认知无线电环境下一种基于信任度的双门限协同频谱感知算法-外文文献翻译

1、摘要该文提出了一种在认知无线网络控制信道带宽受限条件下基于信任度的双门限协同频谱感知算法。 首先每个认知用户基于双检测门限独立进行频谱感知，但只有部分可靠的认知用户通过控制信道向认知无线网络基站发送本地感知结果。当所有的用户都不可靠时， 选取信任度最高的认知用户发送本地感知结果进行判决。理论分析和仿真表明， 同常规能量检测算法相比较， 该算法能够在控制信道带宽受限条件下，以较少的网络开销获得更好的频谱感知性能。关键词：认知无线电；频谱感知；信任度；双门限1 引言随着无线通信技术的飞速发展，有限的频谱资源与不断增长的无线通信需求的矛盾越来越突出。 然而根据现有的固定分配频谱资源策略，绝大多数频谱

2、资源得不到有效利用。 据fcc 的调查统计， 70%的已分配频谱资源没有得到有效利用1。为了提高频谱资源的利用率，认知无线电技术由joseph mitola 提出并得到了广泛的关注 52。频谱感知技术是认知无线电网络的支撑技术之一。通常它又可以分为能量检测法、匹配滤波器法和循环平稳特征法4。能量检测算法因为应用简单且无需知道任何授权用户信号的先验知识成为研究热点。认知用户在接入授权频带之前，必须首先感知该频带空闲即授权用户没有工作，否则会对授权用户造成干扰。 一旦授权用户重新工作， 认知用户必须退避， 实现在不对授权用户产生干扰的情况下对频谱资源的共享。 由于实际信道中的多径和阴影效应，单个认

3、知用户频谱感知的性能并不乐观，针对这个问题d.cabric等人提出了协同频谱感知算法5-6。协同频谱感知算法性能较好，但是当认知用户数量很大的时候，控制信道的带宽将不够用。文献7中提出了一种在控制信道带宽受限条件下的基于双检测门限的频谱感知算法，该算法能够以较小的网络开销，获得接近普通单门限频谱检测算法的性能。针对认知无线电频谱感知的需要， 本文提出了认知无线电环境下一种基于信任度的双门限协同频谱感知算法。 该算法中每个认知用户基于双检测门限独立进行频谱感知，但只有部分可靠的认知用户通过控制信道向认知无线网络基站发射感知报告。当所有的用户都不可靠时， 选取信任度最高的认知用户发射感知报告进行判

4、决。本文对该算法进行了性能分析并通过仿真表明，本文方法比较常规能量检测算法，在减小网络开销的同时提高了检测性能。2 系统模型假设一个认知无线电网络有n 个认知用户和一个认知无线网络基站，如图1 所示。认知无线网络基站负责管理和联系n 个认知用户，在收到认知用户的检测报告后做出最终判决。图1. 认知无线电网络示意图频谱感知的实质是一个二元假设问题，即01( ),( )( )( )( ),n thx th ts tn th（1）其中x(t)代表认知用户接收到的信号， s(t)表示授权用户的发送信号， h(t)代表授权用户与认知用户之间信道的衰落因子。0h 代表授权用户没有工作，1h 代表授权用户正

5、在工作。设是认知用户接收信号的能量，根据能量检测理论8，服从以下分布：122022),(,hxhxmm（2）其中表示瞬时信噪比，并且其服从均值为_的指数分布，22mx表自由度为2m 的中心卡方分布，)(22mx代表自由度为m2非中心参数为的卡方分布， m表示时间带宽积。在能量检测算法本地判决中，每个认知用户把接收到的能量跟预设的门限进行比较，如图 2（a）所示。当时，本地能量检测器做出本地判决1d，表示授权用户在工作，否则判决d为 0。而双门限能量检测算法本地判决如图3(b)所示，本地能量检测器判决规则如下：（3）其中 nd 表示认知用户接受到的能量值不可靠，认知用户不作出任何判决， 发送感知

6、报告给认知无线电网络基站。如果出现所有认知用户都不作出判决的情况，则选择信用度最高的认知用户依据单门限能量检测算法作出本地判决。并发送感知报告给认知无线电网络基站。图 2.（a）一般能量检测算法本地判决示意图（b）双门限能量检测算法本地判决示意图信用度获取方法采取文献9的方法：在最开始阶段，认知无线电网络基站把每个认知用户数目的可信度设为0，当某认知用户本地判决结果与认知无线电网络基站的最终判决结果一致时，该认知用户可信度加1，否则减 1。假设认知用户i 的可信度是i，则其更新过程如（ 4） ：uuiii)1(（4）其中 u是认知用户传送给认知无线电网络基站的判决结果，1u是认知无线电网络基站

7、的最终判决结果。本地判决d=0 本地判决d=1 (a) (b) 0 0本地判决d=0 本地判决d=1 nd 0 1221211, 1,0,0ndd据文献8可知，认知用户在高斯信道下的平均检测概率、平均漏检概率和平均虚警概率如下所示：),2()|(1mdqhpp(5)dmphpp1)|(1(6) )()2/,()|(0mmhppf(7) 出于对授权用户的保护，认知无线电网络基站最终采用or 准则作出判决。3 频谱感知性能分析3.1 网络开销在 1bit 量化条件下，avgk代表归一化平均感知位数，kt和knt分别代表 k 个已向认知无线电网络基站发送数据和n-k 个未向认知无线电网络基站发送报告

8、。则：kikptp)(1 21，kniknptp)(21。设11hpp和00hpp，则划归一划平均感知位数acgk如式 8 所示：nknkknkknkavghtphtpknkphtphtpknkpk11111000|(8) 定义：)|(02110hp，)|(02110hp则：11001ppkavg(9) 由 9 式可得：1avgk可知：基于双门限的协同频谱检测算法的网络开销始终小于常规的能量检测算法。3.2 检测性能分析设)(f和)(g别表示在假设0h 和1h下的概率分布，则根据文献10可知: 00)()2/,(1)|()(mmdhff(10) 01)|()(dhfg=202_2)1(2201

9、_22)1(2!11211mnmnmneene(11) 显然)()(210ff，)()(121gg。假设0，1分别代表在授权用户在工作和授权用户未工作情况下没有认知用户发送感知报告，即当 k=0 时， 则：nnffhkp01200)()(|0(12) nngghkp11211)()(|0(13) 基于双门限的频谱感知算法在瑞利信道下的虚警概率fq ，漏检概率mq 和检测概率dq 分别为：)1)(1( 1,|1|1|0, 10000afpkhuphkphkupq(14) mq =dqhkup1|0,01(15) )1)(1( 1,|1|1|0,11111pakhuphkphkupqd(16) 其

10、中：daqhkupp1|0, 01=nkknkfffkn1121)()()(=02)(nf(17) 12)(gpb(18) 则：)(1)(1(020nffq(19) )(1)(1 (121nfgq(20) 由上式可知当0=0 时，此算法与常规算法相同。当参与协同的认知用户数目n 较大时，00，则基于双门限的频谱检测算法的检测性能与常规能量算法的检测性能近似，可知在控制信道带宽受限制的情况下以较小的性能损失大大降低了网络开销。4 仿真及分析本节通过计算机仿真来评估所提出的基于信任度的双门限协同频谱感知算法的性能。仿真参数设置如表1 所示。表1 仿真参数设置参数数值认知用户数目10n平均信噪比db

11、10_时间带宽积5m授权用户占用信道概率5 .0op授权用户不占用信道概率5.01p图 3 给出了在1 .00的情况下算法的检测性能。 可以看出同常规能量检测算法相比较，本文所提出算法的检测性能得到了明显的改善。例如当001.0fq时，基于信任度的双门限协同频谱感知算法的检测概率dq 比常规能量检测算法高出0.019 。图3检测性能示意图图 4 描述了在不同0的条件下，基于信任度的双门限协同频谱感知算法对网络开销的影响。同常规能量检测算法即0=0 时相比较，本文所提出算法的归一化平均感知位数avgk急剧下降，控制信道带宽与认知用户数量之间的矛盾得到了缓解。例如当01.0fq，0= 0.01 时

12、，基于信任度的双门限协同频谱感知算法的归一化平均感知位数avgk下降了 38%。 当01.0fq，0=0.001时，归一化平均感知位数avgk则下降了 44% 图 4 不同0条件下算法对网络开销的影响5 结束语频谱感知技术是认知无线电网络的支撑技术之一。当认知用户数量很大的时候，控制信道的带宽将不够用。 本文提出了认知无线电环境下一种基于信任度的双门限协同频谱感知算法。 每个认知用户基于双检测门限独立进行频谱感知，但只有部分可靠的认知用户通过控制信道向认知无线网络基站发射感知报告。当所有的用户都不可靠时， 选取信任度最高的认知用户发射感知报告进行判决。本文对该算法进行了性能分析并通过仿真表明，

13、 本文方法比较常规能量检测算法，在减小网络开销的同时提高了检测性能。参考文献1 federal communications commission. spectrum policy task force,rep. et docket no. 02-135 r. nov. 2002. 2 j. mitola and g. q. maguire. cognitive radio: making softwareradios more personalc,ieee personal communication.vol. 6, pp. 13 18,aug. 1999. 3 s. haykin. cog

14、nitive radio: brain-empowered wirelesscommunicationsj. ieee j. sel. areas communication. vol. 23, pp.201 220, feb. 2005. 4 akyldiz if. next generation/dynamic spectrum access/cognitiveradio wireless networks: a surveyj. elsevier computernetworks, 2006(50):2127-2159. 5 d. cabric, s. m. mishra, and r.

15、 w. brodersen. implementationissues in spectrum sensing for cognitive radiosc/ in proc. ofasilomar conf. on signals, systems, and computers, pacific grove,ca, usa, nov. 7-10, 2004, pp. 772 - 776. 6 a.ghasemi and e. s. sousa. collaborative spectrum sensing foropportunistic access in fading environmen

16、tsc/ in proc. 1st ieeesymp. new frontiers in dynamic spectrum access networks,baltimore, usa, nov. 8 11, 2005, pp. 131 136. 7 chunhua sun, wei zhang, letaief k.b. cooperative spectrumsensing for cognitive radios under bandwidth constraintsc/ in proc.ieee wcnc, march 11-15, 2007, pp. 1-5. 8 h.urkowit

17、z. energy detection of unknown deterministic signalsc. proceedings of ieee, vol.55, pp. 523-531, april 1967. 9 ruiliangchen, jung-minpark, kaigui bian. robust distributed spectrum sensing in cognitive radio networksc. in proc. ieeeinfocom, april 2008, pp. 1876-1884. 10 f. f. digham, m. -s.alouini, a

18、nd m. k. simon. on the energydetection of unknown signals over fading channelsc. in proc.ieee icc, anchorage,ak, usa, may 11-15, 2003, pp. 3575 3579. 附原文：a new cooperative spectrum sensing algorithm for cognitive radio networksabstractspectrum sensing is a critical phase in building a cognitive radi

19、o network. however, the bandwidth for reporting secondary users sensing results will be insufficient, when the number of secondary user is very large. in this paper, we propose a new cooperative spectrum sensing algorithm to alleviate the bandwidth problem of reporting channel. compared with convent

20、ional method, only the secondary users with reliable information are allowed to report their sensing results.when no user with reliable information, only the secondary user with highest reputation will report its sensing result. simulation results show that our algorithm achieves better sensing perf

21、ormance and the average number of sensing bits decrease greatly. keywordscognitive radio; cooperative spectrum sensing; double threshold; reputation .introduction due to the increasingly development of wireless applications, more and more spectrum resources are needed to support numerous emerging wi

22、reless service. however, recent measurements by federal communication commission (fcc) have shown that 70% of the allocated spectrum in us is not utilized 1. in order to increase the efficiency of spectrum utilization, cognitive radio technology was recently proposed 2, 3.a requirement of cognitive

23、radios is that their transmission should not cause harmful interference to primary users. namely, the secondary users can use the licensed spectrum as long as the primary user is absent. however, when the primary user comes back into operation, the secondary users should vacate the spectrum instantl

24、y to avoid interference with the primary user. accordingly, spectrum sensing is a crucial phase in building a cognitive radio system. one of the great challenges of implementing spectrum sensing is the hidden terminal problem which caused by the fading of the channels and the shadowing effects. in o

25、rder to deal with the hidden terminal problem, cooperative spectrum sensing has been studied to improve the spectrum sensing performance 4, 5.in6, due to control channel for each cognitive radio to report its sensing result is usually bandwidth limited, a censoring method which has two thresholds is

26、 given to decrease the average number of sensing bits to the common receiver. by censoring the collected local observations, only the secondary users with enough information will send their local decisions to the common receiver. in this paper, we present a new double threshold cooperative spectrum

27、sensing method with reputation. inour system, every cognitive user will firstly obtain an observation independently and only the users with reliable information send their local decisions to the common receiver based on double thresholds. if no user is reliable, only the cognitive user with the high

28、est reputation is selected to sense the spectrum. simulation results show that the spectrum sensing performance under awgn channels is improved and the communication traffic is also reduced as opposed to the conventional method. the rest of the paper is organized as follows. in section , system mode

29、l is briefly introduced. sensing performance is analyzed in section . in section , we present the simulation results of our cooperative spectrum sensing method. finally, we draw our conclusions in section . ii. system model in cognitive radio systems, spectrum sensing is a critical element as it sho

30、uld be firstly performed before allowing secondary users to access a vacant licensed channel. cooperative spectrum sensing has been widely used to detect the primary user with a high agility and accuracy. the essence of spectrum sensing is a binary hypothesis-testing problem: 0h :primary user is abs

31、ent; 1h:primary user is present. for implementation simplicity, we restrict ourselves to energy detection in the spectrum sensing. the local spectrum sensing is to decide between the following two hypotheses: 01( ),( )( )( )( ),n thx th ts tn th(1) where)(txis the signal received by secondary user,)

32、(tsis primary users transmitted signal,)(tnis awgn, and)(this the temporary amplitude gain of the channel. according to energy detection theory 7, we have the following distribution: 122022),(,hxhxmm(2) where is the energy value collected by secondary user, isinstantaneous snr and follows exponentia

33、lly distributionwith the mean value_, m is the time bandwidth product ofthe energy detector,22mrepresents a central chi-squaredistribution with 2m degrees of freedom and.)(22mrepresents a non-central chi-square distributionwithm2degrees of freedom and a non-centrality parameter2. in conventional ene

34、rgy detection method, the localdecision is made by comparing the observation with a pre-fixed threshold as fig.1 (a). when the collected energy exceeds the threshold, decision0h will be made.otherwise decision 1hwill be made. in contrast, the systemmodel which has two thresholds of our interest is s

35、hown infig.1 (b). where decision0h and decision 1h representthe absence and the presence of licensed user, respectively. “ no decision means that the observation is not reliableenough and theith cognitive user will send nothing to thecommon receiver. but when all the secondary users dontsend their l

36、ocal decisions, only the cognitive user with thehighest reputation is selected to sense spectrum based onconventional energy detection method, and send its localdecision to the common receiver. reputation is obtained based on the accuracy of cognitiveuser s sensing results. the reputation value is s

37、et to zero atthe beginning. whenever its local spectrum sensing report isconsistent with the final sensing decision, its reputation isincremented by one; otherwise it is decremented by one.under this rule, assuming the ith cognitive users reputationvalue is 1, the last sensing report of cognitive us

38、er isend tocommon receiver isu,and the final decision isiu ,theniisupdated according to the following relation:uuiii) 1(for the cognitive radio users with the energy detector, the average probabilities of detection, the average probabilities of missed detection, and the average probabilities of fals

39、e alarm over awgn channels are given, respectively, by 7: ),2()|(1mdqhpp(3)dmphpp1)|(1(4) )()2/,()|(0mmhppf(5) where)(a,),(baare complete and incomplete gammafunction respectively, and),(baqmis the generalized marcum function. in this paper, we consider cooperative spectrum sensing with 1bit quantiz

40、ation. let avgkrepresent the normalized fig1. (a)conventional detection method (b)double threshold energy detection method average number of sensing bit.letktand kntrepresenthe event that there are k unlicensed users reporting 1-bit decision and n-k users not reporting to the common receiver, respec

41、tively.the kikptp)(121, kniknptp)(21.and then the average number of sensing bits for our method can be derived as: nknkknkknkavghtphtpknkphtphtpknkpk11111000|（6）for simplicity, we define: )|(02110hp，)|(02110hp（7）letnavgkdenote the normalized average number of sensing bits, then, we obtain navgkas d=

42、0 d=1 (a(b) 0 0d=0 d=1 nd 0 12follows：11001ppkavg（8）from (8), it can be seen that, the normalized average number of sensing bits navgkis always smaller than 1. the communication traffic of our method is are deduced as opposed to the conventional energy detection method. iii. the performance analysis

43、 of spectrumsensing in this section, the spectrum sensing performance of the proposed method will be analyzed. assume the control channel between the unlicensed users and the common receiver is perfect, the local decisions are reported without any error. let )(fand )(gdenote the cumulative distribut

44、ion function (cdf) of the local test statisticunder the hypothesis 0h and1h, respectively. then, we have 10: 00)()2/,(1)|()(mmdhff(9) 01)|()(dhfg(10) obviously,)()(210ff，)()(121gg. if no any local decision is reported to the common receiver, i.e., k=0 , we call that fail sensing. for this case, the

45、common receiver will request the user which has the highest reputation to send its local decision based on conventional energy detection method. let0and 1denote the probability of fail sensing under hypothesis 0h and 1h, respectively. here we have: nnffhkp01200)()(|0(11) nngghkp11211)()(|0(12) appar

46、ently,n00and n11.in our scheme, the false alarm probabilityfq ,the detection probabilitydq ,and the missing probabilitymq : )1)(1( 1,|1|1|0, 10000afpkhuphkphkupq(13) mq =dqhkup1|0, 01(14) )1)(1(1,|1|1|0, 11111pakhuphkphkupqd(15) for simplicity, we assume the channel between the unlicensed users and

47、the base station are ideal, the local decision will be reported without any error. soapstand for the probability of the event that under hypothesis0h, all the k users claim 0h and other n-k users make no local decisions. daqhkupp1|0, 01=nkknkfffkn1121)()()(=02)(nf（16）12)(gpb（17）)(_1)(1(020nffq（18）)(

48、_1)(1(121nfgq（19）iv. simulation results in this section, some simulation results are presented toillustrate the system performance of our cooperativespectrum sensing algorithm based on reputation. the resultsof the conventional one threshold energy detection methodare also shown for a comparison. in

49、 our simulation, the common simulation parametersare given as follows: table 1. simulation parameters 10ndb10_5m5. 0op5.01pfig.2 depicts the performance of cooperative spectrum sensing dqandfq .1.00.it can be observed that, compared it with the conventional method, the detection performance has impr

50、oved significantly. for example, while fq = 0.001, our method achieves extra 0.019 detection probability. fig.3 shows the decrease of the normalized transmission bits for different values of fail sensing, i.e.0= 0, 0.001, 0.01, 0.1. compared with conventional method, i.e., when0= 0, the normalized a

51、verage number of sensing bits is dramatically decreased and bandwidth limited problem of the reporting channel is relieved. for example, whenfq = 0.01, almost 44% and 38% reduction of the normalized average number of sensing bits can be obtained for 0= 0.001 and0= 0.01, respectively. in our algorith

52、m,fq is upper bounded and lower bounded because of the probabilityof fail sensing0and the false alarm probability are based on (7), (13). fig 2.dq vs.fq , 1.00fig 3.navgkvs.fq ,0=00,0.001,0.01,0.1 v. conclusion in this paper, a new scheme in cooperative spectrum sensing for cognitive radio networks

53、under bandwidth constraints was proposed. in our method, only the secondary users with reliable information are allowed to report their sensing results. when no user has reliable information, only he secondary user with highest reputation will report its sensing result. we analyzed the closed expression for the probability of the detection and the false-alarm. fro