5g

0018-9545 (c) 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. Seehttp://www.ieee.org/publications_standards/publications/rights/index.html for more information.

This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI10.1109/TVT.2015.2403951, IEEE Transactions on Vehicular Technology

1

Location Information Assisted Joint SpectrumSensing and Power Allocation for Cognitive RadioNetworks with Primary User Outage Constraint

Hang Hu, Student Member, IEEE, Hang Zhang, Member, IEEE, Ning Li

AbstractThe fifth generation (5G) wireless networks areexpected to achieve 1000 times higher capacity compared to thefourth generation (4G) wireless networks. Thus, improving thespectrum efficiency (SE) is a crucial problem, which must beconsidered. Cognitive radio (CR) is considered as an effectiveapproach to alleviate the spectrum scarcity problem. In this pa-per, based on the location information of the primary transmitter(PT) and the CR network, we estimate the distance betweenthe PT and the secondary transmitter (ST) and then proposea joint spectrum sensing and power allocation (JSS-PA) schemeto improve the SE of the CR network. In the JSS-PA scheme,we focus on jointly optimizing the sensing parameters and thetransmit power of the secondary user (SU) such that the SEis maximized while the primary user (PU) outage constraintis satisfied. When cooperative spectrum sensing is employed todetect the PUs status, we analyze two cooperative strategies,i.e., soft information fusion (SIF) and hard information fusion(HIF). Under SIF strategy, the optimization of sensing and power(S-OSP, for short) algorithm is proposed to maximize the SE.Under HIF strategy, the optimization of thresholds (H-OT, forshort) algorithm is proposed, and then the optimization of sensingand power (H-OSP, for short) algorithm is proposed to find theoptimal duration of local sensing, the optimal transmit power ofSU and the optimal final decision threshold. Finally, we presentthe simulation results to evaluate the performance of the proposedJSS-PA scheme and discuss the effects of the optimal parameterson different schemes under SIF and HIF strategies.

Index TermsCognitive radio, location information, spectrumsensing, power allocation, soft information fusion, hard informa-tion fusion.

I. INTRODUCTION

FUTURE wireless networks will face several challenges,such as higher data rates, lower energy consumption,higher spectrum efficiency and so on [1]. The 5G wirelesssystems, which is expected to solve these challenges, hasattracted much attention in recent years [2]-[4]. It is widelyagreed that the system capacity of the 5G network is 1000times higher than that of the 4G network [5]. To achieve thisgoal, we need more bandwidth, higher area capacity, higherspectrum efficiency, etc. Improving the SE is an importanttask since the current spectrum utilization is not quite efficient

Copyright (c) 2015 IEEE. Personal use of this material is permitted.However, permission to use this material for any other purposes must beobtained from the IEEE by sending a request to [email protected]. Hu, H. Zhang (corresponding author), and N. Li are with the

College of Communications Engineering, PLA University of Scienceand Technology, Nanjing 210007, China (e-mail: xd [email protected];hangzh [email protected]; lining [email protected]).This work was supported by the National Natural Science Foundation of

China (Grant No. 61072044).

[6]. Cognitive radio, with the aim of increasing the spectrumefficiency, has been proposed [7]. It enables dynamic spectrumaccess (DSA) by allowing the secondary users to access thespectrum bands which are allocated to the primary users [8].Accordingly, the CR technology has attracted a lot of attentionfrom academia and industry [9].The aim of the IEEE 802.22 wireless regional area network

(WRAN) standard is to allow sharing of geographically unusedspectrum bands allocated to the TV broadcast service. Itis required that no harmful interference is caused to theincumbent operation (i.e., TV users) and low-power licenseddevices [10]. In order to utilize the licensed spectrum bandswithout causing interference to the primary users, the WRANsystem should be cognizant of all the incumbent operationsnearby. The SUs can utilize the licensed spectrum bands viaspectrum sensing or power allocation. In the former scheme,the SUs need to perform spectrum sensing to detect the PUsstatus. Only when the PU is absent, the SUs are allowedto transmit data. However, when the PU is present, the CRnetwork will not be able to utilize the spectrum. We call itonly spectrum sensing (OSS) scheme in this paper. In thelatter scheme, the SUs do not need to perform spectrumsensing and are allowed to transmit data simultaneously withthe PU, as long as the interference power is constrained tobelow a tolerable level. However, the SU needs to estimatethe interference power caused to the PU [11]. We call it onlypower allocation (OPA) scheme in this paper. No matter whichscheme is used, the QoS of the PU should be guaranteed.However, under the condition of PU outage constraint,

which scheme performs better on improving the spectrumefficiency? In this paper, we investigate the effects of theSUs locations on the scheme selection. When the distancebetween the primary network and the secondary network isvery short, the transmission of the ST even with a small valueof transmit power may make the primary receiver (PR) inoutage. In this scenario, the SE of the OPA scheme will below due to the PU outage constraint. The SUs may employthe OSS scheme because the signal-to-noise ratio (SNR) ofthe received signal is high and the SUs can easily detect thePTs status. Thus, the SE can be improved. When the distancebetween the primary network and the secondary network isvery long, the data transmission between the SUs will havelittle interference on the PU transmission. Because of theeffect of path loss, the PU outage constraint may be satisfiedeven when the ST transmits data with its maximum power. Inthis scenario, the spectrum sensing is unnecessary because it



2

introduces additional overhead. Hence, the OSS scheme mayperform worse than the OPA scheme. In the other scenarios,joint spectrum sensing and power control can be used toprotect the PRs. In previous works, authors in [12] proposed ajoint spectrum sensing, access and power allocation scheme toimprove the secondary users throughput. In [13], the SUs areable to dynamically determine whether to employ spectrumsensing or power control under the QoS constraint, so betterdelay QoS provisioning can be achieved and higher throughputof the CR network can be obtained. There are many studies onjoint spatial-temporal sensing optimization which uses the PUlocation information. In [14], to achieve efficient joint spatial-temporal spectrum sharing, the sensing duration is optimizedto maximize the throughput of the secondary system. However,the power allocation of ST is not investigated. In [15], atwo- dimensional sensing framework is proposed to improvethe opportunity detection performance by fusing the sensingresults in a spatial-temporal sensing window. However, theoptimal design of the spatial-temporal sensing window is notinvestigated in this paper. In [16], a location-aware strategy isproposed to minimize the total power consumption subject to aminimum data rate requirement. However, the effect of sensingduration on the total power consumption is not investigated.In our paper, we propose a joint spectrum sensing and power

allocation (JSS-PA) scheme based on the location informationof the PT. In the JSS-PA scheme, when missed detectionoccurs, the SUs coexist with PUs in the same spectrum band.If the ratio of SU signal over PU signal is larger than apreset value, the PR will be in outage. Both the sensingperformance and the ST transmit power can be adjusted toimprove the SE of the CR network while satisfying the PUoutage constraint. Thus, the JSS-PA scheme is expected toobtain higher spectrum efficiency. If the transmit power of theST is increased with the aim of improving the SE of the CRnetwork, the interference to the PR will be larger, hence moreaccurate spectrum sensing technology should be employed tomake the missed detection probability smaller. If the SUs havelimited sensing abilities and the missed detection probabilityis a large value, the ST must control its transmit power toprotect the PR sufficiently.In either the OSS scheme or the JSS-PA scheme, spectrum

sensing will be conducted. However, reliable spectrum sens-ing is not always guaranteed due to the detrimental effectsof fading and shadowing [7]. Hence, cooperative spectrumsensing has been proposed to improve the sensing performance[17]. In this paper, two cooperative strategies will be analyzed,i.e., soft information fusion and hard information fusion [18].In the former strategy, the received signal of each SU isamplified and sent to the fusion center (FC). After the FCcollects all the local sensing information, an energy detectoris employed to indicate that the PT is present or absent. Inthe latter strategy, each SU makes a one bit decision toindicate the PTs status in the local sensing phase. All theone bit decisions are reported to the FC in the reportingphase. Then, according to some fusion rules, the FC makesa final decision on the PTs status. For cooperative spectrumsensing with hard information fusion (HIF) strategy, authorsin [19] proposed weighted decision fusion schemes to improve

the SUs throughput.Under soft information fusion (SIF) strategy, the optimal

weight coefficients are employed to combine the sensinginformation in the FC [20]. In the JSS-PA scheme, we focuson jointly optimizing the duration of local sensing and theST transmit power such that the SE is maximized whilethe PU outage constraint is satisfied. Under HIF strategy,some previous works investigated the k-out-of-N fusionrule based on the assumption that the cooperating SUs havethe same SNR value of the received signal [21]-[22]. In oursystem model, the SUs are uniformly distributed in a circulararea. Thus, the SUs in different locations will have differentSNR values, and the k-out-of-N fusion rule is not applicablein this paper. In the proposed fusion rule, all the one bitdecisions are included in a predefined set, then the sensingperformance can be measured.The main contributions of this paper are summarized as

follows: In order to improve the spectrum efficiency, we chooseOSS scheme or OPA scheme based on the SUs locations.Then, we propose a joint spectrum sensing and powerallocation (JSS-PA) scheme based on the location infor-mation of the PT. The cooperative spectrum sensing withSIF and HIF is investigated. Simulation results show thatthe JSS-PA scheme outperforms both the OSS schemeand the OPA scheme, the SE of the SIF strategy is higherthan that of the HIF strategy.

For cooperative spectrum sensing with SIF, under the PUoutage constraint, it is proved that there exists one optimalduration of local sensing that can maximize the SE of theCR network. In the JSS-PA scheme, in order to maximizethe SE, the optimization of sensing and power (S-OSP,for short) algorithm is proposed to optimize the durationof local sensing and the ST transmit power jointly.

For cooperative spectrum sensing with HIF, the propertiesof the final false alarm probability and the final detectionprobability are analyzed. In the OSS scheme, under HIFstrategy, the optimization of thresholds (H-OT, for short)algorithm is proposed to optimize the energy detectionthreshold and the final decision threshold in the FC.Based on the H-OT algorithm, in the JSS-PA schemeunder HIF strategy, the optimization of sensing and power(H-OSP, for short) algorithm is proposed to find theoptimal duration of local sensing, the optimal transmitpower of ST and the optimal final decision threshold thatcan maximize the SE of the CR network.

The rest of this paper is organized as follows. The systemmodel and problem formulation are presented in Section II.The solutions of the formulation under SIF strategy are givenin Section III. Section IV is devoted to the solutions of theformulation under HIF strategy. Simulation results are shownin Section V. This paper concludes with Section VI.Notation: pfa, pde and pmd respectively denote the false

alarm probability, the detection probability and the misseddetection probability for individual SU; QFA, QDE and QMDdenote the corresponding probabilities of cooperative spectrumsensing in the FC; Subscripts s and h denote SIF and HIFrespectively.



3

PT

PR

PR

PR

CRN 1

CRN 2

CRN 3

CRN 4

PT: PU Transmitter PR: PU Receiver CRN: CR Network

SU: Secondary User

FC: Fusion Center

ST: SU Transmitter SR: SU Receiver

PR

PR

SSl

PSl

L

SPl

L

Fig. 1. System model.

II. SYSTEM MODEL AND PROBLEM FORMULATION

Our system model is illustrated in Fig. 1. In the primarynetwork, PT denotes the PU transmitter, and PR representsthe PU receiver. We suppose that the location of the PT andthe transmit power of the PT are known to the SUs. Thisassumption is reasonable. When the PT is a TV transmitter, itslocation and transmit power are possibly known because theseparameters are fixed. In [23], the fusion center can use theenergy levels sent by the SUs to construct channel gain (CG)maps and estimates the PU locations and the transmit powerlevels. However, the SUs are unable to access the databaseand hence do not have the knowledge of operation time ofthe PT. Thus, spectrum sensing is required to decide whetherthe PT is present. Since the PRs should be protected, a PT-centered boundary will be determined by the minimum SNRof received PT signal. Without loss of generality, the PT isassumed to be located at coordinate (0; 0). The protected areais a circular field, and the radius of the protected boundary isdenoted as L.The CR network consists of a number of SUs and a fusion

center. We consider that the SUs are uniformly distributed in acircular field with a radius of r, the fusion center is assumedto be located in the centre. In order to obtain the locationinformation, the devices in the cognitive radio network areequipped with satellite-based geolocation technology (e.g. G-PS). The SUs detect the PTs status in the local sensing phase.In the reporting phase, all the sensing results are reported to theFC via a common control channel (CCC). Then, the FC makesa final decision to indicate that the PT is present or absent.If the PT is absent, one of the secondary users is allowed toconduct data transmission.We assume that the PT signal is BPSK signal, the noise is

real-valued Gaussian variable with zero mean and variance 2.In the local sensing phase, an energy detector is employed foreach SU to detect the PTs status. The false alarm probability,the detection probability, and the missed detection probabilityat jth SU can be calculated as [24]

pfa;j = Q

j2 1r

tsefs2

!; (1)

pde;j = Q

j2 j 1

stsefs

2(2j + 1)

!; (2)

pmd;j = 1 pde;j ; (3)where j is the threshold of energy detection at jth SU, tseis the duration of local sensing, fs is the sampling frequency,

j denotes the SNR of PTs signal at jth SU. For a givenmissed detection probability, the false alarm probability canbe expressed as

pfa;j = Q p

2j + 1Q1(1 pmd;j) + jrtsefs2

!: (4)

Let QFA, QDE and QMD denote the false alarm probabil-ity, the detection probability and the missed detection proba-bility of cooperative spectrum sensing in the FC respectively.The transmit power of PT is denoted as PPT , the distance

between the PT and the PR is assumed to be l. The average re-

ceived power of PR can be calculated as Pl =PPT gP E

h2P

l ,

where gP is the channel gain between the PT and the PR,hP is the channel response of PT to PR, and is the pathloss exponent. If PR is located on the protected boundary, i.e.,the distance between the PT and the PR is L, the received

power of PR is PL =PPT gP E

h2P

L . The protected boundary

is determined by the threshold SNR th of the PR. Then, theradius of the protected boundary L can be computed by

L =

PPT gP E

h2P

th 21

! 1

; (5)

where 21 is the variance of the noise at PR. In general, thelocations of the PRs are unknown to the secondary users, hencethe PRs can be anywhere within the protected area. When theST is located in the protected area (e.g. CNR 1 in Fig. 1),any transmission of ST even with a small value of transmitpower can make the PR in outage. Therefore, in this case, theOPA scheme can not be employed, the SUs should use OSSscheme and perform spectrum sensing to sufficiently protectthe PR. In this paper, it is assumed that the outage probabilityof the PR should not be larger than a preset value pthout. In theOSS scheme, the SUs cooperatively sense the PTs status, thePR will be in outage when missed detection occurs. Thus, weset QMD pthout as a constraint.

A. OSS Scheme

The transmit power of ST is denoted as PST , the distancebetween the ST and the secondary receiver (SR) is assumed tobe lSS . The received power of SR from ST can be calculatedas PlSS =

PST gS h2SSlSS

, where gS is the channel gain betweenthe ST and the SR, hSS is the channel response of ST to SR.When the PT is actually absent, the transmission rate of theCR network can be computed by

1 = log2(1 + S) = log2

1 +

PlSS22

= log2

1 +

PST gS h2SSlSS 22

;

(6)



4

where S denotes the SNR of the secondary link, 22 is thevariance of the noise at SR. However, in realistic scenario,perfect spectrum sensing without sensing error is not achiev-able, the PTs true status may be incorrectly detected. Thedistance between the PT and the SR is assumed to be lPS ,the received power of SR from PT can be calculated as

PlPS =PPT gP E

h2P

lPS

. When the PT is incorrectly detectedto be absent while its true status is present, the transmissionrate of the CR network can be calculated as

2 = log2(1 + SI) = log2

1 +

PlSS(PlPS +

22)

= log2

1 +

PST gS h2SS lPS22 lPS + PPT gP E

h2P lSS

;

(7)

where SI denotes the signal to interference plus noise ratio(SINR) of the secondary link.It is assumed that the distance between the PT and the

ST is dPS , the coordinate of the ST is (xST ; yST ) and thecoordinate of the SR is (xSR; ySR). Then, we can obtainthat dPS =

px2ST + y

2ST , lPS =

px2SR + y

2SR and lSS =p

(xST xSR)2 + (yST ySR)2.The ST will conduct data transmission in the following two

cases:(i) The PTs true status is absent, and the final decision

of the FC indicates that the PT is absent. The probability ofthis case happening is (1 QFA), where represents theprobability that the PTs true status is absent.(ii) The PTs true status is present, and the final decision of

the FC indicates that the PT is absent, i.e., missed detectionoccurs. The probability of this case happening is (1)QMD,where 1 represents the probability that the PTs true statusis present.Since the duration of the data transmission in one frame is

TtseKtre [25], in case (i), the average SE of the cognitiveradio network can be computed by

1 = (1QFA) 1 T tse KtreT

: (8)

In case (ii), the SUs coexist with PUs in one channel, theprimary signal is an interference to the SR. The average SEof the cognitive radio network can be calculated as

2 = (1 )QMD 2 T tse KtreT

: (9)

Considering the above two cases, the average SE of thecognitive radio network is given by

= 1 +2 =(1QFA) 1

+ (1 )QMD 2 T tse Ktre

T:

(10)

In our system model, the PRs should be protected. When theST is located in the protected area, the SUs use OSS scheme.The missed detection probability should not be larger thanpthout, i.e., QMD pthout. Our goal is to optimize the sensingparameters to maximize the SE under the condition of PUoutage constraint. Mathematically, the problem is written asfollows

OP : max (11)

C1 : QMD pthout (12)C2 : 0 < tse < T Ktre (13)

Different cooperative strategies require different approachesto solve the optimization problem (11). For SIF strategy, theproblem will be solved in Section III; For HIF strategy, theproblem will be solved in Section IV.

B. OPA Scheme

When the location of ST is outside the protected area (e.g.CNR 2 in Fig. 1), the SUs may employ OSS, OPA, or JSS-PAscheme. The OSS scheme has been analyzed in Section II-A.For the OPA scheme, the SUs do not need to perform spectrumsensing, however, the ST should control its transmit power toavoid interference to the PRs. Since the locations of the PRsare unknown to the secondary users, to sufficiently protectthe PRs, the secondary users should suppose that one PR islocated on the protected boundary, and this PR is also locatedon the line between the PT and the ST. Thus, the distancebetween the PT and the PR is L, the received power of PR

from PT is PL =PPT gP E

h2P

L . The distance between the ST

and the PR is assumed to be lSP , the received power of PRfrom ST can be calculated as PlSP =

PST gS h2SPlSP

, where hSPis the channel response of ST to PR, and h2SP is assumed to beexponentially distributed with E

h2SP

= 1. If the ratio of SU

signal over PU signal is larger than a preset value , the PRwill be in outage. In the OPA scheme, the outage probabilityof PR is given by

pout = ProbPlSPPL

>

= Prob

(PST gS L h2SP

PPT gP lSP Eh2P > )

= e PPT gP l

SP E[h2P ]

PST gS L :

(14)

To sufficiently protect the PRs, pout must be equal to orless than pthout, i.e., pout pthout. Then, we can obtain that

PST PPT gP lSP E

h2P

lnpthout

gS L = P ST : (15)In order to maximize the SE of the CR network, the transmit

power of ST should be equal to P ST . In this case, the averageSE of the cognitive radio network can be presented as

= 1 +2 = log21 +

P ST gS h2SSlSS 22

+ (1 ) log2

1 +

P ST gS h2SS lPS22 lPS + PPT gP E

h2PlSS

:

(16)

When the location of ST is quite far away from the PTand lSP is a very large value (e.g. CNR 4 in Fig. 1), thedata transmission between the SUs will have little interferenceon the PU transmission. Due to the effect of path loss,pout may be equal to or less than pthout even when the STtransmits data with its maximum power PST;max. In this case,



5

the spectrum sensing is unnecessary because it introducesadditional overhead, and the OPA scheme is better comparedwith the OSS scheme. It is assumed that when lSP lthSP , theST can transmit data with its maximum power PST;max andpout pthout is also guaranteed. When the distance betweenthe ST and the PR is lthSP and the ST transmits data withPST;max, the received power of PR from ST can be calculatedas PlthSP =

PST;maxgS h2SP(lthSP )

. Then, the outage probability of PRis given by

pout = ProbPlthSPPL

>

= e

PPT gP (lthSP )

E[h2P ]PST;maxgS L = pthout:

(17)Solving the above equation, it is derived that

lthSP = L "PST;max gS ln

pthout

PPT gP E [h2P ]

# 1

: (18)

Let L+ = L+ lthSP , when dPS L+, the ST can transmitdata with its maximum power PST;max to maximize the SEof the CR network.

C. JSS-PA Scheme

When the location of ST is outside the protected area andL < dPS < L

+ (e.g. CNR 3 in Fig. 1), joint spectrum sensingand power control can be used to protect the PRs, i.e., the JSS-PA scheme. In this case, when missed detection occurs, theSUs coexist with PUs in the same frequency band. If the ratioof SU signal over PU signal is larger than a preset value , thePR will be in outage. Thus, in the JSS-PA scheme, the outageprobability of PR can be presented as

pout = QMD ProbPlSPPL

>

= QMD Prob

(h2SP >

PPT gP lSP Eh2P

PST gS L)

= QMD ePPT gP lSP E[h2P ]

PST gS L :(19)

In (19), it can be seen that both the missed detectionprobability QMD and the ST transmit power PST can beadjusted to guarantee that pout pthout. If the transmit powerof the ST is increased with the aim of improving the SE ofthe CR network, the interference to the PR will be larger,hence more accurate spectrum sensing technology should beemployed to make the missed detection probability smaller.If the SUs have limited sensing abilities and QMD is a largevalue, the ST must control its transmit power to protect thePR sufficiently.According to the analyses in Section II-B, if PST > P ST ,

the probability ProbnPlSPPL

> o

will be larger than pthout. Inthis case, the missed detection probability QMD should beadjusted to make pout pthout. According to (19), we canderive that

QMD pthout ePPT gP lSP E[h2P ]

PST gS L : (20)

When the distance between the PT and the ST satisfiesL < dPS < L

+, and the transmit power of the ST islarger than P ST , the SUs can employ JSS-PA scheme. Theoutage probability of PR should not be larger than pthout, i.e.,pout pthout. Our goal is to optimize the sensing parametersand the ST transmit power to maximize the SE underthe condition of PU outage constraint. Mathematically, theproblem is written as follows

OP : max = (1QFA)1+ (1 )QMD 2

T tse KtreT

(21)

C1 : L < dPS < L+ (22)

C2 : pout pthout (23)

C3 : P ST < PST PST;max (24)

C4 : 0 < tse < T Ktre (25)For SIF strategy and HIF strategy, the above optimization

problem will be solved in Section III and Section IV respec-tively.

III. SOLUTIONS OF FORMULATION UNDER SIF STRATEGY

In the CR networks, cooperative spectrum sensing requirescooperation among multiple SUs from different locations.When the SIF strategy is employed, the received signal ofeach SU is amplified and sent to the FC. After FC collects allthe local sensing information, the energy detection techniqueis used to decide that the PT is present or absent.For the jth secondary user, the decision statistic of energy

detection is denoted as Vj . The FC receives V1; V2; ; VKfrom the SUs, where K is the number of SUs in the CRnetwork and Vjs are assumed to be independent and identicallydistributed. According to [24], when the PT signal is BPSKsignal, the noise is real-valued Gaussian variable with zeromean and variance 2, we have

Vj N2; 2

4

tsefs

H0;

Vj N2(j + 1);

24(2j+1)tsefs

H1:

(26)

In the FC, the test statistic for cooperative spectrum sensingwith SIF is Vs = 1V1 + 2V2 + + KVK =

PKj=1 jVj

[20], where j is the weight coefficient of the jth SU. SinceVjs are independent and identically distributed, it is derivedthat

Vs N2PK

j=1 j ;24

tsefs

PKj=1

2j

H0;

Vs N2PK

j=1 j(j + 1);24

tsefs

PKj=1

2j (2j + 1)

H1:(27)

The probability density functions (PDFs) of Vs under H0 andH1 can be respectively written as

fVsjH0(v) =1

22

stsefs

PK

j=1 2j

e tsefs(v

2PKj=1 j)

2

44PKj=1

2j ; (28)



6

fVsjH1(v) =1

22

stsefs

PK

j=1 2j (2j + 1)

etsefs[v2PKj=1 j(j+1)]2

44PKj=1

2j(2j+1) :

(29)

The final false alarm probability, the final detection proba-bility and the final missed detection probability of cooperativespectrum sensing with SIF are computed by, respectively,

QFA;s = ProbVs > sjH0

=

Z 1s

fVsjH0(v)dv

= Q0@0@s

2

KXj=1

j

1As tsefs2PK

j=1 2j

1A ; (30)

QDE;s = ProbVs > sjH1

=

Z 1s

fVsjH1(v)dv

= Q0@0@s

2

KXj=1

j(j + 1)

1As tsefs2PK

j=1 2j (2j + 1)

1A ;(31)

QMD;s = 1QDE;s; (32)where s is the decision threshold in the FC. For a givenfinal missed detection probability QMD;s, the final false alarmprobability QFA;s can be expressed as

QFA;s = Q vuutPKj=1 2j (2j + 1)PK

j=1 2j

Q1(1QMD;s) +KXj=1

jj

stsefs

2PK

j=1 2j

!:

(33)

According to (10), for the cooperative spectrum sensingwith SIF, the average SE of the cognitive radio network isgiven by

s =(1QFA;s) 1+ (1 )QMD;s 2

T tse KtreT

:(34)

A. Optimal OSS Scheme under SIF Strategy

In the OSS scheme in accordance with (11)-(13), the opti-mization problem is maximizing the SE of the CR network sunder the condition of PU outage constraint QMD;s pthout.Theorem 1: Given the duration of local sensing tse, s is

maximized when QMD;s = pthout.Proof: For a given duration of local sensing tse, we take

the first partial derivative of s with respect to QMD;s andobtain

@s@QMD;s

=

dQFA;sdQMD;s

1

+ (1 ) 2 T tse Ktre

T;

(35)

where

dQFA;sdQMD;s

=

dQFA;s=ds

dQMD;s=ds

= dQFA;s=dsdQDE;s=ds

: (36)

Then, it is derived that

dQFA;s=ds = 122

s

tsefs

PKj=1 2j e 12!21 ; (37)

dQDE;s=ds = 122

s

tsefs

PKj=1 2j (2j + 1) e 12!22 ;

(38)where !1 =

s2

PKj=1 j

qtsefs

2PK

j=1 2j

and

!2 =hs2

PKj=1 j(j + 1)

iqtsefs

2PK

j=1 2j (2j+1)

.Substituting (37) and (38) into (36), we have

dQFA;sdQMD;s

= vuutPKj=1 2j (2j + 1)PK

j=1 2j

e 12 (!21!22) < 0: (39)

Thus, it can be concluded that @s@QMD;s > 0. Therefore, sis an increasing function of QMD;s, and s is maximizedwhen QMD;s = pthout. Theorem 1 is proved.Theorem 2: For cooperative spectrum sensing with SIF,

there exists one optimal duration of local sensing tse that canmaximize the SE s.Proof: See the Appendix A.According to [20], the optimal weight coefficient to combine

the sensing information can be calculated as j =

j

j+1.

In order to obtain the optimal duration of local sensing, wecan set @s@tse = 0. Compared with Bi-section and Golden-section algorithms, the Newton-Raphson method has a muchfaster convergence rate [26]. Hence, in this paper, we employNewton-Raphson method to find the root of @s@tse = 0.

B. Optimal JSS-PA Scheme under SIF Strategy

When the location of ST is outside the protected area andL < dPS < L

+ is satisfied, we may use JSS-PA scheme. Inthis case, the duration of local sensing tse and the ST transmitpower PST can be jointly optimized to maximize the SE of theCR network. The PU outage constraint pout pthout has beenconverted to a constraint on the missed detection probability,which is shown in (20). Thus, the optimization problem canbe written as

OPs : max s (40)

C1 : QMD;s pthout ePPT gP lSP E[h2P ]

PST gS L

(41)C2 : P ST < PST PST;max (42)C3 : 0 < tse < T Ktre (43)

For a given pair of PST and tse, it has been proved that sis an increasing function of QMD;s in the Section III-A. Thus,

s is maximized when QMD;s = pthout ePPT gP lSP E[h2P ]

PST gS L .Note that, for a given location of ST, QMD;s is a function

of PST , QFA;s is a function of QMD;s or PST . Based on(34), s is also a function of PST . Due to the complexityof Q() and Q1() functions as well as the coupling effectbetween QFA;s and QMD;s, s may not be a unimodal orconcave function of PST for a given coordinate of the ST



7

( , )ST STx y

SPl

Calculate the distance between the ST and

the PR and .STP

Known

parameters

coordinate of the ST ,

tolerance , , 0m ,max(0)ST STP P

m=m+1

( 1) ( )s sm m ! " #

Given , calculate the

optimal sensing time ;

( 1)STP m

( )set m

Given , find the optimal

transmit power .

( )set m

( )STP m

End

( )ST STP P m ! ( )

se set t m !

,max( )

s sm" ! "

Yes

No

, ,

Fig. 2. The flowchart of the S-OSP algorithm.

(xST ; yST ). Generally speaking, it is very difficult to find anefficient algorithm to obtain the optimal value of PST for agiven duration of local sensing tse. Since the range of PST(P ST < PST PST;max) is known to us and P ST can beobtained from (15), finding the optimal PST is still solvable.For a given tse, the optimal transmit power of ST P ST can beachieved by using an exhaustive search.According to the analyses in Section III-A, s is a concave

function of tse. For a given ST transmit power PST , theoptimal duration of local sensing tse can be achieved by usingthe Newton-Raphson method.Under SIF strategy, the optimization of sensing and power

(S-OSP, for short) algorithm is proposed to find the optimalduration of local sensing tse and the optimal transmit power ofST P ST that can maximize the SE of the CR network. In the S-OSP algorithm, the ~(m) represents ~ atmth iteration, where~ may be the duration of local sensing tse, the ST transmitpower PST or the SE s. Initially, PST and s are set to bePST;max and 0 respectively. When s(m+1)s(m) is lessthan or equal to , the iteration is terminated. The flowchartof the S-OSP algorithm is illustrated in Fig. 2.In the S-OSP algorithm, for a given transmit power, the

optimal duration of local sensing is calculated by using theNewton-Raphson method; For a given duration of local sens-ing, the optimal transmit power is found by using an exhaustivesearch. The optimal pair of duration of local sensing and trans-mit power is obtained by using an efficient iterative algorithm[27]. Therefore, the complexity of the S-OSP algorithm iseasy to calculate, i.e., multiplying the complexity of Newton-Raphson method by the complexity of exhaustive search.The S-OSP algorithm iterates until the SE converges to a

maximum value. Because at the converged values tse andP ST , we have s(t

se; P

ST ) s(tse; PST ) for P ST 0, YM > 0.Proof: If the local decision of the jth SU is Rj = 0 and the

final decision of the FC indicates that the PT is present, then[R1 R2 Rj1 0 Rj+1 RK ] is included in 1. In thiscase, when the values of R1; R2; ; Rj1; Rj+1; ; RKremain unchanged, [R1 R2 Rj1 1 Rj+1 RK ] isdefinitely included in 1 because the FC decides that the PTis present even when Rj = 0. On the condition that the valuesof R1; R2; ; Rj1; Rj+1; ; RK remain unchanged, itmakes no sense that the FC indicates the presence of PT whenRj = 0 while it indicates the absence of PT when Rj = 1.Whereas: if the local decision of the jth SU is Rj =

1 and the final decision of the FC indicates that the P-T is present, then [R1 R2 Rj1 1 Rj+1 RK ]is included in 1. In this case, when the valuesof R1; R2; ; Rj1; Rj+1; ; RK remain unchanged,[R1 R2 Rj1 0 Rj+1 RK ] may not be included in

1 because the jth SU decides that the PT is absent (Rj = 0)in the local sensing phase. Rj = 0 may not contribute to thefinal decision that the PT is present.Since the value of Rj is 0 or 1, according to the equations

(44) and (45), it can be concluded that XM > 0, YM > 0.Theorem 4 is proved.According to (47), analyzing the property of h over tse

is very difficult, and the explicit expression for @h@tse is hardto derive. Generally speaking, it is very difficult to find anefficient algorithm to obtain the optimal value of tse. Sincethe range of tse (0 < tse < T Ktre) is known to us, findingthe optimal tse is still solvable. The optimal duration of localsensing tse can be achieved by using an exhaustive search.Next, we will optimize the final decision threshold M with aspecific value of tse.From theorem 3, it is shown that h is maximized when

QMD;h = pthout. Hence QMD;h will be replaced with p

thout

in the following analysis. For a specific value of tse, (1 )QMD;h 2 TtseKtreT is a fixed value. Therefore, theoptimization problem can be reduced to

OP2 : max (1QFA;h) (50)C1 : QMD;h = pthout (51)C2 : 1 M K (52)

According to [28], this problem can be solved by using theNeyman-Pearson criteria. Thus, the following two formulasshould be satisfied

=

QKj=1(1 pde;j)1RjpRjde;jQKj=1(1 pfa;j)1RjpRjfa;j

M; (53)

1X

R:M

24 KYj=1

(1 pde;j)1RjpRjde;j

35 = pthout: (54)

Then, (53) can be written as

KYj=1

(pde;jpfa;j

Rj1 pde;j1 pfa;j

1Rj)M; (55)

KXj=1

Rj ln pde;j

pfa;j+ (1Rj) ln 1 pde;j

1 pfa;j

lnM:

(56)Theorem 5: Given the duration of local sensing tse (0 0, ln1pde;j1pfa;j < 0.

Proof: See the Appendix C.Theorem 5 and (56) indicate that, for a given M , Rj = 1

will contribute to the final decision that the PT is present whileRj = 0 will contribute to the final decision that the PT isabsent.For large scale CR networks, computing all the j values

will introduce a large amount of overhead. Thus, in thissection, it is assumed that all the SUs have the same energy de-tection threshold, i.e., 1 = 2 = = K = h. Accordingto theorem 4, it can be concluded that QMD;h is an increasingfunction of h. For the fusion rule in the FC, when there isonly one element in the set 1, QMD;h will be maximized.The only possible case is Rj = 1 for j = 1; 2; ;K.In this case, lnM should be chosen as

PKj=1

nln

pde;jpfa;j

o.

Let us denote +h as the energy detection threshold whichcan satisfy 1 QKj=1 pde;j = pthout. For any positive value, QMD;h(+h ) < QMD;h(+h ) = pthout. With energydetection threshold +h in the local sensing phase, it isimpossible to find a 1 that can satisfy the constraint (51).Thus, +h is the lower bound to satisfy (51). When all theR values are included in 1 except for [0; 0; ; 0], QMD;hwill be minimized. In this case, lnM should be chosen asPK

j=1

nln

1pde;j1pfa;j

o+, where is a small positive value. Let

us denote ++h as the energy detection threshold which cansatisfy 1

h1QKj=1(1 pde;j)i = pthout. For any positive

value , QMD;h(++h + ) > QMD;h(++h ) = p

thout. With

energy detection threshold ++h + in the local sensingphase, it is impossible to find a 1 that can satisfy theconstraint (51). Thus, ++h is the upper bound to satisfy(51). Based on the above analysis, the range of h hasbeen reduced to

+h ;

++h

, and the range of lnM has been

reduced to (Z+;Z++], where Z+ =PK

j=1

nln

1pde;j1pfa;j

oand

Z++ =PK

j=1

nln

pde;jpfa;j

o.

Since it is very difficult to analyze the property of h overtse, the optimal value of tse will be obtained by an exhaustivesearch. For a given tse, under HIF strategy, the optimizationof thresholds (H-OT, for short) algorithm is proposed to findthe optimal values of h and M that can satisfy OP2. In thethe H-OT algorithm, the F(i) represents F at ith iteration,where F may be the energy detection threshold h, the finaldecision threshold M or the set 1. Initially, h and M areset to be +h and e

Z++ respectively. The flowchart of the H-OTalgorithm is illustrated in Fig. 3.In the H-OT algorithm, the optimal duration of local sensing

tse and the optimal final decision threshold M are obtained



9

Known

parameters

Only one element is included in ,

to satisfy ,

i=i+1

Calculate with each except for the

elements in , choose with the

largest to be included in .

Yes

No

1,1, ,1 !" #$ % 1(0)&

(0)h h

'(

, ,11

K th

MD h de j outjQ p p

(( ) (*

,

,1

ln

(0)

Kde j

fa jj

p

pM e e

''(

+ ,- -- -- -. /- -- -- -0 1

2( (

Z

3 R

1( 1)i& ) R

31( )i&

Compute which satisfies by

searching from to , calculate

with and .

( )hi ,

th

MD h outQ p

h !

h !!

( )M i

( )hi

1( )i"

2 1Ki #

Select a pair of and that can

maximize .

( )hi ( )M i

,1

FA hQ

Fig. 3. The flowchart of the H-OT algorithm.

by comparing 2K possible values. Thus, the H-OT algorithmis convergent. In the following, the optimization of sensingand power (H-OSP, for short) algorithm is proposed to findthe optimal duration of local sensing, the optimal transmitpower of ST and the optimal final decision threshold that canmaximize the SE of the CR network. Note that the H-OSPalgorithm contains the H-OT algorithm, hence we will analyzethe computational time of the H-OSP algorithm.

B. Optimal JSS-PA Scheme under HIF Strategy

When the location of ST is outside the protected areaand L < dPS < L+ is satisfied, we may use JSS-PAscheme. In this case, the energy detection threshold h, theduration of local sensing tse, the final decision thresholdM and the ST transmit power PST should be optimizedto maximize the SE of the CR network. The PU outageconstraint pout pthout has been converted to a constrainton the missed detection probability, which is shown in (20).Therefore, the optimization problem is maximizing the SE ofthe CR network h subject to (23), (24) and (25). Since his an increasing function of QMD;h, h is maximized when

QMD;h = pthout e

PPT gP lSP E[h2P ]PST gS L .

For a given ST transmit power PST , let us denote h as theenergy detection threshold which can satisfy 1QKj=1 pde;j =pthout e

PPT gP lSP E[h2P ]PST gS L and h as the energy detection

threshold which can satisfy 1 h1QKj=1(1 pde;j)i =

pthout ePPT gP lSP E[h2P ]

PST gS L . In JSS-PA scheme, the range of hwill be

h ;

h

.

( , )ST STx y

SPl

Calculate the distance between the ST and

the PR and .STP

Known

parameters

coordinate of the ST ,tolerance , , , 0n ,max(0)ST STP P

n=n+1

( 1) ( )h hn n ! " #

Given and , calculate

and by using algorithm 2.

End( )ST STP P n

!

( )se set t n !

,max( )

h hn" ! "

Yes

No

!1

(0)2se re

t T Kt" #

( 1)set n # ( 1)

STP n # ( )

hn

( )M n

( )M M n$ "

( should be replaced with )th

outp

2[ ]PT P SP P

ST S

P g l E h

P g Lth

outp e

! !

Calculate the optimal sensing time

with , and .

( )set n

( 1)STP n ( )

hn ( )M n

Compute the optimal transmit power

with , and .

( )STP n

( )set n ( )

hn ( )M n

Fig. 4. The flowchart of the H-OSP algorithm.

Note that, analyzing the property of h over PST is verydifficult since there is no explicit expression for h. Generallyspeaking, it is very difficult to find an efficient algorithm toobtain the optimal value of PST . Since the range of PST(P ST < PST PST;max) is known to us and P ST can beobtained from (15), we can employ an exhaustive search tofind the optimal transmit power of ST P ST .

Under HIF strategy, the optimization of sensing and power(H-OSP, for short) algorithm is proposed to find the optimalduration of local sensing tse, the optimal transmit power ofST P ST and the optimal final decision threshold M

that canmaximize the SE of the CR network. In the H-OSP algorithm,the (n) represents at nth iteration, where may be theduration of local sensing tse, the ST transmit power PST , thefinal decision thresholdM or the SE h. Initially, tse and PSTare set to be 12 (T Ktre) and PST;max respectively. Whenh(n+ 1)h(n) is less than or equal to , the iteration isterminated. The flowchart of the H-OSP algorithm is illustratedin Fig. 4.

To describe the complexity of the H-OSP algorithm accu-rately, we compute the computational time of the H-OSP algo-rithm and compare it with exhaustive search in the simulationpart. Similar to the analysis of S-OSP algorithm, the H-OSPalgorithm is also an iterative algorithm. The H-OSP algorithmiterates until the SE converges to a maximum value, and thismaximum value is the global maximum value.



10

V. SIMULATION RESULTS

A. Simulation Setup

In this section, computer simulations are conducted to eval-uate the performance of the proposed JSS-PA scheme underSIF and HIF strategies. In the simulations, the PT is assumedto be located at coordinate (0; 0). The SUs are uniformlydistributed in a circular field with a radius of r = 800m, theFC is located in the centre. The number of SUs is K = 20.If the PT is detected to be absent, one of the SUs is allowedto conduct data transmission, and there is no collision amongthe SUs. The frame duration is T = 40ms, the individualreporting time is much smaller than T and is set as tre = 10s[25]. fs = 10kHz, = 0:85. To sufficiently protect thePR, pthout = 0:1 unless otherwise stated. The transmit powerof the PT is PPT = 30kW [29]. Because of the hardwarelimitation or other regulations, the maximum transmit powerof ST is assumed to be 6W unless otherwise stated. Supposethat the PR is interfered and will be in outage if the ratioof SU signal over PU signal is larger than 28dB [30]. ThePRs are protected by the minimum SNR th = 5:5dB. Thepath loss exponent is = 3:8. For the sake of simplicity, wesuppose that the noise power is 88dBmW , gP = gS = 1.

B. Results and Discussion

Fig. 5 is simulated to show the SE of the CR network versusdPS for various strategies. When the value of dPS is small,the ST is located within the protected boundary (dPS < L),the SE is 0 when the OPA scheme is employed because anytransmission of ST even with a small value of transmit powercan make the PR in outage. The SE of the OSS scheme is thesame as that of the JSS-PA scheme when dPS < L. When thelocation of ST is outside the protected area and L < dPS o.

Fig. 13 shows the probability versus dPS for differentvalues of PST;max under HIF strategy. The JSS-PA schemeis conducted in this simulation, and the optimal tse and theoptimal M are employed. When dPS < L, the probability is 1. This is because in the protected area, any transmissionof ST even with a small value of transmit power can makethe PR in outage. When dPS L+, the probability is lessthan or equal to 0:1. Due to the effect of path loss, willdecrease as the distance between PT and ST becomes longer.When L < dPS < L+, according to Fig. 12, the optimal STtransmit power is either PST;max or P ST . When P

ST is chosen

as the transmit power, the probability will be equal to pthout,which validates the analyses in section II. When the optimaltransmit power is PST;max, the probability is larger than0:1. In this case, the SUs need to conduct spectrum sensing



13

1.2 1.4 1.6 1.8 2 2.2 2.4 2.6

x 1040

510

2

2.5

3

3.5

4

4.5

5

dPS (m)

tse

(ms)

SE (b

its/s/

Hz)

poutth

=0.1

poutth

=0.05

poutth

=0.01

Fig. 14. The SE vs both the duration of local sensing and dPS for variousvalues of pthout.

to satisfy the PU outage constraint.Under HIF strategy, Fig. 14 illustrates the SE of the CR

network h versus both the duration of local sensing tse andthe distance dPS for various values of pthout. A 3D graphis illustrated. The larger the value of pthout, the higher thespectrum efficiency. The value of pthout indicates the level ofprotection to the PR, smaller pthout means better protection toPR. Thus, it can be concluded that relaxing the constraint onthe protection to the PR will result in a higher SE of the CRnetwork.

VI. CONCLUSION

With the assistance of the location information of PT andthe CR network, a joint spectrum sensing and power allocation(JSS-PA) scheme is proposed to improve the spectrum efficien-cy. Under SIF and HIF strategies, the sensing parameters andthe ST transmit power are jointly optimized to maximize theSE of the CR network. Then, efficient algorithms are proposedto obtain the optimal values. It has been shown that the JSS-PA scheme outperforms both the OSS scheme and the OPAscheme, the SE of the SIF strategy is higher than that of theHIF strategy. In order to maximize the SE of the CR network,the duration of local sensing, the ST transmit power and thefinal decision threshold in FC are important parameters whichshould be optimized. In addition, relaxing the constraint onthe protection to the PR will result in a higher SE of the CRnetwork.Based on the system model in this paper, the protected area

can not be estimated accurately by using the PT location andits transmit power. Hence, the PRs may be interfered by thesecondary transmission. Therefore, more accurate estimationof the protected area will be investigated in our future work. Inaddition, we will investigate the energy efficiency (EE) of theCR network due to user device requirements and environmentconcerns. Which scheme performs better on improving theEE? We will study this problem as well as consider the trade-off between the SE and the EE.

APPENDIX A: PROOF OF THEOREM 2

From theorem 1, it is shown that s is maximized whenQMD;s = p

thout. Hence QMD;s will be replaced with p

thout in

the following analysis. We take the first partial derivative ofs with respect to tse and obtain

@s@tse

= 1 dQFA;sdtse

T tse KtreT

(1QFA;s)1 1T (1 )pthout2

1

T:

(57)

According to (33), dQFA;sdtse can be derived as

dQFA;sdtse

= 14

KXj=1

jj s

fs

tsePK

j=1 2j

e 12"sPK

j=12j(2j+1)PK

j=12j

Q1(1pthout)+PK

j=1 jj

rtsefs

2PKj=1

2j

#2

< 0:(58)

Substituting (58) into (57), we have

limtse!0

@s@tse

= +1; (59)

limtse!TKtre

@s@tse

= (1QFA;s)1 1T

(1 )pthout2 1

T< 0:

(60)

The above two equations indicate that when tse ! 0, s isincreasing in tse, and when tse ! T Ktre, s is decreasingin tse. Thus, when 0 < tse < T Ktre, there must exist amaximum value for s. In the following, it will be provedthat the maximum point of s is unique. We take the secondpartial derivative of s with respect to tse and obtain

@2s@t2se

= 1d2QFA;sdt2se

T tse KtreT

+21dQFA;sdtse

1T:

(61)According to (58), d

2QFA;sdt2se

is derived as

d2QFA;sdt2se

=1

8tse

KXj=1

jj s

fs

PK

j=1 2j

"

1ptse

+

PKj=1 jjPKj=1

2j

vuutfs2

KXj=1

2j (2j + 1)

Q1(1 pthout) +ptsefs(

PKj=1 jj)

2

2PK

j=1 2j

#

e 12"sPK

j=12j(2j+1)PK

j=12j

Q1(1pthout)+PK

j=1 jj

rtsefs

2PKj=1

2j

#2:

(62)

Thus, d2QFA;sdt2se

> 0. Then, @2s@t2se

< 0, and s is a concavefunction of tse. Therefore, the maximum point of s is unique,for cooperative spectrum sensing with SIF, there exists oneoptimal duration of local sensing tse that can maximize theSE s.



14

APPENDIX B: PROOF OF THEOREM 3For a given duration of local sensing tse, we take the first

partial derivative of h with respect to QMD;h and obtain

@h@QMD;h

=

dQFA;hdQMD;h

1

+ (1 ) 2 T tse Ktre

T;

(63)

wheredQFA;hdQMD;h

= dQFA;h=dpde;j

dQDE;h=dpde;j

=

dQFA;h=dpfa;j

dpfa;j=dpde;jdQDE;h=dpde;j

= X

M

YM dpfa;jdpde;j

= XM

YMdpfa;j=dj

dpde;j=dj

:(64)

Then, it is derived that

dpfa;j=dj = 122

rtsefs

e 12!23 ; (65)

dpde;j=dj = 122

s

tsefs(2j + 1)

e 12!24 ; (66)

where !3 =j2 1

qtsefs2 and !4 =

j2 j 1

qtsefs

2(2j+1). Substituting (65) and (66)

into (64), we havedQFA;hdQMD;h

= XM

YMp2j + 1 e 12 (!23!24): (67)

It will be proved that XM > 0 and YM > 0 in Theorem 4.Thus, we can conclude that @h@QMD;h > 0. Therefore, h isan increasing function of QMD;h, and h is maximized whenQMD;h = p

thout.

APPENDIX C: PROOF OF THEOREM 5According to (1) and (2), pfa;j = Q(!3), pde;j = Q(!4),

where !3 and !4 have been defined in the Appendix B. Let

= !3 !4 =j2 1r

tsefs2

j2 j 1

stsefs

2(2j + 1):

(68)

We take the first partial derivative of with respect to jand obtain

@

@j=p

2j + 1 1 1

2

stsefs

2(2j + 1)> 0: (69)

Since

limj!0

=q

2j + 2j + 1p2j + 1

s tsefs2(2j + 1)

> 0;

(70)it can be concluded that > 0, hence !3 > !4. Based on thefact that Q(x) is a decreasing function of x, we have pfa;j 1 pde;j . Therefore, ln pde;jpfa;j > 0,ln

1pde;j1pfa;j < 0.

REFERENCES

[1] S. Chen, J. Zhao, The requirements, challenges, and technologies for 5Gof terrestrial mobile telecommunication, IEEE Commun. Mag., vol. 52,no. 5, pp. 36-43, May 2014.

[2] J. Mitola, III, J. Guerci, J. Reed, Y.-D. Yao, Y. Chen, T. Charles Clancy,J. Dwyer, H. Li, H. Man, R. McGwier, Y. Guo, Accelerating 5G QoEvia public-private spectrum sharing, IEEE Commun. Mag., vol. 52, no.5, pp. 77-85, May 2014.

[3] P. Demestichas, A. Georgakopoulos, D. Karvounas, K. Tsagkaris, V.Stavroulaki, J. Lu, and J. Yao, 5G on the horizon: key challenges forthe radio-access network, IEEE Veh. Technol. Mag., vol. 8, no. 3, pp.47-53, Sep. 2013.

[4] Q. Li, H. Niu, A. T. Papathanassiou, and G. Wu, 5G network capacity:key elements and technologies, IEEE Veh. Technol. Mag., vol. 9, no. 1,pp. 71-78, Mar. 2014.

[5] C.-X. Wang, F. Haider, X. Gao, X.-H. You, Y. Yang, D. Yuan, H.M. Aggoune, H. Haas, S. Fletcher, E. Hepsaydir, Cellular architectureand key technologies for 5G wireless communication networks, IEEECommun. Mag., vol. 52, pp. 122-130, Feb. 2014.

[6] R. Q. Hu, Y. Qian, An energy efficient and spectrum efficient wirelessheterogeneous network framework for 5G systems, IEEE Commun.Mag., vol. 52, no. 5, pp. 94-101, May 2014.

[7] Y.-C. Liang, K.-C. Chen, G. Y. Li, and P. Mahonen, Cognitive radio net-working and communications: an overview, IEEE Trans. Veh. Technol.,vol. 60, no. 7, pp. 3386-3407, Sep. 2011.

[8] S. Eryigit, S. Bayhan, and T. Tugcu, Energy-efficient multichannelcooperative sensing scheduling with heterogeneous channel conditionsfor cognitive radio networks, IEEE Trans. Veh. Technol., vol. 62, no. 6,2690-2699, July 2013.

[9] L. Wang, K. Wu, J. Xiao, M. Hamdi, Harnessing frequency domain forcooperative sensing and multi-channel contention in CRAHNs, IEEETrans. Wireless Commun., vol. 13, no.1, pp. 440-449, Jan. 2014.

[10] C. R. Stevenson, G. Chouinard, Z. Lei, W. Hu, S. J. Shellhammer, W.Caldwell, IEEE 802.22: the first cognitive radio wireless regional areanetwork standard, IEEE Commun. Mag., vol. 47, no. 1, pp. 130-138,Jan. 2009.

[11] E. Peh, Y.-C. Liang, and Y. Zeng, Sensing and power control incognitive radio with location information, in Proc. IEEE InternationalConference on Communication Systems, pp. 255-259, 2012.

[12] H. Mu, J. K. Tugnait, Joint soft-decision cooperative spectrum sensingand power control in multiband cognitive radios, IEEE Trans. on SignalProcess., vol. 60, no. 10, pp. 5334-5346, Oct. 2012.

[13] Y. Wang, P. Ren, F. Gao, Z. Su, A hybrid underlay/overlay transmissionmode for cognitive radio networks with statistical quality-of-serviceprovisioning, IEEE Trans. Wireless Commun., vol. 13, no.3, pp. 1482-1497, Mar. 2014.

[14] P. Chen, Q. Zhang, J. Yu, Y. Zhang, and B. Cao, Sensing-throughputtradeoff in joint spatial-temporal sensing based cognitive radio networks,in Proc. IEEE/CIC International Conference on Communications inChina, pp. 727-732, Aug. 2013.

[15] Q. Wu, G. Ding, J. Wang, and Y.-D. Yao, Spatial-temporal opportunitydetection for spectrum-heterogeneous cognitive radio networks: two-dimensional sensing, IEEE Trans. Wireless Commun., vol. 12, no. 2,pp. 516-526, Feb. 2013.

[16] T. Xue, Y. Shi, and X. Dong, A framework for location-aware strategiesin cognitive radio systems, IEEE Wireless Commun. Lett., vol. 1, no. 1,pp. 30-33, Feb. 2012.

[17] B. Cao, Q. Zhang, J. W. Mark, L. X. Cai, H. V. Poor, Towardefficient radio spectrum utilization: user cooperation in cognitive radionetworking, IEEE Network, vol. 26, no. 4, pp. 46-52, 2012.

[18] D. Duan, L. Yang, and J. C. Principe, Cooperative diversity of spectrumsensing for cognitive radio systems, IEEE Trans. on Signal Process., vol.58, no. 6, pp. 3218-3227, June 2010.

[19] E. Peh, Y.-C. Liang, Y. L. Guan, and Y. Zeng, Cooperative spec-trum sensing in cognitive radio networks with weighted decision fusionschemes, IEEE Trans. Wireless Commun., vol. 9, no.12, pp. 3838-3847,Dec. 2010.

[20] J. Ma, G. Zhao, Y. Li, Soft combination and detection for cooperativespectrum sensing in cognitive radio networks, IEEE Trans. WirelessCommun., vol. 7, no.11, pp. 4502-4507, Nov. 2008.

[21] E. Peh, Y.-C. Liang, Y. L. Guan, and Y. Zeng, Optimization ofcooperative spectrum sensing in cognitive radio networks: a sensing-throughput tradeoff view, IEEE Trans. Veh. Technol., vol. 58, no. 9, pp.5294-5299, Nov. 2009.



15

[22] S. Atapattu, C. Tellambura, and H. Jiang, Energy detection basedcooperative spectrum sensing in cognitive radio networks, IEEE Trans.Wireless Commun., vol. 10, no.4, pp. 1232-1241, April. 2011.

[23] S.-J. Kim, E. D. Anese, and G. B. Giannakis, Cooperative spectrumsensing for cognitive radios using kriged kalman filtering, IEEE J. Sel.Topics Signal Process., vol. 5, no. 1, pp. 24-36, Feb. 2011.

[24] Y.-C. Liang, Y. Zeng, E. Peh, and A. T. Hoang, Sensing-throughputtradeoff for cognitive radio networks, IEEE Trans. Wireless Commun.,vol. 7, no. 4, pp. 1326-1337, Apr. 2008.

[25] E. Peh, Y.-C. Liang, Y. L. Guan, and Y. Pei, Energy-efficient coopera-tive spectrum sensing in cognitive radio networks, in Proc. IEEE GlobalCommunications Conference, pp. 1-5, 2011.

[26] Steven C. Chapra, Raymond P. Canale. Numerical Methods for Engi-neers, sixth edition. McGraw-Hill, 2010.

[27] R. W. Yeung. A First Course in Information Theory, 1st edition.Springer, 2002.

[28] E. L. Lehmann, and J. P. Romano, Testing Statistical Hypotheses, thirdedition. Springer, 2005.

[29] X. Dou, Discussion of the influence on the periphery electromagneticenvironment of middle-small sized radiated television tower, Environ-mental Protection and Technology, vol. 15, no. 2, pp. 22-25, 2009.

[30] FCC, Second memorandum opinion and order, Report ET Docket No.10-174, Sep. 2010.

Date post:	11-Sep-2015
Category:	Documents
Upload:	royalraecher
View:	215 times
Download:	0 times

5g

Documents