• <tr id="yyy80"></tr>
  • <sup id="yyy80"></sup>
  • <tfoot id="yyy80"><noscript id="yyy80"></noscript></tfoot>
  • 99热精品在线国产_美女午夜性视频免费_国产精品国产高清国产av_av欧美777_自拍偷自拍亚洲精品老妇_亚洲熟女精品中文字幕_www日本黄色视频网_国产精品野战在线观看 ?

    A Physical Layer Algorithm for Estimation of Number of Tags in UHF RFID Anti-Collision Design

    2019-11-07 03:13:06ZhongHuangJianSuGuangjunWenWenxianZhengChuChuYijunZhang4andYiboZhang
    Computers Materials&Continua 2019年10期

    Zhong HuangJian SuGuangjun WenWenxian ZhengChu ChuYijun Zhang4 and Yibo Zhang

    Abstract:A priori knowledge of the number of tags is crucial for anti-collision protocols in slotted UHF RFID systems.The number of tags is used to decide optimal frame length in dynamic frame slotted ALOHA(DFSA)and to adjust access probability in random access protocols.Conventional researches estimate the number of tags in MAC layer based on statistics of empty slots,collided slots and successful slots.Usually,a collision detection algorithm is employed to determine types of time slots.Only three types are distinguished because of lack of ability to detect the number of tags in single time slot.In this paper,a physical layer algorithm is proposed to detect the number of tags in a collided slot.Mean shift algorithm is utilized,and some properties of backscatter signals are investigated.Simulation results verify the effectiveness of the proposed solution in terms of low estimation error with a high SNR range,outperforming the existing MAC layer approaches.

    Keywords:UHF RFID,anti-collision,cluster algorithm.

    1 Introduction

    Anti-collision algorithms are carried out in multi-access UHF RFID systems to reduce collisions as well as to increase channel efficiency.As one of the most popular anticollision algorithms for RFID system,dynamic framed slotted ALOHA(DFSA)algorithm employs a mechanism similar with time division multiple access(TDMA).Synchronized frame is divided into several time slots for random access.Reader dynamically adjusts the number of time slots based on the estimated number of tags.It is well known that DFSA reaches the optimal system throughput when the number of slots equals to the number of tags waiting to be identified.There are two main problems in DFSA design,estimation of the number of tags and frame length adjustment[Chen,Liu,Ma et al.(2018)].

    Various researches are proposed to improve the estimation accuracy so as to improve the channel efficiency of RFID system.The number of tags is estimated by multiplying the number of collided slots in a frame by the expected number of tags per collided slot(=2.39),which is a constant for all frames regardless of number of tags and frame size[Schoute(1983)].In the mechanisms proposed by Vogt et al.[Vogt(2002);Chen and Lin(2006);Cha and Kim(2005)],estimation of the number of tags is based on successful,collided,and empty probability in a frame.Probability theory is utilized in these works.The number of tags is estimated in Chen[Chen(2009)]by multiplying the number of collided time slots by a well-defined factor,which is found by an iterative algorithm.A posteriori probability distribution-based method is proposed in Eom et al.[Eom and Lee(2010)]to further improve the accuracy.However,the computation complexity is higher than the others.A Bayesian method[Annur,Srichavengsup,Nakpeerayuth et al.(2015)]is used to update the posterior probability distribution of number of users' slot by slot so as to estimate the number of tags.

    Since RFID readers usually adopt in-phase and quadrature information of tag signals,tag recovery and estimation methods of the number of tags based on signal processing are developed to solve collision problem.Tag recovery is able to turn a collided slot into a successful slot while estimation of the number of tags based on physical layer process is able to enhance performance of anti-collision algorithm in upper layer.Most researches require multiple antennas receiver[Angerer,Langwieser and Rupp(2010)],specific tag signal strength[Fyhn,Jacobsen,Popovski et al.(2011)],modified coding mechanisms[Parks,Liu,Gollakota et al.(2014)],etc.Bipartite Grouping(BiGroup)[Ou,Li and Zheng(2017)]is the first one proposed to parallelly decode multiple CTOS tags with the help of both time domain and constellation domain information in physical layer.Algorithms based on signal processing in time domain are proposed to reduce SNR requirements.Collided signals are transformed to time-scale domains and LS criterion is utilized for tag signal separation in Zeng et al.[Zeng,Wu,Yang et al.(2017)].An edge transition scheme is proposed to recover collision and decode tag signals in Benbaghdad et al.[Benbaghdad,Fergani and Tedjini(2016)].These works focus on tag signal recovery issues in physical layer and pay less attention on MAC layer design.Tan et al.[Tan,Wang,Fu et al.(2018)]proposed a collision detection and signal recovery method and combine them with DFSA algorithm.Optimal frame length is calculated based on a collision recovery probability coefficient,which is obtained by simulations of its physical layer design.A novel closed-form solution is further proposed in Ahmed et al.[Ahmed,Salah,Robert et al.(2018)]for optimal FSA frame length decision,in which collision recovery probabilities are provided arbitrarily.An accurate tag number estimation algorithm is still needed and not addressed.

    In this paper,we propose to estimate the number of tags in a collided time slot in physical layer.Different from the existing work,we focus on improving estimation accuracy under CTOS tag assumption.Estimation error is reduced compared with MAC layer design and SNR range is expanded compared with current physical layer design.In this scheme,the number of tags in single time slot is determined by the number of clusters located in constellation domain.Signal samples are firstly scaled based on the baseband noise level.After that,mean shift algorithm is utilized to divide the data into several clusters.In the end,a cluster adjustment process is carried out to get better performance.Simulation results verify the effectiveness of the proposed solution in terms of low estimation errors and high SNR range,outperforming the existing MAC layer-based approaches.This paper is organized as following.Section II describes signal model of UHF RFID system and drawback of clustering algorithms.After that,design consideration and cluster-based algorithm is proposed in Section III,which is evaluated with numeric simulations in Section IV.Finally,conclusion follows.

    2 Signal model and cluster algorithm

    2.1 Backscatter signal and its distribution

    In a UHF RFID system,reader energizes tags by transmitting continuous wave.Passive tags backscatter the radio to communicate with the reader.After that,backscatter signals are down converted in the reader side.As shown in Fig.1(a),baseband signal in receive path of a reader consists of three components,backscatter signal of tags,self-jammer and noise.Eq.(1)shows the detailed format of those signals.

    The first one indicates signals of multiple tags,the second one indicates the signals of self-jammer caused by RFID system.Where i indicates the index of collided tags in the same time slot,Aiandθiare respective transmitted data and initial phase of the i-th tag signal.θ0andμis the amplitude and phase of self-jammer.n(t)represents white noise in the receiver.

    Figure 1:RFID framework and backscatter signal model

    Backscatter signals have both in-phase and quadrature components,with which we could plot the signal samples in a two-dimension coordinates.Every sample is represented by a point on the I-Q plane.Samples of the same transmitting status are dispersed and scattered around a centroid position,forming a cluster.As shown in Fig.1(b),there are 4 clusters,representing 4 transmitting status of 2 tags.The number of clusters is decided by the number of tags.Every tag has two transmission status,the size of the full status space is 2n,where n is the number of tags.Three properties could be obtained from Eq.(1).

    1.Since noise in in-phase axis and in quadrature axis are different due to the existence of self-jammer,every cluster is shaped as an eclipse.

    2.We assume that noise in both axes obeys gaussian distribution.As a result,around 95% samples of one cluster are located in a circle of radius of two times standard deviation of noise.The higher the noise is,the bigger the circle is.When SNR goes down,the radius gets bigger,and when it is larger than the distance between cluster centers,the clusters overlap.Fig.1(c)shows an example when clusters are overlapped.

    3.The clusters are always located in pairs,symmetrizing to the center of all samples.For every cluster,reverse all the tag status,a symmetric cluster is obtained.Their symmetric center is located at the center of the whole graph.Eq.(2)shows the coordinates of the center point,wherensam,xcenterandycenterdenotes the number of samples,x axis coordinate and y axis coordinate.

    2.2 Density based clustering algorithm

    A straight forward way to find the number of tags is to divide signal samples into clusters with a cluster algorithm.Mean shift algorithm and DBSCAN algorithm are both widely used method to identify multiple dimension data without a prior knowledge of the number of clusters.They both make use of sample density to make decisions.Mean shift algorithm updates every cluster center based on the vector sum of all samples in it until convergence.Bandwidth is setup to decide the range of clusters.On the other hand,DBSCAN choose samples with large number of neighbors as core samples.Connection between core samples are calculated and clusters are divided.Parameters distance is required to decide neighbors and minimum points are required to decide core samples.

    Table 1:Parameter settings of cluster algorithms

    Both algorithms are tested with simulations.Parameter settings are shown in Tab.1.Fig.(2)shows the cluster division of two algorithms.Samples in one cluster are encircled by a circle.DBSCAN divides the signals into 6 clusters while mean shift divides them into more than 20 clusters.

    Figure 2:Performance of two algorithm in different SNR scenarios

    3 Physical layer algorithm for estimation of the number of tags in single time slot

    3.1 Design consideration

    Mean shift algorithm and DBSCAN are available for clustering in some cases with proper parameter settings.However,both algorithms are not designed for detection of the number of tags.DBSCAN requires uniform distributed samples in one cluster,which is not preferred in our case.Density of samples in one cluster decreases along with distance to the center.Furthermore,it is not able to identify overlapped clusters.Mean shift algorithm outperforms DBSCAN in low SNR scenario.Cluster centers are updated towards a denser direction,it always finds the densest points.However,isolated samples are identified as a cluster in some cases.In this paper,we make use of samples distribution information and its properties to improve performance of mean shift algorithm.The improvement is based on the following considerations.

    First of all,our purpose is to find number of clusters,not accurate cluster division.Wrong assignment to clusters may cause bit error in signal recovery case,but not in number detection case.As a result,we only consider a small core area of clusters,which is defined as samples within distance of two times standard deviation of noise.Find it and we get a valid cluster center.The noise samples are ignored naturally.

    Second,as described in Section II,all clusters are shaped like an eclipse.On the meantime,mean shift algorithm calculates updated vector based on a Euclid distance,which means samples in a perfect circle are all considered.It brings a big performance decrease.It is fitting and proper to adjust the scale of in-phase and quadrature signal magnate by noise level.After that,the core area of clusters is formed as a perfect circle.It is better for identification.

    Finally,clusters are distributed in pairs.Every cluster have a symmetrical one with similar number of samples.Their cluster centers are also symmetrical to the whole center of samples.This property makes us able to discard isolated noise cluster identified by mean shift algorithm.

    3.2 Algorithm details

    The detailed algorithm is shown as Algorithm.1,we first transform samples to make clusters shaped as perfect circle other than eclipse.After that,mean shift algorithm is carried out.Some random points are selected as initial centers.These centers are updated based on samples in their neighborhood.After they converge,an adjustment scheme is carried out to discard isolated noise clusters.

    4 Simulation and performance evaluation

    In this section,we evaluate the performance of proposed algorithm under different scenarios.Success rate is firstly proposed for accuracy of detecting number of tags to evaluate performance in single time slot scenario.After that,total success rate and estimation error are derived based on probability to evaluate performance in multiple time slots scenario.Success rate is compared with DBSCAN algorithm and estimation error is compared with probability-based methods.

    4.1 Performance in single time slot scenario

    The performance of proposed algorithm is evaluated by accuracy or success rate,which is defined as the number of successful experiments over the total number of experiments.In order to improve reliability of simulations,simulations are carried out multiple times in different SNR scenarios.Furthermore,scenarios with different number of tags are evaluated separately due different performance in these cases.Scenarios when the number of tags is larger than 4 is not considered here because it rarely happens in practice.

    The simulation runs as the following steps.

    In the first step,we initialize the system parameters,i.e.,number of tags and SNR.

    In the second step,10000 experiments are executed.In each experiment,pseudo signals are generated based on Eq.(1).Signal strength and initial phase of tags are randomly selected in every experiment.Proposed algorithm and DBSCAN are used to determine the number of tags.Both actual number of tags and determined number of tags are recorded for performance evolution.

    In the third step,switch to the next parameter and execute Step 2 for another time.

    In the fourth step,success rate in each system parameter set are calculated.Before that,a performance indicatorbased on conditional probability is calculated by real number and determined number,as shown in Eq.(5).Wheredenotes detected number of tags whilendenotes actual number of tags.It is apparent that success rate when the number of tags is n equals probability ofp(n|n).

    Fig.3 shows success rate of proposed algorithm in different conditions of SNR and the number of tags.Success rate of proposed algorithm is greater than DBSCAN in all conditions.When SNR is larger than 18 dB,success rate of proposed algorithm is larger than 0.9 in 2 and 3 tags scenarios.Success rate is relatively lower when there are 4 tags,still over 0.8 when SNR is large enough.

    Figure 3:Success rate comparison in different conditions

    4.2 Performance of multiple time slots

    Performance of proposed algorithm in multiple time slots is evaluated by two indicators,total success rate and estimation error.When there are multiple time slots and unknown number of unidentified tags,number of tags in one time slot follows a binomial distribution,as shown in Eq.(6).

    whereB(r)denotes the probability ofrtags in one slot,ndenotes number of tags to be identified in the read range,Ldenotes frame length,i.e.,number of time slots.

    Total success rate takes distribution of number of tags in one slot into consideration,which is defined as Eq.(7).Total success rate shows an average performance under specific condition of frame length and the number of tags.

    Similar with other estimation researches,estimation error is a good indicator for performance evaluation.Here it is defined in a probabilistic way in Eq.(8).

    whereE(n)denotes average number of tags in one time slot,whiledenotes average estimated number of tags in every time slot.It is calculated by Eq.(9).

    wheredenotes expectation of estimated number of tags on condition of the number of tags in one time slot,shown as Eq.(10).

    Figure 4:Total success rate in different conditions of the number of tags

    Fig.4 shows total success rate in different conditions of the number of tags when number of time slots are set to 128.Success Rate decreases along with the number of tags increases because possibility is larger when the number of tags is higher.Apparently,SNR effects total success rate.Total success rate decreases to 0.5 when SNR is 15 dB and the number of tags reaches 300.However,total success rate is higher than 0.8 in most high SNR conditions(higher than 20).

    Figure 5:Estimation error under different number of tags and SNR conditions

    Fig.5 shows estimation error comparison between proposed algorithm and two other methods.Estimation error of our proposal decreases while SNR goes up.In 15 dB SNR condition,our proposal beats S+2.39C method when the number of tags is more than 220.In 20dB SNR condition,estimation error of our proposal is close to method proposed in Chen.(2009).Estimation error gets lower when SNR is higher than 20 and outperforms both two methods.

    5 Conclusion

    This paper focuses on tag number estimation method for RFID anti-collision purpose.A clustering algorithm is proposed to detect the number of tags in physical layer.The proposed algorithm makes full use of in-phase and quadrature information to get better performance in low SNR scenarios.Simulation results show that it is good in a large range of SNR conditions.It is better than DBSCAN in terms of success rate and better than probability-based methods in terms of estimation errors.

    Acknowledgement:This work was supported in part by the National Natural Science Foundation of China under project contracts[NOS.61601093,61791082,61701116,61371047],in part by Sichuan Provincial Science and Technology Planning Program of China under project contracts No.2016GZ0061 and No.2018HH0044,in part by Guangdong Provincial Science and Technology Planning Program of China under project contracts No.2015B090909004 and No.2016A010101036,in part by the fundamental research funds for the Central Universities under project contract No.ZYGX2016Z011,and in part by Science and Technology on Electronic Information Control Laboratory.

    免费久久久久久久精品成人欧美视频| 在线观看免费视频网站a站| 亚洲精品美女久久av网站| 亚洲精品aⅴ在线观看| 日本一区二区免费在线视频| 黑人欧美特级aaaaaa片| 精品国产超薄肉色丝袜足j| 久久狼人影院| 亚洲一码二码三码区别大吗| 亚洲精品国产一区二区精华液| 美国免费a级毛片| 亚洲美女搞黄在线观看| 中文字幕人妻丝袜一区二区 | 欧美日韩国产mv在线观看视频| 亚洲精品在线美女| 一本一本久久a久久精品综合妖精| 日韩人妻精品一区2区三区| av又黄又爽大尺度在线免费看| 色播在线永久视频| 韩国精品一区二区三区| 精品人妻熟女毛片av久久网站| 国产成人精品久久二区二区91 | 中文字幕亚洲精品专区| 十八禁网站网址无遮挡| 只有这里有精品99| 黑人欧美特级aaaaaa片| 国产无遮挡羞羞视频在线观看| 国产午夜精品一二区理论片| 亚洲精品久久久久久婷婷小说| 国产精品免费大片| 国产 精品1| 精品一区二区三卡| 电影成人av| 久久久国产一区二区| 免费在线观看视频国产中文字幕亚洲 | 天天影视国产精品| 国产成人精品久久二区二区91 | 如何舔出高潮| videosex国产| av片东京热男人的天堂| 亚洲一区中文字幕在线| 欧美av亚洲av综合av国产av | 少妇人妻精品综合一区二区| 亚洲国产精品999| 这个男人来自地球电影免费观看 | 岛国毛片在线播放| 国产一区二区在线观看av| 啦啦啦视频在线资源免费观看| 丝袜在线中文字幕| 日韩,欧美,国产一区二区三区| 人人妻人人澡人人爽人人夜夜| 中文欧美无线码| 国产人伦9x9x在线观看| 如何舔出高潮| 一级毛片黄色毛片免费观看视频| 精品午夜福利在线看| 国产精品嫩草影院av在线观看| 国产亚洲av片在线观看秒播厂| 久久国产精品大桥未久av| 视频区图区小说| 国产av国产精品国产| 美女视频免费永久观看网站| 久久久国产精品麻豆| 国产国语露脸激情在线看| 中文字幕av电影在线播放| 如何舔出高潮| 香蕉丝袜av| 午夜老司机福利片| 欧美乱码精品一区二区三区| 亚洲国产看品久久| 亚洲人成电影观看| xxxhd国产人妻xxx| 十八禁高潮呻吟视频| 最近中文字幕2019免费版| 菩萨蛮人人尽说江南好唐韦庄| 嫩草影院入口| 一个人免费看片子| 1024视频免费在线观看| 男女高潮啪啪啪动态图| 国产无遮挡羞羞视频在线观看| 十八禁网站网址无遮挡| 亚洲美女视频黄频| a级毛片黄视频| 在线观看一区二区三区激情| 久久99一区二区三区| 久久久国产一区二区| 欧美日韩精品网址| 亚洲国产精品999| 免费观看a级毛片全部| 我的亚洲天堂| 国产精品蜜桃在线观看| 美女高潮到喷水免费观看| bbb黄色大片| 丝袜美足系列| 免费少妇av软件| 人人妻人人澡人人看| 欧美日韩视频高清一区二区三区二| 妹子高潮喷水视频| 人人妻人人澡人人爽人人夜夜| 日本vs欧美在线观看视频| 国语对白做爰xxxⅹ性视频网站| tube8黄色片| 亚洲av日韩在线播放| 日本猛色少妇xxxxx猛交久久| 秋霞在线观看毛片| 久久女婷五月综合色啪小说| 一边亲一边摸免费视频| 性色av一级| 国产在线视频一区二区| 一本色道久久久久久精品综合| 亚洲四区av| 秋霞伦理黄片| 国产精品偷伦视频观看了| 人人妻人人添人人爽欧美一区卜| 亚洲少妇的诱惑av| 一级,二级,三级黄色视频| 久久午夜综合久久蜜桃| 伊人久久大香线蕉亚洲五| 亚洲人成77777在线视频| 七月丁香在线播放| 在线 av 中文字幕| 亚洲精品美女久久久久99蜜臀 | 国产1区2区3区精品| 久久人人97超碰香蕉20202| 国产男女超爽视频在线观看| 国产亚洲午夜精品一区二区久久| 激情视频va一区二区三区| 成年人免费黄色播放视频| 亚洲av成人精品一二三区| 美女午夜性视频免费| 亚洲人成77777在线视频| 久久久久国产一级毛片高清牌| 老司机在亚洲福利影院| 99国产精品免费福利视频| 操出白浆在线播放| 免费在线观看视频国产中文字幕亚洲 | 中文字幕高清在线视频| 在线观看人妻少妇| 香蕉国产在线看| 中文字幕高清在线视频| 亚洲精品国产av成人精品| 国产在线一区二区三区精| 成年美女黄网站色视频大全免费| 黑人欧美特级aaaaaa片| 精品亚洲成国产av| 十八禁人妻一区二区| av国产精品久久久久影院| 国产乱人偷精品视频| 久久99一区二区三区| 国产精品一区二区在线不卡| 十八禁高潮呻吟视频| 人成视频在线观看免费观看| 我的亚洲天堂| 亚洲国产看品久久| 亚洲欧洲国产日韩| 日韩 亚洲 欧美在线| 在线精品无人区一区二区三| av网站在线播放免费| 国产乱来视频区| 国产一区二区激情短视频 | 久久久精品国产亚洲av高清涩受| av电影中文网址| 成年动漫av网址| 中文欧美无线码| 丰满少妇做爰视频| 亚洲国产精品999| 久久这里只有精品19| 老司机影院成人| 久久毛片免费看一区二区三区| 亚洲熟女精品中文字幕| 欧美日韩精品网址| 最近中文字幕2019免费版| 欧美另类一区| 丁香六月天网| 国产精品一国产av| a 毛片基地| 亚洲在久久综合| 色婷婷av一区二区三区视频| 热re99久久精品国产66热6| 90打野战视频偷拍视频| 制服人妻中文乱码| 伊人久久大香线蕉亚洲五| 国产女主播在线喷水免费视频网站| 人人澡人人妻人| 女人高潮潮喷娇喘18禁视频| 久久99精品国语久久久| 超碰成人久久| 成人亚洲精品一区在线观看| 曰老女人黄片| 啦啦啦 在线观看视频| 国产一区亚洲一区在线观看| 青春草国产在线视频| 波野结衣二区三区在线| 久久午夜综合久久蜜桃| 一级爰片在线观看| 久久久久精品久久久久真实原创| 青春草国产在线视频| 成人亚洲欧美一区二区av| 男女边吃奶边做爰视频| 晚上一个人看的免费电影| 精品酒店卫生间| 午夜激情久久久久久久| 免费黄网站久久成人精品| 亚洲一码二码三码区别大吗| 99久久99久久久精品蜜桃| 亚洲人成网站在线观看播放| 中文字幕色久视频| 国产爽快片一区二区三区| 永久免费av网站大全| 天美传媒精品一区二区| 国产又爽黄色视频| 又大又黄又爽视频免费| 在线免费观看不下载黄p国产| 看免费成人av毛片| 国产亚洲最大av| 嫩草影视91久久| 老鸭窝网址在线观看| 高清不卡的av网站| 久久天堂一区二区三区四区| xxx大片免费视频| 亚洲av综合色区一区| 国产熟女欧美一区二区| 久久精品久久精品一区二区三区| 少妇被粗大的猛进出69影院| 久久久久视频综合| 亚洲精品中文字幕在线视频| 1024视频免费在线观看| 国产精品熟女久久久久浪| 捣出白浆h1v1| av天堂久久9| 日日啪夜夜爽| 男的添女的下面高潮视频| 欧美日韩福利视频一区二区| 男男h啪啪无遮挡| 成人三级做爰电影| 免费观看a级毛片全部| 欧美精品av麻豆av| 日韩中文字幕欧美一区二区 | 日韩电影二区| 亚洲人成电影观看| 热re99久久国产66热| 少妇精品久久久久久久| 久久天躁狠狠躁夜夜2o2o | 一本大道久久a久久精品| 精品视频人人做人人爽| 久久性视频一级片| 热re99久久国产66热| 精品国产一区二区三区四区第35| 婷婷色麻豆天堂久久| 国产激情久久老熟女| 狂野欧美激情性bbbbbb| 欧美激情极品国产一区二区三区| 十分钟在线观看高清视频www| 老司机深夜福利视频在线观看 | 99久久综合免费| 熟妇人妻不卡中文字幕| 久久女婷五月综合色啪小说| 99热网站在线观看| 亚洲欧美清纯卡通| tube8黄色片| 国产免费福利视频在线观看| 777米奇影视久久| 午夜激情av网站| kizo精华| 亚洲人成电影观看| 色视频在线一区二区三区| 欧美日韩亚洲高清精品| 国产一区二区三区综合在线观看| 看非洲黑人一级黄片| 嫩草影院入口| 欧美精品高潮呻吟av久久| 国产成人精品久久二区二区91 | 黄网站色视频无遮挡免费观看| a级片在线免费高清观看视频| 国产极品天堂在线| 久久久久久久久免费视频了| 蜜桃国产av成人99| 黄色一级大片看看| 亚洲av成人精品一二三区| 一本久久精品| h视频一区二区三区| 午夜av观看不卡| 久久精品国产a三级三级三级| 乱人伦中国视频| 天天添夜夜摸| 欧美国产精品一级二级三级| 免费黄色在线免费观看| 不卡视频在线观看欧美| 久久久精品国产亚洲av高清涩受| 老司机亚洲免费影院| 国产成人午夜福利电影在线观看| 亚洲天堂av无毛| 亚洲欧美一区二区三区久久| 国产淫语在线视频| 国产精品偷伦视频观看了| 在线观看免费高清a一片| 欧美精品高潮呻吟av久久| 亚洲激情五月婷婷啪啪| 校园人妻丝袜中文字幕| 老熟女久久久| 亚洲精品aⅴ在线观看| 亚洲情色 制服丝袜| 亚洲av成人不卡在线观看播放网 | 国产精品久久久久久久久免| 色网站视频免费| 欧美国产精品一级二级三级| 成年av动漫网址| 亚洲精品自拍成人| 在线观看免费高清a一片| 亚洲国产看品久久| 免费人妻精品一区二区三区视频| 国产av国产精品国产| 天天躁狠狠躁夜夜躁狠狠躁| a级毛片在线看网站| av在线播放精品| 欧美精品高潮呻吟av久久| 只有这里有精品99| 国产在视频线精品| 久久毛片免费看一区二区三区| 久久久精品国产亚洲av高清涩受| 中文乱码字字幕精品一区二区三区| 深夜精品福利| 国产97色在线日韩免费| 97精品久久久久久久久久精品| 五月开心婷婷网| 国产成人a∨麻豆精品| 久久婷婷青草| 亚洲av成人精品一二三区| 老司机亚洲免费影院| 国产毛片在线视频| 亚洲av男天堂| 丝袜脚勾引网站| 成年人免费黄色播放视频| 又粗又硬又长又爽又黄的视频| 中文字幕亚洲精品专区| 亚洲一码二码三码区别大吗| 大片免费播放器 马上看| 熟女av电影| 色综合欧美亚洲国产小说| 国产深夜福利视频在线观看| 18在线观看网站| 2021少妇久久久久久久久久久| 国产午夜精品一二区理论片| 汤姆久久久久久久影院中文字幕| 午夜日韩欧美国产| 久久天堂一区二区三区四区| av片东京热男人的天堂| 韩国高清视频一区二区三区| 国产熟女欧美一区二区| 青春草亚洲视频在线观看| 晚上一个人看的免费电影| 日韩av在线免费看完整版不卡| 亚洲婷婷狠狠爱综合网| 久久精品aⅴ一区二区三区四区| 国产精品偷伦视频观看了| 美女福利国产在线| 精品一区二区三卡| 最黄视频免费看| 最近的中文字幕免费完整| 国产成人精品在线电影| 久久久久精品国产欧美久久久 | 国产一区二区三区综合在线观看| 国产男人的电影天堂91| 久久久久久久大尺度免费视频| 久久精品国产亚洲av高清一级| 国产精品一区二区在线不卡| 新久久久久国产一级毛片| 视频在线观看一区二区三区| 好男人视频免费观看在线| 国产一区二区三区综合在线观看| 精品久久久久久电影网| 欧美日韩av久久| 亚洲国产日韩一区二区| 在线观看三级黄色| 三上悠亚av全集在线观看| 亚洲国产精品一区二区三区在线| 免费观看性生交大片5| 久久婷婷青草| 午夜老司机福利片| 国产精品免费视频内射| 国产一区二区激情短视频 | 久久女婷五月综合色啪小说| 国产人伦9x9x在线观看| 波多野结衣av一区二区av| 美女大奶头黄色视频| av国产精品久久久久影院| 成人午夜精彩视频在线观看| 国产国语露脸激情在线看| 97人妻天天添夜夜摸| 日韩,欧美,国产一区二区三区| 多毛熟女@视频| 丁香六月天网| 婷婷色综合大香蕉| 国产精品一区二区精品视频观看| 丝袜美腿诱惑在线| 久久青草综合色| 国产午夜精品一二区理论片| 日韩中文字幕视频在线看片| 国产精品蜜桃在线观看| 日韩人妻精品一区2区三区| 一本久久精品| 精品酒店卫生间| 亚洲情色 制服丝袜| 一区福利在线观看| 午夜老司机福利片| 久久久精品区二区三区| 永久免费av网站大全| 一级毛片我不卡| 久久精品国产亚洲av涩爱| 夫妻午夜视频| 国产亚洲av片在线观看秒播厂| 亚洲一卡2卡3卡4卡5卡精品中文| 国产片特级美女逼逼视频| 欧美日韩综合久久久久久| 国产视频首页在线观看| 男女无遮挡免费网站观看| 国产精品秋霞免费鲁丝片| 性色av一级| 久久99精品国语久久久| 纯流量卡能插随身wifi吗| 亚洲欧洲国产日韩| 少妇精品久久久久久久| 免费久久久久久久精品成人欧美视频| 久久亚洲国产成人精品v| 久久人人爽av亚洲精品天堂| 大香蕉久久网| 美国免费a级毛片| 高清黄色对白视频在线免费看| 亚洲四区av| 美女大奶头黄色视频| 麻豆乱淫一区二区| 操出白浆在线播放| 亚洲国产毛片av蜜桃av| www.熟女人妻精品国产| 不卡视频在线观看欧美| 欧美日韩国产mv在线观看视频| 精品免费久久久久久久清纯 | 久久久精品免费免费高清| 久久久久精品国产欧美久久久 | 女的被弄到高潮叫床怎么办| 99精品久久久久人妻精品| 嫩草影视91久久| 中国三级夫妇交换| 午夜福利一区二区在线看| 久久热在线av| 看十八女毛片水多多多| 国产精品熟女久久久久浪| 久久天躁狠狠躁夜夜2o2o | 日本黄色日本黄色录像| 日韩欧美精品免费久久| 国产成人精品久久二区二区91 | 成人毛片60女人毛片免费| 亚洲精品第二区| 成人黄色视频免费在线看| 亚洲av成人不卡在线观看播放网 | 一边亲一边摸免费视频| 久久精品aⅴ一区二区三区四区| 在线观看国产h片| 午夜福利,免费看| 卡戴珊不雅视频在线播放| 欧美激情高清一区二区三区 | 别揉我奶头~嗯~啊~动态视频 | 久久久久久久大尺度免费视频| 少妇人妻久久综合中文| 亚洲精品久久成人aⅴ小说| 搡老乐熟女国产| 岛国毛片在线播放| 久久精品亚洲熟妇少妇任你| 超色免费av| 久久久久久人妻| av又黄又爽大尺度在线免费看| 热99久久久久精品小说推荐| 久久 成人 亚洲| 免费观看av网站的网址| 国产精品国产三级国产专区5o| 99精品久久久久人妻精品| 久久久国产欧美日韩av| 日本猛色少妇xxxxx猛交久久| 高清视频免费观看一区二区| 午夜91福利影院| 欧美精品人与动牲交sv欧美| 婷婷色av中文字幕| 性高湖久久久久久久久免费观看| 日本欧美国产在线视频| 久久精品亚洲av国产电影网| 日韩中文字幕欧美一区二区 | 精品国产一区二区久久| 国产爽快片一区二区三区| av又黄又爽大尺度在线免费看| 久久天堂一区二区三区四区| 热re99久久国产66热| 亚洲一码二码三码区别大吗| 亚洲av综合色区一区| 国产成人91sexporn| 久久精品熟女亚洲av麻豆精品| 两个人看的免费小视频| 国产成人精品在线电影| 国产精品99久久99久久久不卡 | 少妇精品久久久久久久| 青春草亚洲视频在线观看| 在线精品无人区一区二区三| 无限看片的www在线观看| 久久天躁狠狠躁夜夜2o2o | 欧美最新免费一区二区三区| av女优亚洲男人天堂| 国产 一区精品| 精品亚洲乱码少妇综合久久| 久久久精品94久久精品| www.av在线官网国产| 中文字幕精品免费在线观看视频| 一区二区三区精品91| 晚上一个人看的免费电影| 最近最新中文字幕免费大全7| 国产av一区二区精品久久| 国产成人免费无遮挡视频| av.在线天堂| 国产精品免费视频内射| 久久这里只有精品19| 两个人免费观看高清视频| av在线播放精品| 国产福利在线免费观看视频| 女人被躁到高潮嗷嗷叫费观| 啦啦啦在线观看免费高清www| 91精品三级在线观看| 99久国产av精品国产电影| 中文欧美无线码| tube8黄色片| 天堂8中文在线网| 一级毛片黄色毛片免费观看视频| 久久精品久久久久久久性| 亚洲av男天堂| 亚洲男人天堂网一区| 中文字幕精品免费在线观看视频| 如日韩欧美国产精品一区二区三区| 国产野战对白在线观看| 久久综合国产亚洲精品| 51午夜福利影视在线观看| 激情视频va一区二区三区| 国产精品亚洲av一区麻豆 | 99久久99久久久精品蜜桃| 麻豆av在线久日| 日韩制服丝袜自拍偷拍| 国产视频首页在线观看| 国产精品av久久久久免费| 亚洲精品成人av观看孕妇| 日韩制服骚丝袜av| 蜜桃在线观看..| 9191精品国产免费久久| 国产黄色免费在线视频| 国产野战对白在线观看| 国产成人av激情在线播放| 中文字幕人妻丝袜制服| 热re99久久精品国产66热6| 国产一区亚洲一区在线观看| 成年人免费黄色播放视频| 美女国产高潮福利片在线看| 中文字幕人妻熟女乱码| 亚洲国产看品久久| 91精品伊人久久大香线蕉| 日韩一本色道免费dvd| 国产在线一区二区三区精| 青春草亚洲视频在线观看| 免费女性裸体啪啪无遮挡网站| 国产成人欧美| 亚洲欧洲日产国产| 国产男女超爽视频在线观看| 性色av一级| 自拍欧美九色日韩亚洲蝌蚪91| 岛国毛片在线播放| 免费黄色在线免费观看| 最新在线观看一区二区三区 | 精品国产一区二区三区久久久樱花| 国产精品国产三级国产专区5o| 久久久久视频综合| 国产在线免费精品| 亚洲精品成人av观看孕妇| 成人国产av品久久久| 国产亚洲午夜精品一区二区久久| 人人妻人人澡人人爽人人夜夜| 国产乱来视频区| 精品午夜福利在线看| 欧美日本中文国产一区发布| 亚洲欧美中文字幕日韩二区| 菩萨蛮人人尽说江南好唐韦庄| 亚洲,欧美精品.| 国产精品嫩草影院av在线观看| 视频区图区小说| 久久亚洲国产成人精品v| 在现免费观看毛片| 亚洲av成人精品一二三区| 亚洲,欧美精品.| 在线 av 中文字幕| 精品一区二区三区四区五区乱码 | 国产精品av久久久久免费| 婷婷成人精品国产| 黑人猛操日本美女一级片| 在线天堂最新版资源| 欧美成人精品欧美一级黄| 男女国产视频网站| 黄频高清免费视频| 成年av动漫网址| 成人漫画全彩无遮挡| 久久精品国产亚洲av高清一级| 嫩草影视91久久| 久久久久久久久免费视频了| 亚洲av欧美aⅴ国产| 国产亚洲精品第一综合不卡| 亚洲 欧美一区二区三区| 麻豆av在线久日| 免费少妇av软件| 精品卡一卡二卡四卡免费| 国产精品一国产av| 赤兔流量卡办理| 热99国产精品久久久久久7| 一区二区av电影网|