Research Article Event-Based Clustering Architecture for ...downloads.hindawi.com › journals ›...

13
Research Article Event-Based Clustering Architecture for Power Efficiency in Wireless Sensor Networks Kai-Ting Yang, 1 Wei Kuang Lai, 1 Shu-Min Li, 1 and Yuh-Chung Lin 2 1 Department of Computer Science and Engineering, National Sun Yat-Sen University, Kaohsiung, Taiwan 2 Department of MIS, Tajen University, Pingtung, Taiwan Correspondence should be addressed to Kai-Ting Yang; [email protected] Received 11 August 2013; Revised 11 February 2014; Accepted 2 March 2014; Published 4 May 2014 Academic Editor: Periklis Chatzimisios Copyright © 2014 Kai-Ting Yang et al. is is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. In order to set up WSN in various rigorous environments, the size and power constraints are stricter due to the high demands for convenience and reliability. erefore, power efficiency is very important for a WSN. For this, a novel architecture is presented in this paper. e proposed architecture categorizes sensors into different clusters by events. In each cluster, a minimum spanning tree is constructed for intracluster routing. e hierarchical architecture is useful in reducing the power consumption. In each intracluster routing tree, only leaf nodes are responsible for periodical detection. Data transmissions only occur when abnormal events are detected. An abnormality will be reported to the data center only if the majority of cluster members sense the same event. By reducing unnecessary data transmissions and shortening transmission distances, the proposed mechanism significantly reduces the power consumption and prolongs the network lifetime without influencing the accuracy of event response. e simulations show that the proposed architecture has about an 18-fold improvement rate in the device lifetime and avoids the false positive caused by the erroneous alarm of a single sensor. e proposed architecture is feasible, practical, and highly applicable to many applications. 1. Introduction Because of the remarkable advances of wireless technologies, many applications have been developed which make human life more and more convenient. Users are free from tangling with wired networks such that they can enjoy various network services no matter where they are. In addition to entertain- ment services, many other services can be provided by way of wireless technologies such as the application on environmen- tal detections which can assist in improving our daily lives. One of the most notable examples is wireless sensor network [1, 2]. In wireless sensor network, low-cost and low-power- consumption sensors are applied to detect specific events and return gathered information to data centers via wireless links. Data centers analyze the information and give corresponding responses which are preset by users. Wireless sensor network nowadays has been applied to a wide range of applications such as the detection of weather conditions including tem- perature and humidity, monitoring ocean currents and water pollution, and the invasion detection in military usage. WSN is a good solution to monitor the abnormity in a specific area. While any abnormity occurs, scientists and engineers can be informed rapidly of the benefit of WSN. WSN is also applied to medical care. Recently, a special application called “Body Sensor Network” (BSN) [3, 4] is proposed due to the problem of aging society. In BSN, various sensors are adopted to monitor some physiological conditions such as body temperature, blood pressure, heartbeat, body motion, and so on. In a hospital BSN will be an excellent assistant to the medical care members. It releases the heavy burden on the medical care members. All the sensed data would be transmitted to data centers in the nursing station as usual. Once any abnormal event is detected, the data center can automatically contact the medical center and ask for help. However, there are still some challenges to apply sensor networks to practical applications, such as (1) size constraint: the size of sensors should be as small as to be suitable to deploy a large amount of sensors in the limited sensing area; (2) cost constraint: because of deployment of large amount of sensors, the price and maintaining cost of a sensor will be Hindawi Publishing Corporation International Journal of Distributed Sensor Networks Volume 2014, Article ID 612590, 12 pages http://dx.doi.org/10.1155/2014/612590

Transcript of Research Article Event-Based Clustering Architecture for ...downloads.hindawi.com › journals ›...

Page 1: Research Article Event-Based Clustering Architecture for ...downloads.hindawi.com › journals › ijdsn › 2014 › 612590.pdf · Research Article Event-Based Clustering Architecture

Research ArticleEvent-Based Clustering Architecture for Power Efficiency inWireless Sensor Networks

Kai-Ting Yang,1 Wei Kuang Lai,1 Shu-Min Li,1 and Yuh-Chung Lin2

1 Department of Computer Science and Engineering, National Sun Yat-Sen University, Kaohsiung, Taiwan2Department of MIS, Tajen University, Pingtung, Taiwan

Correspondence should be addressed to Kai-Ting Yang; [email protected]

Received 11 August 2013; Revised 11 February 2014; Accepted 2 March 2014; Published 4 May 2014

Academic Editor: Periklis Chatzimisios

Copyright © 2014 Kai-Ting Yang et al. This is an open access article distributed under the Creative Commons Attribution License,which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

In order to set up WSN in various rigorous environments, the size and power constraints are stricter due to the high demands forconvenience and reliability. Therefore, power efficiency is very important for a WSN. For this, a novel architecture is presented inthis paper. The proposed architecture categorizes sensors into different clusters by events. In each cluster, a minimum spanningtree is constructed for intracluster routing. The hierarchical architecture is useful in reducing the power consumption. In eachintracluster routing tree, only leaf nodes are responsible for periodical detection. Data transmissions only occur when abnormalevents are detected. An abnormality will be reported to the data center only if themajority of cluster members sense the same event.By reducing unnecessary data transmissions and shortening transmission distances, the proposedmechanism significantly reducesthe power consumption and prolongs the network lifetime without influencing the accuracy of event response. The simulationsshow that the proposed architecture has about an 18-fold improvement rate in the device lifetime and avoids the false positivecaused by the erroneous alarm of a single sensor. The proposed architecture is feasible, practical, and highly applicable to manyapplications.

1. Introduction

Because of the remarkable advances of wireless technologies,many applications have been developed which make humanlife more and more convenient. Users are free from tanglingwithwired networks such that they can enjoy various networkservices no matter where they are. In addition to entertain-ment services, many other services can be provided by way ofwireless technologies such as the application on environmen-tal detections which can assist in improving our daily lives.One of the most notable examples is wireless sensor network[1, 2]. In wireless sensor network, low-cost and low-power-consumption sensors are applied to detect specific events andreturn gathered information to data centers via wireless links.Data centers analyze the information and give correspondingresponses which are preset by users. Wireless sensor networknowadays has been applied to a wide range of applicationssuch as the detection of weather conditions including tem-perature and humidity, monitoring ocean currents and waterpollution, and the invasion detection in military usage. WSN

is a good solution to monitor the abnormity in a specificarea. While any abnormity occurs, scientists and engineerscan be informed rapidly of the benefit of WSN. WSN isalso applied to medical care. Recently, a special applicationcalled “Body Sensor Network” (BSN) [3, 4] is proposed dueto the problem of aging society. In BSN, various sensors areadopted to monitor some physiological conditions such asbody temperature, blood pressure, heartbeat, body motion,and so on. In a hospital BSN will be an excellent assistantto the medical care members. It releases the heavy burdenon the medical care members. All the sensed data would betransmitted to data centers in the nursing station as usual.Once any abnormal event is detected, the data center canautomatically contact the medical center and ask for help.

However, there are still some challenges to apply sensornetworks to practical applications, such as (1) size constraint:the size of sensors should be as small as to be suitable todeploy a large amount of sensors in the limited sensing area;(2) cost constraint: because of deployment of large amountof sensors, the price and maintaining cost of a sensor will be

Hindawi Publishing CorporationInternational Journal of Distributed Sensor NetworksVolume 2014, Article ID 612590, 12 pageshttp://dx.doi.org/10.1155/2014/612590

Page 2: Research Article Event-Based Clustering Architecture for ...downloads.hindawi.com › journals › ijdsn › 2014 › 612590.pdf · Research Article Event-Based Clustering Architecture

2 International Journal of Distributed Sensor Networks

an important factor to establish a wireless sensor network;(3) battery power constraint: in order to improve the systemreliability and to reduce the mean-time-to-failure (MTTF)between the replacement of batteries, the power consumptionof sensors must be low enough.

In the case of abnormity monitoring, the reliability ofthe system is the most important consideration due to safetyissues of human life. All the constraints mentioned abovehave to be strictly examined. Many researchers pay a lot ofattention to the development of miniaturized sensors withreasonable maintenance fees which can be put into narrowareas. Under the constraints of size and cost, the batterysize is strictly restricted accordingly. Thereupon the issue ofpower consumption becomes more and more essential tomakeWSN feasible. For power efficiency, a novel architectureof WSN is presented in this paper, which reduces the regularpower consumption of WSN so that the device lifetime andnetwork reliability can be significantly improved.

We summarize four major contributions of the proposedarchitecture as follows.

(i) The proposed architecture reduces redundant datatransmissions without influencing the accuracy ofevent response so that server loads are decreased andthe number of clients the system can accommodate isincreased.

(ii) By reducing unnecessary data transmissions andoptimizing routing path, the proposed mechanismsimultaneously reduces the power consumption andprolongs the network lifetime.

(iii) The proposedmechanism is feasible and highly appli-cable to many applications.

(iv) The proposed mechanism achieves the above threeadvantages through software upgrading without anyextra facility and installation cost.

The remainder of this paper is organized as follows.Section 2 presents some related works. Section 3 introducesour proposed WSN architecture. Experimental simulationresults are reported in Section 4 which shows the feasibilityand advantages of our work. Finally, a brief conclusion isgiven in Section 5.

2. Related Work

As the secure and efficient issues on the power consumptionof body sensor network obtain increasing research interest,some previous works [5–7] focus on the optimization ofpower consumption of sensors to ensure long term biologydetection capability. These works also demonstrate that thecollision probability increases if the number of sensors ina BSN increases. Higher collision probabilities cause higherpacket loss probabilities which increase the number ofretransmission. Therefore, these works also propose someschemes to reduce the collision probability which alleviatesthe waste of power caused by retransmissions.

A new transmission scheduling scheme, called Dis-tributed Queuing Body Area Network (DQBAN) [5], has

been proposed to reduce extra power consumption ofretransmissions. DQBAN utilizes cross-layer fuzzy-rule andenergy-aware radio activation policy to set the transmissionpriority for all sensors in BSN. With the consideration ofQoS (Quality of Service) and remaining power of sensors,DQBANdetermines the transmission priority for each sensorin order to reduce collision and latency. A novel MACprotocol, called Energy Efficient Medium Access Protocol(EEMAP) [6], has been proposed to construct piconets inWSN. Every piconet has amaster sensor andmany other slavesensors.Themaster coordinates all other slaves’ transmissiontimings so that the power consumption and the additionallatency caused by retransmissions can be reduced effectively.In addition, the coordinating function of the master alsoarranges the durations of idle mode and active mode forslave sensors so as to reduce the power consumption ofsensors. An adaptive power conserving algorithm has beenproposed in [7]. In [7], each sensor adopts a coding processto merge its data with the received data from other sensorsto reduce the total number of transmissions. After thesynchronization among sensors, a sensor can only transmitdata in its own individual transmission cycle to reduce thecollision probability.

There are some works [8–11] which can save energy byadjusting the sampling rates of sensors. An adaptive samplescheme has been proposed in [8] to keep the sampling ratesat the low bound which does not cause excessive loss in sens-ing resolution. A Markov Decision Process-based SamplingMethod (MDPS) has been proposed in [10] which optimizesthe sample rates for all sensors by a global coordinateapproach. However, it would suffer extra power consumptionfor negotiations among all sensors. Therefore, they laterpropose theReinforcement LearningAverageApproximation(RLAA) [11] which replaces the global coordinate approachby a local coordination scheme for reducing the complexity.

Some researches [12, 13] demonstrate that the externaltransmission is the major cause of the power consumption ofBSN and try to shorten the external transmission distancesof sensors to save power. The work in [12] partitions the dutyregions of base stations within a building by the Voronoidiagram. When users move around within the building, sen-sors worn by users will dynamically attach to the nearby basestation.Therefore, the power consumption of longer-distancetransmissions can be reduced. The authors of [13] proposea path optimization scheme in which sensors carried by auser would adopt internal transmissions in BSN to transmitdata to the sensor closest to the base station.Then, the sensorperforms external transmissions to transmit data to the basestation. Therefore, the scheme can reduce the transmissionpower by minimizing the transmission distance. However,the improvement in power consumption is not significantbecause the distances from the base station to sensors are notvery different.

To solve the power consumption of external transmis-sions to the base stations with longer distances, a hierarchicalnetwork topology, called DexterNet, has been proposed in[14]. DexterNet applies the concept of mobile base station(MBS). In DexterNet, users carry MBSs for gathering datafrom sensors worn by users and forwarding data to remote

Page 3: Research Article Event-Based Clustering Architecture for ...downloads.hindawi.com › journals › ijdsn › 2014 › 612590.pdf · Research Article Event-Based Clustering Architecture

International Journal of Distributed Sensor Networks 3

Table 1: The summary of related works.

Main idea for power efficiency ItemDecrease sampling rates [8–11]Select the sensor closest to the base station [12–14]Control transmission power at the lower bound ofsuccessful transmission [15]

Data aggregation to reduce the trade volume [16, 17, 21]Refine circuit [18, 19]Tree based or hierarchical architecture to reduceexternal transmission [14, 20, 21]

servers. The advantage is that sensors only deliver data to theMBS with short-range transmissions. Therefore, DexterNetcan reduce the power consumption of sensors by the help ofMBS.

The Body-Posture-based Dynamic Link Power Control(BDLPC) has been proposed in [15]. The BDLPC sets thetransmission power to the lower threshold of transmissionpower that can successfully transmit data. By the lowertransmission power, the interference between sensors canalso be significantly reduced.

In [16, 17], authors focus on the techniques of humanmotion recognition. Since more sensors are required formotion recognition than other applications, lower powerconsumption becomes a critical issue to keep the motionrecognition mechanisms functioning. The work in [17]presents a priority-based transmission scheme, named DataCompression by Temporal and Spatial Correlation (DCTSC),to reduce the power consumption of motion recognition.Thedata gathered from neighboring sensors formotion detectionare dependent. Therefore, sensors in DCTSC transmit datain the order of their predefined priorities. Sensors with lowerpriorities receive the data transmitted by sensors with higherpriorities, stuff the data with the different parts, and sendout the new data with compression. DCTSC involves theprinciple of “incremental store” to combine related data sothat the amount of data transmitted by all sensors can besignificantly decreased. In addition to the above state-of-the-art works, some works [18, 19] improve the power efficiencyby ameliorating the circuit design of sensors. Similar to theproposed scheme, the tree-based architectures have beenproposed in [20, 21]. The Collection Tree Protocol (CTP)proposed in [20] provides reliable and loop-free transmissionpaths from leafs to the root of a tree. Benefiting from thetree-based architecture, the number of external transmissionscan be reduced so that the power efficiency can be improved.To improve power efficiency further, in [21], data from childnodes would be aggregated at their parent nodes so that thetotal data volume can be decreased. The related works aresummarized in Table 1.

In this paper, we propose a novel scheme which isdifferent from the above works. It constructs internal routingtrees to reduce the number of sensors periodically sensingand transmitting data. Different from other schemes, theproposed scheme directly reduces the power consumption bydecreasing the number of periodical senses and transmissions

simultaneously. In our proposed scheme, only leaf nodes inthe constructed routing tree have to sense and transmit dataperiodically. Compared to [21], the proposed scheme furtherreduces the number of transmissions and sensing originatingfrom parent nodes.The performance of the proposed schemeis greatly improved in terms of power consumption andsurvival time.

3. Event-Based Clustering Architecture

InWSN, compact and lightweight sensors would be installedin some narrow areas. In addition to size and weight con-straints, sensors should also be under strict power constraint.Optimizing power consumption of WSN can bring threeadvantages: (1) reducing sensing interrupts due to batteryexhaustion, (2) diminishing maintenance fees of replacingbatteries, and (3) lessening trouble in replacing the batteriesof the sensors installed in a pathless area. All these motivateus to find effective ways to reduce the power consumption ofWSN. In this paper, we aim at the issue of power efficiencyand propose a novel mechanism to reduce regular powerconsumption of WSN, which optimizes the transmissionpaths and decreases the numbers of transmissions and mea-surements.

Figure 1 shows a general case of WSN, where multi-ple sensors are deployed to measure many environmentalconditions. All sensors in WSN will periodically transmitmeasured data to a neighboring base station through wirelesstechnologies, and then the base station transfers those data toa specific data server to recognize whether any extraordinaryphenomenon exists or not. IEEE 802.15.4 is one of thefamous standards adopted in such a scenario. Following thearchitecture in Figure 1, each sensor individually reports itsdata to the server via the attached base station. Therefore,we denote the architecture as Individual Report (IR) in thispaper. However, the major cost of IR will be the powerconsumption to periodically transmit data to the base station.The other way to save power is to reduce sampling rates ofsensors. However, it may lead to lower accuracy of abnor-mality detection. To solve this tradeoff dilemma, we proposea new WSN architecture, called Event-based Clustered bodysensor network (EBC) in this paper, which decreases thenumber of sensors that periodically sense and transmit datainstead of the sampling rate.

EBC algorithm clusters sensors into groups according tothe events. One specific event is identified if the majorityof the corresponding sensors in the event group are in con-sensus. After clustering, internal routing tree is constructedfor every event-driven cluster. As shown in Figure 2, withinan internal routing tree, only leaf nodes are responsible forperiodical sensing. A parent node is triggered to evaluatethe specific abnormal event only if its children sense theabnormality and report it. Following this methodology, theabnormal event is evaluated and reported upward level bylevel within the internal routing tree. Finally, the root nodesimilarly evaluates and transmits this abnormal event to thedata server. In the proposed EBC algorithm, the number ofsensors responsible for periodical sensing is determined by

Page 4: Research Article Event-Based Clustering Architecture for ...downloads.hindawi.com › journals › ijdsn › 2014 › 612590.pdf · Research Article Event-Based Clustering Architecture

4 International Journal of Distributed Sensor Networks

Basestation

Figure 1: Sensors individually transmit data to the base station.

Periodically sense data

Report to server when abnormity occurs

BS

Sense data and report to their parents when abnormity occurs

Figure 2: The architecture of event-based clustered body sensor network.

the number of leaf nodes. No periodical sensing is neededfor other nodes except the leaf nodes and no periodicaldata transmission occurs for all nodes in EBC. In addition,the external transmission from the sensor to the BS, thefarthest transmissionwith themost power consumption, onlytakes place when the abnormality is detected. Therefore, thepower consumption can be significantly reduced by EBC. Forexample, in a binary balance tree with 𝑁 nodes, the numberof leaf nodes, 𝑁

𝐿, is

𝑁𝐿

= 2[log2(𝑁+1)]−1

=𝑁 + 1

2. (1)

It means that only 𝑁𝐿nodes need to periodically sense data

in EBC. It is less than a half of the number of nodes thatperiodically sense and transmit data in original IR.

In our proposed architecture, an abnormality will bereported to the data center only if the majority of clustermembers sense the same event. It can avoid the false positivecaused by the erroneous alarmof a single sensor. For example,

in the case of temperature monitoring, the false positive willbe alarmed by a single sensor when any heat source or coldsource passes by occasionally.Therefore, the consensus in thecluster is adopted to report an abnormal event.

There are three phases in our proposed scheme, includingSensor Clustering, MST Routing Tree Construction, andSensing/Reporting Procedure.

3.1. Sensor Clustering. Firstly, sensors in EBC are categorizedto different clusters by events. Each cluster is responsiblefor the detection of a specific event. An abnormality will bereported as an active event only if most of the sensors inthis cluster sense and report it. Suppose that there exist 𝑥

kinds of events and 𝑁 sensors in a WSN and the numberof sensors which corresponds to the 𝑥 events is representedas {𝑛1, 𝑛2, . . . 𝑛𝑥}. Equation (2) holds when there exist some

sensors belonging to two or more events. 𝑠𝑖(𝑙𝑖, 𝐸𝑖) represents

the allocated location (𝑙𝑖) and the corresponding cluster (𝐸

𝑖)

of the sensor 𝑖. The information of (𝑙𝑖, 𝐸𝑖) can be assigned by

Page 5: Research Article Event-Based Clustering Architecture for ...downloads.hindawi.com › journals › ijdsn › 2014 › 612590.pdf · Research Article Event-Based Clustering Architecture

International Journal of Distributed Sensor Networks 5

Algorithm Sensor Clustering and Routing Tree Construction

Input:𝑁: the number of sensor nodes𝑋: the number of events𝐺: the set of event tags𝑔: the event tag of a node𝐶: the set of clusters

Output:𝑇: the set of routing trees for events

for (𝑖 = 1; 𝑖≤ 𝑁; 𝑖 + +) {

for (𝑗 = 1; 𝑗 ≤ 𝑋; 𝑗 + +) {

if (𝑔𝑖

== 𝐺[𝑗]) {

𝐶[𝑗] = 𝐶[𝑗] ∪ 𝑖;}

}

for (𝑘 = 1; 𝑘 ≤ 𝑋; 𝑘 + +) {

𝑇[𝑘] := MST(𝐶[𝑘]); // construct minimum spanning tree for each cluster.}

Return 𝑇;

Algorithm 1: Pseudocodes of clustering and the construction of intracluster routing trees in EBC algorithm.

the WSN designer manually or the data center automatically,where 𝐸

𝑖∈ {𝑒𝑖,1

, 𝑒𝑖,2

, . . . , 𝑒𝑖,𝑚

} are the events that the sensor𝑖 is interested in. In other words, the sensor 𝑖 may join oneor more clusters. At the beginning of the formation of WSN,each sensor reports to the data center about its function, id,and other related information. Therefore, to construct theclusters in the proposed EBC, initiation by the data center isan efficient and flexible method.The data center can specify acluster by transmitting a unique tag to the sensors which arecategorized into the cluster. Thus only 𝑛

𝑖sensors receive the

𝑖th tag to form the cluster 𝐸𝑖. Once sensors receive the tags,

they would broadcast the received tags so that members of acluster are known to each other. Consider

𝑁 ≤

𝑥

∑𝑖=1

𝑛𝑖. (2)

3.2. Routing Tree Construction. After clustering sensors, intr-acluster routing trees would be constructed for each clusterin the second phase. Sensors in a cluster will construct atree-based architecture for the negotiation and hierarchicalreport. The tree-based architecture can efficiently reducethe communication overheads among cluster members. Inaddition, the architecture can decrease the required amountof simultaneously active sensors so as to achieve the goal ofpower saving.

In EBC, we adopt the Prim Algorithm to construct theminimum spanning tree (MST) for intracluster routing. Thedata structures such as Fibonacci heap, adjacency list, andso on can be efficiently applied to implement the PrimAlgorithm.The pseudocodes of the sensor clustering and therouting tree construction are represented in Algorithm 1.

3.3. Sensing/Reporting in EBC. In EBC, the hierarchicalarchitecture is utilized to reduce the power consumption. Ineach intracluster routing tree, only leaf nodes are responsiblefor periodical sensing with the period of 𝐷

𝑠. No data

transmission occurs when no abnormal event is detected.Once an abnormality is sensed by a leaf node, it would wakeup its parent and report the abnormality. When receivingthe abnormality report, the parent node will sense the eventitself to confirm the occurrence of the abnormality. If itsenses the abnormality, it would upwardly report to its parentnode. Otherwise, if no abnormality is sensed, it will decidewhether the received report from its child should forwardupwardly or not by the rule of Majority Verdict. All nonleafnodes act as the parent node described above including theroot node. The root node is responsible for reporting to theremote server via the attached base station. In other words,the external transmission only takes place on the root node.The operations of each role in EBC are described as follows.

3.3.1. The Operation of Leaf Nodes. In EBC, leaf nodes areresponsible for periodical sensing with the period of 𝐷

𝑠.

A leaf node would report to its parent only if it senses anabnormality. In addition to the abnormality report, a leafnode informs its parent of its existence by sending a hellomessage with a longer period 𝐷

𝑒(𝐷𝑒

≫ 𝐷𝑠) to ensure

the reliability of the system. Therefore, the proposed EBCwould not suffer the false-negative caused by the failures ofleaf nodes. As shown in Figure 3, a leaf node will contactits parent only when an abnormality is sensed or the 𝐷

𝑒

timer timeouts. No transmission is required as no abnormalevent is detected. By reducing the number of transmissions,

Page 6: Research Article Event-Based Clustering Architecture for ...downloads.hindawi.com › journals › ijdsn › 2014 › 612590.pdf · Research Article Event-Based Clustering Architecture

6 International Journal of Distributed Sensor Networks

Algorithm LN (Leaf Node)

Input:𝐷𝑠: the period of measurement (Sampling Period)

𝐷𝑒: the timer for reporting the status to the parent

𝑅: Measured Result𝑓: the parent node of a leaf node

Output:𝑅: Measured Result

while (1) {

𝑅 =Measure(𝐷𝑠); // measures the event with period 𝐷

𝑠.

If (!𝑅) { // if an abnormality occurs.Transmit(𝑓, 𝑅); // transmits the measured result to its parent node 𝑓.

}

If (𝐷𝑒

= 0) {

Status Report(𝑓); // report status to the parent node 𝑓.Reset(𝐷

𝑒);

}

}

Algorithm 2: The pseudocode of the leaf node algorithm.

the leaf nodes which consume most of the power in EBC cansave more power than other schemes. The pseudocode of theoperation for a leaf node in EBC is shown in Algorithm 2.

3.3.2. The Operation of Parent Nodes. For reducing powerconsumption, parent nodes in EBC are usually in idle mode.The transition state of a parent node is represented inFigure 4. Similarly, for system reliability, a parent node mustwake up and contact its parentwith the period𝐷

𝑒so its parent

can know that it is still functioning. In addition, a parentnode may be wakened by its child when its child senses anabnormality and reports it. After being wakened, it will sensethe event by itself to confirm the report from its child. Ifthe sensed result is also abnormal, the parent node wouldwake up its own parent and upwardly report the event. Onceits sensed result is normal, the parent node must count thenumber of its descendants which sense the abnormality todecide whether it should report upwardly or not. Supposethere are 𝑑

𝑎descendants reporting the abnormality and the

total number of its descendants is 𝑑. The parent node shouldreport upwardly when (3) is satisfied. If the rule of MajorityVerdict is adopted, 𝑇

𝑟in (3) should be 0.5. By considering

the diversity of sensed events, the value of 𝑇𝑟can be different

for each event. The more urgent the event is, the less the 𝑇𝑟

would be set. For example, 𝑇𝑟might be set to 0 for the event

which affects human life immediately, for example, the rateof a heartbeat. It means that the data center can receive thealarm even though only one sensor in the cluster senses theabnormal event. Consider

𝑑𝑎

𝑑> 𝑇𝑟. (3)

Based on the constructed routing tree, each parent nodecan easily obtain the total number of its descendants (𝑑) andthe number of its descendants that detect abnormal events

(𝑑𝑎) which are needed to calculate (3). The information, 𝑑

and 𝑑𝑎, will disseminate to its ascendants. Therefore, all the

intermediate nodes have information to determine if thedetected abnormal event should be reported upwardly basedon (3). If (3) is not satisfied, for human safety, the parent nodewould keep sensing with a period of𝐷

𝑠during a time interval

𝐷𝑘to trace the event. If no abnormality is sensed during 𝐷

𝑘

or a report is sent to its parent node, the node returns to idlemode. The pseudocode of the operation for a parent node inEBC is shown in Algorithm 3.

3.3.3. The Operation of Root Node. The root node is alsoone of the parent nodes in the routing tree. Therefore thebehavior of the root node is similar to a parent node.The onlydifference is that the root node is responsible for reportingto the data center via the attached base station rather thanreporting to another sensor. Although the transmissiondistance is longer, the root node still has a longer lifetimethan the other members due to the fewer transmissions. Thetransition state of the root node can be seen in Figure 5.

The power efficiency of a leaf node in EBC is dominatedby the probability of abnormality, 𝑃

𝑎. Even though 𝑃

𝑎is equal

to 1, the worst case in EBC, the total power consumptionof EBC is still less than the general architecture in Figure 1by shortening the transmission distance. The external trans-mission, with the longest transmission distance, only takesplace in root node. Instead of external transmission, othernodes contact their parent nodes by way of the internal pathsof the routing tree. Thereupon the power consumption bytransmissions can be reduced in EBC.

Suppose that 𝑤𝑚, 𝑤𝑠, and 𝑤

𝑏represent the power con-

sumptions of each measurement, each transmission betweensensors, and each transmission between a sensor and itsattached base station, respectively. In addition, the occur-rence probability of an interested event is 𝑝, the period of

Page 7: Research Article Event-Based Clustering Architecture for ...downloads.hindawi.com › journals › ijdsn › 2014 › 612590.pdf · Research Article Event-Based Clustering Architecture

International Journal of Distributed Sensor Networks 7

Sensing Report

Pn

Pa

1

Informexistence

PDe

1

Pa : the probability of sensing an abnormality

Pn : the probability of sensing a normality

PDe: the probability that the De timer is timeout

Figure 3: The transition state of a leaf node.

Sensing

Report

Idle

Informexistence

Pa : the probability of sensing an abnormalityPn : the probability of sensing a normalityPR: the probability of receiving a reportPDe

PDk: the probability that the Dk timer is timeout

PTr: the probability that da/d > Tr

Pa + Pn · PTr

P

1

1

R

(1 − PDk) · Pn(1 − PTr )

PDkPDe

: the probability that the De timer is timeout

Figure 4: The transition state of a parent node.

the measurement is 𝑡𝑚, and the period of transmitting a

hello message in EBC is 𝑡ℎ. Suppose that the total number

of sensors is 2𝑘 − 1 where 𝑘 is a positive integer. Forsimplification, we assume that EBC constructs a binarybalance tree for these sensors. The power consumptions ofEBC and IR within an observation time interval, 𝑇, can besummarized in Table 2.

As shown in Table 2, sensors consume power when theyare measuring or transmitting data. Furthermore, sensors inEBC consume additional power to transmit hello messageto their parents for system reliability. Therefore, the powerconsumption of a leaf node in EBC within a time interval,𝑇, can be represented as follows:

𝐸𝐿

𝐴= 𝐸𝐿

𝑚+ 𝐸𝐿

𝑡+ 𝐸𝐿

ℎ,

=𝑇

𝑡𝑚

⋅ 𝑤𝑚

+𝑇

𝑡𝑚

⋅ 𝑝 ⋅ 𝑤𝑠

+𝑇

𝑡ℎ

⋅ 𝑤𝑠.

(4)

Sensing

Report

Idle

Informexistence

Pa : the probability of sensing an abnormalityPn : the probability of sensing a normalityPR: the probability of receiving a reportPDe

PDk: the probability that the Dk timer is timeout

PTr: the probability that da/d > Tr

Pa + Pn · PTr

PR

(1 − PDk) · Pn(1 − PTr )

PDkPDe

: the probability that the De timer is timeout

1

1

Figure 5: The transition state of the root node.

The power consumptions of a parent node and a root node inEBC within a time interval, 𝑇, can be represented as (5) and(6), respectively:

𝐸𝑃

𝐴= 𝐸𝑃

𝑚+ 𝐸𝑃

𝑡+ 𝐸𝑃

=𝑇

𝑡𝑚

⋅ 𝑝 ⋅ 𝑤𝑚

+𝑇

𝑡𝑚

⋅ 𝑝 ⋅ 𝑤𝑠

+𝑇

𝑡ℎ

⋅ 𝑤𝑠,

(5)

𝐸𝑅

𝐴= 𝐸𝑅

𝑚+ 𝐸𝑅

𝑡+ 𝐸𝑅

=𝑇

𝑡𝑚

⋅ 𝑝 ⋅ 𝑤𝑚

+𝑇

𝑡𝑚

⋅ 𝑝 ⋅ 𝑤𝑏

+𝑇

𝑡ℎ

⋅ 𝑤𝑏.

(6)

As mentioned previously, the transmission of hello messageis for reliability. To ensure that a sensor is functional during

Page 8: Research Article Event-Based Clustering Architecture for ...downloads.hindawi.com › journals › ijdsn › 2014 › 612590.pdf · Research Article Event-Based Clustering Architecture

8 International Journal of Distributed Sensor Networks

Algorithm PN (Parent Node)

Input:𝑅: measured result𝑓: the parent node of a node𝑆: the set of child nodes𝐷𝑘: the predefined time interval

𝑑: the total number of descendants𝑑𝑎: the total number of descendants which alarmed the abnormality

𝑇𝑟: the predefined threshold

Output:𝑅: Measured Result

while (1) {

𝑏 = Listen(𝑆); // listen to its child nodes𝑅 =Measure( ); // measures immediately

If (!𝑅 ‖ 𝑑𝑎/𝑑 > 𝑇

𝑟) { // abnormality occurs or the alarm ratio exceeds the threshold

Transmit(𝑓, 𝑅); // transmits the measured result to its parent node 𝑓.}

else{

while (𝑐𝑢𝑟𝑟𝑒𝑛𝑡 𝑡𝑖𝑚𝑒 ≤ 𝑅.𝑡𝑖𝑚𝑒 + 𝐷𝑘) { // during the time interval of 𝐷

𝑘

𝑅 =Measure( );If (!𝑅) { // if an abnormality occursTransmit(𝑓, 𝑅); // transmits the result to its parent node 𝑓.Return;

}

}

}

}

Algorithm 3: The pseudocode of the parent node algorithm.

Table 2: The comparison in power consumption between EBC and IR.

Architecture EBC IRroles Leaf node Parent node Root nodeNo. of nodes 2𝑘−1 2𝑘−1 − 2 1 2𝑘 − 1

Power consumption in measurement 𝐸𝐿

𝑚=

𝑇

𝑡𝑚

⋅ 𝑤𝑚

𝐸𝑃

𝑚=

𝑇

𝑡𝑚

⋅ 𝑝 ⋅ 𝑤𝑚

𝐸𝑅

𝑚=

𝑇

𝑡𝑚

⋅ 𝑝 ⋅ 𝑤𝑚

𝐸𝑚

=𝑇

𝑡𝑚

⋅ 𝑤𝑚

Power consumption in transmission 𝐸𝐿

𝑡=

𝑇

𝑡𝑚

⋅ 𝑝 ⋅ 𝑤𝑠

𝐸𝑃

𝑡=

𝑇

𝑡𝑚

⋅ 𝑝 ⋅ 𝑤𝑠

𝐸𝑅

𝑡=

𝑇

𝑡𝑚

⋅ 𝑝 ⋅ 𝑤𝑏

𝐸𝑡

=𝑇

𝑡𝑚

⋅ 𝑤𝑏

Power consumption for hello 𝐸𝐿ℎ

=𝑇

𝑡ℎ

⋅ 𝑤𝑠

𝐸𝑃ℎ

=𝑇

𝑡ℎ

⋅ 𝑤𝑠

𝐸𝑅ℎ

=𝑇

𝑡ℎ

⋅ 𝑤𝑏

N/A

the interarrival time of an event, we can simply set the periodof transmitting a hello message, 𝑡

ℎ, as follows;

𝑡ℎ

≜𝑡𝑚

𝑝. (7)

According to (7), we can rewrite (4), (5), and (7) as (8), (9),and (10), respectively. Consider

𝐸𝐿

𝐴=

𝑇

𝑡𝑚

⋅ (𝑤𝑚

+ 2 ⋅ 𝑝 ⋅ 𝑤𝑠) , (8)

𝐸𝑃

𝐴=

𝑇

𝑡𝑚

⋅ 𝑝 ⋅ (𝑤𝑚

+ 2 ⋅ 𝑤𝑠) , (9)

𝐸𝑅

𝐴=

𝑇

𝑡𝑚

⋅ 𝑝 ⋅ (𝑤𝑚

+ 2 ⋅ 𝑤𝑏) . (10)

According to Table 2, we also can represent the powerconsumptions of any node in IR within a time interval, 𝑇, asfollows:

𝐸𝐴

=𝑇

𝑡𝑚

⋅ (𝑤𝑚

+ 𝑤𝑏) . (11)

Observing (8) and (11), we can find that 𝐸𝐴is larger than

𝐸𝐿𝐴when 𝑤

𝑏> 2 ⋅ 𝑝 ⋅ 𝑤

𝑠. For common applications of

WSN such as temperature monitors, 𝑝 is very small and evenapproaches zero for a long termobservation. Suppose that thetransmission power is inversely proportional to the square ofthe transmission range. A transmission range between twosensors inWSNwould be about severalmeters which ismuchshorter than a transmission range between a sensor and itsattached base station. Therefore, 𝑤

𝑠is less than 𝑤

𝑏.

Page 9: Research Article Event-Based Clustering Architecture for ...downloads.hindawi.com › journals › ijdsn › 2014 › 612590.pdf · Research Article Event-Based Clustering Architecture

International Journal of Distributed Sensor Networks 9

Table 3: Simulation parameters.

Parameter ValueTopology size 800 ∗ 800Number of nodes 10, 15, 20, 25, 30Number of events 4, 5, 6, 7, 8Transmission power (sensor to BS) 1 JTransmission power (sensor to sensor) 0.5 JMeasure power 0.05 JTotal power of a battery 10000 J

So the following condition is always true for commonapplications of WSN and 𝐸𝐿

𝐴is less than 𝐸

𝐴. Consider

𝑤𝑏

> 2 ⋅ 𝑝 ⋅ 𝑤𝑠. (12)

Similarly, by observing (9), (10), and (11), we can find that 𝐸𝑃

𝐴

and 𝐸𝑅𝐴are alsomuch less than 𝐸

𝐴.Therefore, nomatter what

role a sensor acts in EBC, its power consumption is much lessthan that of the sensor in IR. Benefiting from the extremesmall occurrence probability and the shorter transmissionrange, EBC can have lower power consumption than IR.

In addition to the factors mentioned above, EBC canbenefit from another important factor, the number of activenodes, and gain notable overall power saving. As shown inTable 2, the number of active nodes in EBC, the leaf nodes,is only 2

𝑘−1 in a binary balance tree. It is less than half of thenumber of active nodes in IR.

4. Experimental Results

In order to verify the effectiveness of our proposed EBC(Event-based Clustered body sensor network) scheme, weconduct the simulation setup as follows: we randomly gen-erate 𝑁 sensor nodes in a 800∗800 area. The value of 𝑁 isfrom 10 to 30, and the number of events 𝑋 is from 4 to 8.The simulation parameters are summarized in Table 3. Asmentioned in Section 3.2, EBC constructs internal routingtree by MST (Minimum Spanning Tree) algorithm. In ourMST algorithm, the costs of links are determined by theirtransmission distances. The longer the distance is, the higherits cost is. In contrast, sensor nodes in general WSN schemesindividually detect and report to data servers, called Individ-ual Report (IR).

At the beginning of a simulation, each node randomlygenerates a number from 1 to 𝑋, where 𝑋 denotes the totalnumber of events.This number determines howmany eventsa node is interested in. Therefore, the number of involvednodes for each event may be different. Figure 6 shows atopology generated for our simulation. In Figure 6, the nodesresponsible for the same event are marked with the samecolor. And the nodewith a circle around it is a root of a tree. Inour MST algorithm, the node with the smallest ID is selectedto be the root.

First, we compare the proposed EBC with IR and thescheme in [21] in terms of total power consumption. Thesimulation time is set to one day and the number of eventsis 3. As shown in Figure 7, EBC would have lower total power

Figure 6: A sample topology with 𝑁 = 20, 𝑋 = 3.

1

10

100

1000

10000

5 10 15 20 25 30 35

Tota

l pow

er co

nsum

ptio

n (J

)

Number of nodes

EBCScheme in [21]IR

Figure 7: The relationship between the number of nodes and thetotal power consumption (the number of event is 3).

consumption than IR and the scheme in [21]. The primaryreasons are that fewer sensors are assigned to periodicallysense events, and it is unnecessary to periodically return thesensed data to data centers in EBC. The parent nodes in [21]will periodically sense events and aggregate the data with thatfrom their child nodes and then send to their own parents.The trade volume is less than IR but more than EBC.

Aswe can see fromFigure 7, the total power consumptionincreases with the number of nodes no matter whether EBCor IR is adopted. Gratifyingly, the increasing rate of EBC isless than IR and the scheme in [21]. The reason is obvious.The more nodes are involved, the more measurements andtransmissions take place. In EBC, only a fraction of nodes

Page 10: Research Article Event-Based Clustering Architecture for ...downloads.hindawi.com › journals › ijdsn › 2014 › 612590.pdf · Research Article Event-Based Clustering Architecture

10 International Journal of Distributed Sensor Networks

5 10 15 20 25 30 35

Number of nodes

3

4

5

6

7

8

9

10

Aver

age n

umbe

r of l

eaf n

odes

EBCScheme in [21]

Figure 8: The relationship between the number of nodes and thenumber of leaf nodes (the number of event is 3).

0

1

2

3

4

5

6×10

5

Surv

ival

tim

e of t

he fi

rst e

xhau

sted

node

(s)

EBCScheme in [21]IR

5 10 15 20 25 30 35

Number of nodes

Figure 9:The relationship between the time duration when the firstexhausted node appears and total number of nodes (the number ofevent is 3).

(leaf nodes), instead of all nodes in IR, would be active andthe increasing rate is smoother than IR.

It can be further demonstrated by the result shown inFigure 8, for example, that as the total number of nodesincreases from 10 to 30, the average number of leaf nodesonly increases from 3.7 to 9.1. In contrast to IR and thescheme in [21], which increases 20 nodes to periodicallysense and transmit data, only 5.4 additional nodes in EBCare responsible for periodical sensing as the total number ofnodes increases from 10 to 30.

0

2

4

6

8

10

12

14

16

18

×103

3 4 5 6 7 8 9

Tota

l pow

er co

nsum

ptio

n (J

)

Number of events

EBC

Figure 10: The relationship between the number of events and thetotal power consumption (the number of nodes is 20).

Secondly, we turn our attention to the lifetime of nodes.As shown in Figure 9, the lifetime of the first exhaustednode in EBC is much longer than IR and the scheme in[21]. Because regular transmissions between sensors andbase stations are unnecessary, EBC has about an 18-foldimprovement rate in the lifetime of the first exhausted node.We would like to note that the power consumption oftransmissions is more serious than that of measurements.The total number of nodes would not affect the lifetime ofthe first exhausted node. The lifetimes of all nodes are quitesimilar in IRwhere nodes have the same role. In EBC, the firstexhausted node may possibly be a leaf node, whose role is forperiodical measurements and transmissions. In the schemein [21], the first exhausted node may possibly be a root nodewhich periodically senses data and transmits to a base station.

In the following simulations, the total number of nodesis fixed to 20. We try to find the effects of various eventnumbers. Because IR is not event-based, it would not appearin the following simulations. As shown in Figure 10, thetotal power consumption would rise with the increase of thenumber of events. It is obvious that more events means thatmore tree constructions are needed, which lead to more leafnodes. As discussed previously, the number of leaf nodes isthe primary factor in terms of total power consumption inEBC so that the total power consumption would be affectedby the number of events. Figure 11 shows the relationshipbetween the number of events and the number of leaf nodes.Fortunately, the increase rate is not proportional to thenumber of events as shown. Finally, Figure 12 shows that thelifetime is independent of the number of events. More eventswould lead to more leaf nodes.

The lifetime of a leaf node is mainly dominated by itsnumber of operations so that it is nearly unrelated with thenumber of events. Note that a node cannot be a leaf node ofdifferent clusters simultaneously in our simulation.

Page 11: Research Article Event-Based Clustering Architecture for ...downloads.hindawi.com › journals › ijdsn › 2014 › 612590.pdf · Research Article Event-Based Clustering Architecture

International Journal of Distributed Sensor Networks 11

3 4 5 6 7 8 9

Number of events

EBC

1

2

3

4

5

6

7

8

9

10

Aver

age n

umbe

r of l

eaf n

odes

Figure 11: The relationship between the average number of leafnodes and the number of events.

541000

541200

541400

541600

541800

542000

542200

542400

Surv

ival

tim

e of t

he fi

rst e

xhau

sted

node

3 4 5 6 7 8 9

Number of events

EBC

Figure 12:The survival time of the first dead nodewhen the numberof events is between 4 and 8.

We would note that EBC and the scheme in [21] havemore control overheads than IR due to the tree formation.In [21], the root node would be replaced due to its powerexhaustion and then tree reformation is needed. In EBC, thestrategy of selecting the root node is to select the node closestto the base station. Because the root node in EBC wouldhave longer survival time than other nodes, the possibility ofthe tree reformation is much lower than the scheme in [21].During 14 days simulation, the number of control messagesdue to the tree formation in EBC and the scheme in [21] is5417 and 29834, respectively.

The results of the simulations can prove that our proposedEBC can have better power efficiency than a general architec-ture. It would be useful for improving the practicability andreliability of sensor networks.

5. Conclusion

This paper proposed an event-based Clustered architecture toimprove the power efficiency of sensor networks. It clusterssensors into groups according to the events so that onespecific event is identified if themajority of the correspondingsensors in the event group are in consensus. After clustering,internal routing tree is constructed for every event-drivencluster.

Based on the tree architecture, we can diminish the num-ber of nodes which are assigned to periodically sense data tothe number of leaf nodes in a tree. External transmissionsbetween a sensor and its attached base station take place atroot node whenmost nodes in the tree report the same event.By reducing the number ofmeasurements and transmissions,the proposed architecture can have lower power consumptionand longer node lifetime. It can improve the practicability andthe reliability of wireless sensor networks.

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper.

References

[1] S.-H. Choi, B.-K. Kim, J. Park, C.-H. Kang, and D.-S. Eom, “Animplementation of wireless sensor network,” IEEE Transactionson Consumer Electronics, vol. 50, no. 1, pp. 236–244, 2004.

[2] O. B. Akan and I. F. Akyildiz, “Event-to-sink reliable transportin wireless sensor networks,” IEEE/ACM Transactions on Net-working, vol. 13, no. 5, pp. 1003–1016, 2005.

[3] N. Raveendranathan, S. Galzarano, V. Loseu et al., “Frommodeling to implementation of virtual sensors in body sensornetworks,” IEEE Sensors Journal, vol. 12, no. 3, pp. 583–593, 2012.

[4] Y.-C. Kan and C.-K. Chen, “A wearable inertial sensor node forbody motion analysis,” IEEE Sensors Journal, vol. 12, no. 3, pp.651–657, 2012.

[5] B. Otal, L. Alonso, and C. Verikoukis, “Highly reliable energy-saving mac for wireless body sensor networks in healthcaresystems,” IEEE Journal on Selected Areas in Communications,vol. 27, no. 4, pp. 553–565, 2009.

[6] O. Omeni, A. C. W. Wong, A. J. Burdett, and C. Toumazou,“Energy efficient medium access protocol for wireless medicalbody area sensor networks,” IEEE Transactions on BiomedicalCircuits and Systems, vol. 2, no. 4, pp. 251–259, 2008.

[7] S.-L. Chen, H.-Y. Lee, C.-A. Chen, H.-Y. Huang, and C.-H. Luo, “Wireless body sensor network with adaptive low-power design for biometrics and healthcare applications,” IEEESystems Journal, vol. 3, no. 4, pp. 398–409, 2009.

[8] C. A. Boano, M. Lasagni, K. Romer, and T. Lange, “Accuratetemperature measurements for medical research using bodysensor networks,” in Proceedings of the 14th IEEE InternationalSymposium on Object/Component/Service-Oriented Real-TimeDistributed Computing Workshops (ISORCW ’11), pp. 189–198,March 2011.

[9] Y. He, W. Zhu, and L. Guan, “Optimal source rate allocation inbody sensor networks with energy harvesting,” in Proceedings ofthe 12th IEEE International Conference onMultimedia and Expo(ICME ’11), pp. 1–6, July 2011.

Page 12: Research Article Event-Based Clustering Architecture for ...downloads.hindawi.com › journals › ijdsn › 2014 › 612590.pdf · Research Article Event-Based Clustering Architecture

12 International Journal of Distributed Sensor Networks

[10] S. Liu and A. Panangadan, “Evaluation of a Markov decisionprocessbased coordinated sampling method,” in Proceedings ofthe 8th ACM/IEEE International Conference on Information,2009.

[11] D. Reem, “An algorithm for computing Voronoi diagrams ofgeneral generators in general normed spaces,” in Proceedings ofthe 6th International SymposiumonVoronoiDiagrams in Scienceand Engineering (ISVD ’09), pp. 144–152, June 2009.

[12] K. Kim, J. Kim, I.-S. Lee, H. Lee, and M. Yoon, “An efficientrouting protocol based on position information in mobilewireless body area sensor networks,” in Proceedings of the1st International Conference on Networks and Communications(NetCoM ’09), pp. 396–399, December 2009.

[13] P. Kuryloski, A. Giani, R. Giannantonio et al., “DexterNet: anopen platform for heterogeneous body sensor networks and itsapplications,” in Proceedings of the 6th International Workshopon Wearable and Implantable Body Sensor Networks (BSN ’09),pp. 92–97, June 2009.

[14] M. Quwaider, J. Rao, and S. Biswas, “Body-posture-baseddynamic link power control in wearable sensor networks,” IEEECommunications Magazine, vol. 48, no. 7, pp. 134–142, 2010.

[15] S. Liu, A. Panangadan, A. Talukder, and C. S. Raghavendra,“Learning a policy for coordinated sampling in body sensornetworks,” in Proceedings of the 8th International Conference onBody Sensor Networks (BSN ’11), pp. 77–82, May 2011.

[16] S.-Y. Chen, W.-T. Lee, H.-C. Chao, Y.-M. Huang, and C.-F. Lai,“Adaptive reconstruction of human motion on wireless bodysensor networks,” in Proceedings of the International Conferenceon Wireless Communications and Signal Processing (WCSP ’11),pp. 1–5, November 2011.

[17] C.-H.Wu and Y.-C. Tseng, “Data compression by temporal andspatial correlations in a body-area sensor network: a case studyin pilates motion recognition,” IEEE Transactions on MobileComputing, vol. 10, no. 10, pp. 1459–1472, 2011.

[18] X. Zhang, H. Jiang, L. Zhang, C. Zhang, Z. Wang, and X. Chen,“An energy-efficient asic for wireless body sensor networks inmedical applications,” IEEE Transactions on Biomedical Circuitsand Systems, vol. 4, no. 1, pp. 11–18, 2010.

[19] Q. Gao and O. Yadid-Pecht, “A low-power block-based CMOSimage sensor with dual VDD,” IEEE Sensors Journal, vol. 12, no.4, pp. 747–755, 2012.

[20] R. Fonseca, O. Gnawali, K. Jamieson, S. Kim, P. Levis, and A.Woo, “The Collection Tree Protocol,” Tech. Rep. TEP 123, 2006.

[21] X. Zhang, J. He, and Q. Wei, “Secure and energy-efficientrouting for wireless sensor networks,” Journal of Networks, vol.6, no. 9, pp. 1288–1295, 2011.

Page 13: Research Article Event-Based Clustering Architecture for ...downloads.hindawi.com › journals › ijdsn › 2014 › 612590.pdf · Research Article Event-Based Clustering Architecture

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttp://www.hindawi.com Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014

Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014

Active and Passive Electronic Components

Control Scienceand Engineering

Journal of

Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014

International Journal of

RotatingMachinery

Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014

Hindawi Publishing Corporation http://www.hindawi.com

Journal ofEngineeringVolume 2014

Submit your manuscripts athttp://www.hindawi.com

VLSI Design

Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014

Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014

Shock and Vibration

Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014

Civil EngineeringAdvances in

Acoustics and VibrationAdvances in

Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014

Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014

Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawi Publishing Corporation http://www.hindawi.com

Volume 2014

The Scientific World JournalHindawi Publishing Corporation http://www.hindawi.com Volume 2014

SensorsJournal of

Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014

Modelling & Simulation in EngineeringHindawi Publishing Corporation http://www.hindawi.com Volume 2014

Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014

Chemical EngineeringInternational Journal of Antennas and

Propagation

International Journal of

Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014

Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014

Navigation and Observation

International Journal of

Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014

DistributedSensor Networks

International Journal of