On the Performance of Cooperative Spectrum Sensing of Cognitive Radio Networks in AWGN and Rayleigh Fading Environments

Transcription

1 754 wasan et al.: On the performance of Coperatve Spectrum Sensng of Cogntvr Rado Networks On the Performance of Cooperatve Spectrum Sensng of Cogntve Rado Networks n AWGN and Raylegh Fadng Envronments Wasan Kadhm Saad a,b,*, Mahamod Ismal a, Rosdadee Nordn a and Ayman A. El-Saleh c a Department of Electroncs, Electrcal and System Engneerng, Faculty of Engneerng and Buld Envronment, Unverst Kebangsaan Malaysa, 436 Bang, Selangor, Malaysa. [e-mal: wasan, mahamod, adee@eng.ukm.my] b Department of Communcatons Engneerng, Najaf Techncal College, Foundaton of Techncal Educaton, Baghdad, Iraq. c Faculty of Engneerng, Multmeda Unversty, 63 Cyberjaya, Selangor, Malaysa. [e-mal: ayman.elsaleh@mmu.edu.my] *Correspondng author: Wasan Kadhm Saad Receved Aprl 4, 3; revsed July 7, 3; accepted August 8, 3; publshed August 3, 3 Abstract For the purpose of enhancng the spectrum effcency, cogntve rado (CR) technology has been recently proposed as a promsng dynamc spectrum allocaton paradgm. In CR, spectrum sensng s the key capablty of secondary users n a cogntve rado network that ams for reducng the probablty of harmful nterference wth prmary users. owever, the ndvdual CRs mght not be able to carry out relable detecton of the presence of a prmary rado due to the mpact of channel fadng or shadowng. Ths paper studes the cooperatve spectrum sensng scheme as means of optmzng the sensng performance n AWGN and Raylegh channels. Results generated from smulaton provde evdence of the mpact of channel condton on the complementary recever operatng characterstc (ROC). Based on the results, t was found that wth constant local SNRs at the secondary users, the probablty of mssed detecton (P m ) of cooperatve spectrum sensng n a cogntve rado network, calculated usng a closed form expresson, can be sgnfcantly mnmzed. Thus, the paper llustrates that mprovement of the detecton performance of the CR network can be acheved by establshng a centralzed cooperaton among neghborng cogntve rado users. Fnally, verfcaton of the valdty of the fuson schemes utlzed for combnng the ndvdual CR decsons s provded. Keywords: Cogntve rado, Spectrum sensng, Dynamc spectrum access, Cooperatve sensng, AWGN channel, Raylegh fadng channel.

2 KSII TRANSACTIONS ON INTERNET AND INFORMATION SYSTEMS VOL. 7, NO. 8, Aug Copyrght c 3 KSII. Introducton Due to the rapd advances n a wreless communcaton system, there has been an ncreasng demand for the new wreless servces n both the used and unused frequency spectrum. owever, ths ncreasng demand faces a great barrer whch s the lmtaton of rado resources. In attemptng to overcome ths challengng problem, cogntve rado (CR) has been cted n as one of the most promsng technologes that can offer a support for the ncreasng demand for spectrum avalablty and t s capable of ncreasng the spectral effcency []. Thus, provdng opportuntes for researchers n obtanng further understandng and explorng the opportuntes presented by dle frequences as the frst concept of opportunstc spectrum access based on the CR technology. A CR s also capable of adaptng to the dynamc rado envronment and the network parameters wth the am of maxmzng the extent to whch the lmted rado resources are utlzed, whle at the same tme makng wreless access more flexble. CR technology s recommended for researchng the unlcensed use of free bands. Thus, one of the core functons of CR s detectng the free bands through the spectrum sensng functonalty. Spectrum sensng s known as a key enablng practcalty n cogntve rado networks (CRNs) whch performs detecton of exstence of prmary user (PU) sgnals n the concerned bands, as well as dentfcaton the avalable channels that are useable. In determnng the free and employed bands, the unlcensed systems analyze the sgnals receved from the lcensed system [-3]. In cogntve rado wth spectrum sensng, when the lcensed user (LU) abruptly wshes to have an access to the frequency band already allocated to the LU, the cogntve user (CU) starts the process of searchng for the dle spectrum agan. Accordng to [4], t s suggested that enhancng rado RF front-end senstvty or dgtal sgnal processng technques ncludng energy detecton, matched flterng, and cyclostatonary feature detecton can be one way of mprovng the executon of spectrum sensng. To acheve hgh performance for cogntve rado, collaboratve spectrum sensng s requred to mprove the detecton probablty and dmnsh the detecton tme, thereby mprovng the senstvty of the cogntve recever [5]. n [6], the examned optmzng the cooperatve spectrum sensng usng energy detecton for the purpose of mnmzng the total error rate n cogntve rado networks (CRNs) was studed. Another nvestgaton of the cooperatve spectrum sensng usng an mproved energy detector n multple antenna based CRNs along wth Raylegh fadng prmary user (PU-CR) lnks and mperfect reportng channels for enhancng the relablty n detectng a spectrum hole was also carred out by [7]. owever, the study descrbed n the current paper used dfferent fuson rules (e.g. AND and OR) and closed form expressons as to provde a more realstc pcture of cooperatve energy detecton n Raylegh fadng channel. Ths paper also attempted to provde an explanaton of the performance degradaton of the CRNs n fadng and low SNR envronments. Thus, ths paper s organzed as follows: Secton offers the analyss of the deployment of CRNs ncludng the spectrum sensng concept and energy detector. In Secton 3, we derve of detecton and false alarm probabltes for local sensng. Secton 4 provded a dscusson on the performance of the energy detecton wth numercal and smulaton results. Secton 5 presents bref concludng remarks of the study.. Deployment of Cogntve Rado Network Ths secton ntroduces nformaton about the spectrum sensng concept and then, t provdes a survey of the energy detecton scheme so that analyzng how the probablty of detecton and probablty of false alarm are related can be descrbed n the next secton.

3 756 wasan et al.: On the performance of Coperatve Spectrum Sensng of Cogntvr Rado Networks. Spectrum Sensng Concept Fg. dsplays the system model of the present study, and t s notced that t s probable for some of the CR users to make detecton of the prmary sgnal whereas some other CR users cannot detect the presence of the prmary sgnal due to the mpact of deep fadng and shadowng. Fg.. Cooperaton n cogntve rado network. As shown n Fg., there s a potental realzaton of enablng CR users who are WRAN users or secondary users (SUs) to opportunstcally access to unused TV bands of prmary users (PUs). In such a case, the possblty of enhancng the sgnal detecton probablty through the cooperatve sgnal detecton s hgh. In realty, detectng the PUs whch receves the data n the communcaton range for the user s consdered as the most effectve means of detectng the spectrum holds. It becomes dffcult for any SU to drectly measure the channel between the prmary transmtter and recever, and therefore, the detecton of the prmary transmtter has become the major focus of the recent work s dependent on the local nodes of users. Thus, each CR has to make a dstncton between employed and non- employed spectrum bands [8]. In the cooperatve spectrum sensng scheme, every secondary user SU s able to make several executons of the local spectrum sensng and then, sends a bnary local decson to the base staton. Followng ths, the base staton fuses the local decsons and makes a fnal decson n order to make determne the absence or presence of the prmary user PU. In general, the sampled receved sgnals of the secondary users contan two dfferent hypotheses, and n CRNs [9]. The functon of the energy detector s measurng the exstng energy on the lcensed channel durng a notce nterval and announces a whte space f the measured energy s less than a threshold. Thus, the spectrum sensng problem can be desgned as a bnary hypothess problem llustrated as follows: : prmary user PU does exst and : prmary user PU does not exst For smplcty of the mplementaton, the work s lmted to the energy detecton n the spectrum sensng. The local spectrum sensng provdes optons to be selected between the followng two hypotheses []: n ( t), y ( t) () hp x( t) n ( t), Where y (t) refers to the sgnal whch s receved by the secondary user (SU), x(t) refers to the sgnal transmtted to the prmary user s (PU s), n (t) s the addtve whte Gaussan nose (AWGN) receved by -th SU, and h p s the channel gan. Thus, after the energy detecton s

4 KSII TRANSACTIONS ON INTERNET AND INFORMATION SYSTEMS VOL. 7, NO. 8, Aug Copyrght c 3 KSII used for cooperatve spectrum sensng, the results of the sensng for the secondary users (SUs) are transferred to a fuson center by usng decson fuson. Therefore, each secondary user has to make a decson on the prmary user actvty, and all these decsons produced by the secondary users are reported to the fuson center by usng the reportng channel. The fuson rules generalzed as the k-out-of-n rule (where k s the number of the users utlzed for cooperaton and n s the total number of users n the network). Where f there s k or more cogntve relays that separately decde the presence of the prmary actvty, therefore, the fuson centre decde the presence of prmary user (PU). Where, f k = (e.g., the central unt decdes a PU s utlzng the channel f more than one SU's result are ), k = n (e.g., the central unt decdes that the observed channel s occuped by a PU f all sensng results for the SU's should be ), and k = n/ (e.g., the fuson central decdes a PU s utlzng the channel f half of or more secondary users results s ), the k-out-of-n rule represents OR rule, AND rule, and Majorty rule, respectvely [-].. Energy Detector Energy detecton s represent the most well-known spectrum sensng schemes, ams at determnng whether or s true; ths s acheved by sensng the energy of sgnal y. Fg. dsplays the block dagram of the typcal energy detector. Y(n) BPF ( ) ntegrator Test statstc (a) Y(n) A/D M pont FFT Y(t) ( ) Sum of N samples Test statstc (b) Fg.. The block of tradtonal energy detector. (a) In tme doman and (b) In frequency doman. Fg. (a) llustrates the tradtonal energy detector n tme doman, and as reported by [3], the applcaton of a band-pass flter to the objectve sgnal s made frst and ths s followed by squarng the receved sgnal and ntegratng t n the ntegrator so that the test statstc can be obtaned. Fg. (b) shows how the A/D converter replaces the band pass flter whle mantanng the M ponts of FFT and the others as they are at the same tme. After that, the last result s compared wth the threshold and gave the decson. So, the output of energy detector s: j N Z y j ( t) () N Thus, based on the central lmt theorem, when N s large enough (e.g. N > ), the value of Z approxmates Gaussan dstrbuton. The mean and varance of Z s gven as [3]:, E( Z) P, (3)

5 758 wasan et al.: On the performance of Coperatve Spectrum Sensng of Cogntvr Rado Networks V a r 4, N ( Z) 4 4 P, N N (4) N Where E(.) and Var(.) denote mean and varance, and P h x( t) s the sgnal j N energy detected by the cogntve sensng node. Although Gaussan dstrbuton provdes a good approxmaton n general, t mght not be practcally the choce as assumng large number of samples, N, mght not be correct when short sensng tme s targeted. Ths s because n tme-varyng fadng envronments, the sensng tme cannot be too long (and thus the number of samples cannot be too large) as ths makes the process of montorng PU actvtes neffcent. Therefore, n the subsequent sectons, ch-square dstrbuton s used and s not approxmated by a Gaussan dstrbuton to ensure a more practcal scenaro that can beneft the readers. 3. Dervaton of Detecton and False Alarm Probablty for Local Sensng At each SU, the decson statstc of energy detecton, Z, has the followng dstrbuton: Z x x m m (5) Where Z means the collected energy by a cogntve user, m refers to the tme- bandwdth product of the energy detector. For smplcty, t s assumed to be an nteger, and x m represents a central ch-square dstrbuton wth m degrees of freedom whle x m ( ) represents a non-central ch-square dstrbuton wth m degrees of freedom and a non-centralty parameter for and s the nstantaneous SNR receved at the -th SU [4-5]. In ths paper, we use the ch-square dstrbuton n the subsequent dscusson. The probablty densty functon (pdf) of Z can be formulated as the followng: f Z ( y) m m y y m m e e y y I m y (6) Where () s the gamma functon and I () s the uth-order modfed Bessel functon of the u frst knd. There are two probabltes regardng n spectrum sensng: under hypothess, probablty of detecton, accordng to whch, the probablty of the algorthm correctly

6 KSII TRANSACTIONS ON INTERNET AND INFORMATION SYSTEMS VOL. 7, NO. 8, Aug Copyrght c 3 KSII detectng the exts of prmary sgnal, and under hypothess, the probablty of false alarm, whch defnes the probablty of the algorthm as a process of ncorrectly pronouncng the presence of the prmary sgnal. From the prmary user s PU s perspectve, the hgher probablty of detecton means the best protecton receved by t. From the secondary user s SU s perspectve, the lower probablty of false alarm ndcates that the secondary users have more opportuntes to use t when the frequency bands are avalable. It s obvous that n obtanng a better detecton algorthm, t s mportant that the probablty of detecton to be as hgh as possble when the probablty of false alarm s low [5, 6]. The followng equatons llustrate probabltes of detecton, mss detecton, and false alarm for SU over non-fadng channels n the case of the CR users wth the energy detector when the detectng channels are presumed to be the AWGN channels [7]: P P Z > ) Q (, ) f ( x) d( ) (7) d, ( m x x P m, P( Z ) Pd, P (8) ( m, / ) P( Z > ) f ( x) d( ) (9) ( m) f, x where refers to the detecton threshold for the -th SU whch s, for smplcty, assumed to be the same for all users. The probablty of detecton can be obtaned from Eq. (6) to evaluate Eq. (7). Ths evaluaton of the cumulatve dstrbuton functon (cdf) of Z for even degrees of freedom, whch n our case s m, can be llustrated as follows: F Z ( y) Q (, y) () m Therefore, P Q (, ) () d, m P m, Pd, () Usng Eq. (6) to evaluate Eq. (9), the probablty of false alarm over an AWGN channel s gven as []: ( m, / ) Pf, (3) ( m) We note that Eq. (9) s derved due to the fact that ( m, / ) / ( m) s freelancer of, () and (, ) are complete and upper ncomplete gamma functon, the nstantaneous sgnal-to-nose-rato of the detectng channel (SNR), f (x) s the pdf of under certan fadng model, and ( a, b) refers to the generalzed Marcum Q-functon defned as follows: Q m

7 76 wasan et al.: On the performance of Coperatve Spectrum Sensng of Cogntvr Rado Networks Q m x a a, b) x exp I m ( ax dx m a (4) b m ( ) Where I m refers to the modfed Bessel functon of the frst type and order ( m ). When the complex receved sgnal conssts of a large number of plane waves, for some types of scatterng envronments, the receved sgnal has a Raylegh dstrbuton. would have an exponental dstrbuton under Raylegh fadng as follows: l f ( ) e l (5) Besdes, the nstantaneous SNR of a wreless channel between transmtter and recever s also dentfed as: SNR h Es / N (6) Where, Es s the transmt energy, N s the varance of addtve whte Gaussan nose (AWGN) and h s the channel gan. Thus, the average SNR s referred as: SNR E (7) avg ( h Es / N) E( h ) Es / N The detectng channels n wreless propagaton envronments are affected by fadng, thus, the sensng executon of a sngle SU should be denoted by the average Pd of a sngle SU over Raylegh fadng detectng channel and a closed-form formula for the probablty of detecton over Raylegh fadng channel can be found by substtutng Eq. (5) n Eq. (7), P d, e k! l e m k m m ( l ) k e k k! ( l k (8) Snce P f, s assumed for the case so that, P, of Eq. (3) remans the same [8]. f and hence free-lancer of the SNR of detectng channel,

8 KSII TRANSACTIONS ON INTERNET AND INFORMATION SYSTEMS VOL. 7, NO. 8, Aug Copyrght c 3 KSII 4. ard Fuson Schemes for Cooperatve Spectrum Sensng Lettng M denote the number of users who are cooperatng, for smplfyng ths, all M users are presumed to experence ndependent and dentcally dstrbuted (..d) fadng/shadowng wth same average SNR. At the base staton, all -bt decsons are combned together n accordance wth the followng logc rule. M k, E Z (9) < k, The CR base staton recevng the decsons from M other users are performed and decdng when there s at least (k-out-of-m) cogntve rados (CRs) whch carres out, otherwse, the base staton decdes [9]. For dstncton, conflatng the local decsons and formng global decson n cooperatve spectrum sensng for CRs can be conducted by some classcal algorthms such as OR rule and AND rule, whch are utlzed n the common recever as to mnmze the harmful nterference to prmary user. Afterward, the probablty of detecton and the probablty of false alarm of the fnal decson are presented by [5-6], respectvely. 4. OR Fuson Rule In OR fuson rule, the assumpton that the fnal decson s true f the absence of the prmary user PU s ndcated by all the secondary users SUs whereas s true f the presence of a prmary user s pronounced by at least out of M secondary users. Ths means that evaluaton of the OR fuson rule can be carred out by settng (k = ) n expresson Eq. (9), where as there are M users whch employed n cooperatve spectrum sensng among k users, Where M k. In assumng that all decsons are freelancer, so, for the fnal decson, the cooperatve probablty of detecton Q d (M ),the cooperatve probablty of mssed detecton (M ) and the cooperatve probablty of false alarm Q f (M ) can be evaluated, respectvely, as: Q m 4. AND Fuson Rule M Q ( M ) ( ) () Q d P d, M ( M ) () m P m, M Q ( M ) ( ) () f P f, In AND fuson rule, the fnal decson s true only f the presence of the prmary user PU s ndcated by all the M secondary users SUs. Otherwse, the absence of the prmary user s

9 76 wasan et al.: On the performance of Coperatve Spectrum Sensng of Cogntvr Rado Networks assumed. In other words, the AND rule s correspondent to the case of (k = M) n expresson Eq. (9). If assumng that all decsons are freelancer, then, for the fnal decson, the cooperatve probablty of detecton (M ), the cooperatve probablty of mssed detecton Q d Q m (M ) and the cooperatve probablty of false alarm (M ) can be evaluated as, respectvely: M Q f Q ( M ) ( ) (3) d P d, M Q ( M ) ( ) (4) m P d, M Q ( M ) ( ) (5) f P f, Where M s the number of SUs subscrbng cooperaton, and detecton and false alarm of the P, and d P, the probabltes of f SU derved from Eq. () and Eq. (7). 5. Smulaton Results and Dscusson Assessment of the performance of cooperatve spectrum sensng s usually performed through ts complementary recever operatng characterstc (ROC) curve P vsp ), spectrum usage ( f d and SNR requrements for dfferent stuatons of nterest. In the computer smulaton by usng the Matlab program, the number of secondary users s set as (M = 6) and the samplng frequency of the receved sgnal s assumed to be MZ and sensng tme s set at ms. Fg. 3 provdes explanaton of the complementary ROC curves for local spectrum sensng for varous SNR values and the tme-bandwdth product of the energy detector (m = ) under AWGN and Raylegh fadng channels. It was notced that as the SNR decrease, there s a gradual decrease of the probablty of detecton for a fxed probablty of false alarm under both AWGN channel and Raylegh fadng channel. Probablty of mssed detecton, Q m AWGN, SNR = -5 db AWGN, SNR = db AWGN, SNR = 5 db Raylegh, SNR = -5 db Raylegh, SNR = db Raylegh, SNR = 5 db Probablty of false alarm, Q f Fg. 3. Complementary ROC curves for local spectrum sensng under AWGN and Raylegh channel wth dfferent values of SNR and (m=).

10 KSII TRANSACTIONS ON INTERNET AND INFORMATION SYSTEMS VOL. 7, NO. 8, Aug Copyrght c 3 KSII Fg. 4, shows the executons of computer smulatons to the complementary recever operatng characterstc (ROC) curves for Raylegh fadng channel over the average SNR, when P f vares from. to, where the sensng performance of one secondary user for dfferent values of SNR, and the secondary user s SNR s are assumed to be SNR = [-8, -, -, -4] db. Probablty of mssed detecton, Qm - - SNR= -8 db SNR= - db SNR= - db SNR= -4 db - - Probablty of false alarm, Qf Fg. 4. Cooperatve spectrum sensng performance wth dfferent SNR under the Raylegh fadng channel. As dsplayed by Fg. 4, t can be notced that when there s an ncrease n the local SNR, there s a steady decrease n the probablty of mssed detecton. Moreover, t can be seen that the overall mprovement of the performance s sgnfcant whereas roughly four tmemprovement s detected from (-4) db to (-8) db. There wll be also a degradaton of the spectrum sensng executon when the SNR decreases. It can be observed that as there s a decrease n the SNR, the probablty of mss detecton becomes larger, and ths wll be the case when the SU tolerates heavy shadowng or fadng, whch wll cause low SNR. Fg. 5, shows how the cooperatve spectrum sensng s executed to dfferent numbers of SU wth SNR = -5 db. It ndcates that there s a rapd degradaton of the detecton performance when there s an ncrease n the number of SU. Therefore, the ncrease n the number of SUs can enhance the detecton capablty dramatcally. Moreover, a great decrease n the probablty of mssed detecton can be seen as the secondary users (SUs) are cooperatng for a gven probablty of false alarm. Thus, n case when the number of cooperatve users M ncreases, the performance of the cooperatve sensng performance can be enhanced under Raylegh fadng channel. The am of usng the complementary ROC curves of one SU sensng over Raylegh fadng detectng channel s to evaluate the fadng mpact of detectng channel on the executon of local spectrum sensng at a one SU, and ths s provded by the curves wth M equal to (,, 3, and 6). It ndcates that fadng on the detectng channel results nto severely degradng the executon of the local spectrum sensng.

11 764 wasan et al.: On the performance of Coperatve Spectrum Sensng of Cogntvr Rado Networks Probablty of mssed detecton, Qm M= M= M=3 M=6 - - Probablty of false alarm, Qf Fg. 5. Cooperatve spectrum sensng performance for dfferent number of secondary users under the Raylegh fadng channel. Accordng to Fg. 5, and wth ncreasng the number of M, the cooperatve sensng performance can be mproved under the Raylegh fadng channel, also, the probablty of mss detecton ( Qm ) s greatly reduced f the secondary users are cooperated. Based on Fg. 6, t s obvous wth the ncrease n the probablty of false alarm, the curve of detecton probablty wll ncrease, and detecton probablty s hghly mproved..9.8 Probablty of detecton, Pd Pf =.. Pf =. Pf =.. Pf =. Pf = SNR/dB Fg. 6. Cooperatve spectrum sensng performance for dfferent values for probablty of false alarm. Fg. 7, shows the complementary ROC curves of cooperatve spectrum sensng for varous decson fuson rules when the conventonal energy detector (ED) usng OR and AND decson fuson rules are appled for the case M = 6 and SNR =5 db. It s observed that, for the statonaryq, the probablty of mssng OR decson rule s the smallest n comparson to f AND decson rule for the case M = 6, whereas t s the same for M =.

12 KSII TRANSACTIONS ON INTERNET AND INFORMATION SYSTEMS VOL. 7, NO. 8, Aug Copyrght c 3 KSII Fg. 7. Performance comparson of cooperatve spectrum sensng between OR and AND rules over an AWGN channel when (M=&6). It also shows that OR fuson rule can lmt the nterference to the prmary user. owever, ths paper concentrates on descrbng the recever executon through ts ROC curves Q vsq ) or ( m f complementary ROC curves ( Q vs Q ) for dssmlar stuaton of nterest. ( d f Fg. 8 shows the complementary ROC over Raylegh channel for dfferent average SNR values and the number of cooperatng users M equal to (,, 6, and 8). From Q Q curve, low slopes for Q <.can be deduced. Wth each step ncrease n SNR values startng from f Probablty of mssed detecton, Qm ED wth OR, M= ED wth AND, M= ED wth OR, M= 6 ED wth AND, M= Probablty of false alarm, Qf to db, t can be notced there s mprovement n the probablty of mssed detecton ( Q ) m. m f Probablty of mssed detecton, Qm SNR = -db, M = SNR = -4dB, M = SNR = -8dB, M = 6 SNR = -db, M = Probablty of false alarm, Qf Fg. 8. Complementary ROC curves for the Raylegh channel at dfferent SNR and M values. As dsplayed n Fg. 9 and Fg., t s clear that the detecton performance of cooperatve

13 766 wasan et al.: On the performance of Coperatve Spectrum Sensng of Cogntvr Rado Networks spectrum sensng utlzng OR-rule and AND-rule at SNR value are equal to 5 db and the tme-bandwdth product of the energy detector (m = ) under both AWGN and Raylegh fadng channels. Probablty of mssed detecton, Q m Fg. 9. Complementary ROC curves for cooperatve spectrum sensng usng OR-rule under AWGN and Raylegh channel. Moreover, a great mprovement n the detecton performance of cooperatve spectrum sensng for Raylegh fadng channel compared wth AWGN channel can be observed. For the cooperatve spectrum sensng (e.g., M = 7 and ), t does not show any knd of mprovement as compared to the detecton performance of local sensng (e.g., M=). AWGN, N = AWGN, N = 7 AWGN, N = Raylegh, N = Raylegh, N = 7 Raylegh, N = Probablty of false alarm, Q f Probablty of mssed detecton, Q m AWGN, N = AWGN, N = 7 AWGN, N = Raylegh, N = Raylegh, N = 7 Raylegh, N = Probablty of false alarm, Q f Fg.. Complementary ROC curves for cooperatve spectrum sensng usng AND-rule under AWGN and Raylegh channel. The popular k-out-of-n fuson rule s taken nto consderaton n the cooperatve spectrum sensng for the decson fuson strategy, and n partcular, the focus s on the OR-rule and AND-rule. Whle OR-rule always outperforms AND-rule, and n detectng, t s also more capable than AND-rule wth error-free reportng channels.

14 KSII TRANSACTIONS ON INTERNET AND INFORMATION SYSTEMS VOL. 7, NO. 8, Aug Copyrght c 3 KSII 6. Concluson Ths paper provded an evaluaton of the detecton performance for local spectrum sensng and cooperatve spectrum sensng usng OR-rule and AND-rule under AWGN and Raylegh fadng channel. Based on ths evaluaton, t was found that the cooperatve spectrum sensng can hardly enhance the detecton performance n low SNR envronment. In CR perspectves, t s probable that the mssed detecton of PUs (lcensees) by CR users cause severely nterference whle releasng false alarms by the CR users. Ths wll defntely mpact the spectrum accessblty, thus, reducng the throughput of the CR network. In ths paper, the energy detector executon of spectrum sensng was evaluated under AWGN and Raylegh fadng envronments. It has been found that the cooperatve sgnal detecton ensures the mprovement of the detecton executon by usng dfferent data fuson rules. Moreover, the channel behavour can be more closely modelled by usng a complex dstrbuton whch provdes a descrpton of shadowng and multpath fadng. In these scenaros, ROC curve for the Raylegh case provdes a comprehensve pcture of the detecton performance of the cooperatve spectrum sensng system. Ths paper also assessed the detecton performance of a local-sensng CR user n low SNR envronments. Ths assessment showed that the cooperaton among CR users can result nto sgnfcant mprovement on the detecton performance and compensatng the degradaton of the spectrum sensng executon caused by the possbly weak PU sgnals. Fnally, the paper provded a verfcaton of the valdty of the OR- and ANDfuson schemes whch were used for combnng the ndvdual decsons of CR users, where the deleterous mpact for the fadng effectvely can be cancels by usng these fuson decsons of varous secondary users. References [] Lan F. Akyldz, Brandon F. Lo and Ravkumar Balakrshnan, Cooperatve spectrum sensng n cogntve rado networks: A survey, Physcal Communcaton Journal, vol. 4, pp. 4-6,. Artcle (CrossRef Lnk) [] Yunfe Chen and Beauleu N.C., Performance of Collaboratve Spectrum Sensng for Cogntve Rado n the Presence of Gaussan Channel Estmaton Errors, IEEE transactons on communcatons, vol. 57, no. 7, pp , 9. Artcle (CrossRef Lnk) [3] Zhang Yu, Yang Wen-dong and Ca Yue-mng, Cooperatve Spectrum Sensng Technque, n Proc. IEEE Internatonal Conference on Wreless Communcatons, Networkng and Moble Computng (WCom), pp. 67-7, 7. [4] Cabrc D., Mshra S.M. and Brodersen R.W., Implementaton Issues n Spectrum Sensng for Cogntve Rados, n Proc. IEEE Conference Record of the Thrty-Eghth Aslomar Conference on Sgnals, Systems and Computers, vol., pp , 4. [5] Yng-Chang Lang, Yonghong Zeng, Peh E.C.Y and Anh Tuan oang, Sensng-Throughput Tradeoff for Cogntve Rado Networks, IEEE Transactons on wreless communcaton, vol. 7, no. 4, pp , 8. [6] We Zhang, Mallk R.K. and Letaef K., Optmzaton of Cooperatve Spectrum Sensng wth Energy Detecton n Cogntve Rado Networks, IEEE Transacton on Wreless Communcatons, vol.8,no.,pp ,9. Artcle (CrossRef Lnk) [7] Sngh A., Bhatnagar R.M. and Mallk K.R., Cooperatve Spectrum Sensng n Multple Antenna Based Cogntve Rado Network Usng an Improved Energy Detector, IEEE Communcatons

15 768 wasan et al.: On the performance of Coperatve Spectrum Sensng of Cogntvr Rado Networks letters,vol.6,no.,pp.64-67,. Artcle (CrossRef Lnk) [8] Dlp S. Aldar, Centralzed Integrated Spectrum Sensng for Cogntve Rados, Internatonal Journal of Computer Scence & Communcaton (IJCSC), vol., no., pp ,. [9] Peh E.C.Y., Yng-Chang L., Yong-Lang Guan, Yonghong Zeng, Cooperatve Spectrum Sensng n Cogntve Rado Networks wth Weghted Decson Fuson Scheme, n Proc. of IEEE 7 st Vehcular Technology Conference (VTC -Sprng ), pp. -5,. [] Dgham F.F., Aloun M.S. and Smon Marvn K., On the Energy Detecton of Unknown Sgnals over Fadng Channels, IEEE Transactons on Communcatons, vol. 55, no., pp. -4, 7. Artcle (CrossRef Lnk) [] Atapattu S., Tellambura C. and a Jang, Energy Detecton Based Cooperatve Spectrum Sensng n Cogntve Rado Networks, IEEE Transactons on Wreless Communcatons, vol., no. 4, pp.3-4,. Artcle (CrossRef Lnk) [] Ln Xao, Ka Lu and Ln Ma, A Weghted Cooperatve Spectrum Sensng n Cogntve Rado Networks, n Proc. of IEEE Internatonal Conference on Informaton Networkng and Automaton (ICINA), vol., pp. V-45 - V-48,. [3] Zh Quan, Shuguang Cu and Sayed A.., Optmal Lnear Cooperatve for Spectrum Sensng n Cogntve Rado Networks, IEEE journal of selected topcs n sgnal processng, vol., no., pp. 8-4, 8. [4] Le Zhang and Shuquan Xa, A New Cooperatve Spectrum Sensng Algorthm for Cogntve Rado Networks, n Proc. of IEEE- ISECS Internatonal Colloquum on Computng, Communcaton, Control, and Management (CCCM), vol., pp. 7-, 9. [5] Wang Yue, Chunyan Feng, Zhmn Zeng, and Cal Guo, A Robust and Energy Effcent Cooperatve Spectrum Sensng Scheme n Cogntve Rado Networks, n Proc. of IEEE- th Internatonal Conference on Advanced Communcaton Technology (ICACT), vol., pp , 9. [6] Pramod K. Varshney and C.S. Burrus, Dstrbuted Detecton and Data Fuson, Sprnger Verlag, ISBN ,997. Artcle (CrossRef Lnk) [7] a-peng Yao, Zheng Zhou, Xuan Sun and Bn L, Locaton based spectrum sensng performance analyss over fadng channels n cogntve rado networks, Scence Drect Journal, vol. 7, no. 5, pp. 6-3,. [8] Yaqn Zhao, Shuyng L, Nan Zhao, and Zhlu Wu, A Novel Energy Detecton Algorthm for Spectrum Sensng n Cogntve Rado, Informaton Technology Journal, vol. 9, no. 8, pp ,. Artcle (CrossRef Lnk) [9] Jaq Duan and Yong L, Performance Analyss of Cooperatve Spectrum Sensng n Dfferent Fadng Channels, n Proc. of IEEE nd Internatonal Conference on Computer Engneerng and Technology (ICCET), vol. 3, pp. V3-64 V3-68,.

16 KSII TRANSACTIONS ON INTERNET AND INFORMATION SYSTEMS VOL. 7, NO. 8, Aug Copyrght c 3 KSII Wasan Kadhm Saad receved her B.Sc. n Electrcal and Electronc Engneerng/ Electronc& Communcaton from Unversty of Technology, Baghdad, Iraq, and her M.Sc. n Satellte Communcaton Engneerng from Unversty of Technology, Baghdad, Iraq. She s currently PhD student n Department of Electronc, Electrcal and System Engneerng, Faculty of Engneerng and Bult Envronment, Unversty Kebangsaan Malaysa (UKM). er man research nterest s Cogntve rado networks, MIMO system, and Cogntve rado-mimo system. Mahamod I. s a full professor n the faculty of engneerng at Unverst Kebangsaan Malaysa (Malaysa). e receved hs B.Sc. (ons) from Unversty of Strathclyde, UK, M.Sc. from UMIST, Manchester, UK, and s PhD from Unversty of Bradford, UK. s man research nterest s n moble, personal and satellte communcaton, and wreless networkng. Rosdadee Nordn receved hs B. Eng. from Unverst Kebangsaan Malaysa n and Ph.D. from Unversty of Brstol, Unted Kngdom n. e s currently a lecturer n Department of Electrcal, Electroncs and System Engneerng n Unverst Kebangsaan Malaysa. s research nterests nclude Multple-Input Multple-Output (MIMO), Orthogonal Frequency-Dvson Multple Access (OFDMA), resource allocaton, green rado, ntercell nterference, cooperatve dversty and ndoor wreless localzaton. Ayman A. El-Saleh receved hs B.Sc. degree n Communcatons Engneerng from Omar El-Mukhtar Unversty (OMU), Lbya, n 999, hs M.Sc. n Mcroelectroncs Engneerng and Ph.D. n Wreless Communcatons both from Unverst Kebangsaan Malaysa (UKM), n 6 and, respectvely. e joned the Faculty of Engneerng, Multmeda Unversty (MMU) n October 6 at whch he s currently a Senor Lecturer. e s also a member of IEEE, IET, ICICE and IACSIT. s research nterests nclude cogntve rado networks, cooperatve spectrum sensng, resource allocaton, FPGA-based dgtal system desgn, and applcatons of artfcal ntellgence and evolutonary algorthms n wreless communcatons.