WARFARE REMOTE STATERGY

ACQUIRING AND IMPLIMENTATION OF

SOLDIER SYSTEM

Abstract

In todays world enemy warfare is an important factor in any nationssecurity. The national security mainly depends on army (ground), navy(sea), air-force (air).The important and vital role is played by the armysoldiers. There are many concerns regarding the safety of these soldiers.As soon as any soldier enters the enemy lines it is very vital for the armybase station to know the location as well as the health status of all soldiers.In our project we have come up with an idea of tracking the soldier aswell as to give the health status of the soldier during the war, which en-ables the army personnel to plan the war strategies. Also the soldier canask for directions to the army base unit in case he feels that he is lost. Byusing the location sent by the GPS the base station can guide the soldierto safe area. What is soldier tracking system:-: Consider an exampleof the terrorist attack, in this situation the soldiers and the commandershould be in contact .The Commander should know the exact positionand condition of soldiers and should track each soldier. Our project givesthe, Position of the all soldiers to the commander, Status of each soldier,Pro- vides continuous communication between soldier and commander,Alert to soldiers and commander. This system is very useful to betterplan- ning and synchronization so as create strategy and denitely helps forcompletion of mission. This unit is mounted on the soldier body. It hasmainly four segments: Global Positional System(GPS) Receiver, Biomedi-cal(Temperature) sensor,Passive Infrared (PIR)Sensor,RF transmitter. Achannel model for time-variant multi-link wireless body area networks(WBANs) is proposed in this paper, based on an extensive measurementcampaign using a multi-port channel sounder. A total of 12 nodes wereplaced on the body to measure the multi-link channel within the createdWBAN. The resulting empirical model takes into account the receivedpower, the link fading statistics, and the link auto- and cross-correlations.The distance dependence of the received power is investigated, and thelink fading is modeled by a log-normal distribution. The link autocor-relation function is divided into a decaying component and a sinusoidalcomponent to account for the periodical movement of the limbs causedby walking. The cross-correlation between dierent links is also shownto be high for a number of specic on-body links. Finally, the model isvalidated by considering several extraction- independent validation met-rics: multi-hop link capacity, level crossing rate (LCR) and average fadeduration (AFD). The capacity aims at validating the path-loss and fading1 model, while the LCR and AFD aim at validating the temporal behavior.For all validation metrics, the model is shown to satisfactorily reproducethe measurements, whereas its limits are pointed out.

1

1 Introduction

Recent years have witnessed a growing interest in the search for low-power, low-data-rate and body-centric wireless communications. Academic and industrialeorts in this research eld have led to the emergence of a new kind of promisingnetworks, called the wireless body area networks (WBANs). WBANs consistof a set of wearable biosensor nodes, which collect or relay physiological andcontextual signals proling the human body activities. Typical applications in-clude real-time monitoring of heart activity, blood pressure, breathing rate, orskin temperature; continuous diagnostics; and remote medical treatment of apatient. The close proximity of sensors to one another and to the human bodydemands challenging optimizations to limit the radiated power, the consumptionas well as the interference with other sensors and coexistent networks. Hence,a detailed knowledge of on-body multi-link propagation is crucial to properlydesign relevant systems at the physical and access network layers. There havebeen immense advancements in eld of technology over the past years. Smallsize, low cost sensor networks have been developed which are capable of col-lecting, analyzing and transmitting information to a central processor node forfurther analysis and report generation. Wireless Sensor Network is a term fordensely deployed sensor nodes which are capable of collecting real time infor-mation. These networks provide advantages over traditional sensing devices asthey provide low cost network deployment and are fault tolerant and robust. Asensor network consists of large no of sensor nodes deployed in random topology.Sensor networks are deployed close to the phenomenon which is to be observed.Instead of sending collected data directly to processor node, sensor nodes usesits processing capabilities and transmit partially processed data to task man-ager node via satellite for further processing as . The communication powerof transceiver is limited therefore information is communicated to Base stationor Sink by multihop path . Wireless Sensor Network (WSN) has become a vi-tal research area, due to their wide ranging applications. WSN based systemshave been deployed widely in many applications including civilian, industrial,agricultural, and military applications . A sensor network is composed of sen-sor nodes which are small in size, low in cost, and have short communicationrange. A sensor node usually consists of four sub-systems: a. A computingsubsystem: this is responsible for functions such as execution of the communi-cation protocols and control of sensors, b. A sensing subsystem: this subsystemis responsible for sensing the environmental characteristics, such as using tem-perature, humidity, or acoustic sensor, c. A communication subsystem: thisconsists of a short radio range used to communicate with neighboring nodes,d. A power supply subsystem: this includes a battery which provides energyto sensor node. Researchers have focused on dierent aspects of WSN, suchas hardware de- sign, routing, data aggregation, and localization. One of thecritical issues which needs to be taken into consideration is localizing objectsthrough distributed sensor network. Node localization is the problem of ndingthe geographical location of each target node (the object with unknown location)based on other reference nodes (nodes with known location). Localizing sensornodes is one of the fundamental and dicult problems that must be solved forWSN. Track- ing and localization systems have been deployed to track civilian,soldiers, and animals. depicts the idea of tracking mobile targets through WSNs,and trans-

2

mit the localization information to a base station. The WSN based local-ization and tracking issues have received much attention recently driven by theneed to achieve high localization accuracy with the minimum cost. This is be-cause: Sensor eld Local monitoring a. In several applications, the location itselfis the information of interest, b. Transferring sensors measurements without in-curring the sensors locations is an ineective task, c. Several routing protocolsare based on the locations of sensor nodes. There are many issues aecting theeciency of the tracking system includ- ing: the cost of extra localization hard-ware, number of reference nodes in the tracking area, and the communicationrange for the target and reference nodes. These issues have to be taken intoaccount before developing a WSN-based tracking system. A wide range of nodelocalization and tracking systems have been proposed recently. According to ,most of the WSN-based localization solutions are either range-based or range-free . Range-based approaches deploy various techniques (ultrasound, infrared,and GPS devices) to rst determine the distance between reference and targetnodes, and then compute the location using geometric principles. In order tocalculate the distance between the target and reference nodes, an additionalhardware is usually required to be attached to each wireless sensor device. Onthe other hand, in range-free approaches, distances are not computed directly,as the number of hops between the target and reference nodes is calculated. Assoon as the hop counts are calculated, dis- tances between nodes (reference andtarget nodes) are computed based on the average distance per hop. And then,geometric principles are used to calculate the targets location. Range-free tech-niques are not accurate as range-based ones and often require a high number ofmessages to be transmitted before cal- culating the targets locations . Since,range-based systems require attaching additional hardware to each target andreference nodes, which increases both the cost and power consumption for sen-sor nodes. But, range-based approaches oer better localization accuracy thanthe range-free systems. Furthermore, range free approaches require scatteringa large number of reference nodes in order to oer ecient localization accuracy .Therefore, this paper focuses on range based systems as they oer better local-ization accuracy than range free systems. The existing WSN-based localizationsystems are discussed in details . However, in this paper, we aim to discuss theexisting WSN based localization approaches which could be used for Threats de-tection and tracking in military applications, and point out the key issues whichneed to be taken into consid- eration before designing and implementing a WSNbased localization approach for Threats detection and tracking. The idea of theresearchs goal is depicted . Our contribution lies on the following aspects: i.Study the existing WSN-based localization and tracking systems, ii. Present,discuss, and compare WSN based localization and tracking methods which couldbe deployed in military appli- cations, iii. Present the key issues which needto be taken into consideration before designing and developing a WSN-basedlocalization system for Threats detection and tracking applications. This pa-per is organized as follows: presents a classication of WSN-based tracking andlocalization. discusses the strengths and weaknesses of the existing localizationand tracking systems. Recommen- dations for WSN developers in order to de-sign and implement a WSN-based tracking system for military applications are. And nally, conclusions and future work presented.

3

2 LITERATURE SURVEY

WSN is one of the fastest growing technologies in ubiquitous networking to-day. Standardization eorts, such as IEEE 802.15.4 , are geared to reduce costs,provide device customizability for diverse applications and create standards forinteroperability. The IEEE 802.15.4 standard was developed to address a de-mand for low-power and low-cost in low-rate wireless personal area networks(LR-WPAN). Dealing with low data rates, IEEE 802.15.4 oers very long bat-tery life (months or even years) and very low complexity. The IEEE standard802.15.4 denes the physical layer (PHY) and medium access control (MAC) sub-layer specications for LR-WPAN in the 2.4 GHz and 868/915 MHz bands. Afree license to use the industrial, scientic and medical (ISM) 2.4 GHz band isavailable worldwide, while the ISM 868 MHz and 915 MHz bands are only avail-able in Europe and North America, respectively. A total of 27 channels withthree dierent data rates are allocated in IEEE 802.15.4, including 16 channelwith a data rate of 250 Kbps in the 2.4 GHz band, 10 channels with a data rateof 40 Kb/s in the 915 MHz band and 1 channel with a data rate of 20 Kb/sin the 868 MHz band. Channel sharing is achieved using carrier-sense multipleaccess (CSMA), and acknowledgments are provided for reliability. Addressingmodes for 64-bit (long) and 16-bit (short) addresses are provided with unicastand broadcast capabilities. The main characteristics of WSN devices are smallphysical size, low-power consumption, limited processing power, short-rangecommunication capability and small storage capacity. A number of studies andprojects have focused on novel ubiquitous healthcare systems utilizing WSNtechnology to simplify methods of monitoring and treating patients. A case inpoint is the MobiHealth project, which developed a system for ambulant patientmonitoring over public wireless networks based on a body area network (BAN). Another example is the Ubiquitous Monitoring Environment for Wearable andImplantable Sensors project (UbiMon) [33] at Imperial College London, whichaims to provide a continuous and unobtrusive monitoring system for patients tocapture transient, but life-threatening events. CodeBlue was designed to oper-ate across a wide range of devices, including low-power motes, PDAs and PCs,and it addresses the special robustness and security requirements of medicalcare settings. Rapid technological development, ease of use and falling costshave made mobile devices increasingly popular, producing great changes in to-days lifestyle. During the past decade, the development of wireless mobile andinformation technologies (IT) has helped to extend the concept of ubiquitouscoverage to new segments of society. Thus, many applications that were initiallyavailable at a xed location only have been transformed into ubiquitous applica-tions, to be used wirelessly and exibly at anytime, anywhere. The same trendhas been observed in the medical eld. Over the years, many telemedicine andhealthcare related societies and authorities have turned to wireless technolo-gies to overcome the poor mobility of desktop PC-based healthcare monitor-ing systems. As a result, the possibility to monitor biomedical signals using amobile device is no longer an unachievable dream. In this work, an Android-based smartphone is used as monitoring terminal due to its smart functions andcomputer-like features. Compared to other smartphone operation systems, theAndroid device has many advantages, such as openness. Moreover, all appli-cations are equal, there are no boundaries between applications and its devel-opment is fast. Furthermore, the Android smartphone is currently one of the

4

most popular smartphones on the market.Android, provided by Google andthe Open Handset Alliance (OHA), is an open-source software stack for mobilede- vices that includes an operating system, middleware and key applications.OHA is a group of approximately technology and mobile companies, which com-bine their eorts toward the goal of accelerating innovations in mobile networksand toward oering a better mobile experience to the customer. Android is builton an open Linux kernel that includes such core system services as security,memory management, process management, network stack and drivers. Fur-ther, the kernel acts as an abstraction layer between hardware and the softwarestack and can be extended to incorporate new cutting edge technologies. Usinga Linux kernel as a hardware abstraction layer, allows Android to be ported toa wide variety of platforms. Eclipse is a multi-language software developmentenvironment comprising an integrated development environment (IDE) and anextensible plug-in system. For the development of applications, a software de-velopment kit (SDK) is provided with the necessary tools and API. Eclipse SDKis meant for Java developers, who can extend its abilities by installing plug-inswritten for the Eclipse platform, such as development toolkits for other program-ming languages, as well as create their own plug-in modules. All applicationsare written in the Java programming language. The application layer includesa set of core applications preinstalled on every Android device, including email,maps, contacts, web browser, phone dialler, calendar, text message and AndroidMarket. Further, Android applications can utilize the functionalities of otherapplications and services. A service is an application component without a userinterface that runs in the background for an indenite period of time.

5

3 EXISTING METHOD

DISCUSSION AND CRITICAL ANALYSIS WSN-based tracking system canbe used to oer critical tasks for military applications, such as localizing andtracking mobile targets (vehicles, and soldiers). Practically, designing and de-veloping a localization method for WSN is a complicated task due to the limitedcapabilities of sensor devices. The analysis of the existing WSN localization andtracking systems is presented in this section.

3.1 Analysis of the Existing WSN based Approaches

In this section, we study and compare the eciency of the existing systems interms of their strengths and weaknesses. We start with GPS-based approaches.GPS-based systems oer reasonable localization accuracy and have been de-ployed in several applications including military, civilian and industrial. How-ever, using GPS-based systems for Threats detection and tracking tasks is an in-ecient for several reasons: First, each threat object is required to be at- tachedwith a GPS receiver and transceiver, and this is not applicable in mili- taryThreats detection and tracking applications. Second, GPS systems require astraight line of sight between the receiver and GPS satellites, and this re- quire-ment is not valid for indoor tracking, due to the obstacles and walls foundbetween the senders and receivers. And third, attaching additional hardwareto each sensor node will increase both the sensors size and cost. Therefore,GPS-based systems are considered as inecient tracking system for Threats de-tection and tracking applications. Lets move to the second approach discussedin the literature, the camera-based approach. Camera-based systems oer ef-cient localization accuracy, as these systems track positions and identities ofmobile targets, without the need for attaching any device to suspect objects.The same with GPS approaches, camera-based systems suer from a high cost,in addition to the requirement of a straight line of sight to be existed betweenthe camera (reference node) and target object, in order to compute the targetslocations. One the other hand, camera-based approaches require to be used inday time in order to be able to detect the targets positions. A night camera canbe used to track objects; however this solution is an ineective in terms of cost.Consequently, using the camera approach in WSN-based tracking system is nota practical tracking solution for several reasons: a. The requirement of installa-tion and maintenance, including periodic lens cleaning, b. Performance aectedby inclement weather such as fog, rain, and snow, c. Reliable night-time signalactuation requires street lighting, d. Some models are susceptible to cameramotion caused by strong winds or vibration of camera mounting structure. RF-based approach was the third approach taken into consideration for de- tectingand tracking Threats. RF-based systems can be deployed to track targets withlow cost, as there is no need to attach special hardware to each reference andtarget wireless sensor nodes. These approaches oer cost-wise localization solu-tions, but oering low localization accuracy. RF-based systems dont re- quireattaching any additional device or sensor to each target node except the trans-mitter and receiver found at each sensor device; however the sensors cost isnot expensive as in the camera and GPS-based systems. The last approach dis-cussed in the literature was the acoustic-based approach. This kind of sys- temsoer ecient localization information and considered as cost wise systems,as they

6

require installing an inexpensive microphone to each reference node. Though,each target object requires emitting a voice in order to be sensed by referencenodes. As a result, reference nodes might then be able to detect the position ofthat target object.

3.2 Comparing the Existing WSN based Approaches

In this section, we compare the exiting WSN based approaches which havebeen discussed in the literature. Table 1 compares the existing localizationtechniques in terms of accuracy, the requirement of additional hardware (HW),cost, den- sity, deployability, and power consumption. In terms of trackingGPS and camera-based systems oer reasonable localization accuracy, but GPStracking systems require attaching an expensive hardware to each target object,which increases cost, complexity, and power consumption for each sensor node.The to additional hardware is a critical requirement in Threats detec- tionand tracking systems. GPS, and RF-based approaches require each suspect(target) object to be attached with an additional hardware, in order to be local-ized. While, there is no need for any additional hardware to be attached tothe target node when the camera or acoustic-based systems are used. In termsof , GPS and camera-based systems are considered as high cost localizationsolutions. GPS systems require attaching a GPS receiver and transceiver toeach target object, while an additional camera sensor is required to be addedto each reference node in the camera-based tracking systems. In acoustic-basedsystems, an additional hardware is needed (acoustic sensor), however its con-sidered as a low cost sensor compared to GPS receiver, transceiver, and camerasensors. RF-based systems oer low cost localization system, as there is norequirement to attach any special device (microphone, ultrasonic, or camera) toreference and target wireless sensor nodes, as this kind of systems depends onlyon the radio signals transmitted from reference nodes .GPS tracking systems donot require a high density of sensor nodes to be deployed in the tracking area ofinterest, as the targets coordinates computed through the values received fromthe satellites. However, vision-based systems require a high density of referencenodes. The density of reference nodes in the RF-based systems is based onthe transmission range. In the acoustic-based systems, the density of referencenodes is based on the voice emitted by the target node.

3.3 Challenges of WSN Threats Detection and TrackingSystems

The constrained computation power, battery power, storage capacity, and com-munication bandwidth of the tiny sensor devices pose challenging problems inthe design and deployment of Threats detection and tracking systems. Threatsdetection and tracking systems pose many challenges when deployed for mili-tary applications. One of the key technological challenges is how to track threatobjects without attaching any additional device (sensor) to those targets. InRF and GPS-based tracking system, its essentially to attach a device (GPSreceiver and transceiver, or RF module) in order to be able to detect and trackThreats positions. One more challenge is how to track Threats with the lowestcost possible. Attaching additional hardware (camera, or GPS) to each threatobject is inecient too, because it rises up the tracking systems cost. Using

7

acoustic sensor might reduce the tracking complexity and cost. Acoustic-basedsystems require attaching a simple acoustic sensor to each reference node, incontrast to GPS and camera-based systems. Now, using an acoustic method todetect and track the position of target objects might overcome some of thosechallenges.

8

4 KEY ISSUES

Through discussing and analyzing the existing approaches for military tasks,there are many recommendations must to be taken into consideration beforedesigning and implementing a military tracking system. As result, developersneed to consider the following issues: 1. Accuracy: the designed system mustoer reasonable localization accuracy for Threats detection and tracking sys-tems, as obstacles and walls could be existed in the tracking area of interest,and hence achieves low localization accuracy. 2. Cost: adding a supplementaryhardware to hundreds or thousands of nodes is a highly inecient solution interms of cost. The localization technique must be cheap in cost in order to beattractive solution for WSN systems . 3. Power consumption: in WSN, energyis mainly consumed by three subsystems: signal processing, data transmission,and hardware operations. Consuming less energy in WSN based tracking sys-tems is a primary objective in designing a WSN application, as each sensor nodeis usually equipped with batteries which could be hard to replace. 4. Coverage:the localization system should cover the tracking area of interest in order tosense any suspicious object, and then achieve the goal that was designed for . 5.Density of reference node: a high density of reference nodes will increase boththe cost and power consumption for WSN system. Therefore, the total numberof reference nodes should be as minimum as possible 6. Delay time: the sinknode (administrator) must be informed about the threats position as soon asone of the reference nodes detects its position 7. Deployability : usually, sen-sor nodes are scattered or deployed using airplanes. The tracking system mustbe easy to deploy with no need for a hard installation . 8. Accessibility: thedesigned tracking system must be able to work indoors and outdoors. As thesystem could be deployed where a number of obstacles (such as tree, wall, orvehicle) found in the tracking area of interest. 9. Form factor: attaching addi-tional hardware or sensor device to each sensor node will increase the sensorssize. In Threats detection and tracking system, its critical to keep the sensornodes size as tiny as possible in order to be invisible for Threats objects.

9

5 SYSTEM ARCHITECTURE

The KNOWME network employs a three-tier architecture , depicted in Fig.1. The rst tier is the WBAN layer or sensor layer, which wirelessly providesphysiological signals. The second tier is the mobile phone, which acts as a datacollection hub for the external sensor data. The mobile phone also processesdata locally and provides simple feedback to the user instantly. The last layeris a back-end server that can provide additional processing as well as data stor-age. The WBAN layer is comprised of the on-body sensors and the mobiledevice. As noted earlier, due to the eld deploymentrequirement, we used onlyo-the-shelf sensors and mobile devices to build KNOWME. KNOWME con-sists of a Nokia N95, as well as a Bluetooth-enabled oximeter (OXI) and anelectrocardiograph (ECG) from Alive Technologies arranged in a star topologywith the mobile phone being the hub. While there are newer and more energy-ecient wireless protocols such as Zigbee, we were restricted to using Bluetoothfor communication with the sensors since the N95 supports only the Bluetoothwireless protocol for sensor interfacing.1 Additionally, sensor data are collectedfrom in-built N95 sensors: an ACC and a global positioning system (GPS).The mobile application must gather data from multiple sensors with minimaluser intervention and with no interruption to regular mobile device functional-ity. To achieve continuous long-term data collection (e.g., 12 h/day for multipleweeks), mobile application robustness is necessary. In addition to applicationrobustness, voluntary user participation is essential for data collection. Hence,satisfying the users primary purpose of using a mobile phone takes priority overKNOWME. KNOWMEs execution priority is lower than other higher-prioritytasks, such as incoming and outgoing calls. Whenever there is resource con-tention with higher-priority tasks, the mobile phone will simply terminate theKNOWME application.The mobile application is divided into two components:a background process (KMCore) and a client interface application (KMClient).The KMCore is comprised of seven components arranged in a four-layered hi-erarchy: Device manager (bottom) Data collector Data analyzer, local storagemanager, data transmitter Service manager (top) Figure shows how variouscomponents in the KMCore interact with each other. There is one thread persensor, providing robustness to errors from individual failing sensors as couldoccur with a single manager for all devices. The data collector thread receivesand synchronizes sensor data from each device manager, resulting in a singlehealth record; health records are collected, buered, and sent to the local stor-age, the transmitter, and the analyzer. The local storage manager writes thedata to ash storage and handles conguration data as well. The trans- mittermodule that transfers data to the back-end handles data compression and en-cryption for privacy and energy saving. The analyzer modules imple- ment asimplied version of the physical activity detection methods detailed later; whilethe back-end server currently implements the full-blown classier. We observethat much eort was employed in designing and operating KM- Core key ele-ments that are a direct consequence of the implemented Bluetooth standardand the time-division multiple access (TDMA) strategy. In fact, if a code-division multiple access strategy were employed, most of the functionality ofthe device manager and data collector could be absorbed into the data ana-lyzer. Non-functioning sensors would be determined during analysis and wouldnot change any of the data collection or formatting. Due to the fact that only a

10

modest number of sensors are being employed, long spreading sequences wouldbe unnecessary. The complexity of multi-user detection is comparable or lessthan that of the activity detection methods that we have already implementedon the mobile phone and are described in the sequel. Finally, the KMCore ser-vice manager communicates with the KMClient graphical user interface (GUI).The framework is fairly complex and resource intensive, but not critical to datacollection. If mobile phone resources (memory, computation power) are limited,the KMClient is shut down without aecting the KMCore since the GUI anddata collection systems are separate.

11

Figure 1: KNOWME system architecture flow diagram and screen shot fromthe sedentary behavior analyzer implemented on the mobile platform.JPG

.

12

6 LEARNED FROM SYSTEM DESIGN

Energy Consumption Energy consumption to support KNOWME operationsis signicant and motivates the design of a host of schemes to improve batterylife. In KNOWME, the sensors simply transmit data to the mobile phone fusioncenter; the Nokia N95 performs all the coordination, processing, and computa-tion tasks. The energy consumption (in Joules) of the three sensors followed bytheir sampling rates and their transmission rates (samples/sec) using Bluetoothover a 10 minute interval were: phone ACC (37.8 J, 30, 30); ECG (114.8 J, 300,4); OXI (1374 J, 100, 10). As the sensors were not programmable, task schedul-ing and sensor data compression were impossible. However, in KNOWME, theprimary energy bottleneck was the mobile phone. Sensors operated comfortablyduring the course of a day; as such, our energy eciency research was centeredon the mobile phone. If data are collected from all sensors (ECG, OXI, ACC,GPS) and written to a local ash drive on the N95 without buering, the bat- terylife is 4 h. This is in sharp contrast to the N95s 10 h of rated talk time and200 standby hours. By using a combination of data buering, adaptive sensorthrottling, and dynamic selection of data transmission methods, battery life canbe improved by nearly 200 percent [9]. In KNOWME, the most complex dataanalysis function is user state detection. State detection can take place eitheron the phone or on the back-end server, which incurs no computation cost to thephone, but does incur a transmission energy cost. Figure 2 shows the energycost associated with physical activity detection for local and remote compu-tation based on 10 minutes of ECG and ACC data. We see that the energyconsumption of back-end computation is a function of the three transmissionoptions (EDGE, 3G, and WiFi) coupled with compression costs at the mobilephone. When WiFi is available, it is energy ecient to perform remote compu-tation. One anomaly worth noting is that when using compression and WiFitransmission, the energy cost is higher than sending uncompressed data. Thereason for this discrepancy is that the energy cost of compressing on the phonefar outweighed the reduced communication energy on WiFi radio. When theuser is roaming, local computation can be better. Through this experiment, wedemonstrate that there is no single, static, best choice when it comes to tradingo the energy costs of computation with communication; the choice of remote orlocal is a complex function of compression, computation and transmission costs.Application Stability An inherent challenge is developing a mobile appli-cation to reside on a mobile device not originally designed for use in a WBAN.Limited memory and computational resources of the mobile phone present majorchallenges to system stability e.g., an incoming call may receive higher prior-ity, competing for system memory with the KNOWME application, resulting ina non-repeatable memory allocation failure for KNOWME. Debugging crashesduring complex system interactions suggested the design approach of separatingcritical data collection from visualization/data analysis functions. The choicesof available programming paradigms on mobile phones are also limited. Forinstance, the N95 supports Python, J2ME, or Native Symbian. Each paradigmprovides a trade-o between programmer productivity and execution overhead.Due to limited debugging capability, we employed an emulator, which may notfaithfully capture mobile phone behavior. Hence, most of the system design eortwas focused on the WBAN design with the primary goal of providing robustnessunder unpredictable operating conditions. Functional Support Typical signal

13

processing methods employ complexoating-point computations that are not currently supported by the N95 hard-

ware. Such operations are executed as a software routine, consuming bothsignicant power as well as time. Thus, naive implementation of signal process-ing algorithms on the mobile phone can cause dramatic application slowdowns.We used either approximations or pre-computed values to reduce this impact.As shown in Fig. 3a, the KNOWME Network automatically recognizes phys-ical activities by fusing multimodal sensor signals as well as multidomain sub-systems. Machine learning methods are employed in order to perform accuratephysical state detection. We have designed and analyzed a signicant number ofnovel features, extracted from the biometric signals. Within these feature sets,we have assessed the most informative features. We have employed personalizedmodels tailored to individuals resulting in further performance enhancement.Finally, we underscore that our approaches account for the inherent variabilityfound within a single individuals behavior due to variations in context.

14

7 ENERGY-EFFICIENT SENSOR SELECTION

As noted, our selected set of o-the-shelf biometric sensors is Bluetooth en- abled.Bluetooth is an access protocol/technology for the exchange of data over shortdistances between both xed and mobile platforms. Several modulation formatshave been considered since the inception of the standard; currently 8- dierentialphase shift keying is possible to achieve a 3 Mb/s data rate. To achieve a desiredspectral mask, frequency- hopped spread spectrum is em- ployed, although notexploited for multiple access. The Bluetooth protocol is packet-based and en-ables a master-slave system wherein a single master may have up to seven slavesin a piconet. It is this master-slave piconet that com- prises our WBAN. Thesystem works in a time-division multiple access mode wherein slaves communi-cate with the master in a round-robin fashion; although it is the master (themobile phone) that determines with which slave it will communicate. As previ-ously noted, in contrast to the traditional view of a sensor network, it is the cellphone fusion center that is the energy bottleneck of our system.Coordinating andlistening to the Bluetooth transmissions from the biometric sensors consumesmuch more energy for the cell phone than the transmission of those signals fromthe sensors. As such, to maximize system life, we need to optimally determinewhich sensors to listen to and for how long this is the sensor selection prob-lem. In particular, we have considered the problem of allocating a xed numberof transmission samples across the sensors while balancing between classicationerror and energy consumption. Further- more, the evolution of physical activitystate during the day can be modeled as a dynamic stochastic system. We ex-plicitly considered these properties to determine sequential allocations exploitingstochastic control methods. Finally, a unique feature of WBANs with hetero-geneous sensors is that each sensor has dierent discriminative properties as wellas dierent energy costs. Ideally, we would derive an optimal sampling protocolbased on the SVM classier. Such an approach faces two important challenges:There is no closed form expression for the performance of the SVM to optimize.The top performing SVM uses a large feature set and complexity of optimalfeature selection would be prohibitive for implementation on a mobile device.Thus, we considered a single exemplary feature per sensor, exploited Gaus- sianmodels for the sensor measurements, and further approximated the proba- bil-ity of classication error in order to convert a combinatorial integer-programmingproblem into a continuous vector-valued optimization. This approximation ef-fort is largely motivated by the need to implement the sampling strategy on themobile phone. When the optimal sampling strategy derived in this manner isapplied to the SVM classier, signicant energy savings are experienced with noperformance loss. We extended our work in by modeling our allocation prob-lem as a partially observable Markov decision process (POMDP) to capture thesystems se- quential nature. POMDPs are well matched to our problem as wecan model the dynamics of physical state change via a Markov chain (Fig. 4b).Dy- namic programming (DP) and greedy search strategies are employed toopti- mize the trade-o between classication error and energy cost, and signicantenergy gains are obtained . We observe that other sensor selection strategiesfor WBANs often assume correct knowledge of the current PA state .presentsa simplied form of our proposed sensor selection scheme. The individ- ual isalternating between a set of physical activities and during this process, a setof measurements is generated and communicated to the fusion center, which

15

then estimates the underlying activity. This estimate is used to update thefusion centers belief on the underlying true activity (belief state) that in turndetermines the subsequent samples allocation.

16

8 Related Work

From early measurements, it became quickly obvious that body movementsplay an important role in WBAN fading characteristics and should therefore beac- counted in order to develop accurate and reliable channel models. Despitethe huge number of publications related to body communications, only few pa-pers characterize the time-variant properties of the channel, most likely owingto technical diculties for performing measurements. In 2007, the IEEE 802.15Task Group 6 (TG6) was formed to address specic communication standardsfor WBANs: the associated on-body channel model is known as CM3. In achannel sounder was used for measuring 10 channels congured in a star topol-ogy. The lognormal distribution seemed to match best in static scenarios whilethe Nakagami-m and Weibull distributions showed good agreement in moderateto severe fading conditions. This work was extended . Authors in also pointedout the scenario-dependent behavior of WBAN propagation, and argued thata single comprehensive channel model would not be suitable to describe theparticular features of each transmission scenario. In addition to IEEE 802.15TG6, a mixed-parameter distribution and a Nakagami-m distribution [8] weresuggested by Cotton et al. formodeling on-body fading statistics. In , a prelim-inary analysis of time series is presented using autoregressive transfer functions.Other statistical parameters, including level crossing rate (LCR) and averagefade duration (AFD), are also extracted from these measurements . More re-cently, has investigated the eect of arm waving using a dynamic phantom tomimic human walking and running motions. However, the study is focused onon- and o-body single links in the horizontal plane. By contrast, is focused onmeasuring multi-link cross-correlations in a wireless BAN using ZigBee sensornodes at 2.4 GHz. While little information is given on the other channel pa-rameters, the correlation properties are discussed as a function of the on-bodysensor locations and the various motion patterns. Correlation levels are in gen-eral very low. In , multi-link correlations are statistically characterized basedon full wave simulations of a voxel model, but no experimental validation isproposed. Finally, the activities of the Special Interest Group E (SIG-E) ofthe COST 2100 action were focusing on dierent aspectsof the dynamic on-bodychannels. The contributing authors presented a statistical model of about 20time-variant channels for dierent scenarios [16]. By applying a sliding timewindow, a correlated lognormally-distributed large-scale fading was separatedfrom the small-scale fading component. The latter was shown to be Riceanor Rayleigh distributed depending on the motion mode. Moreover, an analyt-ical model based on a threecylinder representation was introduced in , whileDoppler and correlation issues are experimentally characterized . In both cases,multi-link correlations were found to vary signicantly as a function of the mo-tion pattern and the involved nodes. Contributions In this paper, a modelfor dynamic multi-sensor WBANs is de- rived from an extensive measurementcampaign, which considers a (nearly) full mesh topology of 12 nodes. Multi-linkchannels are measured with tangentially and orthogonally polarized antennas.To the best of the authors knowledge, this is one of the very rst measurementscampaign to measure a large amount of links of the WBAN simultaneously forlong periods of time. The main features of the model are the following: 1) ascenario-based approach is used, where the received power and fading statisticsare characterized on a per-link basis; 2) the link auto-correlation is modeled,

17

taking into account the in uence of the body movement; 3) the multi-link cross-correlation is also modeled on a perlink basis. Finally, the proposed modelis compared with the original measurements by considering several validationmetrics.

18

Figure 2: a Picture of the measurement b Node placement on body.JPG

.

19

9 MEASUREMENT CAMPAIGN

A measurement campaign was performed to measure the multi-link dynamicbody area propagation channel. The WBAN consisted of 12 nodes, placed asshown in Fig. 1(a) and referred to as shown in Fig. 1(b). The measurementswere taken in a room of approximately 5 by 10 m. All furniture of the roomhad been removed as to create an empty space of size approximately 5 by 5m. In these conditions, the main re ections were from the walls, as well asfrom the channel sounder used for the measurement (3 metallic boxes of sizem). During the measurements, the person was walking freely around the room.The nodes were terminated by SMT-3TO10M-A Skycross antennas. Thesesantennas have a return loss of around 10 dB in the considered bandwidth, anda omnidirectional radiation pattern in the azimuth plane. Two polarizationswere measured, vertically and normally to the body (see Fig. 2). The personused for the measurement was a male subject (1.87 m, 85 kg). A MIMO 8 8Elektrobit channel sounder, whose transmit and receive units shared a commonclock to avoid phase drift, was used to measure the wireless channel [19]. Themeasurements were performed with 1 (or 2) node(s) transmitting and 7 (or 6)nodes receiving simultaneously, the positions of the transmit and receive nodesbeing changed between measurements to get a statistical characterization ofalmost all of the 12 12 channels. The parameters of the sounder are givenin Table I. Each node was connected to the channel sounder with a 6 mSMAcable. A pre-measurement calibration showed that the cables did not in uencethe channelmeasurements.Note that the center frequency of 4.2 GHz is due tothe channel sounder center frequency. The channel burst sample rate is about 20Hz. Each burst consist of four successive measurements that will be averagedto increase the measurement SNR. For each measurement run, 3000 channelmeasurements were recorded (representing about 35 s for each dataset). Notethat given the approximate speed of movement of our subject, a burst rate of20 Hz corresponds to a measurement approximately every . As a consequence,the sample rate might be too low to capture some of the deeper fades thatoccur in this scenario. The model described in this paper should thus be usedcautiously when considering deep fades in WBANs. In conclusion, for eachposition of the transmitter, 7 (or 6) channels are measured simultaneously, forabout 35 s. This enables to characterize the time-variant channel for each link,the narrowband channel between transmitter and receiver being denoted as ,as well as the correlation between links with an identical transmitter but withdierent receivers. center frequency 4.2GHz bandwidth 100GHz Transmit Power6dBm Channel Sample Rate 169.70Hz Channel Samples/ burst 4 Burst rate20Hz

20

10 MULTI-LINK ON-BODY CHANNELMODEL

In WBAN, the radio channel depends on a large number of parameters: bodycondition and movements, tissue properties, frequency, antenna localization andpolarization, surrounding environment, etc. All these mechanisms will have adierent impact on the channel when the human body is moving: antenna ori-entation and separation changes, direct transmissions and/or multipaths appearor vanish. Subsequently, the WBAN channel exhibits large time-variations su-perimposing both shadowing and interference eects.As it is almost impossible todistinguish between these eects, we will not try to model them separately. Thepresent section illustrates the various components of the model, which can besummarized as follows: 1) each individual link is assigned an average receivedpower (a power- distance law can be used for this, although the high spreadsaround the re- gression curve suggest that the average power should be consid-ered to be a deterministic value); 2) each link is given a fading mechanism (thestatistical distribution of the fading is identical for all links, albeit with dierentparameter values); 3) each link is given a temporal autocorrelation function, tocorrelate suc- cessive time samples of the channel; 4) nally, the dierent linksare cross-correlated (it will be shown from the measurements that certain linksshow high (anti-)correlation, that cannot be neglected from a system point ofview.)

10.1 SECURITY CHALLENGES AND BIOMETRICS

To ensure security of the overall system, BAN/BASN must be protected, for ex-ample, against eavesdropping, injection, and modication of packets. Se- curityissues in BANs/BASNs for telemedicine and m-health are particularly importantbecause sensitive medical information must be protected from unau- thorizeduse for personal advantage andfraudulent acts that might be hazardous to ausers life (e.g., alteration of system settings, drug dosages, or treatment pro-cedures). Although the security issues in networks are always considered toppriority, studies carried out in this area for BANs/BASNs were few. The verylimited work found in this area is not applicable to BASN in telemedicine andmhealth because the biosensors in a BASN must operate with extremely strin-gent constraints, a requirement that was relatively relaxed in generic BANs.Less threatening than eavesdropping and tampering problems, but as impor-tant, is avoidance of interference between BASNs of dierent individuals, be-cause communications of BASNs could easily cross over to each other whenmany people have their own BASNs in the future. Therefore, the problem ofanti-interference must also be taken care of in the development of a BASN.Even though the two challenges seem to be unrelated to each other, the prob-lems could be converged into one simple question: how can sensors or nodesof a BASN know that they belong to the same individual? In this article weintend to present and investigate the system performance of a novel biometricsolution to this question. Biometrics is a technique commonly known as the au-tomatic identication or verication of an individual by his or her physiologicalor behavioral characteristics. In order to be a practical biometrics system,it ispostulated that the utilized characteristics should be : Universal: possessed bythe majority, if not the entire population Distinc- tive: suciently dierent in anytwo individuals Permanent: suciently in- variant, with respect to the matching

21

criterion, over a reasonable period of time Collectable: easily collected and mea-sured quantitatively Eective: yield a biometric system with good performance;that is, given limited resources in terms of power consumption, computationcomplexity, and memory storn age, the characteristic should be able to be pro-cessed at a fast speed with recognized accuracy Acceptable: willingness of thegeneral public to use as an identier Invulnerable: relatively dicult to reproducesuch that the biometric system would not be easily circumvented by fraudulentacts Instead of applying bio- metrics generally in user authentication of com-mon cryptosystems, we intend to develop a technique that would authenticatesensors nodes and/or secure cipher key transmission among them in BASN. Asillustrated in Fig. , since the human body is physiologically and biologicallywell-known to consist of its own trans- mission or transportation systems (e.g.,the blood circulation system), we would like to investigate how to make use ofthese secured communication pathways available specically in BASN but notother wireless networks. It is believed that if used properly, these systems canbe naturally secured conduits for in- formation transmission within a BASN,where other techniques (e.g., hardware or software programming) must be usedin generic wireless networks to achieve the same purpose. The idea is particu-larly practical in securing BASN with a telemedicine or m-Health application,as nodes of these BASN would already comprise biosensors for collecting med-ical data, which could be physiological characteristics uniquely representing anindividual. If these intrinsic character- istics can be used to verify whether twosensors belong to the same individual, the multiple usages of the recorded phys-iological signals will certainly save re- sources while adequate security measuresare employed.

22

Figure 3: Capacities are represented by the stacked chart.JPG.

.

23

Figure 4: Level crossing rate for measurement and model for vertical polariza-tion.JPG.

Figure 5: average fade duration for measurement.JPG

24

11 EXPERIMENTAL TESTING ANDRESULTS

We used data collected previously from two experiments, of which the origi-nal purpose was to simultaneously capture electrocardiogram (ECG) and pho-toplethysmogram (PPG) data for the estimation of blood pressure. In-housecircuits were designed for capturing ECG and PPG via stainless steel electrodesand infrared optical sensors. In one experiment 14 healthy subjects were re-cruited and two PPGs captured from the index ngers of the two hands of eachsubject, and an ECG captured from three ngers of the subjects were recordedsimultaneously for 23 min. In another experiment [12], 85 subjects were re-cruited, and within a 2-month period, ECG and PPG were captured for 23 minon 3 4 days. ECG and PPG were both sampled at 1000 Hz. An algorithm wasdeveloped to mark the peak of the R-wave of ECG and the foot of the PPGpulse. We divided the ECG and PPG pairs into segments such that the cor-responding ECG segment contained exactly 68 R-waves (resulting in 67 IPIs). Atotal of 838 data segments were obtained from the 99 subjects. As shown in Fig.3, IPIs were obtained independent from the dierent physiological signals, and thesequence of IPIs in one segment was coded into a binary sequence b of 128 bits.Since the Hamming distance was used to evaluate the dissimilarity be- tweentwo binary sequences b and c, the encoding method was carefully selected suchthat the Hamming distance of the binary codes corresponding to any two similarIPIs would be small. Two binary sequences would only be considered matchedif the Hamming distance was smaller than a threshold. Ideally, b and c shouldmatch only if they were recorded from the same person during the same periodof time. Similar to the generic biometric verication systems, we evaluated theperformance of the proposed biometric approach by two types of errors: Falserejection rate (FRR), the rate of which b and c measured from the same personduring the same period of time were unmatched (i.e., corresponding to a nodein the same BASN being rejected by the judging node) False acceptance rate(FAR), the rate of which b matched c measured from a dierent person or ata dierent time (i.e., corresponding to a node of another BASN or an impostorbeing accepted as a legal node). In addition to the conditions stated above, weinvestigated the performance of the operating system when the number of IPIsused for each b was reduced from 67 IPIs to 34 IPIs so that b would have 64 bitsinstead of 128 bits. We also studied the scenario when both ECG and PPG weredownsampled from 1000 Hz to 200 Hz. The left and right panels of Fig. 4 show,respectively, the half total error rate (HTER = 1/2(FAR + FRR)) against thedistance, and the genuine acceptance rate against the false acceptance rate forthe dierent test conditions.

25

12 Threats Detection and Tracking Systems forMilitary Applications using WSNs

A Sensor node consists of following components A. Power Unit A sensor nodeconsumes power for sensing, processing and transmitting data. A power sourcecan be both a chargeable and non chargeable battery. For chargeable powersource solar cells can be used. Power unit is an important component of sensornode as lifetime of node depends on lifetime of its battery. B. Sensors A sensoris a small hardware device which is capable of generating response to change inphysical environment. Although sensors are of dierent type which are applica-tion specic but desired characteristics of a sensor node are small size and lowpower consumption. C. ADC Convertor Sensor node produces analog signalsso ADC convertors are embedded to convert analog signals to digital signals.These digital signals now can be sent for further processing. D. Storage andProcessing Unit Sensor nodes are designed for processing data before transmit-ting. A little processing is done at node level. For processing data, storage unitis required. Memory requirement is too application dependent. But usually ashmemories are used because they are cost eective. E. Transceiver Transceivers areneeded for co-ordination with other nodes. To send and receive data transceiveris needed. To minimize energy consump- tions transceivers are usually turned orather remaining in idle mode because idle mode consumes almost same poweras the transmitting mode. For trans- mitting data in wireless sensor networkRF (radio frequency), Infrared or optical communications (LASER) are possiblechoice

12.1 USING WBANS TO DETERMINE ENERGY EX-PENDITURE

As alluded to in the Introduction, the actual physical state of an individualis typically not the end-goal of health monitoring. In the context of pediatricobesity specically and obesity in general, one is concerned with how much en-ergy the individual is expending. Energy expenditure is notoriously challengingto assess over the long term due to the expense and bulkof typical measure-ment strategies. A common gold standard for energy measurement is the rateof consumption of oxygen (O2). In our studies [14], calibration of our energyexpenditure prediction methods was done against a MedGraphics Cardio IImetabolic cart, which measures the rates of O2 and CO2 consumption and ex-piration. Metabolic carts are expensive and bulky. In a typical experiment, thetest subject wears a mask that nearly obscures the entire face; a large tube fromthe mask feeds into the measurement system. Clearly, this technique does notscale well to free-living studies. More portable carts exist; however, they are stillcumbersome and prohibitively expensive. On-body sensors, particularly iner-tial sensors, represent a cost-eective alternative without sacricing accuracy. Theparticular problem we studied was the computation of energy expenditure dueto walking using body-worn inertial sensors. An inertial sensor is a device thatcaptures the movement of the object to which it is attached. Our fundamentalhypothesis was that the movement descriptors captured using inertial sensorscould be used to estimate caloric expenditure. Since no Bluetooth-based inertialsensor was available at the time, we developed our own prototype model to test

26

our algorithms. A single, custom-developed hip-mounted inertial sensor con-sisting of a triaxial ACC and a triaxial gyroscope was employed . An impor-tant component in our study [14] was to compare the ecacy of gyroscope-basedmodels against ACC-only solutions with respect to predicting energy expendi-ture. Gyroscopes are much more useful in tracking dynamic activities and donot suer from the problems of gravitational bias as do accelerometers. Our fur-ther innovation was the development of data-driven kinematic motion modelsmapping movement to energy for walking, which exploited the inherently cycli-cal nature of walk. We designed three prediction methods that show signi-cant improvement over simple linear regression tting: Least Squares Regression(LSR), Bayesian Linear Regression (BLR), and Gaussian Process Regression(GPR). Many accelerometry-based studies on physical activity in the area ofhealth focus on the use of uniaxial ACCs. Not surprisingly, triaxial informationis more accurate than uniaxial information. Our data-driven statistical modelsallowed us to bypass count-based techniques and the use of thresholds yieldingimprovements over . However,a surprise from our study was the fact that gyro-scopic information yielded prediction accuracy equivalant to, if not better, thanACCs. This result was important because it demonstrated the use of a sensoralternative to accelerometers in measuring energy expenditure.

12.2 LESSONS LEARNED FROM ENERGY EXPENDI-TURE ESTIMATOR DESIGN

Our analysis showed that LSR-based approaches are prone to outlier sensitiv-ity and overtting. Nonlinear regression methods showed better prediction accu-racy, but required an order of magnitude increase in runtime. Our study showedhow probabilistic models in conjunction with joint modeling of triaxial accel-erations and rotational rates could improve energy expenditure prediction forsteady-state treadmill walking, closely matching ground truth. Another sig-nicant contribution of our work was that our sensors transmitted data via Blue-tooth to secondary devices as opposed to local storage. We have success- fullyconnected sensors to both traditional PCs and Android-based smartphones toreceive streaming data. Range and the number of simultaneously transmit-ting Bluetooth devices limited the maximum data transmission rate using Blue-tooth. We are currently examining the integration of these inertial sensors intoKNOWME.

27

Figure 6: communication infrastructure.JPG

.

28

Figure 7: Sensor node architecture.JPG

.

29

13 CONCLUSION

In this paper, time-variant multi-sensor WBANs were investigated. An exper-imental campaign was performed to measure the dierent links of a full-meshWBAN simultaneously for long periods of time, using a multi-port channelsounder and two polarizations (vertical and normal). A stochastic channel modelwas then derived from the measurements. The power of the dierent links can-not be modeled satisfactorily with a simple power-distance law. Therefore, ascenario-based approach is chosen, where the powers of each link are modeledas deterministic values. The channel fading was modeled with a log-normaldistribution. The autocorrelation of each link was modeled by combining anexponential decay and a sinusoidal component, the latter being related to theperiodical movement of the arms and legs when walking. The parameters of theautocorrelation model were extracted from the measurements with a MMSEestimator. The correlation between the dierent links were measured, and wereshown to be high for certain links, especially for links with nodes on the limbs.The channel model was eventually validated with a number of validation met-rics. The multi-link capacity was used to determine whether the channel modelwas able to reproduce the received power and fading statistics of each link, in-dividually and taken in a multi-link setup. It was shown that there is a goodmatch between the measurements and the model. The multi-link capacity alsorevealed that, for certain nodes, the fact of using one (or two) relay(s) may sig-nicantly improve the link reliability. Finally, the time-variant behavior of themodel was successfully validated by comparing the LCR and the AFD for bothmeasurements and model, except for very low received power levels, therebyillustrating the limits of the log-normal distribution for modeling the fadingstatistics. Our main objective was to study the exiting localization approacheswhich can be deployed in WSN and could serve military applications, in orderto bring out few key issues for designing and implementing a tracking system todetect and track Threats through border security area. And have outlined howthe tracking approach can meet these requirements. It stands for integratingthe acoustic device with wireless sensor nodes in order to be able to detect andtrack the positions of threat objects. The major contribution of this paper in-cludes classication the exiting WSN-based tracking and localization systems.For future work, we aim to develop and design an acoustic-based system forThreats detection and tracking. This kind of systems is based only on emittinga voice by a threat object. Moreover, acoustic sensors are considered as an ef-fective cost solution. Further improvements would be considered to enhance thelocalization accuracy for the proposed system, by integrating ultrasonic sensorto track the positions of silent Threats. Wireless Sensor Networks are emergingas a great way of data collection of monitoring. Its applications are not limitedto Military Services but extend to vast area of human activities. The exibility,fault tolerance, high sensing delity, low-cost and rapid deployment character-istics of sensor networks create many new and exciting application areas forremote sensing. In the future, this wide range of application areas will makesensor networks an integral part of our life.

30

14 REFERENCES

1.Stephane van Roy, Francois Quitin, LingFeng Liu, Claude Oestges,Dynamicchannel modeling for multy-sensor body area networks.IEEE TRANSACTIONSON ANTENNAS AND PROPAGATION, VOL. 61, NO. 4, APRIL 2013

2.Carmen C. Y. Poon and Yuan-Ting Zhang, The Chinese University of HongKong,A novel biometrics method to secure wirless body area sensoe networktelemedicine and M-Health.IEEE Communications Magazine April 2006

3.Urbashi Mitra, B. Adar Emken, Sangwon Lee, and Ming Li,A case studyin wireless body area sensor network design. IEEE Communications MagazineMay 2012

4.International journal of advanced in computer science and software engi-neering.Volume 2, Issue 11, November 2012.

5.A study on threats detection and tracking system for military applica-tion using WSNs.International Journal of Computer Applications (0975 8887)Volume 40 No.15, February 2012.

6. P. S. Hall and Y. Hao, Antennas and Propagation for Body-CentricWireless Communications. Norwood, MA, USA: Artech House, 2006.

7. K. Y. Yazdandoost and K. Sayraan-Pou, Channel model for body areanetwork (BAN). IEEE p802.15-08-0780-09-0006, Apr. 2009, IEEE 802.15 Work-ing Group Document, Tech. Rep..

8. M. Kim and J.-I. Takada, Statistical property of dynamic ban channel gainat 4.5 GHz,. IEEE p802.15-08-0489-01-0006 Sep. 2008, IEEE 802.15WorkingGroup Document, Tech. Rep..

9. K. Takizawa, K. Y. Yazdandoost, T. Aoyagi, N. Katayama, J.-I. Takada,T. Kobayashi, H.-B. Li, and R. Kohno, Preliminary channel models for wear-able WBAN. IEEE 802.15-08-0416-02-0006, Mar. 2008, IEEE 802.15 WorkingGroup Document, Tech. Rep..

10. T. Aoyagi, J.-I. Takada, K. Takizawa, H. Sawada, N. Katayama, K.Y.Yazdandoost, T. Kobayashi, H.-B. Li, and R. Kohno, Channel models for wear-able and implantable WBANs. IEEE 802.15-08-0416-03-0006 Sep. 2008,IEEE802.15 Working Group Document, Tech. Rep..

11. D. Miniutti, L. Hanlen, D. Smith, A. Zhang, D. Lewis, D. Rodda,and B. Gilbert, Narrowband channel characterization for body area networks.IEEEp802.15-08-0421-00-0006 Jul. 2008, IEEE 802.15 Working Group Docu-ment, Tech. Rep..

12. S. L. Cotton, W.G. Scanlon, and J.Guy, The distribution applied to theanalysis of fading in body to body communication channels for re and rescuepersonnel, IEEE Antennas Wireless Propag. Lett., vol.7, pp. 6669, 2008.

13. W. Scanlon and S. Cotton, Understanding on-body fading channels at2.45 GHz using measurements based on user state and environment, in Proc.Loughborough Antennas and Propag. Conf. (LAPC), 2008

14. S. Cotton, G. Conway, and W. Scanlon, A time-domain approach to theanalysis and modeling of on-body propagation characteristics using synchro-nized measurements at 2.45 GHz, IEEE Trans. Antennas Propag., vol. 54, no.4, pp. 943955, Apr. 2009.

15. Z. Hu, Y. Nechayev, P. Hall, C. Constantinou, and Y. Hao, Measure-ments and statistical analysis of on-body, channel fading at 2.45 GHz, IEEEAntennas Wireless Propag. Lett., vol. 6, pp. 612615, 2007.

31

16. S. Cotton and W. Scanlon, An experimental investigation into the inuence of user state and environment on fading characteristics in wireless bodyarea networks at 2.45 GHz, IEEE Trans.Wireless Commun., vol. 8, pp. 612,2009 . 17. K. Ogawa and K. Honda, BAN shadowing properties of an arm-waving dynamic phantom, in Proc. 6th Eur. Conf. Antennas Propag. (EU-CAP), Apr. 2012.

18. M. Kim, K. Wangchuk, and J.-I. Takada, Link correlation property inWBAN at 2.4 GHz by multi-link channel measurement, in Proc. 6th Eur.Conf. Antennas Propag. (EUCAP), Apr. 2012.

19. C.Oliveira,M.Mackowiak, andL.M. Correia,Correlation analysis in on-body communications, in Proc. 6th Eur. Conf. Antennas Propag. (EUCAP),Apr. 2012.

20. R. Verdone and A. Zanella, Pervasive Mobile and Ambient WirelessCommunicationsCOST Action 2100. New York, NY,USA: Springer, 2012.

21. R. DErrico and L. Ouvry, A statistical model for on-body dynamicchannels, Int. J. Wireless Inf. Networks, vol. 17, no. 34, pp. 92104, 2010.

22. L. Liu, R. DErrico, L. Ouvry, P. De Doncker, and C. Oestges, Dynamicchannel modeling at 2.4 GHz for on-body area networks, Advances in Electron-ics and TelecommunicationsRadio Communication Series: Recent Advancesin Wireless Communication Networks, vol. 2, no. 4, Dec. 2011.

23. R. DErrico and L. Ouvry, Doppler characteristics and correlation pro-prieties of on-body channels, in Proc. 5th Eur. Conf. Antennas Propag.(EUCAP), Apr. 2011.

24. K. Kalliola and P. Vainikainen, Characterization system for radio chan-nel of adaptive array antennas, in Proc. Personal, Indoor and Mobile RadioCommun. Waves of the Year 2000. PIMRC 97., 8th IEEE Int. Symp., Sep.1997, vol. 1, pp. 9599, vol.1

32

Item QuantityWidgets 42Gadgets 13

Table 1: An example table.

14.1 How to add Tables

14.2 How to write Mathematics

LATEX is great at typesetting mathematics. Let X1, X2, . . . , Xn be a sequence ofindependent and identically distributed random variables with E[Xi] = andVar[Xi] =

2