Study and Optimization of Cooperative Spectrum Sensing in OFDM Cognitive Radio Networks

Transcription

1 IJCSI Internatonal Journal of Computer Sene Issues, Vol. 9, Issue 6, o 3, ovember ISS (Onlne): Study and Optmzaton of Cooperatve Spetrum Sensng n OFDM Cogntve Rado etworks Yngxue LI, Shuqun SE and Qua WAG Shool of Eletron Engneerng, Bejng Unversty of Posts and eleommunatons Bejng, 876, Chna Shool of nformaton and ommunaton engneerng, Bejng Unversty of Posts and eleommunatons Bejng, 876, Chna Abstrat In ths paper, a ooperatve sensng algorthm for OFDM s proposed n ths paper based on the performane analyss of OPD algorthm, then the system weght vetor s optmzed and the losed form expresson of optmal weght vetor s obtaned usng the maxmum mproved defleton oeffent. heoretal analyss and smulaton results demonstrate the remarkable mprovement of proposed algorthm on deteton performane ompared wth the lassal ooperaton deteton algorthm. eywords: Cyl prefx; Cooperatve sensng; Defleton oeffent; OFDM; Cogntve rado Introduton Orthogonal Frequeny Dvson Multplexng (OFDM) has beome the ore modulaton tehnque soluton for next generaton wreless ommunaton. It has been used by prmary users (PU) due to the harater of multarrer and hgh spetrum effeny. Reently the study of deteton of OFDM sgnal n ogntve rado (CR) system has drawn onsderable attenton. he deteton algorthms of sngle ogntve user an be roughly dvded nto two ategores: the algorthm based on Plot ones (P) [] and the algorthm based on Cyl Prefx (CP) [-5], for the reason that the P mode s able and dffult to obtan through estmaton, more researh s fous on the algorthm based on CP. he orrelaton of CP has been alulated to detet the PU sgnal n the tme of a sngle OFDM blok n [], and then a new algorthm was proposed by extendng the orrelaton deteton to a seres of onseutve OFDM bloks n [3]. After analyss of [,6], the Optmal EYMA-PEARSO Detetor (OPD) was proposed based on EYMA-PEARSO rteron n []. And ths OPD algorthm does not need any pror knowledge about synhronzaton nformaton wth low omplexty. hrough related lterature analyss, we found that most of OFDM sgnal detetors utlze frequeny doman or tme doman to help deteton but pay lttle attenton to the spae resoures. hs may lead to the dden termnal problem and shadow fadng problem. Utlzng spae dversty tehnology, mult-user ollaboratve spetrum deteton tehnology an solve hdden termnal problem and shadow fadng problem of spetrum sensng n ogntve rado and mprove the deteton performane effeny[7-]. It s the orgnal ntenton of the paper that how to ombne the OFDM detetor and the ooperatve deteton algorthm effetvely n CR networks. Based on the above onsderatons, we propose a new ooperaton deteton algorthm for OFDM sgnal. he ontrbuton of ths paper s as followng: Frst, we proposed a new ooperatve spetrum sensng algorthm for OFDM sgnal wth soft ombned deson fuson n ogntve rado networks; seond, we utlze the maxmum modfed defleton oeffent to mprove the deteton performane and gve the losed form expressons of optmal weghted vetor. System Model. Loal Sensng Algorthm he detetor we proposed for sngle node of seondary network s based on the OPD derved n []. Utlzng the feature of OFDM sgnal, OPD s sutable for mult-cr deteton n ogntve networks wth low omplexty. Eah CR node samples sgnal n the same nterested frequeny band to detet the OFDM sgnal of PU durng the sensng perod. We assume that eah OFDM blok obtans data wth length of d and CP wth length of. Wthout loss of generalty, we sample from + OFDM bloks n order to ensure the Copyrght () Internatonal Journal of Computer Sene Issues. All Rghts Reserved.

2 IJCSI Internatonal Journal of Computer Sene Issues, Vol. 9, Issue 6, o 3, ovember ISS (Onlne): total number of OFDM bloks. he reeved vetor x onssts of onseutve OFDM symbols. After * defnng the sample value produt r x x d, we have R r k( d) k,,,. When PU s present, there wll be onseutve values of R that have a dfferent dstrbuton from the other d values, otherwse all the averaged sample value produts R ( d ) are dentally dstrbuted. Gven the synhronzaton msmath, S denotes the set of onseutve ndes, () denotes takng the real part. he OPD detetor s gven as follow: C R () S he dstrbuton of R n, has been gven by [], but some mportant parameter we need has not been gven. Due to the mutual ndependene of Rn, S, we an obtan the dstrbuton of the deteton result C R n, of the th CR s as follows S C n,, h s n, / h s n, h s, () From the dstrbuton,we an get the expetaton and ane of C n the two knds of test ondtons and respetvely.,, E C, E C h s C C, n,. OFDM Cooperatve Sensng Model d (3) s n, / h s n, he sensng system of ogntve networks s shown n fgure, CR nodes detet PU sgnal and use the lnear weghtng mode n fuson enter. Sensng proess s dvded nto sensng stage and reportng stage. After sensng stage the reeved sgnal n CR s gven as follow: h () n x (5) hs n,, Where x denotes the reeved sgnal vetor by the th CR node, and h denotes the attenuaton fator of the sensng hannel from CR to PU, s denotes a sequene of onseutvely transmtted OFDM symbols, and n s a nose vetor, and n (, n, I ) where n, s the nose ane of the th sensng hannel and represents the omplex Gaussan dstrbuton. Eah CR use the OPD algorthm to detet the reeve sgnal samples durng the sensng stage, and then the aumulatve results of OPD are reported to the FU through the Completely orthogonal report hannels. he nformaton FU reeved from the th CR s y C v, then FU uses the lnear weghted method to ombne the data from all the CRs, and judges whether there s PU sgnal or not after omparng wth a predefned hreshold. OFDM sgnal CR CR CR S S R, R, S R, Fg the OFDM ooperatve sensng algorthm struture C C C n v v v y y y Fuson enter FU reeves the deteton data C from dfferent CRs through the report hannels and fuses these data n order to mprove the deteton effeny. aordng to the Fg and type (3), the reeved data y by FU obey the Gaussan dstrbuton, ther expetatons are the same as C for the reason that E y E C and anes an be obtaned as n,, y v, h s n, / h s n,, y v, y After reevng (6), the FU judges whether the PU sgnal s present or not through lnear fuson as follow y y y (7) Where s threshold of the global judgment, and the s the weght vetor,,,. an dretly reflet the sze of CR s ontrbuton to the fuson result, whh relates to the state of the sensng hannel and report hannel. Aordng to the formula () (5) (6) and (7), we an nfer that y follows the Gaussan dstrbuton, the expetaton and ane of and Copyrght () Internatonal Journal of Computer Sene Issues. All Rghts Reserved.

3 IJCSI Internatonal Journal of Computer Sene Issues, Vol. 9, Issue 6, o 3, ovember ISS (Onlne): s gven as respetvely y y E y E y μ Σ Σ Where μ s -dmensonal olumn vetor, h s,, h s μ, both Σ and Σ dagonal matrx, Σ dag,,,,, Σ dag,,,,. he false alarm probablty and deteton probablty of the whole ooperatve algorthm were gven respetvely as follow. Pf Pd Q Σ Q μ Σ s (8) (9) () 3 Optmzaton Algorthm Based on the Improved Defleton Coeffent As the sngle parameter to reflet the performane of bnary deteton algorthm, the defleton oeffent s dfferent from the false alarm probablty and deteton probablty, sne the two parameters an reflet deteton performane only when they are onsdered jontly. In general, the bgger value of defleton oeffent means the better deteton performane. Conversely, the smaller value of defleton oeffent, the relatvely poorer performane of the deteton algorthm []. We use maxmzaton defleton oeffent method to optmze the weghted vetor of global judgment. he alulaton formula of the defleton oeffent s of two knds[]: the denomnator s y and y respetvely. We propose the mproved defleton oeffent ombnng two knds of denomnator n ths paper, and the denomnator adopts p y p y n the paper. he optmzaton goal s to maxmze the mproved defleton oeffent d through hoosng the rght. We an obtan the mathematal model as follow. E y E y argmax d, p y p y μ p Σ p Σ p () he vetor s the only able n the model wth, and d s the ontnuous funtons of, t s not dffult to fnd that d value when satsfes the ondton of we an wrte t as follow. d μμ pσ p Σ pσ p Σ has a maxmum μ pσ p Σ pσ p Σ he expresson of weghted vetor an be alulated as follow d, () opt satsfyng () pσ p Σ opt pσ p Σ μ μ pσ p Σ he value of s a μ postve real number, and t doesn t affet the value of opt whh wll be normalzaton at last, so after normalzaton Where opt an be wrtten as follow () opt pσ p Σ μ (3) pσ p Σ μ We an fnd opt an be easy to obtan from formula (3), and we analyss the mpat of p value on the performane of ooperatve deteton n Fg 3. Smulaton Results In ths seton omputer smulaton and analyss s proposed n order to verfy the orretness and effetveness of the proposed algorthm. All the smulatons are obtaned by Monte Carlo method, the sensng hannels are set to be Gaussan fadng. he SR of the sensng hannels are generated randomly wth a average SR. we set the FU reeve nformaton from CR wthout fadng and the smulaton model are desgned as Fg.. Copyrght () Internatonal Journal of Computer Sene Issues. All Rghts Reserved.

4 IJCSI Internatonal Journal of Computer Sene Issues, Vol. 9, Issue 6, o 3, ovember ISS (Onlne): Fg he ROC urves of the optmzaton algorthm wth ous CP number and ous OFDM blok number Fg shows the omparson of the smulaton performane when the CP number and OFDM blok number are dfferent. the CR number s 5, and the average SR s SR db n the smulaton. the CP number of PU OFDM sgnal s set to, 5, respetvely and the OFDM number of PU sgnal s set to,, 8 respetvely. From Fg, t an be seen that the bgger the CP number and OFDM number of PU are set, the better the performane of algorthm an be obtaned. It s remarkable that the CP number seems to play the bgger role than OFDM number n the deteton algorthm. Fg the ROC urves of optmzaton algorthm, MRC algorthm and energy deteton algorthm Fg shows that the ROC urves of the optmzaton algorthm of OPD, Maxmal-Rato Combng (MRC) of OPD and the ooperaton energy deteton algorthm wth ous CR number. we set SR db, 5, 5 n ths smulaton, and the ooperaton energy deteton algorthm also uses MRC fuson algorthm n FU. he smulaton results n Fg shows that the deteton performane of the proposed optmzaton algorthm s better than the MRC ollaboratve deteton of OPD and ollaboratve energy deteton, and the bgger the CR number s, the better the deteton performane s. It's mportant to note that when average SR s small enough, the performane dfferenes between p and p are tny, and the deteton performane urves of p and p onde wth eah other. Conlusons Fg 3 he ROC urves of the optmzaton algorthm wth ous values of p he dfferent value of p n formula (3) wll lead to the dfferent value of weght vetor and then wll ause the ous deteton performane. Fg 3 shows the deteton performane aused by the dfferent value of p( p ). we set SR db, 5, 5, 5 n ths smulaton. From Fg 3, t an be found that the deteton performane of the proposed algorthm an be mproved nonlnearly wth the nrease of the value of p, and the deteton performane s optmum when p=. In the paper, the deteton performane of the OPD algorthm s analyzed and a multuser ooperaton deteton algorthm for OFDM sgnal s proposed, and the performane parameters of the algorthm have been alulated. Furthermore, we have derved the lose form of optmal weght vetor usng the maxmum mproved defleton oeffent. he smulaton results show that the proposed algorthm s better than other ooperatve deteton algorthm wth good performane. Referene [] Chen,. S., Gao, W. & Daut, D. G. Spetrum Sensng for OFDM Systems Employng Plot ones. Ieee Wrel Commun, 9, 8(), [] Chaudhar, S., ovunen, V. & Poor,. V. Autoorrelaton-Based Deentralzed Sequental Deteton of OFDM Sgnals n Cogntve Rados. Sgnal Proessng, IEEE ransatons on, 9, 57(7), [3] Bokharaee, S., guyen,.. & Shwedyk, E. Blnd Spetrum Sensng for OFDM-Based Copyrght () Internatonal Journal of Computer Sene Issues. All Rghts Reserved.

5 IJCSI Internatonal Journal of Computer Sene Issues, Vol. 9, Issue 6, o 3, ovember ISS (Onlne): Cogntve Rado Systems. Ieee Veh ehnol,, 6(3), [] Axell, E. & Larsson, E. G. Optmal and Sub-Optmal Spetrum Sensng of OFDM Sgnals n nown and Unknown ose Varane. Ieee Journal on Seleted Areas n Communatons,, 9(), 9-3. [5] Sebesta, V., Marsalek, R. & Fedra, Z. OFDM Sgnal Detetor Based on Cyl Autoorrelaton Funton and ts Propertes. Radoengneerng,, (), [6] wang, C.., La, G. L. & Chen, S. C. Spetrum Sensng n Wdeband OFDM Cogntve Rados. Ieee Sgnal Proes,, 58(), [7] Atapattu, S., ellambura, C. & Jang,. Energy Deteton Based Cooperatve Spetrum Sensng n Cogntve Rado etworks. Ieee Wrel Commun,, (), 3-. [8] Ma, J., Zhao, G. D. & L, Y. Soft Combnaton and Deteton for Cooperatve Spetrum Sensng n Cogntve Rado etworks. Ieee Wrel Commun, 8, 7(), [9] Quan, Z., Cu, S. G. & Sayed, A.. Optmal Lnear Cooperaton for Spetrum Sensng n Cogntve Rado etworks. Ieee J-Stsp, 8, (), 8-. [] Zhao, Y. X., Song, M. & Xn, C. S. A weghted ooperatve spetrum sensng framework for nfrastruture-based ogntve rado networks. Comput Commun,, 3(), [] kay, S. M. Fundamentals of Statstal Sgnal Proessng: Deteton heory 998. Vol. Prente all. Copyrght () Internatonal Journal of Computer Sene Issues. All Rghts Reserved.