Performance Evaluation of the Dependable Properties

of a Body Area Wireless Sensor Network

Venki Balasubramanian and Andrew Stranieri

Centre for Informatics and Applied Optimisation
School of Science, Information Technology and Engineering
University of Ballarat,
Ballarat, Australia
{v.balasubramanian, a.stranieri}@ballarat.edu.au

Abstract—Body Area Wireless Sensor Networks (BAWSNs) nations already plagued by the high-rising costs of healthcare
are self-organizing networks capable of monitoring health owing to the large number of chronic patients are further
intrinsic data of a patient. BAWSNs extended with a healthcare challenged to resource regular monitoring. Therefore, in-house
application can be used to perform medical assessments by remote healthcare monitoring applications provide an
remotely monitoring patients. The accuracy of medical
assessments fundamentally depends on the correctness of the
alternative by delivering healthcare to a dispersed population
data received from the BAWSN. However, data errors may arise having a variety of diseases requiring regular monitoring and
at the sensor or during transmission across the wireless sensor care. This is especially true in suburban or rural areas where
network. Therefore, it is imperative to measure the health expert doctors, nurses, electronic equipment and other
intrinsic data of a patient precisely. The formulated measurable infrastructures are often inadequate or difficult to access in
properties in our work precisely measure the performance of the time.
BAWSN in a remote Healthcare Monitoring Application (HMA). One of the major challenges in remote healthcare
In this paper, we collated various performances using the monitoring application (rHMA) involves conducting medical
measurable properties in our real-time test-bed and presented a
assessments based on remotely sensed patient’s health data.
comprehensive evaluation of these properties in a BAWSN.
The accuracy of the health data such as sensed time, unaltered
Index Terms—BAWSN, ehealth, reliability, availability, critical data content and identity of the person should be intact to avoid
time, remote healthcare monitoring application false diagnoses. However, ensuring data is accurate is difficult
in a Body Area Wireless Sensor Networks (BAWSN) because
I. INTRODUCTION sensors are not always reliable or accurate and data
transmission over a wireless network is also not perfect. We
In Australia, the number of people over 80 years old has can expect BAWSN’s to be increasingly accurate with
increased from 260,000 (recorded in 1981) to 560,000 advances in sensor and wireless technologies; however an
(recorded in 2001) and, it is expected to surpass 790,000 by innovation that will always be useful is a metric that provides
2021 [1]. As the home of one of world’s largest population, the an estimate of how accurate a BAWSN can be expected to be.
population of India will continue to grow through 2050. By A set of dependable properties for a BAWSN were proposed
2030, it is estimated that 45.4 % of the Indian population and formulated in our work in [7]. In the current work, we
would be older adults [2]. Furthermore, in India, the sixty-plus show how these properties can be used to estimate the accuracy
age group will grow more than threefold from 8.4 % to 22.6 % of a BAWSN in a real-time test-bed.
over fifty years, while those aged eighty and older will expand The paper is arranged in the following sections. Section II
nearly fourfold from 0.8 % to 3 % [2]. This trend is not unique will describe a remote healthcare monitoring application and
to Australia and India, rather it is a global issue because it has our background work. In Section III, we briefly describe
been estimated that the global population of 65 years old and properties that make a BAWSN dependable. Section IV will
above will double from 357 million (recorded in 1990) to 761 give the details of the experimental setup, while the
million by 2025 [3]. performance of the operational properties will be given in
In addition to the growth in the ageing population, there has Section V. Finally, Section VI will conclude this paper.
been an explosion in the prevalence of chronic conditions.
Chronic conditions including diabetes and cardiovascular II. REMOTE HEALTHCARE MONITORING APPLICATION
disease cannot be cured and must be closely monitored for the The rHMA is an application system made up of two related
emergence of complications. However, many chronic diseases applications, the healthcare application (HA) and the BAWSN,
show notable symptoms of the disease or complications only in the monitoring application component:
their later stages [4–6]. Continuous monitoring of a health
condition is a proven method for early detection of many
conditions [4]. However, many developed and developing
1. The BAWSN consists of state-of-the-art self-organizing any alert messages in the case of an emergency and to send the
wireless sensors capable of sensing vital health data of a pre-processed data to the HA for further medical diagnoses. It
patient, a base station, which aggregates the health data is evident that the rHMA depends heavily on its monitoring
from the sensors, and a local processing unit (LPU), which component, the BAWSN, for the continuous generation of the
can be a smart phone or a tablet. health data. The main focus of our work is to evaluate the
performance of the BAWSN, while the reliability and
2. The HA should be considered as a system that uses availability of the AMC application is outside the scope of this
information technology for collecting, maintaining, paper.
analysing and disseminating the patient’s health data for
clinical needs. Although, maintaining and disseminating
the patient’s data are achieved through the use of various III. THE DEPENDABLE PROPERTIES
technologies, to measure and collect the health-related data
of the patient, the HA uses a manual/traditional method To build a BAWSN, the underlying wireless sensors
that requires the patient’s attendance at the hospital. network (WSN) should be extended with an LPU, that is, a
smartphone, for further processing of the data. In general, for
Our real-time test-bed Active Care Loop Framework any underlying networks, one might use Quality of Service
(ACLF) is an rHMA is designed and built using various (QoS) as a property to evaluate WSN performance. However,
wireless technologies and the Internet as the data transfer QoS is a broad term used to denote the level of predictability
medium. It consists of the Assistive Maternity Care as it’s HA and manageability of the services supplied by one or more
and the BAWSN, the monitoring application component. service providers. Moreover, the QoS refers to resource
reservation control mechanisms rather than the achieved
service quality.
In computer networks, the QoS is the ability to provide
different priorities to different applications, users or network
data flows or to guarantee a certain level of performance for a
data flow [9]. The QoS is more applicable for the underlying
network in a given application system. Although the
underlying wireless networks form the foundation for a
BAWSN, the BAWSN itself forms an integral application
component in an rHMA system. In addition, the functionalities
in a BAWSN are also well defined and termed as critical in an
rHMA system because it has the onus of collecting health data
intrinsic to the patient. Therefore, providing different priorities
Fig. 1. Active Care Loop Framework to different applications may not be appropriate.
Further, the properties that define QoS for a BAWSN [9]
The AMC application enables practitioners (such as doctors do not emphasise the security aspects that are defined as
and midwives) and pregnant women to access health records important for the BAWSN [10–13]. In addition, the
from beyond the physical location of a hospital zone not only construction and the maintenance of components are less
through desktop computers, but also through smartphones as explored areas, even though they are vital for achieving
shown in Fig. 1. In the AMC application, a pregnant woman accurate collection and transmission of data and fault detection.
enters her blood pressure (BP) value using a smartphone or The studies in [11, 12] also show that the QoS in a BAWSN is
desktop regularly at home or at work, for instance, to allow her not scalable in terms of power consumption and cannot be used
health care providers to monitor her condition. Depending on as a single solution for all kinds of BAWSN applications.
the gravity of updated data, the database notifies the AMC Based on these shortcomings in using only the QoS for the
application that would then automatically send an alert performance evaluation, the properties that define the
message with specific information on the patient to the care dependability of the component are considered. Highly sought
staff, either as an SMS or an email [7]. properties for WSN, the node power and node
The AMC application described above requires the manual destruction/repair, is encompassed by the availability property
entry of the health data by the pregnant women. To automate [14-16] from the dependability property list. Therefore, the
the entry of the health data, the AMC application is extended dependability properties comprise: reliability, availability,
with a BAWSN to form the ACLF. For instance, in the AMC, security, maintainability and safety. These form an effective set
it is inconvenient for a pregnant woman to continuously enter for evaluating the performance of a BAWSN.
health data, such as the BP, heart rate and glucose levels, for Apparently, not every property of dependability can be
monitoring purposes. Therefore, a BAWSN is used to sense the measured numerically, because certain properties such as
required health data at regular time intervals. The distinct maintainability and safety can only be realized by assessing
functions of a BAWSN are to authenticate the patient for their flexibility in the design during development and their
continuous monitoring, to sense the health data from the adaptability to the environment during its real-time
patient, to pre-process the health data of the patient for sending
deployment. However, properties such as availability, arrival
reliability and security can be quantified. In our research work Error time, inoperative time of application owing
[7] the reliability and availability properties were formulated to erroneous packets
based on derived time-specific parameters. The time-specific Buffering error time, inoperative time of
parameters along with the critical time are described in the application owing to loss of the packet in the base
following sub-section. station buffer
Total error time for the sensor Sn in a
A. Time-critical applications BAWSN
Total buffering error time for the sensor Sn in
The drive for the rHMA is to provide for continuous remote a BAWSN
monitoring of the health condition of the patient. For this to Total error and buffering error time for the sensor
happen, the sensors in the monitoring component of the
BAWSN should continuously send data to the HA through the Sn in a BAWSN
LPU. To provide continuous health data, the BAWSN relies on Total critical lack time in a BAWSN
the collective data provided by all the source sensor nodes and
not on any individual data, because medical diagnosis based on
health data is rarely based on any one specific parameter [17– Based on the above time-critical parameters total unreliable
19]. In addition, the arrival time of the data should be within time of the sensors and the total unavailable time of the sensors
the expected time (εtm) because any arrival of health data in a BAWSN are given by the below Eq. 1 and Eq. 2
outside the expected time cannot be assumed to indicate the respectively.
actual health condition of the patient. For example, a high
blood pressure and a high heart rate when a person is at rest is a
clinical incident and should be taken care of immediately;
whereas, the same condition for a person during substantial
movement may not be a clinical incident. It should be noted (1)
that the arrival time of health data, both the blood pressure and
a heart rate is necessary for medical diagnosis. Therefore, in
our work [7], we defined and implemented a novel critical time
parameter for the BAWSN operations. Because the BAWSN (2)
has stringent timing requirements concerning the arrival data
rate from the sensor nodes within the defined critical time The operational availability of the BAWSN states that a
(εtm), we characterized the rHMA as a time-critical application. system is highly available if the fraction of its downtime is
Based on the time stringent requirements and to address very small, either because failures are rare or because it can
generic measures such as delay, jitter and packet loss, the restart very quickly after a failure, while the operational
following time critical parameters have been selected [7]: reliability of the BAWSN is defined as seamless delivery of
error free health data within the critical time. By using the Eq.
1. The critical lack time (TLACK), is the elapsed time in the 1 and Eq. 2 the operational availability and reliability of the
application for which the medical diagnosis may not be BAWSN is given by Eq. 3 and Eq. 4 respectively.
accurate owing to the delayed packet arrival.
2. The error time (TE), is the defective time in the application
caused by erroneous packets.
3. The buffering error time (TB), is the defective time in the (3)
application caused by the loss of packet.

The parameter used in the formulation of reliability and

availability properties is given in the following table: (4)
To show how these properties can be used to evaluate the
TABLE I. TIME-SPECIFIC PARAMETERS performance of the BAWSN in real-time is just as important as
Expected time interval for a BAWSN to receive formulating these properties for the application developer. The
data from the sensors. The critical time rate for following sections describe the experimental facilities and the
the BAWSN application that could be used for results are analysed using the time-critical parameters.
monitoring purpose. IV. EXPERIMENTAL FACILITIES
Index for counting sensors in a BAWSN
Critical lack time, elapsed time for which the The experiment was conducted in the BAWSN, as shown in
application is inoperative owing to delayed packet Fig. 2. The BAWSN includes heterogeneous components
differing not only in processing power and storage capacity,
but also having a diverged operating platform. The sensor Where,
components consist of four crossbow-technology-based sensor ‐ S is the size of the maximum base sensor buffer (bytes)
motes [20] that operate using the TinyOS [21]. The end-point ‐ C is the current level of base sensor buffer (bytes)
is the local processing unit, a smartphone connected with a ‐ N is the number of sensors used in a BAWSN
Stargate gateway board [20], which is used as the base station ‐ d is the data rate of the sensors (bytes/sec).
to interconnect the 802.15.4/ZigBee-based crossbow family
motes and the 802.11_based smartphones. A Stargate can be The above inequality means that data must be requested by
used as a base station only after housing a base sensor mote in the end-point (smartphone) before the buffer of the aggregation
the Stargate board. Please note that Fig. 2 shows six mote point (base station) is full: otherwise, it is lost. This inequality
sensors; however, during our initial experiment to calculate the is programmed into the smartphone to mitigate the loss of the
availability and reliability of the BAWSN, only four sensors packets in the base station buffer. After the initial request, the
were used. To implement and test the proposed security smartphone computes the request time interval tR (1 / fR ) on the
scheme in the BAWSN, six sensors were used in which the fly, at which time the next request is sent, based on the current
sensors marked 5 and 6 in Fig. 2 have been programmed with a buffer level C, whereas the other factors S, N and d are kept
different identification set. These two sensors are used to test constant for a given BAWSN [7]. Note that the smartphone’s
the behaviour of the security scheme in the BAWSN, which is request time interval tR can be achieved merely by using the
outside the scope of this paper. lesser value compared to the programmed time in the sensor —
The mote sensor and the Stargate (base station) have less than 10 seconds in our scenario. This proved to be highly
hardware on/off switches, with the sensors being programmed inefficient in terms of sending a request to the base station,
to transfer data packets every 10 seconds to the base station even if the buffer is empty. Consequently, these inefficient
when the switch in the BAWSN is turned ‘on’. The user requests considerably reduce the smartphone’s battery power.
initiates the continuous monitoring application Obi-MATE1 Moreover, the application developer can configure the
using a smartphone. The application starts the protocol that monitoring application to compute the request time interval
registers the smartphone with the base station. Once the based on N, C and d instead of hardcoding the request time
smartphone has been registered, the gathered data at the base interval in the monitoring application, because tR depends on
station can be fetched through the Stargate gateway. Because the requirement of the continuous monitoring and the data rate
the base sensor mote attached to the Stargate has a limited of the sensors used in the BAWSN. The inequality in Eq. 5
buffer memory for storing the sensed data received from all the enables the end-point to send a request to the base station just
sensor nodes, it is essential that data are fetched from the base before the buffer is full and empties the buffer. This timely
station before the buffer overflows and causes a loss of data. request from the smartphone gradually reduces the base station
Therefore, we have modelled the smartphone to request the buffer overflow and also ensures that no data is lost owing to
Stargate at a specific frequency, ; the number of requests sent implementation/system faults. The following section presents
from the smartphone over a selected time interval should the results for reliability and availability obtained by
satisfy the following inequality: conducting the experiments in the real-time test bed [7].

 (5) V. EXPERIMENTAL RESULTS

The packet size for all the sensor motes was configured to
29 bytes, including the header. During the initial phase, the
data packets of the sensors were not encrypted. To study the
impact of inequality on the performance of the BAWSN and
also to calculate the operational properties, a series of
experiments were conducted in our BAWSN under similar
conditions with each a duration of 60 minutes. The average
result of the five experiments is plotted as shown in Fig. 3. The
properties reliability and the availability values in the
experiment are calculated for every 10-minute interval within
the actual duration of the experiment.

Fig. 2. Experimental setup for the BAWSN

1
The concept of developing assistive mobile application for the pregnant
women and the midwives in ACLF has been given the name of Obi-MATE
(a) (d)

Fig. 3. Performance of a BAWSN (a) Operational Reliability (b) Operational

Availability (c) Operational Throughput (d) Number of error and lost
packets during operational time

These experiment results demonstrate a gradual increase in the

operational properties throughout the duration of the
experiment. The decreased reliability and the availability
during the initial interval is caused by the system calculating
the critical time during this initial phase and also, the
programmed inequity in the smartphone learns about the buffer
level only after a few requests, because there were some
dropped packets.

Following this, the experiments demonstrate that the loss of the

(b) packets in the base station buffer is gradually reduced, thereby
mitigating . A significant reduction in is also noted
owing to the timely retrieval of the packets from the base
station. It is noted that, because of the decreased and
, the operational properties of the BAWSN foster and
stabilize with time, because of the negligible and .
Even though negligible, the and are achieved; the
operational properties remain basically unchanged, because a
consequential is noted from time to time throughout the
experiment. In Fig. 3(a) and (b), the operational properties
show very little increase in the latter half of the experiment.
The consequential during the experiment is caused by the
use of various wireless technologies, such as ZigBee and Wi-
Fi. It is evident from the Fig. 3(d), the packet drop reduced to
(c) zero in the later part of the experiment whereas there is always
the presence of error packets. Intense research is required [22]
to establish the actual cause of error packets that can be
controlled by the user, so that the reliability and availability of
the BAWSN can be improved. On the other hand, the
possibility of attaining maximum operational properties is to
 [8] Quality of Service, Available:
obtain a negligible . This can be achieved by providing a http://en.wikipedia.org/wiki/Quality_of_service, accessed 24
common hardware switch for the sensors, so that all the sensors January 2009.
can be turned ‘on’ simultaneously. The Fig. 3(c) shows the [9] D. Chen and P. K. Varshney, “Qos support in wireless sensor
comparison of an ideal throughput and the actual throughput networks: A survey,” in Int. Conference on Wireless Networks
(from for sensors) for the operational time. The purpose of this (ICWN 2004), CSREAPress, June. 2004.
graph is to substantiate that the throughput measure in the [10] B. Latre ́, B. Braem, I. Moerman , C. Blondia, and P. Demeester,
BAWSN is of a lesser significance to measure its performance. “A Survey on Wireless Body Area Networks," Journal of
Because, during the initial phase of the experiment, the actual Wireless Network, vol. 17, Issue. 1, pp. 1 -18, January. 2011.
throughput is less than the ideal because of the dropped/lost [11] I. F. Akyildiz, W. Su, Y. Sankarasubramaniam, and E. Cayirci,
packets; however, it is in line with an ideal throughput in the “A survey on sensor networks,” IEEE Communications
later part that demonstrates that system is working flawlessly. Magazine, vol. 40, no. 8, pp. 102–114, Aug. 2002.
But this is not the case, because the occurrence of an error [12] A. Bhargava and M. Zoltowski, “Sensors and wireless
packet is also considered during the throughput measurement. communication for medical care,” in Database and Expert Sys-
Also, Fig. 3(d) shows the number of dropped packets during tems Applications, 2003, in Proceeding. 14th International
the operational time that might be an inappropriate measure for Workshop on, Sep. 2003, pp. 956–960.
evaluating the performance of the system. Most of the packets [13] S. Cherukuri, K. K. Venkatasubramanian, and S. K. S. Gupta,
drop at the same time instant during the initial part of the "BioSec: A Biometric Based Approach for Securing
experiment, so the system is not performing well only at that Communication in Wireless Networks of Biosensors
time instant, not for that complete interval. Implanted in the Human Body," in Proceedings of the
2003 International Conference on Parallel Processing
Workshop (ICPPW’03), 6-9 Oct. 2003, pp. 432-439.
VI. CONCLUSION [14] L. Xian, "On the Availability of Wireless Sensor Networks," in
Proceedings of the 10th IEEE Wireless Communications &
The paper exemplifies the importance of a BAWSN in a Networking Conference, Las Vegas, USA, 2008 , pp. 3081–
remote healthcare monitoring application. The justification for 3086.
dependable properties for a BAWSN along with the overview [15] P. Sazia, S.K. Dong, and S.P. Jong," Towards Optimal Software
about the time-specific parameters used in the dependable Rejuvenation in Wireless Sensor Networks using Self-
properties is also presented. As a major focus the performance Regenerative Components," in Proceedings of 2nd International
evaluation of the operational properties particularly the Conference on Information Security and Assurance, Busan,
reliability and the availability of the dependable application Korea, 2008,pp. 295–300,
with experiment results conducted using the real-time test-bed. [16] P. Sazia, S.K. Dong, and S.P. Jong, "Towards Survivable
Finally, it is important to note that the performance evaluation Sensor Networks using Self-Regenerative Rejuvenation and
shown in this paper does not advocate the accuracy of usage in Reconfiguration," in Proceedings of 1st International
any medical diagnosis. Conference on Information Security and Assurance ,
Manchester, UK, 2007,pp. 546–549.
REFERENCES [17] What is Prostate Cancer, Available:
[1] Human Rights and Mental Illness, Available: http://www.prostate.org.au/articleLive/pages/What-is-Prostate-
http://apo.org.au/research/national-inquiry-concerning-human- Cancer.html, accessed 03 October 2009.
rights-people-mental-illness, accessed 20 Dec 2013. [18] E. Ergezen, R.W. Hart, R. Philip, and R.M. Lec, "A high
[2] The Health Of Aging Populations In China and India, Available: frequency thickness shear mode (TSM) sensor for detection of
http://content.healthaffairs.org/content/27/4/1052.full, accessed biomarkers for prostate cancer," Frequency Control Symposium,
25 Dec 2012 2008 IEEE International , 19-21 May. 2008, pp. 341-345.
[3] N. Hooyman and H. Kiyak, Social gerontology: A [19] M. Caron, G. Choquet-Kastylevsky, and R. Joubert-Caron,
multidisciplinary perspective, 6th ed., Allyn and Bacon, 2002. “Cancer Immunomics Using Autoantibody Signatures for
Biomarker Discovery,” Molecular and Cellular Proteomics, pp.
[4] K. Kancherla, R. Chilkapatti, S. Mukkamala, J. Cousins, and 1115-1122, 2007.
C. Dorian, "Non Intrusive and Extremely Early Detection of
Lung Cancer Using TCPP," Computing in the Global [20] Crossbow Technology Inc., Motes, Wearable Sensors,
Information Technology, ICCGI '09, Fourth International Multi- Available : http://www.xbow.com/ , accessed 24 January 2010.
Conference , 23-29 Aug. 2009, pp.104-108. [21] TinyOS Embedded Operating System, Available:
[5] World Health Organization, Available: http://www.tinyos.net , accessed 24 January 2010.
http://www.who.int/mediacentre/factsheets/fs317/en/index.htm, [22] S. Shin, S. Choi, H.S. Park, and W.H. Kwon," Packet error rate
accessed 24 January 2009. analysis of IEEE 802.15.4 under IEEE 802.11b Interference," in
[6] International Diabetes Federation (IDF), Available: Proceedings of the 3rd International Conference on
http://www.idf.org, accessed 24 January 2009. Wired/Wireless Internet Communications, LNCS Springer,
Xanthi, Greece, 2005, pp. 279–288.
[7] V. Balasubramanian, and D. B. Hoang, “Reliability Measure
Model for Assistive Care Loop Framework using Wireless
Sensor Networks.” International Journal of Healthcare
Engineering, vol. 1, no.2, pp. 239–254, 2010.