- 1 -
中国科技论文在线
A Multiple-hop Synchronization Protocol with Packet
Reconstitution based on Body Sensor Networks
Chen Zifei, Liu Lan
(School of Information Engineering, Wuhan University of Technology, WuHan 430070)
Abstract: This paper proposes a multi-hop synchronization protocol for multiple physiological
information transmission over the body sensor network (BSN). A packet reconstitution mechanism was
designed to achieve synchronous transmission. The experimental results validated the efficiency of the
protocol on continuous real-time monitoring of multiple physiological parameters over multi-hop
BSNs.
Keywords:BSN; Protocol; Synchronization
0 Introduction
According to demographic survey, the number of elderly people and those who live alone is
increasing rapidly [1]. Therefore continuous real-time health monitoring is in great demand.
However, this requires the development of bio-signal acquisition and processing devices that are
capable of supporting people, especially patients, in their everyday life [2]. Body sensor network
(BSN) allows low-cost, continuous and ambulatory health monitoring with real-time updates of
medical records. Attribute to the development of low power integrated circuits and wireless
communication, BSN systems show up a favorable adaptability to a variety of applications. Much
work has been done on BSN and most of them aimed at making BSN systems power-efficient,
intelligent and reliable. For example, the work in [3, 4] proposed a context-aware sensing, allowing
BSN to react to different events and adapt its monitoring behavior. The system [5] reduced
considerable power consumptions by using flexible and low power design.
The paper [6] which was published by my coworkers proposed a BSN platform for pervasive
healthcare. Various physiological parameters were collected with this BSN platform and
transmitted wirelessly to the base station. The base stations sent data hop-by-hop to the data hub as
shown in . The synchronization was not considered.
The changes of node number and locations could make a great impact on data transmission if
a time synchronization mechanism is inappropriate. Due to limited energy, memory and
computational resources available on typical BSN nodes, the traditional synchronization protocols
used in wired networks are usually not suitable for BSN (., the Network Time Protocol (NTP)
or the Precision Time Protocol (PTP) [7, 8]). Some accurate wireless protocols such as the IEEE
1588v2 and the IEEE are not designed for BSN systems [9]. In the past few years, several
WSN-specific synchronization protocols have been proposed such as the Flooding Time
Synchronization Protocol (FTSP) [10], which mainly aimed at maximizing the synchronization
accuracy while reducing protocol complexities. An adaptive strategy to perform periodic
synchronizations for WSN nodes on the basis of both target accuracy and in-tolerance probability
has been described in [11]. Comparing with WSN, the related position of BSN nodes is fixed as
shown in , thus it is possible to adopt a less complex synchronization mechanism for BSN
systems.
- 2 -
中国科技论文在线
Fig. 1 Monitoring system
In this paper, we present a synchronization protocol for BSN without any acknowledgement
message. Based on a novel packet reconstitution mechanism, a common time reference has been
produced to ensure the synchronization of physiological data. By dividing the common time into
several time slices with the number of nodes, each node sent data on their own time slices, which
effectively avoids the conflicts of data transmission. In the following sections, a description of the
protocol is provided, including packet format design and communication architecture.
1 Protocol Design
Due to the characteristics of various physiological parameters, the communication protocol
has demands on real-time, data integrity, validity, continuity and synchronization.
Packet Format
A packet format with fixed length was adopted upon our hardware platform [6]. To satisfy the
demands of data transmission, state control and various request messages, the packet was designed
with a universal format. The definitions of the first 8 bytes in the data packet are illustrated in
and are described as follows:
Format of packet
z Packet Type: mark used to identify various function of a packet. The types of packet
included synchronization request packet, synchronization answer packet, various
physiological data packet, control packet.
z Source Address: mark of which node or base station sent the packet. The original source
address was the sensor node identity (ID). If the packet was received directly by central
base station (CBS), the source address would not be altered. On the contrary, if the
packet was transmitted layer by layer to the CBS as shown in , the source address
would be replaced by the ID of the relay station which had received the packet.
z Destination Address: mark of destination base station which would receive the packet.
The original destination address was the CBS ID. And it would be changed layer by
layer during the transmission as the source address.
- 3 -
中国科技论文在线
z Packet State: the identifying mark used to distinguish the original packet and the relay
packet. Every BS received the original packets which were sent directly by nodes. In
addition, each of nodes received the relay packets of which destination address was
accordant with their own IDs.
z Node ID: ID of the sensor node which collected the physiological parameters. Node ID
was unique in the BSNs.
z Synchronization SN: serial number of the SBN synchronization.
z Packet SN: serial number of the packet used to reject the same packet. The packet
numbers varied with the acquisition time of data.
z Data Length: number of physiological parameters that the packet carried. Different
physiological parameters need different data length to carry, thus the data length varied
with different physiological packet.
BSN Synchronization
Due to the fading and time varying characteristics of the wireless channel, it is difficult to
guarantee the transmission quality in wireless resource allocation. To reduce the disturbance
between signals and to improve the reliability of data transmission in wireless channel, time
sharing multiplex such as roll polling was a conventional method to be considered. However, the
cost of request time reduced the transmission efficiency. The synchronization mechanism
proposed in this paper made the nodes work in their own slices independently without any request
message.
Synchronization mechanism
The synchronization mechanism provided a common time reference. Time synchronization
had three possible cases: original synchronization, periodic synchronization and exception
synchronization. In the following sections, the description of the three cases was provided:
Original synchronization
At the beginning of synchronization, any node and the BS do not know the situation of the
BSN. As such, each of nodes sent their own synchronization requests messages (SR) periodically
in the BSN. During the intervals of sending, nodes received the SRs from other nodes as shown in
-a. Then each of nodes acquired the information of the BSN (., number of nodes, IDs of
other nodes, sampling frequency of other nodes, and so on) soon.
After that, the smallest ID node as a synchronization node sent a synchronization answer
message (SA) in the BSN as shown in -b. Although nodes in a BSN were so close that the
transfer time of wireless channel almost could be neglected, nodes still spending some time on
sending, receiving and processing the packet. In fact, the time delay offset
T
existed in all nodes
- 4 -
中国科技论文在线
except the synchronization node as described in . As so, to keep a common time reference with
other nodes, the synchronization node can delay a period as long as offset
T
.
' offsetT T T= + (1)
By assuming that all of nodes had received the SA, there would be a common time reference
in the BSN. After a while if SQ signals still existed in the BSN, we could know that some of nodes
have not received the SA. Consequently the synchronization node sent the SA again.
Periodic synchronization
Due to the different sampling frequency, sending frequency and processing speed between
various physiological nodes, the phenomenon of out of sync possible appeared over time. In such
a case, periodic synchronization was necessary. The periodic synchronization can be launched by
synchronization node with a SA.
Exception synchronization
Since the time slices were determined by the number of nodes, a BSN was impossible to
function most effectively with the time varying number of nodes. Exception synchronization was
used to process the change number of nodes:
Increase of nodes: Each of newly added nodes sent SRs in BSN just like the period of
original synchronization, which would stimulated a new round of original synchronization.
Decrease of nodes: Each of nodes was in the receiving mode beyond their own sending time.
Thus each of nodes has the ability to detect the change of number of nodes. And the decrease of
nodes would also trigger a new round of original synchronization.
Data Transmission
To avoid the conflicts of data transmission, each of nodes sent data on their own time slices.
Besides the respective node time slices, we set aside common offset time slices for lost packet.
As far as the notation is concerned, in the rest of this paper the following symbols will be
used:
z iN : total number of time slices in a period (., 1 second) as shown in .
z nL : the minimum number (in bytes) that can carry a single physiological parameter. nL
varied with the types of physiological parameters that collected by the node whose ID
was n .
z nF : sampling frequency of the node whose ID was n . Monitoring of different
physiological parameters needs different sampling frequency.
z pL : data number (in bytes) carried by a single packet. In this paper, the value of
pL was 24 bytes.
z offsetN : number of offset time slices for lost packet. In the period which was divided into
many time slices for all nodes, we set aside several time slices for sending lost packet.
1 1 2 2 n n
i offset
p p p
L FL F L FN N
L L L
⎡ ⎤ ⎡ ⎤ ⎡ ⎤⋅⋅ ⋅= + + + +⎢ ⎥ ⎢ ⎥ ⎢ ⎥⎢ ⎥ ⎢ ⎥ ⎢ ⎥⎢ ⎥ ⎢ ⎥ ⎢ ⎥
" (2)
As shown in sending time was divided into a series of T and a single T included
node time slices and offset time slices. The number of time slices was calculated by (2). For a
- 5 -
中国科技论文在线
single node, first we worked out the bytes number that can carry all the data collected in a T .
Second, according to the number of bytes and p
L
we got the minimum packets number that can
carry all the bytes. In the end, we counted up all the packets numbers and offset
N
, we got iN the
total number of time slices in a period.
Sending time slices
The synchronization node determined the time slices according to the SR. The information of
time slices was carried by the SA. Thus after received SA, each of nodes sent data one by one in
their own time slices. If CBS received the discontinuous packets, it would send the resending
request of lost packet in the BSN via BS. After received the resending request, the corresponding
node resent the lost packet in offset time slices.
Data Synchronization
For the data hub, the data synchronization is very important. It ensured the data hub would
fuse and analyze the data which was collected at the same time. Assume that two sensor nodes
was in BSN, one was Node X and the other was Node Y. Both of nodes separately collected two
serial data (“X1X2X3X4X5X6”,”Y1Y2Y3Y4”) and sent them in disparate time slices. The
process of data acquisition and sending was showed in :
Data acquisition and sending
From the discussion on the packet format above, we can know clearly that the information of
synchronization serial number, packet serial number was kept in the packet. Besides, the sampling
frequency was known. Therefore, we can use the three parameters to restore the time relationship
of data of sensor nodes. showed the process of data synchronization at the data hub:
- 6 -
中国科技论文在线
t
t
X1X2X3X4X5X6
Y1Y2Y3Y4
Synchronization 1
Node X
Node Y
X 1 X 2 X 3 X 4 X 5 X 6
Y 1 Y 2 Y 3 Y 4
Synchronization
Packet ID
Sampling Frequency
Synchronization
Packet ID
Sampling Frequency
Data fusion
Data synchronization of data hub
Synchronization SN provided a time reference between sensor nodes. 1 / nF was the length
of time interval. Then we can get a serial data acquisition point on the time line. And Packet SN
told us which point was the time that the data was collected. Finally, we can restore the time
relationship of data of sensor nodes uniquely.
2 Protocol Implementation
The protocol proposed in this paper fulfilled the requirement of BSN on real-time continuous
monitoring and satisfied the demands of synchronization of multiple physiological parameters.
Moreover, the communication mechanism ensured the integrity of data. Therefore, in order to
fully evaluate the performance of the protocol, we did some experiments on synchronization of
multiple parameters, real-time performance and analyzed the transmission efficiency of the
protocol according to packet loss rate.
Environment
A customized BSN development platform was used to evaluate our protocol. The
development platform introduced in the paper [6] consisted of various physiological sensor nodes
and a base station.
Results
Firstly, to simulate the performance of synchronization mechanism, we use a set of identical
sensor nodes instead of different types of sensor nodes, this way we would get a set of similar
waveforms over time axis.
In our experiment the nodes worked at the data acquisition rate of 40Hz. showed the
waveforms of three acceleration sensor nodes that moved in the same manner. The experimental
results showed that the time difference between various nodes less than 10ms. Thus it can be seen
multiple physiological parameters which were collected at the same time would maintain the
uniformities over time axis.
- 7 -
中国科技论文在线
Multiple parameters with synchronization mechanism
Furthermore, we calculated the packet loss rate (PLR) in the conditions of a distance of 5
meters between base station and sensor nodes. Experimental results showed the efficiency and
effectiveness of our protocol on keeping the integrity of physiological parameters. Specially, for
different number of nodes and different data acquisition rate, there was a best value of offset
N
on
transmission. If the offset
N
was too small, the controllability and flexibility of system might
decrease. If otherwise, the channel capacity cannot obtain the full use. Moreover, appropriately
shortening the synchronizing cycle enhanced the transmission reliability effectively.
Tab. 1 Packet Loss Rate
Synchronizing
Cycle: 10
Second
Synchronizing
Cycle: 20
Second
Synchronizing
Cycle: 30
Second
Node
Numbe
r
offsetN PLR offsetN PLR offsetN PLR
3 1 % 1 % 1 %
3 2 % 2 % 2 %
6 3 % 3 % 3 %
6 4 % 4 % 4 %
For real-time display of multiple physiological parameters, the transmission and processing
of data would brought a delay, furthermore, lost packets brought a longer delay because of packet
reconstitution. As shown in the transmission delay had two peaks, one caused by processing
while another caused by transmission loss. Moreover, the order of time slices would have effect on
delay time. Most delays were less than 2 seconds. It demonstrated that real-time monitoring can be
achieved.
In fact, the first successful transmission rate determined the delay time. If synchronizing
cycle gets longer, then it may actually have a negative effect on the transmission reliability and
consequently lead to more lost packet that cost more time on resending. Therefore, synchronizing
cycle and delay time are also closely related.
- 8 -
中国科技论文在线
1 2
delay
(s)
packet number
100
200
300
400
500
600
Transmission delay
3 Conclusion
This paper presented a protocol using a synchronization mechanism to create a common time
reference. Thus the multiple physiological parameters which were collected at the same time
would keep the uniformities over time axis. In addition, the common time which divided the
sending time into a series of time slices for various nodes play a crucial role in avoiding of
transmission owing to that each node sent data in their own time slices. Finally testing results
indicated that the protocol can meet the needs of application in practice.
4 Acknowledgements
Thanks to my old coworkers from Shenzhen Institutes of Advanced Technology.
References
[1] Guang-Zhong Yang. “Body sensor networks” [M] foreword by Sir Magdi Yacoub, New York :Springer, 2006,
pp. 399 – 402.
[2] Volmer, Achim and Orglmeister, Reinhold. Wireless Body Sensor Network for low-power motion-tolerant
syncronized vital sign measurement[C]. Conf. Rec. 2008 IEEE Int. Conf. Engineering in Medicine and
Biology Society. pp. 3422 – 3425.
[3] Thiemjarus, S.; Lo, B. and Yang, G.-Z. “A spatio-temporal architecture for context aware sensing” [C] in Proc.
in Conf. 2006. International Workshop. Wearable and Implantable Body Sensor Networks. – 194.
[4] Li, Huaming and Tan, Jindong, “Body Sensor Network Based Context Aware QRS Detection”[C] Pervasive
Health Conference and Workshops, 2006. pp. 1 – 8.
[5] Shih-Lun Chen and Ho-Yin Lee, “A Wireless Body Sensor Network System for Healthcare Monitoring
Application,” [C] Conf. Rec. 2007 IEEE Int. Conf. Biomedical Circuits and Systems. pp. 243 - 246.
[6] Wang, B.; Wang, L.; Lin, .; Wu, D.; Huang, .; Zhang, .; Yin, Q.; Chen, W. “A Body Sensor
Networks Development Platform for Pervasive Healthcare”[C]. Conf. Rec. 2009 ICBBE Int. Conf.
Bioinformatics and Biomedical Engineering. pp. 11-13.
[7] . Mills, “Precision Synchronization of Computer Network Clocks,” ACM Computer Comm. Review, Vol.
24, , pp. 28-43, Apr. 1994.
[8] IEC 61588:2004 - IEEE 1588:2002, “Precision clock synchronization protocol for networked measurement
and control systems,” Geneva, Switzerland, Sep. 2004.
[9] . Garner, F. Feng, K. den Hollander, J. Hongkyu, K. Byungsuk, L. Byoung-Joon, J. Tae-Chul, . Joung.
IEEE AVB and its Application in Carrier-Grade Ethernet[J]. IEEE Communications Magazine, Vol. 45,
n. 12, pp. 126-134, Dec. 2007.
[10] M. Maróti, B. Kusý, G. Simon, A. Lédeczi. “The Flooding Time Synchronization Protocol”[C] Proc. 2nd
ACM International Conference on Embedded Networked Sensor Systems (SenSys), pp. 39-49, Baltimore,
Maryland, USA, Nov. 2004.
[11] Ageev, A.; Macii, D.; Flammini, A. Precision Synchronization of Computer Network Clocks[C]. Conf. Rec.
ISPCS 2008 Int. Conf. Precision Clock Synchronization for Measurement, Control and Communication. pp.
115 – 120.
- 9 -
中国科技论文在线
基于重传机制的人体传感网多跳同步协议
陈子飞，刘岚
（武汉理工大学信息工程学院，武汉 430070）
摘要：本文介绍了一种应用于人体传感网络多生理参数传输的多跳同步协议。通过一种数据
重传机制实现同步传输。实验结果证明了此协议在多跳的泛人体传感网络中对多生理参数的
连续实时监测的有效性。
关键词：人体传感器网络；通信协议；同步
中图分类号：
