This article has been accepted for publication in a future issue of this journal, but has not been

fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JSEN.2016.2523461, IEEE Sensors
Journal

JOURNAL OF LATEX CLASS FILES, VOL. XX, NO. X, DECEMBER XXXX 1

An Efficient Cross-Layer Reliable Retransmission

Scheme for the Human Body Shadowing in IEEE
802.15.6-Based Wireless Body Area Networks
Hsueh-Wen Tseng, Member, IEEE, Ruei-Yu Wu, and Yi-Zhang Wu

Abstract—In recent years, a number of middle-aged and a short period and need long-term hospitalization. As a re-
elderly people with chronic diseases are increasing. In addition, sult, using wireless body sensor networks (WBSNs) for self-
patients with chronic diseases do not cause health harm in a care becomes an ideal way. Physiological data recorded by
short period and need long-term hospitalization. Thus, wireless
body area network (WBAN) is the best scheme for daily care. sensor nodes not only provide users immediate reminders or
According to the human physiological information, it not only warnings, but also reduce the occurrence of delayed medical
sends real-time notifications to users but also reduces delays of the situation.
notification regarding patients’ conditions. The physician early Over the past decade, the wireless sensor network (WSN)
identifies the cause of disease and applies remedies according to is also one of the most exciting techniques. It has a significant
indications. Daily activities of a person affect the transmission
signal of sensor nodes such as walk or run. Thus, the sensor impact on almost all major industries as well as our home
nodes which have to retransmit the failed frames are continuously lives [3]. Recent advances in integrated circuits, development
interfered a period time by human body shadowing, so the in microelectronics, system on chip, wireless communication,
retransmission procedure should be deferred. If the sensor nodes and wearable sensors have enabled to show a new type of
immediately retransmit the failed frames, they may suffer from networks known as WBSN. The WBSN offers many new
the human posture interference again. Therefore, we propose
an efficient cross-layer reliable retransmission scheme (CL-RRS) promising applications in the area of remote health monitoring.
in IEEE 802.15.6 without additional control overheads. The In a star topology, the communication is established between
proposed scheme not only detects the information of the sensor sensor nodes and a single central controller. This paper focuses
nodes with failed transmission frames to allocate retransmission on the star network topology. A WBSN may consist of a
resources, but also increases the successful probability of frame large number of sensor nodes, which are scattered densely
retransmission. Simulation results validated by our mathematical
analysis show that CL-RRS significantly improves frame loss in the human body to receive and collect personally medi-
rate and average transmission time, as well as reduces power cal information, such as pacemakers, coronary care, muscle
consumption. stimulants, neurostimulators, and retinal prostheses, etc.. In
Index Terms—Wireless body area network, frame retransmis- these cases, the messages of human body information are
sion, human body shadowing, IEEE 802.15.6. important and should be transmitted by the reliable frame
transmission. Different sensor devices may have different
transmission timing according to their priorities. In addition,
I. I NTRODUCTION
some human body information may have real-time properties
In the United States, people life has increased from 69.8 and must be transmitted periodically in WBSNs.
years in 1960 to 78.6 years in 2011 [1]. The size of popu- In WBSNs, transmission signal strength may attenuate
lation with 60 to 80 years old will climb in the near future. during transmission depending on different reasons such as
Furthermore, overall health care expenditure in the US was transmission distance, transmission power, and an obstacle in
$1.8 trillion in 2004, and this number is expected to be triple between sender and receiver. In addition, sensor nodes transmit
by 2020. Besides, according to the World Health Organization data using lower power for saving energy. Because the sensor
(WHO) report, the cost of health care of the US is 20% gross node is installed in/on the human body, the organs or limbs
domestic product (GDP) [2]. The new challenge of medical may affect signal strength [4–6]. When transmission signal is
and health care systems is the elderly population problem. shadowed, higher frame error rate (FER) may cause by weak
The problem of fatal disease should be diagnosed at an early signal strength. As a result, sensor nodes need to retransmit
date. the failed frame and increase transmission cost.
Furthermore, the numbers of middle-aged and the elderly Several wireless technologies are considered for WBSNs:
people with chronic diseases are increasing. Patients with IEEE 802.15.4 [7], Bluetooth [8], and IEEE 802.15.6 [9].
chronic diseases do not cause immediately health harm in Among the well-known specifications, IEEE 802.15.6 origi-
nally designed for wireless body area networks (WBANs), has
Hsueh-Wen Tseng is with the Department of Computer Science and
Engineering, National Chung-Hsing University, Taichung, Taiwan. become one of the promising candidates for interconnections
Ruei-Yu Wu is with the Department of Management Information Systems, between wireless body sensor nodes. IEEE 802.15.6 compared
Hwa Hsia University of Technology, New Taipei, Taiwan. to other schemes, data transmission rate, number of channels,
Yi-Zhang Wu is with the Department of Computer Science and Engineering,
National Chung-Hsing University, Taichung, Taiwan. and quality of service (QoS) to obtain the most advantageous.
Corresponding author: Ruei-Yu Wu (email:d92007@csie.ntu.edu.tw). Therefore, in this paper, we propose a cross-layer reliable

1530-437X (c) 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JSEN.2016.2523461, IEEE Sensors
Journal

JOURNAL OF LATEX CLASS FILES, VOL. XX, NO. X, DECEMBER XXXX 2

retransmission scheme (CL-RRS) to solve the retransmission

problem with the human body shadowing in IEEE 802.15.6. B EAP1 RAP1 MAP EAP2 RAP2 MAP B2 CAP

The proposed CL-RRS is a specially designed scheme in

the physical (PHY) layer to estimate the node addresses more Superframe
accurately from the corrupted frames. Using the cross-layer
Fig. 1. IEEE 802.15.6 superframe structure [12].
operation, the receiver allocates the retransmission resources
from the information of the PHY layer back to the sender, and
then the sender follows the proper retransmission time of the IEEE 802.15.4 specifies PHY and MAC layers for short-
MAC layer to fit the actual network situations. The numerical range wireless communications, low power devices, low cost,
results of simulation experiments show that the proposed and low bit rate networks. The main features of IEEE 802.15.4
scheme can effectively use the suitable retransmission time include 250 Kbps data transfer rate, a variety of network topol-
of the MAC layer for retransmission procedure. The system ogy (e.g. star, mesh, and tree topologies), and 16 channels.
performance of frame loss rate and average transmission time However, with the development applications of WBSN, they
can be significantly improved, and power consumption is also may transmit videos (e.g., endoscopic images or on-line video)
reduced. and request more than 1 Mbps transmission rate. Obviously,
The rest of this paper is organized as follows. Section II IEEE 802.15.4 only provides 250 Kbps and is not suitable
reviews WBSNs in more detail of IEEE 802.15.4, Bluetooth for video applications. In addition, a few of channels may be
low energy, and IEEE 802.15.6 and background study. Our not enough for outpatient services or hospital lobby. Since
CL-RRS scheme is presented in Section III and an analytical the spectrum of IEEE 802.15.4 overlaps with IEEE 802.11,
model is presented in Section IV. The results of extensive microwave, and Bluetooth, the channel interference may occur.
numerical experiments are demonstrated in Section V. Finally, Bluetooth is proposed in 1994, and the key feature is low
the conclusions are described in Section VI. cost. It can be used on some WBSN applications but it con-
sumes higher power [12]. The low energy version (Bluetooth
LE), published in 2010, is designed for applications with
II. R ELATED W ORK AND BACKGROUND S TUDY lower bit rates and duty cycles. Bluetooth LE aims at small
Currently, WBSNs are widely used in self-care of chronic and cheap devices such as wireless sensor nodes. It supports
patients. These patients have generally walking, running, and some applications: home automation, entertainment, sport and
other acts. Thus, human posture changes may cause shadowing fitness, health care and illness treatment, and so on.
phenomenon which results in the transmission failure of sensor Bluetooth LE operates in the 2.4 GHz and supports star
nodes. The failed frames have to retransmit from the sensor topology. It provides 1 Mbps transmission rate, total 40 chan-
nodes. In order to correctly transmit body information, the nels (37 transmission channels and 3 broadcasting channels).
reliable frame transmission has to be implemented. Bluetooth LE has single mode and dual mode. The single
Several transmission schemes in MAC layer are used mode only supports the Bluetooth LE’s special architecture
to access channel such as time division multiple access and is not compatible with previous versions due to the low
(TDMA), carrier sense multiple access with collision avoid- power consumption purpose. In other words, the dual mode
ance (CSMA/CA), and the hybrid scheme [10]. Nodes access is compatible with previous versions. Bluetooth LE supports
the channel in their own time slots without any contention QoS between the master node and slave nodes, and the master
in TDMA. The transmission performance of TDMA scheme node may control several slave nodes at the same time, but
seems wonderful because it gets rid of the overhead of the protocol does not specify the service limitation. Thus, it
collisions. However, when transmission signal is shadowed by results in unmanageable slave nodes. Furthermore, slave nodes
human body at specific time, it increases the error rate of send request packets as they have resource requests. However,
transmitted frames. Bluetooth LE does not provide collision avoidance mechanism
A node using CSMA/CA starts to sense channel only when to solve contention problem for the slave nodes.
the backoff countdown process is complete. The node can ob-
tain a new random backoff value again from double contention A. IEEE 802.15.6
windows (CW) when a frame transmission fails. However, the IEEE working group has defined new PHY and MAC
frame may not be able to guarantee to retransmit within the layers that provide low complexity, low cost, high reliability,
transmission deadline [11]. Thus, the frame may be discarded ultra-low power, and short range wireless communication in
because of larger transmission overhead. Therefore, according or around the human body for wireless body area networks
to the aforementioned descriptions, TDMA and CSMA/CA (WBANs). IEEE 802.15.6 provides 79 channels with data rate
have their advantages and disadvantages, respectively. Thus, 971.4 Kbps and supports one-hop and two-hop star topologies
the mixed mechanism is used to take advantages of both [9, 13]. A superframe can be divided into 256 time slots. Thus,
schemes. it can meet the resource requests of sensor nodes to decrease
Regarding WBSNs, mainly communication standard solu- long waiting time.
tions considered as IEEE 802.15.4 [7], Bluetooth low energy As Figure 1, a beacon frame (B) is transmitted by the hub at
[8], and IEEE 802.15.6 [9]. The most commonly used fre- the beginning of every superframe. The beacon frame contains
quency is 2.4 GHz. We describe the advantages and disadvan- information about timing synchronization, system configura-
tages of three protocols as follows. tion, and so on. In addition, B2 is also the broadcasting control

1530-437X (c) 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JSEN.2016.2523461, IEEE Sensors
Journal

JOURNAL OF LATEX CLASS FILES, VOL. XX, NO. X, DECEMBER XXXX 3

frame and has the same capability as B. In order to decrease improved transmission schedule of sensor nodes. For addi-
the sensor nodes to wake up again and receive additional tional relay nodes, they are suitable to used a long shadowing
control messages, only B is considered in this paper. duration from the human body as with the wear equipment on
The superframe structure is divided into exclusive access the human body, people keeps walking or running posture to
phases (EAP1 and EAP2), random access phases (RAP1 and cause discomfort. As a result, a patient decreases willingness
RAP2), a managed access phase (MAP), and a contention of wear equipment usage. In addition, for walking or running
access phase (CAP). Note that the order of each subperiod posture of a patient, the probability and the duration time
does not be changed in the superframe, and each subperiod of the human body shadowing are decrease. In this case,
is not sure to appear in the superframe. The size of each we use a suitable transmission schedule to transmit the body
subperiod can be adjusted under the superframe size. The information, and it also decreases extra hardware cost.
EAPs are used to transmit high priority or emergency frame, Hence, in this paper, we focus on IEEE 802.15.6 to solve the
and the RAPs and CAP are used to transmit general frame. retransmission problem caused by the human body shadowing.
A slotted CSMA/CA scheme is used to access channel in Although the performance analysis for IEEE 802.15.6 was
EAPs, RAPs, and CAP. The size of CW is set according extensively investigated, little work has been done on the
to the frame priority. The high priority frame uses a small problems of IEEE 802.15.6 transmission efficiency. Therefore,
CW as compared with the low priority frame. The MAP we propose a cross-layer reliable retransmission scheme (CL-
adapts TDMA to transmit the frame. The nodes can access the RRS) to improve the retransmission problem and reduce the
channel in their own time slots without any contention. Based overall energy consumption for IEEE 802.15.6-based WBANs.
on the characteristics of the transmitted frames, the superframe
only includes the EAP, the RAP, and the MAP in this paper. TABLE I
A hub can allocate resources to sensor nodes to guarantee T HE COMPARISON OF PROTOCOLS
the QoS of the transmission frame in the contention-free phase. IEEE 802.15.4 Bluetooth LE IEEE 802.15.6
There are 8 different access categories which indicate priorities Data rate 250 Kbps 1 Mbps 971.4 Kbps
of transmission frame for accessing the medium in contention Channels 16 40 (37 data) 79
access phase. The priority values shall be determined based QoS No Limited Yes
on the designation of frame payloads in the frame [14].
In Table I, we compare with data rate, channels, and QoS
of three protocols. To consider the data rate, IEEE 802.15.4 is
III. C ROSS -L AYER R ELIABLE R ETRANSMISSION S CHEME
lower than the other two protocols. In the number of channel,
IEEE 802.15.6 has obvious advantages to coexist with other In a WBAN, the receiver is a hub to collect body informa-
protocols and avoid interference problem from other using tion from end devices. Our CL-RRS consists of operations
2.4 GHz services. In contention free phase, the hub allocates in both PHY and MAC layers. Upon the occurrence of
time slots to ensure the QoS of the transmission frame. Here, overlapped signals, the preamble of the frame is first detected
the QoS is only considered in contention access phase in from the overlapped signals in the PHY layer. Then CL-RRS
three protocols. In addition, IEEE 802.15.6 increases the slot uses the signal features of the PHY layer during received frame
number of a superframe to 256. It can effectively reduce the to identify the precise reason of failed transmission on the
resource allocation problem. For WBANs, when we consider receiver. Therefore, the retransmission subperiods are used to
data rate, number of channels, and QoS, IEEE 802.15.6 is effectively retransmit the failed frame in the MAC layer.
better than other two protocols. In general, a common approach increases the transmission
In [15], the authors described the perturbations of the energy or information encoding to improve the frame error rate
electromagnetic propagation of radio signals caused by the (FER). However, for the sake of low-energy consumption, the
presence of a moving object in the two-dimensional (2-D) area sensor node cannot amplify transmission energy to increase
near the transmitting/receiving devices. In [16], the authors transmission signal strength. Therefore, IEEE 802.15.6 adopts
statistically analyzed the path loss, shadow fading, and delay the BCH encoding which corrects the bit error to reduce the
dispersion of on-body Ultra-Wideband (UWB) WBAN chan- transmission failure possibility. The BCH has different bit
nel. Humans are always in motion; channel conditions between error rate (BER) according to signal strength [4, 18] indicated
different nodes and the hub are always changing [17]. In that transmission signal strength ratio between the hub and
IEEE 802.15.6, the two-hop relay mechanism with predefined a sensor node at different body postures of normal activities.
relaying nodes is proposed to reduce the possibility of data [6] indicated that human body shadowing causes significant
relaying failure. Pan et al. used a predefined relaying node signal attenuation and increases the FER. Thus, the receiver
to improve the data relaying failure rate and also proposed fails to recognize the received frame due to the low signal
a dynamic scheduling algorithm depending on the packet strength. The average human body shadowing rate is 10%,
delivery rate (PDR) to schedule the relaying node in the two- the body shadowing duration is 80 ms, and the signal strength
hop relay. threshold of IEEE 802.15.6 is −86 dBm in [6].
In reference [6], the authors found that body postures affect Once frames are transmitted during the human body shad-
the strength of transmission signals to decrease the reliability owing, the chain effects of errors occur. As a result, the FER
of frame transmission. To solve a human body shadowing rapidly increases, and the retransmission overhead and power
problem, they added relay nodes on the human body or consumption also increase for the sensor nodes. Therefore,

1530-437X (c) 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JSEN.2016.2523461, IEEE Sensors
Journal

JOURNAL OF LATEX CLASS FILES, VOL. XX, NO. X, DECEMBER XXXX 4

the retransmission policy should not only use a single MAC body shadowing, the time of the unstable signal is much longer
layer mechanism. The random backoff scheme cannot solve than the transmission time of a frame. For the collision, the
the human body shadowing problem effectively. Here, CL- signal strength is fully overlapped from the beginning to the
RRS is proposed to solve the retransmission problem in IEEE end of the frame. This is because senders select the same
802.15.6-based WBANs. backoff value to transmit to the same receiver. Since sensor
In following subsections, first, the frame received mecha- nodes have different transmission distances to the hub, there
nisms of TDMA and CSMA/CA considering the human body are different transmission delays of frames between the sensor
shadowing are described. Then, to distinguish collision and nodes and the hub. Thus, the hub receives a frame with a
human body shadowing problems are discussed. Finally, the very short stable signal period at the beginning (Sstable b ). CL-
transmission modes of a dynamic adjustment in the superframe RRS uses the signal features of the PHY layer during received
is described. frame to identify the precise reason of failed transmission on
the receiver. As the receiver (hub) judges the correct reason
of failed transmission, the receiver uses different superframe
A. TDMA and CSMA/CA Frames Delivery Flow
structures to retransmit the failed frame with the human body
IEEE 802.15.6 standard does not consider the influenced shadowing.
transmissions signal during frame delivery with TDMA and Otherwise, the hub tries to recognize the sender ID of the
CSMA/CA due to the human body shadowing. We determine failed frame header from the MAC layer. If sender ID is
these two types of frame delivery schemes whether the human inconsistent with IDtable recorded by the hub, the hub drops
body shadowing occurs or not. The IDs of sensor nodes are the failed frame. The IDtable is the ID record of sensor nodes
detected from the collided signals in the PHY layer to identify in the hub. On the contrary, the hub records the sender ID and
the sensor nodes. Furthermore, we retransmit the failed frame the allocates MAP1 to retransmit the failed frame. Hence, the
with the suitable subperiods according to the reasons of failure. sender retransmits the failed frame in suitable subperiods, and
Figure 2(a) shows a flow chart of a received frame by using the failed frame can be retransmitted effectively.
TDMA. Let SX present the average signal strength when the The computation complexity of our detection operation on
hub receives a frame from the sensor node X. First, the hub the receiver is analyzed as follows. Let n be the number of
determines body shadowing when it receives a frame. If SX devices supervised by the receiver, and m be the number of
is less or equal to threshold (ST hreshold ), the hub broadcasts device in collision. The length of the PHY frame is assumed
a special negative acknowledgement frame (NACK) to notify to be k bits. First, the preamble (0000...0) is correlated to
and stop the retransmission procedure of X. One flag bit in the the received signal in a sliding window manner, sliding from
ACK frame is used to distinguish the NACK from the standard the beginning to the end of the whole overlapped signal. The
ACK. In addition, if a sensor node wants to transmit/receive length of the whole overlapped signal with m collided frames
a frame, it has to awake. Otherwise, this sensor node enters is at most (m × k − (m − 1)) bits. Then, the procedures
sleep mode for power saving. Hence, the NACK frame is only after preamble is detected are conducted at most m times
received by the awake device. Otherwise, if the hub receives (if the preamble from all m collided frames is all detected,
the frame from X and SX is larger than ST hreshold , it uses which is generally not possible). In conclusion, the overall
cyclic redundancy check (CRC) in the receiving frame for time complexity of the detection preamble is O(m × k).
error detection.
Figure 2(b) shows a flow chart of a received frame by
using CSMA/CA. First, the hub determines the data collision B. Dynamic Superframe
situation when it receives a frame. Upon the occurrence of Based on different priority frames of the sensor nodes, a
overlapped signals, the preamble of the frame is first detected superframe is classified into four modes (A to D) as shown in
from the overlapped signals. CL-RRS uses the signal features Figure 3. In mode A, the superframe is composed of EAP and
of the PHY layer during received frames to identify the RAP, and the frame is transmitted only by contention scheme
precise reason of failed transmission on the receiver. For in different subperiods according to the frame priority.
frame collision, there is a very short stable signal period at In mode C, the MAP is inserted before the EAP in the
the beginning (Sstable b ), an unstable signal strength in initial superframe to periodically provide the frame transmission. If
received signal (Sunstable b ), and an unstable signal strength a frame is unsuccessfully transmitted in the MAP, and the
(Sunstable ) during the frame transmission. failed frame also retransmits by using contention scheme in
There are two situations in which a receiver cannot correctly following the EAP or the RAP. In general, the failed frame
receive a data frame from a sender. They may be frame will be arranged to the MAP of next superframe, and it
collision and human body shadowing. The collision signal occupies transmission resources of other sensor nodes and
strength compared with normal signal strength may be higher, increases the frame drop rate because of long transmission
lower or offset. Thus, it is hardly to judge the reasons of the delay. Thus, C mode increases the successful probability of
failed frame. In addition, the sender ID of the failed frame frame transmission and decreases transmission delay.
does not easily judge from overlapped signals. To compare mode B and D to mode A and C, the MAP1 is
First, the overlapped signals are unstable and they may inserted after the RAP in the superframe. The design avoids
cause by frame collision or human body shadowing. These high priority frames to occupy the transmission resources, and
different reasons lead to different signal intensity. For human starvation occurs in low priority frames. Thus, the low priority

1530-437X (c) 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JSEN.2016.2523461, IEEE Sensors
Journal

JOURNAL OF LATEX CLASS FILES, VOL. XX, NO. X, DECEMBER XXXX 5

Receive Frame

Confirm Frame Collision

If (Sunstable Y
&&Sunstable_b&&Sstable_b
)

Receive Frame N
Analyze Sender ID

Confirm Shadowing by the

N
Body If (IDSender==IDTable)

Send NACK to Stop the Y

Y Confirm Shadowing by the
Sensor (X) Transmitting If (Sx≤SThreshold)
Body
During a Time Period
N Send NACK to Stop the Y
Confirm Correctness of the Sensor (X) Transmitting If (Sx≤SThreshold)
During a Time Period
Frame N
Confirm Correctness of the
Frame
N
If (FCS==true) Drop Frame N
If (FCS==true)
Y
Y
Drop Frame

Complete Receiving Frame Complete Receiving Frame

(a) Flow chart of TDMA. (b) Flow chart of CSMA/CA.

Fig. 2. Flow charts of frames received.

Condition (2)& Condition (1)&

!(Condition (4)) !(Condition (4))
B EAP RAP

Condition (1)
A. Contention Mode
A C
Condition (2)
B EAP RAP MAP1

Condition Condition Condition Condition

B. Contention-Based Retransmission Mode (3) (4) (3) (4)

B MAP EAP RAP

Condition (1)
B D
C. Mix Mode Condition (2)

B MAP EAP RAP MAP1

Condition (2)& Condition (1)&
!(Condition (3)) !(Condition (3))
D. Mix-Based Retransmission Mode
Fig. 4. The state transition diagram for different superframe structures.

Fig. 3. Four types of superframe structures. 1 bit 2 bits 1 bit 4 bits

Non-final Fragment /
Protocol ACK
... Cancel / Scale / Reserved
Vertion Policy
Inactive

frame uses TDMA in MAP1 to access channel, increase

Fig. 5. The Frame Control Field in IEEE 802.15.6.
transmission probability, and decrease failed transmission rate.
Figure 4 illustrates the state transition diagram of different
superframe modes. The diagram consists of four states which the reserve bit is set to 1.
are four kinds of the superframe structures as shown Figure Our proposed scheme can be implemented in the stan-
3. The state transition depends on four conditions as shown dard IEEE 802.15.6 MAC protocol without adding any new
Figure 4. The four conditions are described as follows. message types and modifying the communicating procedure.
1) Periodic data has transmission request. Since the Frame Control Field can be found in all types of
2) Non-periodic data has transmission request. IEEE 802.15.6 messages [9], the reserved bits in the Frame
3) The hub checks the reserved bit of the frame header for Control Field are utilized in Figure 5. The sensor node uses
all sensor nodes with using MAP1. If it is set to 0, the the reserved bits to notify the hub more frames to continuously
data transmission of the node is completed. transmit.
4) A node is marked with the failed frame transmission and The hub checks the body shadowing status of each node

1530-437X (c) 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JSEN.2016.2523461, IEEE Sensors
Journal

JOURNAL OF LATEX CLASS FILES, VOL. XX, NO. X, DECEMBER XXXX 6

to allocate the transmission resources in the subperiods of probability of i − 1 frame transmission. The Psucess is the
the superframe. Once the body shadowing occurs, the sensor probability that the sender successfully transmits a frame
node suspends to retransmit the failed frame. In addition, the without collision. It is equal to (1 − PSC ) ∗ (1 − PF E,x ).
hub considers the awakening status of the sensor node as it The PT xF ail (i) is the probability that the sender has to
allocates MAP and MAP1 to the sensor nodes to retransmit transmit
Pi thej failedi−j frame passing through i times. It is equal
the failed frames. The reason is to avoid that the hub allocates to j=0 PSC ∗ Pf ail .
resource to the sensor node, but it still stays on suspended The PSC is the sum of the collision probability that two
status. On the contrary, the hub allocates transmission resource to P nE frames are simultaneously transmitted. It is equal
nE
to RAP. to i=2 P(pc,i) . The Pf ail is the probability that a frame
transmission fails without data collision, and it is equal to
(nE + 2) ∗ SIF S + BOavg ∗ TCSM Aslot (1 − PSC ) ∗ PF E,x . The P(pc,i) is the collision probability that
NslotEAP = ⌈
TSF slot i contending frames to simultaneously access the channel, and
PnE (1)
2 ∗ max(TEdata1 , TEdata2 , ..., TEdatan ) + i=1 TEdatai
E
it is equal to P(pc,i) = PC,i (BO) ∗ Pnc,i . For example, if three
+ ⌉
TSF slot frames are transmitted simultaneously, the collision probability
is Ppc,3 = PC,3 (BO) ∗ (PcF,1 ∗ pcF,2 + PcF,1 ∗ PcF,3 + PcF,2 ∗
The four superframe structures contain the EAP which
PcF,3 ).
should be allocated carefully. Once the sensor node has not
high priority frames, but the hub still allocates the transmission Pc,k (BO) [19] is the collision probability of k contending
resource to the EAP. Thus, it results in resource waste. In frames with BO, and is derived by recursion as shown in the
addition, IEEE 802.15.6 defines that CW size is set to 1 with following:
the first and the second frame transmission in the EAP. If two 
1 k−1 1
or more nodes simultaneously transmit a frame in the EAP, 
hBO
 + Pc,k−1 (BO) ∗iBO +
frame collision occurs. Thus, in this paper, the slot number of 1 k−2

PC,k−2 (BO) − BO ∗ Ck , k ≥ 3

Pc,k (BO) = (5)
the EAP provides that nodes have a chance to retransmit the  1 ,k = 2

 BO
highest priority frame as frame collision occurs. Hence, we

0, k=1
define NslotEAP , refereed to (1), as the number of allocation
BO denotes the size of the CW for a frame transmission.
slots in the EAP.
Pnc,i is the probability of i contending frames are transmitted
NslotEAP includes the time of twice collisions, the trans- by nE nodes in this period and derived as follows:
mission time of the highest priority frame of the node, the
short interframe space (SIF S), and average backoff inter-
(nE −i+1) (nE −i+1)+1
val. 2 ∗ max(TEdata1 , TEdata2 , ..., TEdatanE ) is the maximum Pnc,i =
X
PcF,a1
X
PcF,a2 . . .
time
PnE of the highest priority frame with twice collisions and a1 =1 a2 =a1 +1
(6)
i=1 TEdatai is the transmission time of the highest priority (nE −i+1)+(i−2)
X
frame. TEdata1 is defined that the transmitting duration of a PcF,ai −2 .
frame with the highest priority. nE is defined that the number ai−2 =ai−3 +1

of the sensor nodes with the highest priority frame. TSF slot According to (1) to derive the requested resource of the
is the time of a slot in the superframe. EAP, the coordinator also has to retain a minimum trans-
BOavg is average backoff interval in the EAP. It is defined mission time of the RAP (NslotRAP min ) because the sensor
as an interval from the time that a sensor node attempts to node can send a resource request to the hub in the RAP. It
access the channel for a frame to the time that the sensor node is composed of SIF S, the backoff time of the maximum
actually transmits the frame over the channel. The backoff contention window size (CWMax ∗ TCSMA slot ), the trans-
intervals, b(0) and b(1), are a successful frame transmission mitting duration of a resource request frame (Tconnec request ),
at the first time and the second time, respectively. BOavg is and the transmitting duration of a assignment resource frame
expressed as follows: (Tconnec assign ). The NslotRAP min can be derived as
T xLimit
X
BOavg = ⌈b(0) + b(1) + PT x (i) ∗ b(i − 1)⌉. (2) SIF S + CWmax ∗ TCSM A slot
i=3 NslotRAP min
=⌈
TSF slot
(7)
b(i) is the expected value of a backoff interval before a Tconnec request + Tconnec assign
+ ⌉.
successful transmission until the i-th times and expressed as TSF slot
i−1 CW
1 X
! ( In each superframe structure, the subperiods are allocated
X CW = 2CWpre , if j = 2, 4
b(i) = k , (3) under (7) after allocating the EAP. We discuss how to allocate
CW k=1 CW = CWpre , if j = 0, 1, 3.
j=0
the resources in the subperiods of each superframe structure. In
PT x (i) denotes the successful probability of a frame transmis- contention mode (mode A), the size of the RAP is the number
sion until continuously i times. It can be expressed as of allocation slots of the superframe minus the transmission
duration for the B, and the number of allocation slots in the
PT x (i) = Psuccess ∗ PT xF ail (i − 1). (4)
EAP. The equation is NslotRAP = NslotSF − NslotBeacon −
This value is derived from the successful probability of NslotEAP .
a frame transmission until i-th times and the unsuccessful In the contention-based retransmission mode (mode B),

1530-437X (c) 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JSEN.2016.2523461, IEEE Sensors
Journal

JOURNAL OF LATEX CLASS FILES, VOL. XX, NO. X, DECEMBER XXXX 7

we allocate MAP1 according to the (Nslotf ail ). Nslotf ail is

defined the number of allocation slots to retransmit the failed Blood
Pressure ECG
frame in the EAP and the RAP of the superframe. Then,
we calculate the number of allocation slots in the RAP. If
the transmission slot of the superframe can not satisfy the
Hub
resource request of the RAP, we first decrease the resource of
the MAP1 with the lowest priority frames until the resource
fits the restriction. The number of allocation slots in the RAP Glucose Drug
is NslotSF − NslotBeacon − NslotEAP − NslotM AP 1 . Monitor Delivery
In the mix mode (mode C), we allocate the MAP according
to the transmission request of TDMA (NslotT DM A ). Then, we
Fig. 6. The scenario for the analytical model.
calculate the number of allocation slots in the RAP. If the
transmission slot of the superframe is not enough, we first
decrease the resource of the MAP with the lowest priority The parameters are defined on the basis of the parameter
frames until the resource fits the restriction. The number setup in [6][11][20][10] and summarized in Table II. ECG
of allocation slots in the RAP is NslotSF − NslotBeacon − is composed of six leads. Each lead samples 500 times per
NslotEAP − NslotM AP . second and 12 bits per sample. Each sample is transmitted two
In the mix-based retransmission mode (mode D), we assign times in order to increase reliability. The superframe duration
the MAP1 and the MAP according to the contention-based is set to 125 ms.
retransmission mode and the mix mode, respectively. Once
the transmission slot of the superframe can not satisfy the
B. Average Transmission Times
resource request of the RAP, we first decrease the transmission
slots of the MAP because the MAP1 is allocated to transmit We now introduce notations used later in our analysis. The
emergency frames. However, if the worst comes to happen, average transmission times of a frame for sensor x, denoted
we decrease the transmission slots of the MAP1. The number by T xAvg as an output measure of our analysis, is defined as
of allocation slots in the RAP is NslotSF − NslotBeacon − the ratio of the total transmission times of transmitted frames
NslotEAP − NslotM AP − NslotM AP 1 . in all subperiods over the total number of transmitted frames
The above resource allocation scheme guarantees to transmit in the superframe. Some notations which used in this analysis
the highest priority frames. Besides, it also minimizes the are listed as follows:
resource waste as the periodic data and aperiodic data contend • T xMAP : The average transmission times of a frame in
the transmission resource. In section V, we show the average the MAP
transmission times and the duty cycle of sensor nodes to • T xEAP : The average transmission times of a frame in
demonstrate the capability of the proposed strategy as well the EAP
as reducing the energy consumption. • T xRAP : The average transmission times of a frame in
the RAP
IV. P ERFORMANCE A NALYSIS • T xMAP 1 : The average transmission times of a frame in

In this section, we propose an analytical model of CL-RRS. the MAP1

• NF rame : The number of transmitted frames in the super-
The model is assumed to be time slotted, and all time durations
are normalized by slot. We first describe our system model, frame
Shadow
• Px : The probability of the sensor x transmission
and then the average transmission times of a frame T xAvg is
analyzed. with human body shadowing
Shadow
• PMAP : The probability of the hub without allocated
resources to the MAP of the next superframe whthin
A. System Model human body shadowing
The wireless body area network is a star topology with N • PF E,x : Frame error rate of the sensor x
Shadow Shadow Shadow
devices as shown in Figure 6. The system includes electrocar- • Pef f ect : The product of Px , PF E,x , and PMAP
diogram sensor (ECG) nodes, glucose monitor sensor nodes, • N slotx : The number of allocation slots to x for trans-
blood pressure sensor nodes, drug delivery sensor nodes, and a mitted a frame
MAP
hub. Different sensor nodes use different transmission scheme • NT x : The number of transmitted frames by the sensor
to transmit body information in different subperiods. The ECG in the MAP
EAP
sensor, the glucose monitor sensor, and the blood pressure • NT x : The number of transmitted frames by the sensor
sensor sense regular information of the human body. These in the EAP
RAP
nodes use hybrid scheme (i.e., TDMA and CSMA/CA) to • NT x : The number of transmitted frames by the sensor
transmit data. The drug delivery sensor senses the level of in the RAP
MAP 1
blood sugar and then the drug is injected based on the level • NT x : The number of transmitted frames by the sensor
of the blood sugar. It only uses CSMA/CA to transmit data. in the MAP1
When the hub successfully receives the data frame from the • T xLimit: The maximal number of transmitted frames
sensor node, it replies an ACK back to the sensor node. as a sensor using CSMA/CA to transmit frames in the

1530-437X (c) 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JSEN.2016.2523461, IEEE Sensors
Journal

JOURNAL OF LATEX CLASS FILES, VOL. XX, NO. X, DECEMBER XXXX 8

TABLE II
S YSTEM PARAMETERS USED IN THE SIMULATION MODEL

ECG Glucose Monitor Blood Pressure Drug Delivery

Data Rate 72 kbps 1 kbps 1.2 kbps 16 kbps
Latency 250 ms 250 ms 250 ms 250 ms
Periodic Data Yes Yes Yes Yes
PShadow,x 7% 5% 10% 5%
Data Priority 6 7 5 7
Frame Error Rate 27.9% 9.8% 29.8% 10.3%

period
MAP
• Nremain : The number of remaining frames after a sensor T xLimit
X
transmitting frames in the MAP T xEAP = NTEAP
x ×( i × PT x (i)+
i=1
(12)
EAP
• Nremain : The number of remaining frames after a sensor
T xLimit × PT xF ail (T xLimit))
transmitting frames in the EAP
RAP
• Nremain : The number of remaining frames after a sensor T xLimit
transmitting frames in the RAP
X
T xRAP = NTRAP
x ×( i × PT x (i)+
i=1
(13)
In the analysis, we analyse the transmission times of frames T xLimit × PT xF ail (T xLimit)).
in each subperiod to calculate the total transmission times of
PT x (i) and PT xF ail (T xlimit) refer to (4) of Section III.
frames in the superframe. In the following, we analyze T xAvg RAP RAP
T xMAP 1 is equal to Nremain . Nremain denotes the number
for CL-RRS scheme and derive as
of frames, and they do not be transmitted in the RAP. These
frames will be transmitted in following MAP1 to satisfy the
T xM AP + T xEAP + T xRAP + T xM AP 1 requests of transmission deadline by using TDMA. Nremain RAP
T xAvg = . (8)
NF rame is expressed as follows:
If frames cannot completely transmit in the MAP, they are T xLimit
X
going to transmit in the following EAP, RAP, and MAP1. First, RAP
Nremain = NTRAP
x × (1 − PT x (i)). (14)
the T xMAP in the MAP is expressed as follows: i=1

  However, frames may not be successfully transmitted in

Shadow
T xM AP = NTMxAP × 1 − Pef MAP 1
f ect . (9) the MAP1 because of the human body shadowing. Nremain
  denotes the number of frames when these frame are unsuc-
NTMAPx × 1 − P Shadow
ef f ect represents that the number of cessfully transmitted in the MAP1. It is expressed as follows:
successfully transmitted frames in the MAP. In other words,
they do not be affected by the human body shadowing. M AP 1 = N M AP 1 × P
Nremain Tx F E,x + (NF rame − T xM AP 1 ). (15)
Shadow
Pef f ect is expressed as follows: EAP RAP
Finally, Nremain is similar to Nremain , it is expressed as
Shadow Shadow Shadow
Pef f ect = Px × PF E,x × PM . (10)  
× 1− T
AP EAP
P xLimit
Nremain = NTEAP
x i=1 PT x (i) . (16)

Equation (10) represents the probability that the sensor node To validate the correctness of the analysis, we perform
Shadow
x transmits a frame with the human body shadowing. PMAP a simple simulation of CL-RRS to verify analytical results.
is defined as the probability that the hub does not allocate the We use a simulation environment as shown in Figure 6. The
resource to the MAP of next superframe because of the human simulation parameters are given based on Table II and applied
body shadowing. It is expressed as to (9) to (16).
Figure 7 shows T xAvg under the results of different sensors
Shadow N slotShadow − (N slotM AP − N slotx )
PM AP = , (11) nodes (i.e., ECG, glucose monitor, blood pressure, and drug
N slotSF
delivery) obtain from analysis and simulation experiments of
where N slotShadow is the number of allocation slots because CL-RRS. In addition, the system is defined as the average
of the human body shadowing. N slotMAP and N slotSF are results of all sensor nodes. From this figure, we observe that
the number of slots in the MAP and the superframe. the analytical results are very close to the experimental results.
Similarly, we calculate T xEAP and T xRAP in the EAP and The figure shows the difference between the analysis and
RAP, respectively. The average transmission times of frames simulation results which are about 0.56%. We find a special
are equal to the number of transmitted frames multiplied by phenomenon for difference between the analysis and simula-
the expected transmission times of a frame. They are expressed tion results. If a sensor node uses CSMA/CA to transmit a
as follows: frame, it shows an obvious difference (e.g., drug delivery) in

1530-437X (c) 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JSEN.2016.2523461, IEEE Sensors
Journal

JOURNAL OF LATEX CLASS FILES, VOL. XX, NO. X, DECEMBER XXXX 9

2
Simulation Results

Average Transmission Times of a Frame (Times)

Analytical Results 100
1.8
Theoretical Value
98 CL−RRS
1.6
96 IEEE 802.15.6 Std.
1.4
94

Goodput (kbps)
1.2
92
1
90
0.8
88
0.6
86
0.4
84
0.2
82
0
ECG Glucose Monitor Blood Pressure Drug Delivery System 80
Different Types of Senders 1 2 3 4 5 6 7 8 9 10
The Number of Times

Fig. 7. Comparison of the analysis and simulation results.

Fig. 8. Effects of the number of running times on system goodput.

Figure 7. This is because in the analysis results, we assume 2

contention window is a fixed value. In fact, contention window CL−RRS

Frame Loss Ratio (%)

is not a fixed value and depends on the priorities of transmitted IEEE 802.15.6 Std.
1.5
frames. In addition, IEEE 802.15.6 defines CW size as 1 with
the first and the second frame transmission in the EAP. In 1
other words, if two or more nodes simultaneously transmit
a frame in the EAP, frame collision occurs. As a result, the
0.5
drug delivery is the highest priority in the EAP, and the frame
encounters more collisions. However, our analytical model has
0
been validated against the simulation experiments. ECG Glucose Monitor Blood Pressure Drug Delivery System
Different Types of Sensors
V. N UMERICAL E XPERIMENTS (a)
500
This section is to evaluate the performance of the proposed Theoretical Value
CL-RRS. The current discrete-event network simulators such 400 CL−RRS
IEEE 802.15.6 Std.
Latency (ms)

as OPNET, NS-2, NS-3, J-Sim, and GloMoSim are unsuitable

to implement a cross-layer solution, since their inner structures 300
are based on a layering design, and each implemented func-
200
tionality run by the simulator engine is tightly coupled with
this structure. Hence, implementing a cross-layer solution in 100
one of these simulators may turn into a non-trivial task. For
this reason, we develop a simulator which easily incorporates 0
ECG Glucose Monitor Blood Pressure Drug Delivery System
cross-layer algorithms and protocols. Different Types of Sensors
Our simulator uses C language to implement IEEE 802.15.6 (b)
specification. For a WBAN, it is a non-light of sight environ-
ment because human body consists of muscle, fat, apparatus Fig. 9. Effects of different sensor nodes on frame loss rate and transmission
and so on. In the PHY layer of our simulation environment, a latency.
realistic channel condition is adopted [18]. For the MAC layer,
the simulator follows the specification of IEEE 802.15.6.
In our simulation model, each sensor node maintains a FIFO rate should be under 5% [9]. Thus, we show that the frame
waiting buffer of infinite frames, transmits a frame at fixed 125 loss rate of CL-RRS meets this criteria. In addition, we show
ms, and the frame length is depending on different sense nodes the performance of CL-RRS and IEEE 802.15.6. The primary
according to Table II. Each sensor node estimates the channel performance metrics are goodput, duty cycle, and the average
gain of its link to the hub from the received signal portion of transmission times of a frame. Duty cycle is defined as the
the preamble sequence. The channel estimation is then applied ratio of the active period of a sensor node over the period
in the PHY layer to detect the remaining data contained in the of all running time. If the duty cycle is low, the sensor node
PHY frame. spends less power consumption and extends working time.
IEEE 802.15.6 standard does not define how to consist Figure 8 shows the effect of the number of running times
the subperiods of a superframe [9]. In our simulation, the on system goodput. The theoretical value is composed of data
superframe is composed of MAP, EAP, RAP, and MAP1. rate of all sensor nodes. We observe that the goodputs of CL-
The superframe duration is set to 125 ms, and the running RRS, IEEE 802.15.6, and the theoretical value are similar.
time is set to 5,000 superframe duration (i.e., 600,000 slots). The error rate of goodput for CL-RRS and the theoretical
Recalled Table II, the application parameters are summarized value is about 0.35% because we consider the frame error rate
from [6][10][11][20]. IEEE 802.15.6 specifies the frame loss and the transmission deadline of a frame in the simulation.

1530-437X (c) 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JSEN.2016.2523461, IEEE Sensors
Journal

JOURNAL OF LATEX CLASS FILES, VOL. XX, NO. X, DECEMBER XXXX 10

60 3
CL−RRS

Average Transmission Times of a Frame (Times)

CL−RRS
IEEE 802.15.6 Std. IEEE 802.15.6 Std.

50 2.5

40 2
Duty Cycles (%)

30 1.5

20 1

10 0.5

0 0
ECG Glucose Monitor Blood Pressure Drug Delivery System ECG Glucose Monitor Blood Pressure Drug Delivery System
Different Types of Sensors Different Types of Sensors

Fig. 10. Effects of different sensor nodes on duty cycle. Fig. 11. Effects of different sensor nodes on average transmission times of
a frame.

0.7
CL−RRS
Although the goodput is similar to the other schemes, CL- IEEE 802.15.6 Std.
0.6
RRS spends less power consumption and also meets the

Power Consumption (mW)

transmission deadline. 0.5

Figure 9 (a) shows the effect of different sensor nodes on the 0.4

frame loss rate for CL-RRS and IEEE 802.15.6, respectively.

0.3
In this figure, the frame loss ratio of CL-RRS is 0.1% greater
than that of IEEE 802.15.6 for glucose monitor and blood 0.2

pressure sensors. In the other word, the frame loss rate of 0.1

all sensor nodes can meet the request of IEEE 802.15.6.

Recalled that glucose monitor and blood pressure sensors use 0
ECG Glucose Monitor Blood Pressure Drug Delivery System
Different Types of Sensors
the hybrid scheme (i.e., TDMA and CSMA/CA) to transmit.
When the transmission is unsuccessful in the first superframe, Fig. 12. Effects of different sensor nodes on power consumption.
they retransmit the failed frame in the special period of next
superframe. The transmission deadline is set to 250 ms. As
a result, if the frame cannot be successfully transmitted in consumption and frame latency. On the contrary, CL-RRS
the interval of two superframes, the frame is dropped. Thus, effectively avoids the human body shadowing based on the
in CL-RRS, the frame loss ratio of glucose monitor and signal strength and decreases the duty cycle.
blood pressure sensors is a slightly higher than that of IEEE Figure 11 shows the effect of different sensor nodes on
802.15.6. the average transmission times of a frame. We observe an
Figure 9 (b) shows the transmission latency of all sensor obvious difference between CL-RRS and IEEE 802.15.6 for
nodes for CL-RRS and IEEE 802.15.6, respectively. The the average transmission times of a frame. The reason is CL-
transmission latency is defined as the interval between the time RRS distinguishes the collision case from all transmission
that a sensor node transmits the frame and the time that the failure cases. Though the probabilities are not 100%, CL-RRS
frame is successfully transmitted to the hub. The theoretical can also improve the retransmission efficiency significantly.
value is set to 250 ms based on Table II. To compare the In CL-RRS, a frame can be delivered successfully with a low
transmission latency of CL-RRS with IEEE 802.15.6, CL-RRS opportunity of failed frame transmission, which decreases the
is lower than IEEE 802.15.6. As a result, their transmission transmission overhead implicitly for a frame transmission.
latency can meet the transmission deadline for all sensor In WBANs, power consumption is a major issue, so that
nodes. The transmission latency of the system for CL-RRS we further show the effect of different sensor nodes on power
is about 42 ms lower than that of IEEE 802.15.6. Summarily, consumption in Figure 12. The power consumption is defined
we observe the transmission latency of CL-RRS satisfies the as the total power consumption for transmission sensor data
transmission deadline of a frame, and obtains better goodput over successful transmitted frames. We find the power con-
and the frame loss rate for IEEE 802.15.6. sumption of CL-RRS is lower than that of IEEE 802.15.6 for
Figure 10 shows the effect of different sensor nodes on all sensor nodes. Thus, CL-RRS effectively avoids the human
the duty cycle. We find the duty cycle of CL-RRS is lower body shadowing and decreases the power consumption.
than that of IEEE 802.15.6 for all sensor nodes. There is
5% difference between CL-RRS and IEEE 802.15.6 for the VI. C ONCLUSIONS
drug delivery sensor node. The maximum difference is 12% A simple and efficient cross-layer reliable retransmission
for the ECG sensor node. The difference of the system is scheme, CL-RRS, has been proposed in this paper to solve
18% between CL-RRS and IEEE 802.15.6. The reason is the retransmission problem of the human body shadowing
that IEEE 802.15.6 continually transmits the failed frame in in IEEE 802.15.6-based WBANs. The proposed scheme first
the transmission duration and encounters the human body detects frame signal and determines the reasons of failed
shadowing. This also results additional bandwidth, power frame transmission in the PHY layer, and then adopts different

1530-437X (c) 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JSEN.2016.2523461, IEEE Sensors
Journal

JOURNAL OF LATEX CLASS FILES, VOL. XX, NO. X, DECEMBER XXXX 11

subperiods in the MAC layer to retransmit the failed frame. [18] D. S. J. Dong, “Cooperative receive diversity for coded gfsk body-
The proposed CL-RRS significantly reduces the chance of area communications,” Electronics Letters, vol. 47, pp. 1098–1100(2),
September 2011.
failed transmission. Our simulation results show that CL-RRS [19] H.-W. Tseng, A.-C. Pang, J. Chen, and C.-F. Kuo, “An adaptive
effectively improves the frame error rate and the average contention control strategy for ieee 802.15.4-based wireless sensor
transmission time, as well as reducing the power consumption. networks,” Vehicular Technology, IEEE Transactions on, vol. 58, no. 9,
pp. 5164–5173, Nov 2009.
Therefore, CL-RRS is a simple and efficient approach for [20] J. Khan and M. Yuce, “Wireless body area network (wban) for medical
enhancing IEEE 802.15.6 protocol in wireless body area applications,” New Developments in Biomedical Engineering, Domenico
networks. Campolo (Ed.), 2010.

ACKNOWLEDGMENT
This work was supported in part by Ministry of Science and
Technology (MOST) under Grant 104-2221-E-005-011 and by
Hwa Hsia University of Technology under Grant HWH105-H-
007.
Hsueh-Wen Tseng received his Ph.D. degrees
in Computer Science and Information Engineering
R EFERENCES from National Taiwan University, Taiwan, in 2009.
Currently, he is an Associate Professor of Computer
[1] “Worldlifeexpectancy,” http://www.worldlifeexpectancy.com/history-of- Science and Engineering in National Chung Hsing
life-expectancy. University. His research interests include networked
[2] “World health organization,” http://www.who.int. embedded system; cloud computing; design, anal-
[3] I. Akyildiz, W. Su, Y. Sankarasubramaniam, and E. Cayirci, “Wireless ysis and implementation of network protocols and
sensor networks: a survey,” Computer Networks, vol. 38, no. 4, pp. 393 wireless communications.
– 422, 2002.
[4] K. Sayrafian-Pour, W.-B. Yang, J. G. Hagedorn, J. E. Terrill, K. Y.
Yazdandoost, and K. Hamaguchi, “Channel models for medical im-
plant communication,” International Journal of Wireless Information
Networks, vol. 17, no. 3-4, pp. 105–112, 2010.
[5] D. Smith, D. Miniutti, T. Lamahewa, and L. Hanlen, “Propagation
models for body-area networks: A survey and new outlook,” Antennas
and Propagation Magazine, IEEE, vol. 55, no. 5, pp. 97–117, Oct 2013.
[6] A. Boulis, D. Smith, D. Miniutti, L. Libman, and Y. Tselishchev, “Chal-
lenges in body area networks for healthcare: the mac,” Communications
Magazine, IEEE, vol. 50, no. 5, pp. 100–106, May 2012.
[7] “Standard for part 15.4: Wireless medium access control (mac) and Ruei-Yu Wu received the B.S. degree in Information
physical layer (phy) specifications for low rate wireless personal area Management from the Fu Jen Catholic University,
networks (lr-wpans),” IEEE Std 802.15.4, Sep 2006. Taiwan, in 1999, the M.S. degree in Computer
[8] “Specification of the bluetooth system version 4.0,” 2010. Science and Information Engineering from the Na-
[9] “Ieee standard for local and metropolitan area networks - part 15.6: tional Chi Nan University, Taiwan, in 2001 and
Wireless body area networks,” IEEE Std 802.15.6-2012, pp. 1–271, Feb the Ph.D. degree in the Department of Computer
2012. Science and Information Engineering at the National
[10] B. Latré, B. Braem, I. Moerman, C. Blondia, and P. Demeester, “A Taiwan University in 2012. Currently, she is an
survey on wireless body area networks,” Wirel. Netw., vol. 17, no. 1, pp. assistant professor in the Department of Manage-
1–18, Jan. 2011. ment Information Systems Hwa Hsia University of
[11] R. Cavallari, F. Martelli, R. Rosini, C. Buratti, and R. Verdone, “A survey Technology. Her currently research interests focus
on wireless body area networks: Technologies and design challenges,” on cloud computing, wireless communications, and algorithms.
Communications Surveys Tutorials, IEEE, vol. 16, no. 3, pp. 1635–1657,
2014.
[12] S. Movassaghi, M. Abolhasan, J. Lipman, D. Smith, and A. Jamalipour,
“Wireless body area networks: A survey,” Communications Surveys
Tutorials, IEEE, vol. 16, no. 3, pp. 1658–1686, Third 2014.
[13] C. Tachtatzis, F. Di Franco, D. Tracey, N. Timmons, and J. Morrison,
“An energy analysis of ieee 802.15.6 scheduled access modes,” in
GLOBECOM Workshops (GC Wkshps), 2010 IEEE, Dec 2010, pp.
1270–1275.
[14] S. Rashwand, J. Misic, and H. Khazaei, “Performance analysis of Yi-Zhang Wu is a software engineer in the QNAP
ieee 802.15.6 under saturation condition and error-prone channel,” in Inc., Taiwan (since March 2015). He received the
Wireless Communications and Networking Conference (WCNC), 2011 M.S. degree in Computer Science and Engineering
IEEE, March 2011, pp. 1167–1172. from the National Chung Hsing University, Taiwan,
[15] V. Rampa, S. Savazzi, M. Nicoli, and M. D’Amico, “Physical modeling in 2015. His research interests include sensor net-
and performance bounds for device-free localization systems,” Signal works, cloud computing and protocol design in wire-
Processing Letters, IEEE, vol. 22, no. 11, pp. 1864–1868, Nov 2015. less personal networks. Now, he works for Network
[16] D. Goswami, K. Sarma, and A. Mahanta, “Experimental determination Video Recorder System for one year in QNAP Inc.,
of path loss and delay dispersion parameters for on-body uwb wban Taiwan.
channel,” in Signal Processing, Informatics, Communication and Energy
Systems (SPICES), 2015 IEEE International Conference on, Feb 2015,
pp. 1–4.
[17] R. Pan, D. Chua, J. Pathmasuntharam, and Y. P. Xu, “An opportunistic
relay protocol with dynamic scheduling in wireless body area sensor
network,” Sensors Journal, IEEE, vol. 15, no. 7, pp. 3743–3750, July
2015.

1530-437X (c) 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.