the
[25]. Moreover, some major issues are found in IoV. For input raw OCT image into five layers accurately. The pro-
example, lack of coordination and communication between posed algorithm initially uses factor analysis technique to
the vehicles in IoV implementation is considered as biggest categorize the well-known and unknown images generated
issue [26], and lack of standards in V2V (vehicle to vehicle) from various sources. Space mapping approach is employed
wireless communication is also considered as challenging in this work to perform the objective extraction. The pro-
task [27]. IoT systems continuously generate huge amount posed image feature extraction with recognition approach
of data, hence, making a decision on data is considered as effectively tested to detect the unfamiliar faces as well as
another challenge. Table 1 depicts the recent development in object recognition, facial expression. LDA, MFA and LPP
wearable sensors for motion recognition. Table 2 depicts the based image feature extraction methods are used in this work
various sensor devices and it uses. to compare and prove the effectiveness of the proposed image
The structure of this manuscript is described as follows: feature extraction with recognition approach. In addition, the
Sect. 1 introduces the motion detection using wearable sensor structure tensor with complex diffusion filtering based image
devices. Section 2 reviews the recent works in motion detec- filtering method is also used to process the OCT images. This
tion for Alzheimer disease. The proposed work is explained image filtering method efficiently removes the noises present
in Sect. 3 in detail. Sections 4 and 5 describes the result and in the raw OCT images. The structure tensor with complex
discussion, and performance evaluation. Finally, Sect. 6 con- diffusion filtering based image filtering method efficiently
cludes the research work. classifies the Alzheimer disease in public health. The detec-
tion of Alzheimer disease is done on the various cellular
layers of the OCT retina images using the STRATUSOCT
2 Related work system.

In the recent years, various image processing algorithms and

machine learning techniques are developed for disease diag- 3 Proposed framework
nosis of Alzheimer disease. For example, Rasta et al. have
reviewed various image enhancement approaches to detect Dynamic time warping (DTW) is used in this paper to mon-
the Alzheimer disease [28]. Corrected red and green com- itor the gait signal data from various patients [44,45]. The
ponents of color retinal images are used to evaluate the gait signal data is collected from various wearable sensor
existing approaches to detect the Alzheimer disease [29]. devices. These IoT devices continuously generate the huge
Sensitivity and specificity are calculated to find the best amount of data [46]. The gait signal data is collected and
approach to detect the Alzheimer disease in public health. processed with the help of DTW algorithm [47,48]. The
Li at al. have identified an approach to deal with complicated essential goal of the DTW algorithm is to compare the var-
image segmentation [26]. Two level segmentation approach ious shapes of gait signals collected from each patients of
is followed in this approach namely background and target Alzheimer disease and warping to align them in time [49].
[30]. The detected retinal layer boundaries are effectively In this paper, the gait signals are referred to the walking pat-
utilized to identify the Alzheimer disease and Glaucoma terns of patients with Alzheimer disease. The time between
[31,32]. Rao et al. have used Linear mixed modeling meth- strides has been reported to differ between the normal indi-
ods to model the influences of axial length, age, TSS, and viduals and the patients who are having Alzheimer disease.
corneal birefringence on Alzheimer disease detection [33]. The major issue with this operation is walking speed of the
The proposed approach randomly selects one image out of individuals over time [50,51]. In general, every individuals
48 eye images to evaluate the performance of the proposed walk with different speed over time. Hence, there is a need to
approach. align them in time. In order to overcome this issue, this paper
Uji et al. have proposed interpolation and super-resolution uses DTW algorithm to compare the various shapes of gait
(SR) algorithms to identify the Alzheimer disease, peak signals collected from normal individuals and people who
signal-to-noise ratio (PSNR), photoreceptor layer status, and are affected by Alzheimer disease. The comparison is done
parallelism in each Oct images [42]. Experimental results with the help of middle level cross identification function
proved the efficiency of the proposed interpolation and [52,53]. This cross levels is used in gait signal classification
super-resolution (SR) algorithms to indentify the Alzheimer for patients who are having Alzheimer disease. Nowadays,
disease. Li at al. have identified an approach to deal with various big data algorithms also play a significant role in

123

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

 Table 1 Recent development in wearable sensors for motion recognition
CT1/RT3 GT3X/GTM AMP 331 IDEEA Step watch Active PAL Sense wear
Cluster Comput (2018) 21:681–690

Monitoring parameters Motion Meteorological Walking speed, Motion types, Steps gait Stepping time, Sleep length,
concentration, motion counts, Steps, beat, gait types features cadence, Motion length
meteorological steps, activity stride length, sedentary and
concentration distance upright time,
level steps,
sit-to-stand
activities,
meteorological
Sensor size in mm 71 × 56 × 28 38 × 37 × 18 71.3 × 24 × 37.5 70 × 54 × 17 75 × 50 × 20 53 × 35 × 7 88.4 × 56.4 ×
24.1
Weight in grams 71.5 27 50 59 38 20 82.2
Sampling rate 0.017–1 Hz 30 Hz Not applicable 32 Hz 128 Hz 10 Hz 32 Hz
No. of accelerometer One One Two 5 One One One
No. of accelerometer axis 1/3 3/1 1 single axis and Two Two One Two
1 double axis
Sensor position Waist Wrist Ankle Feet, chest, thigh Ankle Thigh Upper arm
Data storage 3 hours to 21 40 days Not applicable 7 days 2 months Not applicable Not applicable
days
Type of the battery 1.5V AAA × 1 3.7 V Lithium Not applicable 1 1.5 V AA 750 mAh 3V 1.5 V AAA × 1
Lithium
Accelerometer type Piezoelectric Not applicable Not applicable Piezoelectric Not applicable Piezoresistive Not applicable
Battery life 30 days 20 days Not applicable 60 hours Not applicable 7–10 days 3 days
Data transmission USB USB 916 MHz RF USB USB USB RF/USB
Sensitivity rate Not applicable 0.05–2.5 g Not applicable 5g Not applicable 2g 2g

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

123
683
684 Cluster Comput (2018) 21:681–690

Table 2 Various sensor devices

Classified event Sensor data Classification approach References
and it uses
Fall detection Barometer, microphone, Threshold based kNN [34]
accelerometers algorithm
Fall risk estimation and GaitShoe: accelerometer, SVM, Gaussian process [35]
gait assessment gyroscope, bend sensor, force
sensitive resistor and electric
field sensor
Food preparation and RFID Threshold based [36]
feeding algorithm
Selfcare RFID, accelerometer Proprietary algorithm [37]
House keeping RFID, accelerometer Proprietary algorithm [38]
Activities of daily living Accelerometers k-NN with Gaussian [39]
process
Leisure and EOG, accelerometer, RFID SVM [40]
communication
Energy expenditure Accelerometers Regression model [41]

Sensor (IoT tion. In this paper, DTW algorithm is used to classify the gait
Filter (Signal Segmentaon
moon
preprocessing) signals collected from normal individuals and people who are
detector) affected by Alzheimer disease. In general, speech recognition
features are more similar to the feature of gesture recognition.
DTW algorithm is significantly used to measure the various
Movement Feature Extracon similarity measures in the multivariate time-series data.
detecon (DTW algorithm)
In this paper two gestures such as normal individuals and
people who are affected by Alzheimer disease are to be com-
Fig. 1 Architecture for motion detection using wearable sensor
devices
pared against each other as, two time series X and Y.

X = (x1 , x2 , x3 , . . . xt1 , . . . , x T 1 , ) (1)

disease diagnosis [18,54–58]. The IoT based gait signal mon- Y = (y1 , y2 , y3 , . . . yt1, , . . . , yT 1 , ) (2)
itoring framework discussed in this paper is also used to
monitor the various other physiological signals such as elec- The time series X = (x1 , x2 , x3 , . . . xt1 , . . . , x T 1 ) and Y =
trocardiogram (ECG) and photoplethysmogram (PPG) and (y1 , y2 , y3 , . . . yt1, , . . . , yT 1 ) are considered as multivari-
so on. Figure 1 represents the architecture for motion detec- ate series with huge feature vectors [59–61]. The distance
tion using wearable sensor devices. Figure 2 represents the between the vectors of the time series X = (x1 , x2 , x3 , . . .
example motion detection device and Fig. 3 represents the xt1 , . . . , x T 1 ) and Y = (y1 , y2 , y3 , . . . yt1, , . . . , yT 1 ) are is
Leg movement. defined by,

3.1 Dynamic time warping d: f × f →R>0 (3)

Dynamic time warping is widely used to classify the multi- where,

variate time series data. DTW algorithm is more often used to
classify the speech recognition and hand writing classifica- xt1 , yt2 ∈ f for t1 ∈ [1, t1 ], t2 ∈ [1, t2 ] (4)

Fig. 2 Motion detection device

123

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Cluster Comput (2018) 21:681–690 685

Fig. 3 Leg movement

The above equations state that, the cost measure must be The overall distance function is used measure the overall dis-
small if the time series X = (x1 , x2 , x3 , . . . xt1 , . . . , x T 1 ) tance between the two foot movements X = (x1 , x2 , x3 , . . .
andY = (y1 , y2 , y3 , . . . yt1, , . . . , yT 1 ) are similar and high xt1 , . . . , x T 1 ) and Y = (y1 , y2 , y3 , . . . yt1, , . . . , yT 1 )
if they are very different. The DTW algorithm is to compute the lowest dis-
The cost matrix Ct1 × Ct2 is calculated for the time series tance measure between the two foot movements X =
X = (x1 , x2 , x3 , . . . xt1 , . . . , x T 1 ) and Y = (y1 , y2 , y3 , . . . (x1 , x2 , x3 , . . . xt1 , . . . , x T 1 ) and Y = (y1 , y2 , y3 , . . . yt1, ,
yt1, , . . . , yT 1 ). The Ct1 × Ct2 is used to attain a association . . . , yT 1 ).
mapping elements in X = (x1 , x2 , x3 , . . . xt1 , . . . , x T 1 , ) to The dynamic programming theory is used in this paper to
elements in Y = (y1 , y2 , y3 , . . . yt1, , . . . , yT 1 , ). The map- calculate the distance to each c(k).
ping function is used to calculate the lowest distance measure We define D as the accumulated cost matrix:
between X and Y. The mapping function is defined by, Step 1 Initialize the distance D(1, 1) = d(x1 , y1 )
Step 2 Initialize the distance D(T1 , T2 ) = 2 (Choose n as
F = c(1), c(2), c(3), . . . , c(k), . . . , c(K ) (5) maximum arbitrary number)
Step 3 Calculate D(t1 , t2 ) = min{D(t1−1 , t2−1 ), D(t1−1 ,
where, t2 ), D(t1 , t2−1 )} + d(X t1 , Yt2 )
The distance is calculated based on the following methods
c(k) = c(xk , yk ) (6) The Euclidean distance between the two foot movement
X = (x1 , x2 , x3 , . . . xt1 , . . . , x T 1 , ) and Y = (y1 , y2 , y3 , . . .
The time sequence order of the respective foot movement
yt1, , . . . , yT 1 , ) is calculated as follows:
is mapped into the mapping function. In order to achieve
this task, the following conditions are implemented in the 
 K
proposed framework:     
Step 1 The observation symbols for the initial state and end dmn (X, Y ) =  x k,m − yk,n × xk,m − yk,n (12)
state are aligned as follows: k=1

c (1) = (x1 , y1 ) (7) The absolute distance between the two foot movement X =
c (K ) = (x T 1 , yT 2 ) (8) (x1 , x2 , x3 , . . . xt1 , . . . , x T 1 , ) and Y = (y1 , y2 , y3 , . . . yt1, ,
. . . , yT 1 , ) is calculated as follows:
Step 2 The observation symbols are aligned as increasing
order. The order of observation symbols is defined by, 
K
 
dmn (X, Y ) = xk,m − yk,n 
k1 ≤ k2 ≤ · · · ≤ K (9) k=1

 K
    
Step 3 The observation symbols should not be skipped = xk,m − yk,n × xk,m − yk,n (13)
k=1
ki+1 − ki ≤ 1 (10)

The overall distance function C(F) for the time series The squared distance between the two foot movement X =
X = (x1 , x2 , x3 , . . . xt1 , . . . , x T 1 ) and Y = (y1 , y2 , y3 , . . . (x1 , x2 , x3 , . . . xt1 , . . . , x T 1 , ) and Y = (y1 , y2 , y3 , . . . yt1, ,
yt1, , . . . , yT 1 ) is defined by, . . . , yT 1 , ) is calculated as follows:


K 
K
   
C(F) = c(k) (11) dmn (X, Y ) = xk,m − yk,n × xk,m − yk,n (14)
k=1 k=1

123

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

686 Cluster Comput (2018) 21:681–690

The symmetric kullback leibler metric for the real and posi- changes in the foot movement are identified with the help of
tive foot movement X = (x1 , x2 , x3 , . . . xt1 , . . . , x T 1 , ) and middle level cross function. Figure 6 represents the middle
Y = (y1 , y2 , y3 , . . . yt1, , . . . , yT 1 , ) is calculated as follows: level cross identification of patients who have Alzheimer dis-
ease. The sharp changes in the foot movement of the patients

K
   who have Alzheimer disease is compared with the group of
dmn (X, Y ) = xk,m − yk,n log xk,m − log yk,n (15) ten patients. The comparison results are depicted in Fig. 7.
k=1 In general, patients do not walk at the same rate throughout
the record. Hence, there is an issue in comparing the foot
Dynamic Time Warping Distance Calculation Algorithm
Input: Time series input x and y with the length n and m
Output: Distance between the variable x and y
1. int Dynamic Time Warping Distance(s: array [1..n], t: array [1..m])
2. {
3. Dynamic Time Warping Distance:= array [0 to n, 0 to m]
4. for i := 1 to n
5. Dynamic Time Warping Distance [i, 0] := [0 to ∞ ]
6. for i := 1 to m
7. Dynamic Time Warping Distance [0, i] := [0 to ∞ ]
8. Dynamic Time Warping Distance [0, 0] := 0
9. for i := 1 to n
10. for j := 1 to m
11. cost := d(s[i], t[j])
12. Dynamic Time Warping Distance [i, j] := cost +
minimum(DTW[i-1, j ], // insertion operation
13. Dynamic Time Warping Distance [i , j-1], // deletion
operation
14. Dynamic Time Warping Distance [i-1, j-1]) // match
operation
15. return DTW[n, m]
16. }

4 Result and discussion movement of different patients. In order to overcome this

issue, DTW algorithm is significantly used to compute the
The walking patterns of the individuals who are affected by distance between the segments by warping them and align
Alzheimer disease is collected with the help of wearable IoT them in time. Figure 8 represents the signal alignment of
devices. The walking speed of the individuals over time is foot movement using DTW algorithm. Figure 9 represents the
analyzed with the help of DTW algorithm. In general, every classification result of Alzheimer disease diagnosis based on
individuals walk with different speed over time. Hence, there the foot movement change detection using DTW algorithm.
is a need to align them in time. This paper uses DTW algo-
rithm to compare the various shapes of gait signals collected
from the wearable IoT devices. The foot patterns of the nor-
mal individuals and people who are affected by Alzheimer
disease are compared with the help of middle level cross
identification function. The identified cross levels are used
to classify the gait signal for Alzheimer disease diagnosis.
Force sensitive resistor is placed on the foot of the patient to
observe the force in mill volts (mV). The records are stored as
one minute interval for the left and right foot. Figure 4 repre-
sents the nfoot movement of the patient who have Alzheimer
disease. Every step movement of the patient is continuously
monitored with the help of IoT devices. An example middle
level cross identification is represented in Fig. 5. The sharp Fig. 4 Foot movement of the patient who have Alzheimer disease

123

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Cluster Comput (2018) 21:681–690 687

Fig. 5 Example middle level cross identification Fig. 8 Signal alignment of foot movement using DTW algorithm

Fig. 9 Classification result of Alzheimer disease diagnosis

Fig. 6 Middle level cross identification of patient who have Alzheimer
disease

Fig. 10 ROC analysis for the DTW algorithm

Fig. 7 Comparison of patients’ foot movement with the normal indi-
viduals

5 Performance evaluation

Sensitivity and specificity are calculated to evaluate the clas-

sification model for Alzheimer disease. The classification
generated by the dynamic time warping (DTW) is compared
with the various classification algorithms such as inertial
navigation algorithm (INA), K-nearest neighbor (k-NN) clas-
sifier and support vector machines (SVM). The validations
metric are defined by,

T r ue N egative (T N )
Speci f icit y =
False Positive (F P) + T r ue N egative (T N )
T r ue Positive (T P)
Sensitivit y =
T r ue Positive (T P) + False N egative (F N ) Fig. 11 Comparison of classification rate

123

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

688 Cluster Comput (2018) 21:681–690

Table 3 Performance comparison of the dynamic time warping (DTW) with various classification methods for Alzheimer disease diagnosis
S.No Method Disease No. of Predicted as Predicted as SN = sensitivity (%),
type patients abnormal normal SP = specificity (%)

1 Dynamic time warping (DTW) Abnormal 173 170 13 SN = 95.9

Normal 150 147 13 SP = 94
2 Inertial navigation algorithm (INA) Abnormal 173 169 14 SN = 94.5
Normal 150 145 15 SP = 90
3 K-nearest neighbor (K-NN) classifier Abnormal 173 170 13 SN = 95.9
Normal 150 147 13 SP = 94
4 Support vector machines (SVM) Abnormal 173 171 12 SN = 97.3
Normal 150 146 14 SP = 92

Figures 10 and 11 represent the ROC analysis for the support vector machines (SVM). The experimental results
DTW algorithm and the comparison of classification rate proved the effectiveness of the dynamic time warping (DTW)
respectively. Table 3 depicts the performance comparison of method. In this study, we have observed only the foot move-
the dynamic time warping (DTW) with various classification ment from patient. The future work of this study is to use
methods for Alzheimer disease diagnosis. The experimental various IoT devices to collect various physiological signals
results proved the effectiveness of the dynamic time warping from patient. The physiological signals are used to detect
(DTW) method. early stage of various diseases.

6 Conclusion
References
The human motion recognition is used to detect the early 1. The Hindu: 12. http://www.thehindu.com/news/cities/Hyderabad/
stage of Alzheimer disease. In this paper, IoT devices are many-unaware-of-alzheimers-disease-in-india/article5390719.
used to monitor the human motion in continuous manner. ece. Accessed 27 Jan 2017 (2017)
The walking patterns of the individuals who are affected by 2. Chandra, V., Pandav, R., Dodge, H.H., Johnston, J.M., Belle, S.H.,
DeKosky, S.T., Ganguli, M.: Incidence of Alzheimer’s disease in
Alzheimer disease is collected with the help of wearable IoT
a rural community in India The Indo-US Study. Neurology 57(6),
devices. The walking speed of the individuals over time is 985–989 (2001)
analyzed with the help of DTW algorithm. This paper uses 3. Chandra, V., Ganguli, M., Pandav, R., Johnston, J., Belle, S.,
DTW algorithm to compare the various shapes of gait signals DeKosky, S.T.: Prevalence of Alzheimer’s disease and other
dementias in rural India The Indo-US study. Neurology 51(4),
collected from the wearable IoT devices. The foot patterns
1000–1008 (1998)
of the normal individuals and people who are affected by 4. Pandav, R.S., Chandra, V., Dodge, H.H., DeKosky, S.T., Ganguli,
Alzheimer disease are compared with the help of middle level M.: Hemoglobin levels and Alzheimer disease: an epidemiologic
cross identification function. The identified cross levels are study in India. Am. J. Geriatr. Psychiatry 12(5), 523–526 (2004)
5. Atzori, L., Iera, A., Morabito, G.: The internet of things: a survey.
used to classify the gait signal for Alzheimer disease diag- Comput. Netw. 54(15), 2787–2805 (2010)
nosis. Force sensitive resistor is placed on the foot of the 6. Zhu, N., Diethe, T., Camplani, M., Tao, L., Burrows, A., Twomey,
patient to observe the force in mill volts (mV). The sharp N., Kaleshi, D., Mirmehdi, M., Flach, P., Craddock, I.: Bridging
changes in the foot movement are identified with the help e-health and the internet of things: the sphere project. IEEE Intell.
Syst. 30(4), 39–46 (2015)
of middle level cross function. In general, patients do not 7. Manogaran, G., Thota, C., Lopez, D., Vijayakumar, V., Abbas,
walk at the same rate throughout the record. Hence, there K.M., Sundarsekar, R.: Big Data Knowledge System in Health-
is an issue in comparing the foot movement of different care. InInternet of Things and Big Data Technologies for Next
patients. In order to overcome this issue, DTW algorithm is Generation Healthcare, pp. 133–157. Springer, Berlin (2017)
8. Tsai, C.W., Lai, C.F., Chiang, M.C., Yang, L.T.: Data mining for
significantly used to compute the distance between the seg- Internet of Things: A survey. IEEE Commun. Surv. Tutor. 16(1),
ments by warping them and align them in time. Sensitivity 77–97 (2014)
and specificity are calculated to evaluate the classification 9. Agrawal, S., Das, M.L.: Internet of things—a paradigm shift of
model for Alzheimer disease. The classification generated future internet applications. In: 2011 Nirma University Interna-
tional Conference on Engineering (NUiCONE), pp. 1–7. IEEE, 8
by the dynamic time warping (DTW) is compared with the Dec 2011
various classification algorithms such as inertial navigation 10. Alam, S., Chowdhury, M.M., Noll, J.: Senaas: an event-driven sen-
algorithm (INA), K-nearest neighbor (k-NN) classifier and sor virtualization approach for internet of things cloud. In: 2010

123

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Cluster Comput (2018) 21:681–690 689

IEEE International Conference on Networked Embedded Systems 31. Patel, A.R., Patel, R.S., Singh, N.M., Kazi, F.S.: Vitality of Robotics
for Enterprise Applications (NESEA), pp. 1–6. IEEE, 25 Nov 2010 in Healthcare Industry: an internet of things (IoT) perspective. In:
11. Pasluosta, C.F., Gassner, H., Winkler, J., Klucken, J., Eskofier, Internet of Things and Big Data Technologies for Next Generation
B.M.: An emerging era in the management of Parkinson’s disease: Healthcare, pp. 91–109. Springer, Berlin (2017)
wearable technologies and the internet of things. IEEE J. Biomed. 32. Memon, M., Wagner, S.R., Pedersen, C.F., Beevi, F.H., Hansen,
Health Inform. 19(6), 1873–1881 (2015) F.O.: Ambient assisted living healthcare frameworks, platforms,
12. Gubbi, J., Buyya, R., Marusic, S., Palaniswami, M.: Internet of standards, and quality attributes. Sensors 14(3), 4312–4341 (2014)
things (IoT): a vision, architectural elements, and future directions. 33. Zhang, X., Francis, B.A., Dastiridou, A., Chopra, V., Tan, O.,
Future Gener. Comput. Syst. 29(7), 1645–1660 (2013) Varma, R., Greenfield, D.S., Schuman, J.S., Huang, D., Advanced
13. Botta, A., De Donato, W., Persico, V., Pescapé, A.: On the Imaging for Glaucoma Study Group: Longitudinal and cross-
integration of cloud computing and internet of things. In: 2014 sectional analyses of age effects on retinal nerve fiber layer and
International Conference on Future Internet of Things and Cloud ganglion cell complex thickness by Fourier-domain OCT. Transl.
(FiCloud), pp. 23–30. IEEE, 27 Aug 2014 Vis. Sci. Technol. 5(2):1 (2016)
14. Barnaghi, P., Wang, W., Henson, C., Taylor, K.: Semantics for the 34. Bianchi, F., Redmond, S.J., Narayanan, M.R., Cerutti, S., Lovell,
internet of things: early progress and back to the future. Int. J. N.H.: Barometric pressure and triaxial accelerometry-based falls
Semant. Web Inf. Syst. (IJSWIS) 8(1), 1–21 (2012) event detection. IEEE Trans. Neural Syst. Rehabil. Eng. 18(6),
15. Da Xu, L., He, W., Li, S.: Internet of things in industries: a survey. 619–627 (2010)
IEEE Trans. Ind. Inform. 10(4), 2233–2243 (2014) 35. Bamberg, S.J., Benbasat, A.Y., Scarborough, D.M., Krebs, D.E.,
16. Wang, Y.P., Lin, X., Adhikary, A., Grovlen, A., Sui, Y., Blanken- Paradiso, J.A.: Gait analysis using a shoe-integrated wireless sensor
ship, Y., Bergman, J., Razaghi, H.S.: A primer on 3GPP narrowband system. IEEE Trans. Inf. Technol. Biomed. 12(4), 413–423 (2008)
internet of things. IEEE Commun. Mag. 55(3), 117–123 (2017) 36. Patterson, D.J., Fox, D., Kautz, H., Philipose, M.: Fine-grained
17. Stojkoska, B.L., Trivodaliev, K.V.: A review of internet of things activity recognition by aggregating abstract object usage. In: Ninth
for smart home: challenges and solutions. J. Clean. Prod. 1(140), IEEE International Symposium on Wearable Computers, 2005.
1454–1464 (2017) Proceedings, pp. 44–51). IEEE, 18 Oct 2005
18. Manogaran, G., Thota, C., Kumar, M.V.: MetaCloudDataStorage 37. Im, S., Kim, I.J., Ahn, S.C., Kim, H.G.: Automatic ADL classi-
architecture for big data security in cloud computing. Procedia fication using 3-axial accelerometers and RFID sensor. In: IEEE
Comput. Sci. 31(87), 128–133 (2016) International Conference on Multisensor Fusion and Integration
19. Manogaran, G., Lopez, D.: Spatial cumulative sum algorithm with for Intelligent Systems, 2008. MFI 2008, pp. 697–702. IEEE, 20
big data analytics for climate change detection. Comput. Electr. Aug 2008
Eng. 59(1), 1–25 (2017) 38. Farringdon, J., Moore, A.J., Tilbury, N., Church, J., Biemond, P.D.:
20. Lopez, D., Manogaran, G.: Modelling the H1N1 influenza using Wearable sensor badge and sensor jacket for context awareness.
mathematical and neural network approaches. Biomed. Res. 28(8), In: The Third International Symposium on Wearable Computers,
1–5 (2017) 1999. Digest of Papers, pp. 107–113. IEEE, 18 Oct 1999
21. Manogaran, G., Thota, C., Lopez, D., Sundarasekar, R.: Big data 39. Atallah, L., Lo, B., King, R., Yang, G.Z.: Sensor placement for
security intelligence for Healthcare Industry 4.0. In: Cybersecurity activity detection using wearable accelerometers. In: 2010 Inter-
for Industry 4.0, pp. 103–126. Springer, Berlin (2017) national Conference on Body Sensor Networks (BSN), pp. 24–29.
22. Mulani, T.T., Pingle, S.V.: Internet of things. Int. Res. J. Multidis- IEEE, 7 June 2010
cip. Stud. 2(3) (2016) 40. Bulling, A., Ward, J.A., Gellersen, H., Troster, G.: Eye movement
23. Botta, A., De Donato, W., Persico, V., Pescapé, A.: Integration of analysis for activity recognition using electrooculography. IEEE
cloud computing and internet of things: a survey. Future Gener. Trans. Pattern Anal. Mach. Intell. 33(4), 741–753 (2011)
Comput. Syst. 31(56), 684–700 (2016) 41. Crouter, S.E., Clowers, K.G., Bassett, D.R.: A novel method for
24. Bello, O., Zeadally, S.: Intelligent device-to-device communication using accelerometer data to predict energy expenditure. J. Appl.
in the internet of things. IEEE Syst. J. 10(3), 1172–1182 (2016) Physiol. 100(4), 1324–1331 (2006)
25. Liu, J., Wan, J., Wang, Q., Deng, P., Zhou, K., Qiao, Y.: A survey 42. Uji, A., Abdelfattah, N.S., Boyer, D.S., Balasubramanian, S., Lei,
on position-based routing for vehicular ad hoc networks. Telecom- J., Sadda, S.R.: Variability of retinal thickness measurements in
mun. Syst. 62(1), 15–30 (2016) tilted or stretched optical coherence tomography images. Transl.
26. Li, G., Ma, M., Liu, C., Shu, Y.: Routing in taxi and public transport Vis. Sci. Technol. 6(2):1 (2017)
based heterogeneous vehicular networks. In: Region 10 Conference 43. Li, C., Wang, X., Eberl, S., Fulham, M., Feng, D.: A new energy
(TENCON), 2016 IEEE, pp. 1863–1866. IEEE, 22 Nov 2016 framework with distribution descriptors for image segmentation.
27. Dinesh, M., Sudhaman, K.: Real time intelligent image processing IEEE Trans. Image Process. 22(9), 3578–3590 (2013)
system with high speed secured internet of things: image processor 44. Manogaran, G., Lopez, D., Thota, C., Abbas, K. M., Pyne, S.,
with IOT. In: 2016 International Conference on Information Com- Sundarasekar, R.: Big data analytics in healthcare internet of things.
munication and Embedded Systems (ICICES), pp. 1–5. IEEE, 25 In: Innovative Healthcare Systems for the 21st Century, pp. 263–
Feb 2016 284. Springer, Berlin (2017)
28. Kaur, A., Kaur, P.: A comparative study of various exudate seg- 45. Thota, C., Sundarasekar, R., Manogaran, G., Varatharajan, R.,
mentation techniques for diagnosis of diabetic retinopathy. Int. J. Priyan, M.K: Centralized fog computing security platform for iot
Curr. Eng. Technol. 46(1), 142–146 (2016) and cloud in healthcare system. In: Exploring the Convergence of
29. Jeyabalan, K.: Home Healthcare and Remote Patient Monitoring. Big Data and the Internet of Things. IGI Global, USA
Internet of Things and Data Analytics Handbook, pp. 675–682 46. Yang, C.C., Hsu, Y.L.: A review of accelerometry-based wearable
(2017) motion detectors for physical activity monitoring. Sensors 10(8),
30. Hiremath, S., Yang, G., Mankodiya, K.: Wearable internet of 7772–7788 (2010)
things: concept, architectural components and promises for person- 47. Manogaran, G., Lopez, D.: Health data analytics using scalable
centered healthcare. In: 2014 EAI 4th International Conference on logistic regression with stochastic gradient descent. Int. J. Adv.
Wireless Mobile Communication and Healthcare (Mobihealth), pp. Intell. Paradig. 9(1), 1–18 (2016)
304–307. IEEE, 3 Nov 2014 48. Thota, C., Manogaran, G., Lopez, D.: Architecture for big data stor-
age in different cloud deployment models. In: Segall, R.S., Cook,

123

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

690 Cluster Comput (2018) 21:681–690

J.S., Gupta, N. (eds.) Big Data Storage and Visualization Tech- Gunasekaran Manogaran is
niques. IGI Global currently pursuing Ph.D. in the
49. Sempena, S., Maulidevi, N.U., Aryan, P.R.: Human action recog- Vellore Institute of Technology
nition using dynamic time warping. In: 2011 International Confer- University. He received his Bach-
ence on Electrical Engineering and Informatics (ICEEI), pp. 1–5. elor of Engineering and Master
IEEE, 17 July 2011 of Technology from Anna Uni-
50. Salvador, S., Chan, P.: Toward accurate dynamic time warping in versity and Vellore Institute of
linear time and space. Intell. Data Anal. 11(5), 561–580 (2007) Technology University respec-
51. Baumann, M., Ozdogan, M., Richardson, A.D., Radeloff, V.C.: tively. He has worked as a
Phenology from Landsat when data is scarce: using MODIS and Research Assistant for a project
dynamic time-warping to combine multi-year Landsat imagery to on spatial data mining funded
derive annual phenology curves. Int. J. Appl. Earth Obs. Geoinf. by Indian Council of Medical
28(54), 72–83 (2017) Research, Government of India.
52. Zhang, Z., Tavenard, R., Bailly, A., Tang, X., Tang, P., Corpetti, His current research interests
T.: Dynamic time warping under limited warping path length. Inf. include data mining, big data
Sci. 31(393), 91–107 (2017) analytics and soft computing. He is the author/co-author of papers in
53. Wan, Y., Chen, X.L., Shi, Y.: Adaptive cost dynamic time warping conferences, book chapters and journals. He got an award for young
distance in time series analysis for classification. J. Comput. Appl. investigator from India and Southeast Asia by Bill and Melinda Gates
Math. 1(319), 514–520 (2017) Foundation. He is a member of International Society for Infectious Dis-
54. Lopez, D., Gunasekaran, M., Murugan, B.S., Kaur, H., Abbas, eases and Machine Intelligence Research labs.
K.M.: Spatial big data analytics of influenza epidemic in Vellore,
India. In: 2014 IEEE International Conference on Big Data (Big
Data), pp. 19–24. IEEE, 27 Oct 2014
M. K. Priyan is currently pur-
55. Lopez, D., Gunasekaran, M.: Assessment of vaccination strategies
suing a Ph.D. in the Vellore
using fuzzy multi-criteria decision making. In: Proceedings of the
Institute of Technology Univer-
Fifth International Conference on Fuzzy and Neuro Computing
sity. He received my Bachelor of
(FANCCO-2015) 2015, pp. 195–208. Springer, Berlin
Engineering and Master of Engi-
56. Lopez, D., Sekaran, G.: Climate change and disease dynamics-A
neering degree from Anna Uni-
big data perspective. Int. J. Infect. Dis. 1(45), 23–24 (2016)
versity and Vellore Institute of
57. Lopez, D., Manogaran, G.: Big data architecture for climate change
Technology University, respec-
and disease dynamics. The Human Element of Big Data: Issues,
tively. His current research inter-
Analytics, and Performance. CRC Press, Boca Raton (2016)
ests include Big Data Analyt-
58. Manogaran, G., Lopez, D.: Disease surveillance system for big
ics, Internet of Things, Inter-
climate data processing and dengue transmission. Int. J. Ambient
net of Everything, Internet of
Comput. Intell. 8(2), 88–105 (2017)
Vehicles in Healthcare. He is
59. Wen, J., Chang, X.W.: Success probability of the Babai estimators
the author/co-author of papers in
for box-constrained integer linear models. IEEE Trans. Inf. Theory
international journals and con-
63(1), 631–648 (2017)
ferences.
60. Cai, T.T., Wang, L.: Orthogonal matching pursuit for sparse signal
recovery with noise. IEEE Trans. Inf. Theory 57(7), 4680–4688
(2011)
61. Wen, J., Li, D., Zhu, F.: Stable recovery of sparse signals via Revathi Sundarasekar is cur-
lp-minimization. Appl. Comput. Harmonic Anal. 38(1), 161–176 rently pursuing a Master of Com-
(2015) puter Science and Engineering
in the Priyadarshini Engineer-
R. Varatharajan received his ing College, Vellore, Tamil Nadu,
B.E., M.E. and Ph.D. degrees all India. She received her Bach-
in Electronics and Communica- elor of Engineering in Com-
tion Engineering from Anna Uni- puter Science from Priyadarshini
versity and Bharath University, Engineering College. She is
India. His main area of research the author/co-author of papers
activity is Medical Image pro- in international journals, book
cessing, Wireless Networks and chapters and conferences. Her
VLSI Physical Design. He has current research interests include
served as a reviewer for Springer, Big Data analytics, and Internet
Inderscience and Elsevier jour- of Things.
nals. He has published many
research articles in refereed jour-
nals. He is a member of IEEE,
IACSIT, IAENG, SCIEI and
ISTE wireless research group. He has been serving as Organizing Chair
and Program Chair of several International conferences and in the Pro-
gram Committees of several International conferences. Currently he is
working as a Associate professor in the Department of Electronics and
Communication Engineering at Sri Ramanujar Engineering College,
Chennai, India.

123

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Terms and Conditions
Springer Nature journal content, brought to you courtesy of Springer Nature Customer Service Center GmbH (“Springer Nature”).
Springer Nature supports a reasonable amount of sharing of research papers by authors, subscribers and authorised users (“Users”), for small-
scale personal, non-commercial use provided that all copyright, trade and service marks and other proprietary notices are maintained. By
accessing, sharing, receiving or otherwise using the Springer Nature journal content you agree to these terms of use (“Terms”). For these
purposes, Springer Nature considers academic use (by researchers and students) to be non-commercial.
These Terms are supplementary and will apply in addition to any applicable website terms and conditions, a relevant site licence or a personal
subscription. These Terms will prevail over any conflict or ambiguity with regards to the relevant terms, a site licence or a personal subscription
(to the extent of the conflict or ambiguity only). For Creative Commons-licensed articles, the terms of the Creative Commons license used will
apply.
We collect and use personal data to provide access to the Springer Nature journal content. We may also use these personal data internally within
ResearchGate and Springer Nature and as agreed share it, in an anonymised way, for purposes of tracking, analysis and reporting. We will not
otherwise disclose your personal data outside the ResearchGate or the Springer Nature group of companies unless we have your permission as
detailed in the Privacy Policy.
While Users may use the Springer Nature journal content for small scale, personal non-commercial use, it is important to note that Users may
not:

1. use such content for the purpose of providing other users with access on a regular or large scale basis or as a means to circumvent access
control;
2. use such content where to do so would be considered a criminal or statutory offence in any jurisdiction, or gives rise to civil liability, or is
otherwise unlawful;
3. falsely or misleadingly imply or suggest endorsement, approval , sponsorship, or association unless explicitly agreed to by Springer Nature in
writing;
4. use bots or other automated methods to access the content or redirect messages
5. override any security feature or exclusionary protocol; or
6. share the content in order to create substitute for Springer Nature products or services or a systematic database of Springer Nature journal
content.
In line with the restriction against commercial use, Springer Nature does not permit the creation of a product or service that creates revenue,
royalties, rent or income from our content or its inclusion as part of a paid for service or for other commercial gain. Springer Nature journal
content cannot be used for inter-library loans and librarians may not upload Springer Nature journal content on a large scale into their, or any
other, institutional repository.
These terms of use are reviewed regularly and may be amended at any time. Springer Nature is not obligated to publish any information or
content on this website and may remove it or features or functionality at our sole discretion, at any time with or without notice. Springer Nature
may revoke this licence to you at any time and remove access to any copies of the Springer Nature journal content which have been saved.
To the fullest extent permitted by law, Springer Nature makes no warranties, representations or guarantees to Users, either express or implied
with respect to the Springer nature journal content and all parties disclaim and waive any implied warranties or warranties imposed by law,
including merchantability or fitness for any particular purpose.
Please note that these rights do not automatically extend to content, data or other material published by Springer Nature that may be licensed
from third parties.
If you would like to use or distribute our Springer Nature journal content to a wider audience or on a regular basis or in any other manner not
expressly permitted by these Terms, please contact Springer Nature at

onlineservice@springernature.com