CN109934031A - The method and system of differentiation LOS/NLOS based on RFID system - Google Patents

Abstract

A kind of method and system of the differentiation LOS/NLOS based on RFID system of invention, firstly, devise the matrix that can accurately distinguish the path LOS and NLOS by using the joint variance range of phase and RSS.Second, propose the effective method based on cluster for reducing the negative effect of phase ambiguity on accuracy of identification.Third, only using several channels rather than whole channel and only go to calculate LOS/NLOS Path Recognition with a small number of readings, can be significantly reduced in this way and be delayed and do not lose precision.Precision of method of the invention is high, and real-time is good.

Description

The method and system of differentiation LOS/NLOS based on RFID system

Technical field

Present invention is mainly applied to internet of things field, RFID technique is related to positioning/activity recognition application field, especially A kind of method and system of the differentiation LOS/NLOS based on RFID system.

Background technique

Internet of Things (IOT) applied the fast development in many fields, such as smart city and wisdom factory in recent years [1].As one of most important support technology of Internet of Things, radio frequency identification (RFID) receives the concern much studied in recent years [2]-[10].One typical RFID system is made of a set of RFID reader and a large amount of labels.Reader plan passes through wireless Communication reads data from label and writes data into label.Compared with the traditional technologies such as bar code or two dimensional code, RFID can be Operation in bigger communication range, and more effective solution can be provided for object identification and positioning.Currently, RFID skill Art is widely applied in warehousing management, logistics, intelligent library, the fields such as retailer.Expect the market RFID in 2027 To be more than 20,000,000,000 dollars [11]

Recently, the positioning based on RFID and activity recognition have caused the concern [2] [5] [6] [12]-of many researchs [16], airdrome luggage can be sorted [5] by typical case, [17], and the books [2] of dislocation are found in intelligent library, [13], behavior [18] of the customer in offline shopping is excavated, [19] assess the quality [14] of wound resuscitation process, [20] children With toy [21].These research work by measurement receive signal feature (such as phase and received signal strength (RSS)) come Infer position or the active state of target object.In general, these methods need to measure the signal characteristic under the conditions of sight (LOS) with Realize high-precision.For example, calculate the location algorithm [2] of label position for using phase measurement, [5], [13], reader and Any barrier between label will make the phase value of measurement differ greatly with theoretical value, and lead to big position error.

How non line of sight (NLOS) path is identified to mitigate its negative effect to positioning and activity detection.In actual deployment RFID system in, NLOS signal propagate it is more more common than LOS.In view of intelligent library, bookshelf and mobile personnel may Prevent the signal between reader and label.Indoors in environment, the factors such as multipath and reflection also result in NLOS signal biography It is defeated.If the signal characteristic measured in the case of non line of sight is blindly handled, because they are measured in LOS, it Will lead to it is serious positioning or activity recognition accuracy reduce [2].In order to realize higher positioning accuracy, it would be desirable to will LOS transmission is distinguished with NLOS transmission, and before they are sent into positioning or activity recognition algorithm, is deleted corresponding NLOS reads or corrects to it.

Although having there is many works to be suggested and classified to different the barrer types and size (such as TagScan [22]), but their working principle is that barrier has existed and cannot identify the path LOS/NLOS.In addition, for identification Including Wi-Fi [23], [24], bluetooth [25], cellular tower [26], including [27] and ultra wide band (UWB) based on wireless positioning The method in the path LOS/NLOS of technology) [28] cannot be directly used to RFID system.They rely on fine-grained physical layer information The channel state information of Wi-Fi (for example, be used for) distinguishes the path LOS and NLOS, this requires high RST bandwidth and have high energy The complicated receiver of power.For example, the bandwidth of each channel in Wi-Fi is 20MHz, the bandwidth of UWB is even higher.With their phases Than in RFID system, signal bandwidth is very limited (bandwidth in each channel is only 250 KHz), and receiver label The ability at end is very weak.

Explanation of nouns used in the present invention is as follows:

Sighting distance: referring between reader and label can be mutually in the distance of " seeing " without barrier, and signal is from transmission Hold a kind of mode to receiving end.

Non line of sight: after the signal of transmitting terminal encounters barrier in communication process to be hindered, the propagation path of signal can be sent out Raw to change, signal can reach receiving end by the circulation ways such as reflecting or reflecting.

Summary of the invention

Positioning and activity recognition based on RFID are having dependent on the precise measurement of signal characteristic (for example, phase and RSS) Position or the active state of calculating target object are gone under conditions of sighting distance (LOS).However, in real RFID development system LOS may be interfered often by the barrier between reader and label.The result of its non line of sight (NLOS) signal can be significantly Reduce the accuracy of positioning and activity recognition.It is to guarantee accuracy right and wrong that NLOS how is filtered off in positioning/activity recognition procedure It is often important.The present invention is intended to provide a kind of method and system of the differentiation LOS/NLOS based on RFID system, guarantee identification essence While true property, delay is reduced.

In order to solve the above technical problems, the technical scheme adopted by the invention is that: a kind of differentiation based on RFID system The method of LOS/NLOS, comprising the following steps:

1) reader sends and obtains the phase and RSS information in each channel；

2) using the combination of the phase perhaps RSS information or the phase and the RSS information, identification LOS with NLOS situation.

In step 1), the detailed process for obtaining phase includes:

1) the phase value P={ P1, P2...Pi } in each channel is obtained；

2) P is sent into K mean cluster algorithm, obtains two subsets P1 and P2, calculated the mass center of P1 and P2, be denoted as respectively c1,c2；If | P1 |≤| P2 |, and | | c1-c2 |-π | < 1, then it is determined as phase ambiguity；If the mass center of P1 is less than the mass center of P2, Add π on all readings in P1；Otherwise, if the mass center of P1 is greater than the mass center of P2, π is subtracted from all readings in P1；When | P1 |≤| P2 |, and | | c1-c2 | -2 π | < 0.5, then it is determined as that phase is interrupted；If institute of the mass center of P1 close to 0, in P1 Have and adds 2 π on reading；Otherwise, if the mass center of P1 subtracts 2 π from all readings in P1 close to 2 π；

3) updated P1 and P2 is phase.

The number of channels be 3, on each channel collect data when it is 2 seconds a length of, 2 seconds are to reduce and prolong The time designed late, when channel quantity increases to 3 from 1, accuracy of identification is increased rapidly to 0.97 from 0.93.But work as letter When road number is greater than 4, using more channels, its accuracy of identification improves to be not obvious.In fact, when the number of channel is greater than 8, precision It is almost stable.Therefore, three channels can be randomly selected in we, and collect data 2 seconds on each channel, and use three The data of selected channel execute LOS/NLOS Path Recognition, and delay can be reduced to about 6 seconds by this from about 31 seconds.

The specific implementation process of step 2) includes:

For k-th of channel, RSS variance are as follows:

Wherein N_KIt is the quantity that RSS is read in k-th of channel,It is reading of i-th of RSS channel in this channel,It is the average value that RSS is read in k-th of channel, in the variance VS for calculating all channels_kAfterwards, the average RSS in all channels Variance is；

Wherein M is channel number；After calculating VS, VS is compared with threshold value THRs predetermined；If VS < THRs, then the path is considered as LOS path；Otherwise, which is considered as the path NLOS；Wherein,

The specific implementation process of step 2) includes:

For k-th of channel, the variance of phase place reading are as follows:

WhereinIt is i-th of phase place reading in k-th of channel,It is the average phase value in k-th of channel, The average phase variation in all channels is:

It is LOS by path report if VP < threshold value THRp；It otherwise, is NLOS by path report.

The specific implementation process of step 2) are as follows: judge VPS_fitWhether threshold value is less than, if being then LOS by path report；It is no It then, is NLOS by path report；

Wherein VPS_fitIt is the new module that the present invention defines, refers to phase and the joint variation model of RSS It encloses,Indicate 95% confidence interval of the received signal strength of k-th of channel,Indicate the 95% confidence interval of the phase of k channel.

Correspondingly, the system for distinguishing LOS/NLOS based on RFID system that the present invention also provides a kind of comprising:

Reader, for sending and obtaining the phase and RSS information in each channel；

Identification module, for the knot using the phase perhaps RSS information or the phase and the RSS information It closes, identifies LOS and NLOS situation.

The identification module using the RSS information identification LOS and NLOS situation specific implementation process include:

For k-th of channel, RSS variance are as follows:

Wherein N_KIt is the quantity that RSS is read in k-th of channel,It is reading of i-th of RSS channel in this channel,It is the average value that RSS is read in k-th of channel, in the variance VS for calculating all channels_kAfterwards, the average RSS in all channels Variance are as follows:

Wherein M is channel number；After calculating VS, VS is compared with threshold value THRs predetermined；If VS < THRs, then the path is considered as LOS path；Otherwise, which is considered as the path NLOS.

The identification module includes: using the specific implementation process of the phase identification LOS and NLOS situation

For k-th of channel, the variance of phase place reading are as follows:

WhereinIt is i-th of phase place reading in k-th of channel,It is the average phase value in k-th of channel, The average phase variation in all channels is:

Wherein M is channel number；It is LOS by path report if VP < threshold value THRp；Otherwise, it is by path report NLOS。

The identification module identifies the specific reality of LOS and NLOS situation using the combination of the phase and the RSS information Existing process includes: to judge VPS_fitWhether threshold value is less than, if being then LOS by path report；It otherwise, is NLOS by path report；

Wherein M is channel number；VPS_fitIt is phase and the joint variation range of RSS,Indicate k-th of letter 95% confidence interval of the received signal strength in road,Indicate 95% confidence area of the phase of k-th of channel Between.

Compared with prior art, the advantageous effect of present invention is that: by the present invention in that with the connection of phase and RSS It closes variance range and devises the matrix that can accurately distinguish the path LOS and NLOS, propose that a kind of effective reduce is identifying essence The method based on cluster of the negative effect of phase ambiguity on degree only enumerates several channels rather than whole channel and is only read with minority Reading goes to calculate LOS/NLOS Path Recognition, can be significantly reduced delay in this way and not lose precision；The present invention is the results show that originally The method of invention realizes high-precision in all test cases and recall rate, precision are up to 0.969, and recall rate is up to 0.991. In addition, our method can also distinguish between different types of barrier, accuracy is up to 0.93.

Detailed description of the invention

Fig. 1 is the communication of reader and passive label；

Fig. 2 is the phase measured in LOS and NLOS and RSS image；(a) phase change when LOS； (b)NLOS When phase change；(c) RSS variation when LOS；(d) RSS variation when NLOS；

Fig. 3 is accuracy of identification and F-score based on RSS-T method；(a) precision；(b)F-score；

Fig. 4 is accuracy of identification and F-score based on Phase-T method；(a) precision；(b)F-score；

Fig. 5 is to use VPS_maxAnd VPS_fitThe accuracy of identification and F-score of method；(a) VPS is used_maxPrecision；(b) Use VPS_maxF-score；(c) VPS is used_fitPrecision；(d) VPS is used_fitF-score；.

Specific embodiment

(1) communication between reader and label (LOS)

In RFID system, the communication between reader and label uses reader's session mode of priority.When reader is wanted When being communicated with label, it issues continuous waveform signal first, and there are two purposes: command/data is sent to label, and Energy is provided for label its data back is scattered to reader.(passive) label is collected energy and is adjusted from continuous wave signal Wave processed is to transfer data to reader.This process is as shown in Figure 1.

When signal propagation is LOS, i.e., when there is no barrier between reader and label, receive the feature of signal mainly by The influence of reader-tag distance.The distance between reader and label are indicated with d.The phase for reaching the signal of reader can be with It is modeled as [5]

Wherein λ is the wavelength of signal, and θ div is the diversity item as caused by the hardware deficiency of reader and label.Currently, Many COTS RFID readers as Impinj R420 can be with phase 2 π/4096 ≈ of resolution ratio report received signal 0.0015 radian [5].Similarly, the RSS of the reception signal on reader can be modeled as [31]

Wherein Prx is the RSS at reader, and Ptx is the transmission power of reader, and Gt and Gr are label-side and reading respectively The antenna gain of device side, Tb indicate the backscattering coefficient of label, and value is in [0,1].

(2) communication between reader and label (NLOS)

In an experiment, we place the label of 2 meters of distance before antenna, and measure reader in both cases The phase and RSS:LOS (not having any content between reader and label) and NLOS of upper signal (have a people to stand in reader Between label).For each case, we collect about 31 seconds data, each channel about 2 seconds.It note that both In the case of every other setting be all identical.

Experimental result (as shown in Figure 2) shows the phase in the case of NLOS and the survey in the case of RSS measurement result and LOS Amount result is very different.In LOS, for most of channels, phase value between 0 radian and 1.8 radians, and In the case of NLOS, the phase value of measurement is between 4 radians and 6 radians.Equally, RSS the measured value (- 64dBm in the case of NLOS To -59dBm), also (- 46dBm to -44dBm) is very different with LOS situation.Obviously, this in the case of non line of sight can not neglect Deviation slightly will lead to significant position error.It is above-mentioned the experimental results showed that, we should determine NLOS transmission and reduce NLOS In the case of reading, to guarantee the accuracy of position or activity recognition.

(3) identification of basic LOS/NLOS

We devise method (RSS-T) of two kinds of basic skills i.e. based on RSS threshold value and the method based on phase threshold (phase-T)) identify LOS and NLOS situation: the phase and RSS label of our measuring signals for a period of time, then calculates all The average value of phase and RSS on channel 1.Then we are using determining that signal transmission path is LOS based on the method for threshold value Or NLOS.Details as Follows.

A) method (RSS-T) based on RSS threshold value: in this approach, we calculate the flat of the RSS of all channels first Then mean square deviation judges that corresponding path is LOS or NLOS using threshold value T HRs.If average RSS variance is less than THR, Path report is LOS by we.Otherwise, path report is NLOS by we

For k-th of channel, we calculate its RSS variance and are

Wherein Nk is the quantity that RSS is read in k-th of channel,It is reading of i-th of RSS channel in this channel Number, RSSk are the average value that RSS is read in k-th of channel.After the V Sk for calculating all channels, we calculate all channels Average RSS variance be

Wherein M is channel number.After calculating V S, it is compared by we with threshold value THRs predetermined.If VS < THR, then the path is considered as LOS path；Otherwise, it is considered as the path NLOS.

(a) of Fig. 3 depicts precision (the true LOS/NLOS feelings in the case of the LOS/NLOS of all reports of this method The ratio of condition).It is obvious that accuracy of identification is influenced by THR.For best T HR, the balance quality of LOS and NLOS situation is about 0.74.We also depict the F- score of RSS-T in (b) of Fig. 3 (measurement considers precision and recall rate).It can see It observes, the highest F- score in the case of NLOS is about 0.81, and the highest F- score in the case of LOS is only 0.77.

B) method based on phase threshold (Phase-T): similarly, we distinguish LOS and NLOS using phase difference Path.For k-th of channel, the variance of phase place reading is

Similar with RSS-T, we distinguish LOS using threshold value T HRp and wherein PSEk i is i-th in k-th of channel A phase place reading, and PSEk is the average phase value in k-th of channel.The average phase in all channels changes

The path NLOS: if VP < THRp, path report is LOS by we；Otherwise, we are reported as NLOS.

(a) of Fig. 4 and (b) of Fig. 4 depict the precision of F-score and Phase-T respectively.RSS- is compared in Phase-T execution T is slightly worse.It is about 0.72 to the balance quality of LOS and NLOS.The highest F- score of NLOS case is about 0.74, and LOS case Score be only 0.71.

(4) phase ambiguity and discontinuous algorithm are handled

Although the recognition methods using basis probably can identify LOS/NLOS, the experimental results showed that identification is other Precision only has 0.72, and resolution is not very high.We in phase place reading it can be observed from fig. 2 that observe two kinds of inhomogeneities " exceptional value " of type, reason is phase signal, and there are phase ambiguities and phase to be interrupted, so that the difference between LOS and NLOS Become more difficult.

In order to reduce our identification difficulty, we have proposed a kind of methods based on cluster, can effectively mitigate phase Position obscures and the discontinuous negative effect of phase.For each channel, we will by setting K=2 using K mean cluster algorithm Reading is divided into two groups.If (phase is discontinuous by the distance between mass center of two clusters about π (phase ambiguity) or 2 π Property), we shift one group read another group it is as follows.

Phase ambiguity: if the mass center of two results cluster is different from each other, drift is about π, it is understood that there are phases Fuzziness reading.In this case, the reading in groupuscule is converted to jumpbogroup by us.It is assumed that phase place reading group by P=P1, P2...Pi } it indicates, wherein i is the reading quantity obtained in the i of channel.Then P is sent into K mean cluster algorithm by us, and is obtained Obtain two subsets P1 and P2.Assuming that P1 is less than P2, i.e., | P1 |≤| P2 |.If the mass center of P1 is less than the mass center of P2, we are in P1 In all readings on plus π.Otherwise, if the mass center of P1 is greater than the mass center of P2, we subtract π from all readings in P1. Then the phase variance of channel i is calculated using all new values.

Discontinuous for phase: if the mass center of two results cluster differs about 2 π each other, there are phases in channel Discontinuous reading.It note that phase discontinuously only occurs in a channel.Equally, we are indicated in this channel with P Total indicator reading, and indicate with P1 and P2 the cluster of result, | P1 |≤| P2 |.If the mass center of P1 is close to 0, our institutes in P1 Have and adds 2 π on reading.Otherwise, if the mass center of P1 is close to 2 π, we subtract 2 π from all readings in P1.After reading conversion, We calculate the variance of phase using all phase place readings.

Algorithm idea:

1. obtaining the phase value P={ P1, P2...Pi } in each channel first；

2. P is sent into K mean cluster algorithm, and obtain two subsets P1 and P2, it is assumed that P1 is less than P2, i.e., | P1 |≤| P2 |；

3. the mass center for calculating P1 and P2 is respectively c1 and c2；

4. if | | c1-c2 |-π | it is phase ambiguity if < 1；

5. if the mass center of P1 is less than the mass center of P2, on our all readings in P1 plus π.Otherwise, if the matter of P1 The heart is greater than the mass center of P2, we subtract π from all readings in P1；

6. if | | c1-c2 | -2 π | it is phase interruption if < 0.5；

7. if the mass center of P1 is close to 0, on our all readings in P1 plus 2 π.Otherwise, if the mass center of P1 is close 2 π, we subtract 2 π from all readings in P1；

8. last, we are read using new phase.

(5) precision is improved using series of features

It solves phase ambiguity and phase interruption, improves the resolution of image, but its precision just only has 0.71- 0.74, in order to improve accuracy of identification, we have proposed a new module, it by RSS variance and phase combine, To distinguish the path LOS and NLOS.In fact, new measurement considers stage and the joint variation range of RSS, it is abbreviated as VPS.I Using two methods calculate channel V PS.

Maximum VPS: in the first method, we use the maximum variance range of RSS and the maximum variance range of phase It measures to calculate, and is indicated with VPSmax.For k-th of channel,

WhereinWithIt is the minimum and maximum RSS reading in k-th of channel, andWithIt is the minimum and maximum phase place reading in k-th of channel respectively.After the VPSmax for calculating all channels, we are calculated Its average value is

Suitable VPS: when there is some extreme readings, for LOS situation, VPS_maxMay be also very big, this will affect identification Precision.Our second method calculates measurement using the secrecy interval of RSS and phase place reading, by VPS_fitIt indicates.In detail Ground is said, for k-th of channel, is used95% confidence interval of all RSS readings, is used in combination95% confidence interval of all phase place readings.Then we calculate measurement and are

In the VPS for calculating all channels_fitAfterwards, we use their average value as measurement to distinguish LOS and NLOS Path:

After calculating new measurement, we classify to the path LOS and NLOS using the method based on threshold value, are similar to RSST and Phase-T.The precision and F-score of the two measurements are plotted in Fig. 5.We can observe that with RSST and Phase-T is compared, and new measurement substantially increases accuracy of identification.Use VPS_maxAnd VPS_fitLOS and NLOS balance quality point Not Wei 0.83 and 0.84, much higher than be used only RSS and phase when 0.74 and 0.72.VPS_fitCompare VPS_maxIt is slightly good, because it can Effectively to mitigate the influence extremely read once in a while.When use VPS_maxWhen, the most high F value of LOS and NLOS are 0.86 and 0.88, Use VPS_fitWhen be 0.93 and 0.91, be higher by 10% or more than RSS-T and Phase-T.

By the way that accuracy of identification can be further enhanced using the more features as caused by frequency hopping.In fact, in addition to phase becomes Change, RSS variance, except maximum VPS and fitting VPS, the path LOS and NLOS 2 can also be distinguished using following three features:

1) mean size at the 95% percentile secrecy interval of phase place reading.The 95 percentiles secrecy interval of phase place reading Size is different in LOS with NLOS.We calculate the 95 percentiles secrecy interval of each channel first, then calculate them Mean size on all channels.

2) par of phase ambiguity reading.In general, in non line of sight, phase ambiguity read-around ratio is being regarded More frequently occur away from the case of.We determine whether each channel has fuzziness reading first, as described in Section III-part A, Then the average of the fuzziness reading in all channels is calculated.

3) variance read in a channel.In fact, in LOS, reading rate (i.e. readings per second quantity) phase To stabilization, and in non line of sight, reading rate is widely different.The computer that we calculate each channel first reads speed Then rate uses their average value as measurement to distinguish the path LOS/NLOS.

We have collected one group of test data comprising 2330 LOS cases and 2480 NLOS cases, these data packets Different distance and different types of label containing different settings, including different barriers, between reader and label.I Calculate all above-mentioned 7 features of each case, and be supplied to 4 representational machine learning algorithms, i.e., RandomForest, Bagging, andomCommittee and KStar, and assess using 10 times of cross validations their property Energy.Table 1 gives result.We are it can be found that precision substantially increases 14% compared with the best approach for using V PSf it (0.98vs 0.84)！We have also observed that RandomForest in most cases behaves oneself best all, therefore it is used as Default categories device in remaining experiment.

(6) identification delay is reduced

Our method has had reached 0.98 accuracy of identification, but the time for calculating result but needs 31 seconds, is Achieve the effect that real-time and do not lose precision, we have proposed following methods:

A simple idea for reducing identification delay is only in a channel using measuring, rather than all 16 LOS/NLOS Path Recognition is executed in a channel.Table 1 is depicted using the knowledge in the case of LOS and NLOS when single channel data Other precision.It is observed that using the precision in single channel between 0.90 and 0.93, hence it is evident that lower than using from all The precision of the data in channel.In fact, different channels is similar in the ability distinguished on the path LOS and NLOS, and There is no the optimum channel suitable for both of these case under different scenes.

The rate of precision and recall rate of 1 different classifications device of table

However, it has been found that if we are randomly choosed several channels and are executed LOS/ NLOS using their combination Path Recognition, then obtained precision is by the close precision using all data, but significant reduction postpones.In table 2, when making With K randomly selected channels come when executing LOS/NLOS Path Recognition, we depict the identification essence in the case of LOS/NLOS Degree.As can be seen that accuracy of identification is increased rapidly to 0.97 from 0.93 when channel quantity increases to 3 from 1.But work as the number of channel When greater than 4, accuracy of identification can only be slightly improved using more channels.In fact, precision is almost stable when port number is greater than 8. Therefore, three channels can be randomly selected in we, and collect data 2 seconds on each channel, and use three selected channels Data execute LOS/NLOS Path Recognition.Delay can be reduced to about 6 seconds by this from about 30 seconds.

LOS/NLOS differentiation rate when 2 difference number of samples of table

In fact, we can be further reduced identification delay by reducing the reading in each channel.It is every in order to understand How the quantity read in a channel influences accuracy of identification, our randomly stochastical samplings one from 3 randomly selected channels Group number-reading, and be used to execute LOS/NLOS Path Recognition.As a result it is listed in table 2.

According to table 2 provide as a result, it is observed that when the reading quantity in each channel be greater than 20 when, accuracy of identification is several It is unaffected.Reason be used in our method institute it is functional be all statistical metric criterion, and work as sample size When higher than threshold value, these modules will not be greatly affected.Moreover, for the measurement V PSf of most identification, Because we calculate it using the size at secrecy interval rather than absolute maximum/minimum value, it is possible to inhibit extreme reading Influence.When reading is less than 20, precision is declined slightly.But even if reading times are less than 5, each channel, accuracy of identification Still higher than 0.95.This needs to collect 3*5=15 reading.Since reading rate is about 40 readings per second, this corresponds to Identification delay in about 15=40=0:375 seconds.Below, if clearly not specified, we are using 3 channels and each 5, channel reading is to execute LOS/NLOS Path Recognition.

We have proposed the first LOS/NLOS recognition methods of RFID system.COTS RFID device the experimental results showed that, Our method realizes high-precision in all test cases and recall rate, precision are up to 0.969, and recall rate is up to 0.991.In addition, our method can also distinguish between different types of barrier, accuracy is up to 0.93.The algorithm can be realized Very high accuracy of identification postponed less than 0.4 second.The technology can be used for all positioning based on RFID and behavior identity system, this A little systems need LOS path signal to propagate to realize high-precision.

Bibliography:

[1]Luigi Atzori,Antonio Iera,and Giacomo Morabito.The internet of things:A survey.Computer networks,54(15):2787–2805,2010.

[2]Longfei Shangguan and Kyle Jamieson.The design and implementation of a mobile rfid tag sorting robot.In Proceedings of the 14th Annual International Conference on Mobile Systems,Applications,and Services,pages 31–42.ACM, 2016.

[3]Lei Xie,Yafeng Yin,Athanasios V.Vasilakos,and Sanglu Lu.Managing RFID data:Challenges,opportunities and solutions.IEEE Communications Surveys and Tutorials,16(3):1294–1311,2014.

[4]Qingjun Xiao,Bin Xiao,Shigang Chen,and Jiming Chen.Collisionaware churn estimation in large-scale dynamic rfid systems.IEEE/ACM Transactions on Networking(TON),25(1):392–405,2017.

[5]Lei Yang,Yekui Chen,Xiang-Yang Li,Chaowei Xiao,Mo Li,and Yunhao Liu.Tagoram:Real-time tracking of mobile rfid tags to high precision using cots devices.In Proceedings of the 20th annual international conference on Mobile computing and networking,pages 237–248.ACM,2014.

[6]Jia Liu,Min Chen,Shigang Chen,Qingfeng Pan,and Lijun Chen. Tag- compass:Determining the spatial direction of an object with small dimensions.In Proc.of IEEE INFOCOM,pages 1944–1952,2017.

[7]Jiajue Ou,Mo Li,and Yuanqing Zheng.Come and be served:Parallel decoding for cots RFID tags.In Proceedings of the 21st Annual International Conference on Mobile Computing and Networking(MobiCom),pages 500–511. ACM, 2015.

[8]Shigeng Zhang,Xuan Liu,Jianxin Wang,and Jiannong Cao.Tag Size Profi ling in Multiple Reader RFID Systems.In Proceedings of the 36th Annual IEEE International Conference on Computer Communications(Infocom),pages 1953– 1961,2017.

[9]Xiulong Liu,Bin Xiao,Keqiu Li,Alex X Liu,Jie Wu,Xin Xie,and Heng Qi. Rfid estimation with blocker tags.IEEE/ACM transactions on networking, 25 (1):224–237,2017.

[10]Yuxiao Hou,Yanwen Wang,and Yuanqing Zheng.Tagbreathe:Monitor breathing with commodity RFID systems.In Proceedings of 37th IEEE International Conference on Distributed Computing Systems(ICDCS),pages 404– 413.IEEE, 2017.

[11]Raghu Das.RFID Forecasts,Players and Opportunities 2017-2027, 2017. http://www.idtechex.com/research/reports/r fi d-forecasts-playersand- opportunities-2017-2027-000546.asp.

[12] Jue Wang and Dina Katabi.Dude, where ' s my card?: Rfid positioning that works with multipath and non-line of sight.ACM SIGCOMM Computer Communication Review,43(4):51–62,2013.

[13]Longfei Shangguan,Zheng Yang,Alex X Liu,Zimu Zhou,and Yunhao Liu. Relative localization of rfid tags using spatial-temporal phase profiling.In NSDI, pages 251–263,2015.

[14]Liang Wang,Tao Gu,Xianping Tao,and Jian Lu.Toward a wearable rfid system for real-time activity recognition using radio patterns.IEEE Transactions on Mobile Computing,16(1):228–242,2017.

[15]Jue Wang,Deepak Vasisht,and Dina Katabi.Rf-idraw:virtual touch screen in the air using rf signals.ACM SIGCOMM Computer Communication Review, 44(4):235–246,2014.

[16]Jeffrey Hightower,Roy Want,and Gaetano Borriello.Spoton:An indoor 3d location sensing technology based on rf signal strength.UW CSE 00-02-02, University of Washington,Department of Computer Science and Engineering, Seattle, WA,1,2000.

[17]Tianci Liu,Lei Yang,Qiongzheng Lin,Yi Guo,and Yunhao Liu. Anchor- free backscatter positioning for rfid tags with high accuracy.In INFOCOM, 2014 Proceedings IEEE,pages 379–387.IEEE,2014.

[18]Longfei Shangguan,Zimu Zhou,Xiaolong Zheng,Lei Yang,Yunhao Liu, and Jinsong Han.Shopminer:Mining customer shopping behavior in physical clothing stores with cots rfid devices.In Proceedings of the 13th ACM Conference on Embedded Networked Sensor Systems,pages 113–125.ACM,2015.

[19]Jinsong Han,Han Ding,Chen Qian,Wei Xi,Zhi Wang,Zhiping Jiang, Longfei Shangguan,and Jizhong Zhao.Cbid:A customer behavior identification system using passive tags.IEEE/ACM Transactions on Networking,24(5):2885– 2898,2016.

[20]Siddika Parlak,Ivan Marsic,Aleksandra Sarcevic,Waheed U Bajwa, Lauren J Waterhouse,and Randall S Burd.Passive rfid for object and use detection during trauma resuscitation.IEEE Transactions on Mobile Computing, 15(4):924–937, 2016.

[21]Hanchuan Li,Can Ye,and Alanson P Sample.Idsense:Ahuman object interaction detection system based on passive uhf rfid.In Proceedings of the 33rd Annual ACM Conference on Human Factors in Computing Systems,pages 2555– 2564.ACM,2015.

[22]Ju Wang,Jie Xiong,Xiaojiang Chen,Hongbo Jiang,Rajesh Krishna Balan, and Dingyi Fang.Tagscan:Simultaneous target imaging and material identification with commodity RFID devices.In Proceedings of the 23rd Annual International Conference on Mobile Computing and Networking(MobiCom),pages 288–300, 2017.

[23]Chenshu Wu,Zheng Yang,Zimu Zhou,Kun Qian,Yunhao Liu,and Mingyan Liu.Phaseu:Real-time los identifi cation with wifi.In Computer Communications (INFOCOM),2015 IEEE Conference on,pages 2038–2046.IEEE, 2015.

[24]Zimu Zhou,Zheng Yang,Chenshu Wu,Longfei Shangguan,Haibin Cai, Yunhao Liu,and Lionel M Ni.Wifi-based indoor line-of-sight identification.IEEE Transactions on Wireless Communications,14(11):6125–6136, 2015.

[25]Patrick Lazik,Niranjini Rajagopal,Oliver Shih,Bruno Sinopoli,and Anthony Rowe.Alps:A bluetooth and ultrasound platform for mapping and localization.In Proceedings of the 13th ACM Conference on Embedded Networked Sensor Systems,pages 73–84.ACM,2015.

[26]Marilynn P Wylie and Jack Holtzman.The non-line of sight problem in mobile location estimation.In Universal Personal Communications, 1996.Record., 1996 5th IEEE International Conference on,volume 2,pages 827– 831.IEEE,1996.

[27]Xin Wang,Zongxin Wang,and Bob O’Dea.A toa-based location algorit hmreducingtheerrorsduetonon-line-of-sight(nlos)propagation.IEEE Transactions on Vehicular Technology,52(1):112–116,2003.

[28]Lorenzo Mucchi and Patrizio Marcocci.A new parameter for uwb indoor channel profile identification.IEEE Transactions on Wireless Communications,8(4), 2009.

[29]XinyuLi,YanyiZhang,IvanMarsic,AleksandraSarcevic,andRandallS Burd. Deep learning for rfid-based activity recognition.In Proceedings of the 14th ACM Conference on Embedded Network Sensor Systems CD-ROM,pages 164–175. ACM,2016.

[30]Chuyu Wang,Lei Xie,Wei Wang,Tao Xue,and Sanglu Lu.Moving tag detection via physical layer analysis for large-scale rfid systems.In Computer Communications,IEEE INFOCOM 2016-The 35th Annual IEEE International Conference on,pages 1–9.IEEE,2016.

[31]Lei Yang,Yi Guo,Tianci Liu,Cheng Wang,and Yunhao Liu.Perceiving the slightest tag motion beyond localization.IEEE Transactions on Mobile Computing, 14(11):2363–2375,2015.

[32]Chunhui Duan,Lei Yang,and Yunhao Liu.Accurate spatial calibration of rfi d antennas via spinning tags.In Proceedings of 36th IEEE International Conference on Distributed Computing Systems(ICDCS),pages 519–528.IEEE,2016.

[33]https://www.impinj.com/platform/connectivity/speedway-r420/.

[34]Guangliang Liu,Shigeng Zhang,Jianxin Wang,and Xuan Liu.Arpat: Accurate rfid reader positioning with mere boundary tags.In Parallel and Distributed Systems(ICPADS),2014 20th IEEE International Conference on,pages 880–885. IEEE,2014.

[35]Lionel M Ni,Yunhao Liu,Yiu Cho Lau,and Abhishek P Patil.Landmarc: indoor location sensing using active RFID.Wireless Networks,10(6):701–710, 2004.

[36]Jullawadee Maneesilp,Chong Wang,Hongyi Wu,and Nian-Feng Tzeng.Rfi d support for accurate 3d localization.IEEE Transactions on Computers, 62(7): 1447–1459,2013.

[37]Kai Bu,Xuan Liu,Jiwei Li,and Bin Xiao.Less is more:Efficient rfidbased 3d localization.In Mobile Ad-Hoc and Sensor Systems(MASS),2013 IEEE 10th International Conference on,pages 86–94.IEEE,2013.

[38]Cory Hekimian-Williams,Brandon Grant,Xiuwen Liu,Zhenghao Zhang, and Piyush Kumar.Accurate localization of rfid tags using phase difference.In RFID, 2010 IEEE International Conference on,pages 89–96.IEEE,2010。

Claims (10)

1. a kind of method of the differentiation LOS/NLOS based on RFID system, which comprises the following steps:

1) reader sends and obtains the phase and RSS information in each channel；

2) combination of the phase perhaps RSS information or the phase and the RSS information, identification LOS and NLOS are utilized Situation.

2. the method for the differentiation LOS/NLOS according to claim 1 based on RFID system, which is characterized in that step 1) In, the detailed process for obtaining phase includes:

1) the phase value P={ P1, P2 ... Pi } in each channel is obtained；

2) P is sent into K mean cluster algorithm, obtains two subsets P1 and P2, calculated the mass center of P1 and P2, be denoted as c1, c2 respectively； If | P1 |≤| P2 |, and | | c1-c2 |-π | < 1 is then determined as phase ambiguity；If the mass center of P1 is less than the mass center of P2, in P1 In all readings on plus π；Otherwise, if the mass center of P1 is greater than the mass center of P2, π is subtracted from all readings in P1；When | P1 | ≤ | P2 |, and | | c1-c2 | -2 π | < 0.5 is then determined as that phase is interrupted；If the mass center of P1 is close to 0, all in P1 Add 2 π on reading；Otherwise, if the mass center of P1 subtracts 2 π from all readings in P1 close to 2 π；

3) updated P1 and P2 is phase.

3. the method for the differentiation LOS/NLOS according to claim 1 based on RFID system, which is characterized in that the channel Quantity be 3, on each channel collect data when it is 2 seconds a length of.

4. the method for the differentiation LOS/NLOS according to claim 1 based on RFID system, which is characterized in that step 2) In, the specific implementation process using RSS information identification LOS and NLOS situation includes:

For k-th of channel, RSS variance are as follows:

Wherein N_KIt is the quantity that RSS is read in k-th of channel,It is reading of i-th of RSS channel in this channel, It is the average value that RSS is read in k-th of channel, in the variance VS for calculating all channels_kAfterwards, the average RSS variance in all channels Are as follows:

Wherein M is channel number；After calculating VS, VS is compared with threshold value THRs predetermined；If VS < THRs, The path is considered as LOS path；Otherwise, which is considered as the path NLOS.

5. the method for the differentiation LOS/NLOS according to claim 1 based on RFID system, which is characterized in that step 2) In, the specific implementation process using the phase identification LOS and NLOS situation includes:

For k-th of channel, the variance of phase place reading are as follows:

WhereinIt is i-th of phase place reading in k-th of channel,It is the average phase value in k-th of channel, owns The average phase variation in channel is:

Wherein M is channel number；It is LOS by path report if VP < threshold value THRp；

It otherwise, is NLOS by path report.

6. the method for the differentiation LOS/NLOS according to claim 1 based on RFID system, which is characterized in that step 2) In, the specific implementation process using the combination identification LOS and NLOS situation of the phase and the RSS information includes: judgement VPS_fitWhether threshold value is less than, if being then LOS by path report；It otherwise, is NLOS by path report；

Wherein M is channel number；VPS_fitIt is phase and the joint variation range of RSS,Indicate connecing for k-th of channel 95% confidence interval of signal strength is received,Indicate 95% confidence interval of the phase of k-th of channel.

7. a kind of system of the differentiation LOS/NLOS based on RFID system characterized by comprising

Reader, for sending and obtaining the phase and RSS information in each channel；

Identification module, for the combination using the phase perhaps RSS information or the phase and the RSS information, knowledge Other LOS and NLOS situation.

8. the system of the differentiation LOS/NLOS according to claim 7 based on RFID system, which is characterized in that the identification Module using the RSS information identification LOS and NLOS situation specific implementation process include:

For k-th of channel, RSS variance are as follows:

Wherein N_KIt is the quantity that RSS is read in k-th of channel,It is reading of i-th of RSS channel in this channel, It is the average value that RSS is read in k-th of channel, in the variance VS for calculating all channels_kAfterwards, the average RSS variance in all channels Are as follows:

Wherein M is channel number；After calculating VS, VS is compared with threshold value THRs predetermined；If VS < THRs, The path is considered as LOS path；Otherwise, which is considered as the path NLOS.

9. the system of the differentiation LOS/NLOS according to claim 7 based on RFID system, which is characterized in that

The identification module includes: using the specific implementation process of the phase identification LOS and NLOS situation

For k-th of channel, the variance of phase place reading are as follows:

WhereinIt is i-th of phase place reading in k-th of channel,It is the average phase value in k-th of channel, owns The average phase variation in channel is:

Wherein M is channel number；It is LOS by path report if VP < threshold value THRp；It otherwise, is NLOS by path report.

10. the system of the differentiation LOS/NLOS according to claim 7 based on RFID system, which is characterized in that the knowledge Other module includes: judgement using the specific implementation process of the combination identification LOS and NLOS situation of the phase and the RSS information VPS_fitWhether threshold value is less than, if being then LOS by path report；It otherwise, is NLOS by path report；

Wherein M is channel number；VPS_fitIt is phase and the joint variation range of RSS,Indicate connecing for k-th of channel 95% confidence interval of signal strength is received,Indicate 95% confidence interval of the phase of k-th of channel.