CN101620661B - High-efficiency label counting method in radio frequency recognition system - Google Patents

High-efficiency label counting method in radio frequency recognition system Download PDF

Info

Publication number
CN101620661B
CN101620661B CN2008100292061A CN200810029206A CN101620661B CN 101620661 B CN101620661 B CN 101620661B CN 2008100292061 A CN2008100292061 A CN 2008100292061A CN 200810029206 A CN200810029206 A CN 200810029206A CN 101620661 B CN101620661 B CN 101620661B
Authority
CN
China
Prior art keywords
label
subclass
time
layout
estimation
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN2008100292061A
Other languages
Chinese (zh)
Other versions
CN101620661A (en
Inventor
倪明选
钱辰
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Guangzhou HKUST Fok Ying Tung Research Institute
Original Assignee
Guangzhou HKUST Fok Ying Tung Research Institute
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Guangzhou HKUST Fok Ying Tung Research Institute filed Critical Guangzhou HKUST Fok Ying Tung Research Institute
Priority to CN2008100292061A priority Critical patent/CN101620661B/en
Publication of CN101620661A publication Critical patent/CN101620661A/en
Application granted granted Critical
Publication of CN101620661B publication Critical patent/CN101620661B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Landscapes

  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)
  • Mobile Radio Communication Systems (AREA)

Abstract

The invention relates to a high-efficiency label counting method in a radio frequency recognition system, which solves the problem that the efficiency of calculating the label number in the prior radio frequency recognition system is low. Because of the existence of label signal conflict, the process of 'using a reader-writer to send requests and receiving label replies for calculating the label number is infeasible. The method comprises the following steps: firstly, using a rapid estimation method named as LoF to master the approximate number of a tally set required to be calculated; secondly dividing the tally set into a plurality of subsets (as shown by an attached drawing) through a divide and conquer named as multi-dimensional split; and finally, calculating the number of the labels in the subsets respectively through a PAC accurate counting method, and totaling the numbers to obtain the final result. Compared with the prior method for calculating the number of the labels, the high-efficiency label counting method has the advantages of short processing time, energy consumption conservation, protection on the privacy of a label carrier and the like.

Description

Label counting method efficiently in the radio frequency identification system
Technical field
The number of labels that the present invention relates in the radio frequency identification system is calculated, and belongs to radio RF recognition technology, general fit calculation and wireless communication field.
Background technology
(Radio Frequency Identification, RFID) technology is a kind of contactless automatic identification technology to radio frequency identification.The rfid system of standard mainly is made up of label (Tag) and read write line (Reader), also needs the support of other software and hardwares in actual applications.The RFID label is entrained by relative article, human or animal usually.Read write line comes the data of storing in the acquisition request label by sending radio frequency signal; Label receives after this type of signal, and the product information that relies on energy that induction current obtained to send to be stored in the label chip perhaps relies on power supply initiatively to send a signal to read write line to read write line.The label of each RFID all has an identification number (ID) and some other data, and the data in the label can provide identification to article, the human or animal who carries it.The RFID label can be used in the environment of the mobile object that includes magnanimity owing to its easy manufacture craft and cheap cost.
Along with the development of technology, rfid system is widely used in intelligent space and the general calculation entironment, as aspects such as supply chain management, industrial manufacturing industry, forwarding business, personal management and safety managements.In this type of was used, the total number of labels amount of calculating in certain zone was a very important task for rfid system.For example in an intelligent indoor sport shop, spectators wear the T-shirt with the RFID label, and rfid system reaches the purpose of some people's current control or secure context by the quantity of grasping spectators.
The greatest problem that rfid system runs in communication is: if after read write line sends the request answer signal, receive that at one time a plurality of labels answer read write line, this moment, read write line just was beyond recognition out the information of these labels, also can not calculate number of tags.This situation is referred to as label signal conflict (TagCollision).Therefore, it is infeasible merely using " read write line sends and asks-receive label to be answered " this process to come computation tag quantity.In the research in the past, the researcher proposed the method that two classes solve the label signal conflict, and be respectively: the conflict of 1.ALOHA formula is avoided and 2. tree type conflicts are avoided.But these two class methods need the quite long processing time could correctly calculate number of tags.For example will discern 3000 labels, the time that two class methods need is all more than 20 minutes.They are inapplicable for a lot of application scenarioss, especially the object that moves for statistics---because mobile object has probably during this period of time left the zone of being added up, the result's is inaccurate thereby caused.
Summary of the invention
Technical matters: the present invention is directed to existing radio frequency identification system computation tag number needs and wanted processing time of a specified duration and energy and use also problem than poor efficiency, propose a kind of in radio frequency identification system the method for computation tag quantity efficiently, make that radio frequency identification system can be on the basis that consumes extremely short time and few energy, calculate the number of labels in the appointed area, and can be applicable to plurality of application scenes.
Technical scheme: the present invention proposes a kind of method of the quantity of computation tag efficiently.This method mainly is divided into two steps: first step is to estimate number of labels fast; Second step is accurate counting.In the practical application, must be accurate if do not require the result, rfid system can only adopt first step, thereby obtains the number of labels result of an estimation in the extremely short time.If the system requirements result must be very accurate, then after carrying out the first step end, carry out the calculating of second step again.Accurately counting is compared with estimation and can be consumed more time and energy, but compares with existing counting technology, and efficient is greatly enhanced.
Estimate the detailed process of (LoF) fast:
We are being had and are only having a rfid interrogator by in the zone of adding up at hypothesis, and the communication range of this read write line can cover whole zone.The label of known each RFID all has an identification number (ID), and ID is a binary number of L position.Therefore, the total quantity of label can not be more than n=2 LIndividual.
At first, rfid interrogator radio request signal.And the answer that begins to monitor label.The method of monitoring is for using the ALOHA time frame: layout when an ALOHA time frame comprises L (the mark preface is 1,2 ..., L), the time layout time length be (as 0.1 second) fixed.Label receives after the request, when needing to select L in the layout one, and answer in this time.In at lattice sometime, read write line is only received the answer of a label, and then read write line can identify the information that label sends; If read write line has been received the message that a plurality of labels are beamed back at lattice in the time, label collision will take place, and what is the need breath is not taken the post as in read write line identification, only can know conflict has taken place.
In the present invention, the RFID label adopts following method to come the select time lattice to answer: each label all uses same geometric distributions hash function H.The character of geometric distributions hash function is: if the overall situation has 2 LIndividual different ID (identity number) then has 2 L-1Layout when individual meeting is mapped to the 1st; Have 2 L-2Layout when individual meeting is mapped to the 2nd; Have 2 L-tLayout when individual meeting is mapped to t.The ID that supposes certain label is i, and H (i)=t, and layout was answered read write line when then this label was chosen in t.As shown in Figure 1.
So read write line create one long be the table of bits of L, and the answer of monitoring label in the ALOHA frame under connecing.Not answering if the some time layout has label, is 0 with bit position corresponding in the table of bits then; If there is 1 to answer or clash, be 1 then with the corresponding bit position.After monitoring end, read write line (or the coupled main frame) value of finding is 1 the highest bit, supposes that it is the R position.The total quantity of label is that available following formula is estimated:
The estimated value of the total quantity of label is n ~ = 1.2897 × 2 R - 1
Figure DEST_PATH_GSB00000399534300012
Be the estimated value of number of labels.Its theoretical foundation can be referring to [C.Qian, H-L.Ngan, Y.Liu, " Cardinality Estimation for Large-scale RFID Systems ", in proceedingof IEEE PerCom, 2008.].
We are with this method called after LoF (Lottery Frame) estimation algorithm.
It should be noted that: there is no small error in this kind method of estimation.Only do not need an estimated value in certain error range if do not proceed next step accurate counting, the present invention can adopt and repeatedly repeat estimation approach and can significantly reduce error: be about to above process and repeat to carry out m time, all adopt different hash functions at every turn.As last estimated value, error will reduce m with m time average 1/2Doubly.If repeat 32 times, 15% error then only arranged.
Method of estimation of the present invention is compared with existing method of estimation, is reduced to original about 10% computing time.
Accurately count the detailed process of (PAC):
Among the present invention, accurately counting (PAC, Precise and Anonymous Counting) is made up of two parts, is respectively multidimensional division and counting.
The multidimensional division is (divide-and-conquer) process of dividing and ruling.In the estimation of previous step, read write line and main frame have not only obtained the estimated result of a number of labels, and the label total collection in also should the zone has been divided into the plurality of sub set.As: the label of answering read write line the 1st time in the layout has been formed subclass S 1 The label of answering read write line in the layout in the time of t has been formed subclass S tMain frame also can be estimated the number of tags that each subclass comprises by following relational expression:
Subclass S iInclude approximately n ~ i = 1.2897 × 2 R - i Individual label.
If
Figure S2008100292061D00042
Greater than a threshold value (we are L in this fixed this threshold value, i.e. the figure place of label ID (identity number)), then subclass S iNeed divide once more.With the method that is similar to quick estimation: read write line sends request signal, allows layout answer when all labels of layout are selected in ALOHA frame next time when answering i in the estimation fast for the first time.All the other labels are being kept silent in the ALOHA frame next time.So S iContinuation is split into several littler subclass.This time division is referred to as the two dimension division.If still there is subclass, then proceed three-dimensional division, four-dimensional division greater than L ... till the subclass that does not exist greater than L.As shown in Figure 2.
If the quantity survey (surveying) value of a certain subclass
Figure S2008100292061D00051
Less than threshold value L, then read write line and main frame are accurately counted the label in this subclass.When accurately counting, read write line generates and contains
Figure S2008100292061D00052
The ALOHA frame of layout when individual, and require subclass S iIn label when selecting one at random layout answer.If only there is a label to answer in the some time layout, counter is from increasing one; If conflict has taken place the some time layout, then adopt existing tree type conflict solution.
Tree type conflict solution process is as follows: if conflict has taken place the some time layout, layout when then read write line continues to emit two at the label in these lattice allows the label of conflict select an answer at random in the two.If conflict is still arranged, layout when again each conflict lattice being emitted two again.Till this process knows that all time layouts all do not conflict.The logical organization of Sheng Chenging is the one tree shape at last.For all time layouts that an answer is only arranged, counter adds one one by one.As shown in Figure 3.
Last counter has just obtained the number of labels result of ultimate demand with the results added of each piece gained.
Among the present invention, estimate again earlier for label and accurately to count, than directly concerning all labels directly carry out tree type counting, because we have grasped the estimated information of number of labels in advance, therefore will significantly reduce the possibility that conflict takes place, thereby accelerate the speed of counting, and reduce the energy that consumes.
In the accurate counting method of existing tree type, read write line requires each label all to disclose its ID, and this uses for some, exist potential safety hazard, for example the assailant installs the read write line of a malice, can easily obtain label ID by tree type counting method, and then the tracking tags carrier.And, being in the consideration of privacy, a lot of label carrier do not want to disclose the ID of its label yet.Therefore anonymous agreement has bigger using value in radio frequency system.In addition, because the error in the production run has some label to have the ID identical with other labels probably, this label is called as redundant copy.When having redundant copy, just can't use traditional tree type conflict solution, because same ID can continue to cause conflict.
In order to solve above problem; the present invention uses following mode to protect label carrier's privacy also to detect the existence of redundant copy simultaneously: in manufacture process; on the label except that with its ID; also with MD5 (ID); BF1 (MD5 (ID)), BF2 (MD5 (ID)) and 4 numerical value of BF3 (MD5 (ID)).Wherein MD5 (ID) carries out a result after the mapping of MD5 hash function to ID; BF1 (MD5 (ID)), BF2 (MD5 (ID)) and BF3 (MD5 (ID)) are the mapping result of MD5 (ID) being carried out again three Bloom filtrators (Bloom Filter).In accurate counting process, label is beamed back three the Bloom Filter mapping value of oneself to read write line, rather than the ID of oneself.If read write line finds that three Bloom Filter mapping value of certain several label are identical, then can suspect among these labels has redundant copy, and requires they reports MD5 (ID) value separately.If MD5 (ID) value is still identical, then judging among these labels has redundant copy.In whole process, label is not revealed the ID value of oneself, so the accurate counting method of PAC is a kind of method of counting of anonymity.
Need to prove that realization of the present invention need be carried out transformation on the hardware to label, as write hash function etc. in label.Therefore can't realize method of counting of the present invention with existing label, need in producing the label process, carry out special manufacturing.
Beneficial effect:
Label counting method in the radio frequency identification system that the present invention proposes is compared with the method for existing computation tag quantity, has the following advantages:
● the processing time is short.We compare existing UPE label evaluation method and LOF estimation algorithm; And the accurate counting method of existing tree type compared with the accurate counting method of PAC.Through theoretic analysis and emulation testing, if the ID of label is 16, the processing time can be reduced to original 50%.If the ID of label is 64, the processing time can be reduced to original 18.75%.
● energy efficient.We compare the accurate counting method of existing tree type with the accurate counting method of PAC.Through theoretic analysis and emulation testing, if the number of tags that is counted has 1,000, the processing time can be reduced to original 30%.If the number of tags that is counted has 10,000, the processing time can be reduced to original about 10%.
● the accurate counting method of PAC is a kind of method of counting of anonymity, and bigger using value is arranged in rfid system.
● the accurate counting method of PAC can detect the existence of redundant copy.
Description of drawings
Fig. 1 is the process synoptic diagram of estimating fast.
Fig. 2 is a multidimensional fission process synoptic diagram in accurately counting.
Fig. 3 is a tree type conflict solution process synoptic diagram in accurately counting.
Embodiment
In the actual motion of agreement, the process of estimation is example with Fig. 1 fast, at first, and the rfid interrogator radio request signal.And the answer that begins to monitor label.Layout when the ALOHA time frame of monitoring comprises 7.Label receives after the request that adopt following method to come the select time lattice to answer: each label all uses same geometric distributions hash function H.The character of H is: if the overall situation has 2 LIndividual different ID (identity number) then has 2 L-1Layout when individual meeting is mapped to the 1st; Have 2 L-2Layout when individual meeting is mapped to the 2nd; Have 2 L-tLayout when individual meeting is mapped to t.The ID that supposes certain label is i, and H (i)=t, and layout was answered read write line when then this label was chosen in t.
After monitoring end, read write line (or the coupled main frame) value of finding is 1 the highest bit, promptly the 4th.The total quantity of label is that available following formula is estimated:
n ~ = 1.2897 × 2 R - 1 = 1.2897 × 8 = 10.3176
The layout when method of this quick estimation only needs 7, and the result is rather desirable.
The multidimensional division is (divide-and-conquer) process of dividing and ruling as shown in Figure 2.By Fast estimation, read write line and main frame have not only obtained the estimated result of a number of labels, and the label total collection in also should the zone has been divided into some subclass.As: the time layout 0,1,2 all include the set of a unnecessary label.Main frame can be estimated the number of tags that each subclass comprises by following relational expression:
n ~ i = 1.2897 × 2 R - i Individual label.
If
Figure S2008100292061D00082
Greater than threshold value L (as the time layout 0), subclass S then iNeed divide once more.Adopt the method that is similar to quick estimation again, this time division is referred to as the two dimension division.If the quantity survey (surveying) value of a certain subclass Less than threshold value L (as the time layout 1,2), then read write line and main frame are accurately counted the label in this subclass.As shown in Figure 3, when accurately counting, read write line generates and contains The ALOHA frame of layout when (being 4 among the figure) is individual, and require subclass S iIn label when selecting one at random layout answer.If only there is a label to answer in the some time layout, counter is from increasing one; If conflict has taken place the some time layout, then adopt tree type conflict solution (as the rightest lattice among the figure).
Form 1 provides the class C language false code of multidimensional splitting method.Form 2 provides the class C language false code of accurate counting.
Figure S2008100292061D00085
(a)Kd-Splitting?algorithm?on?tags (b)Kd-Splitting?algorithm?on?the?reader
Form 1
ALOGRITHM：Couting() PROCEDURE while TRUE wait_for_messages()； if there is a probe message from a reader randomly select a slot k from[0. l-1]； transmit a message in time slot k； end if end while ALOGRITHM：Couting(S) INPUT the tag set S PROCEDURE broadcast a request to tags； for i＝0 to l-1 wait_for_responses()； end for for i＝0 to l-1 if time solt i has a single response counter++； else if time solt i is collision TreeSplit(s i) end if end for
(a)counting?algorithm?on?tags (b)counting?algorithm?on?the?reader
Form 2
Last counter has just obtained the number of labels result of ultimate demand with the results added of each piece gained.

Claims (4)

1. label counting method efficiently in the radio frequency identification system, it is characterized in that: estimate fast that by two step: 1-number of labels and the accurate counting of 2-carry out number of labels and determine, adopt distinctive quick estimation LoF mode, send request by read write line during communication, use the ALOHA time frame that the time is divided, each label all uses same geometric distributions hash function H, result according to Hash maps selects the time layout of ALOHA frame to answer, read write line or main frame be according to the ALOHA time frame that listens to, according to the estimation equation of the total quantity of label
Figure FSB00000581587400011
Calculate the estimated value of total number of labels, wherein R is that value is 1 the highest bit, and by adopting different hash functions, with m time average as last estimated value, repeat estimation and reduce error, main frame estimates the roughly number of labels that each subclass comprises by quick estimation number of tags, and the label total collection in also should the zone has been divided into the plurality of sub set; With accurate counting PAC mode, implement the process of dividing and ruling, earlier according to subclass S iEstimation equation
Figure FSB00000581587400012
Estimate each subclass number of tags roughly fast, again by the continuous method that divides, after promptly cutting apart and calculating the quantity of each subclass, if the number of tags that comprises in certain subclass is greater than a threshold value, then this subclass need divide once more, splitting method is with estimation is similar fast, up to the quantity survey (surveying) value of a certain subclass less than threshold value, then read write line and main frame are counted the label in this subclass, during counting, read write line generates the ALOHA frame of layout when containing with identical of this subclass quantity survey (surveying) value, and layout is answered when requiring label in this subclass to select one at random; If only there is a label to answer in the some time layout, counter is from increasing one; If conflict has taken place in the some time layout, then adopt tree-like conflict solution counting, promptly for the some time layout of conflict, layout when read write line continues to emit two at the label in these lattice allows the label of conflict select an answer at random in the two; If conflict is still arranged, layout when again each conflict lattice being emitted two again; This process is not till all time layouts all conflict, and the logical organization of Sheng Chenging is the one tree shape at last, and for all time layouts that an answer is only arranged, counter adds one one by one, thereby obtains the set of quantitative value less than threshold value, and final addition obtains the result.
2. label counting method efficiently in the described radio frequency identification system of claim 1, it is characterized in that: employing repeatedly repeats estimation approach and can significantly reduce error: the process that is about to estimation fast repeats to carry out m time, all adopt different hash functions at every turn, with m time average as last estimated value.
3. label counting method efficiently in the described radio frequency identification system of claim 1 is characterized in that: when first step was estimated number of labels fast, main frame estimated the roughly number of labels that each subclass comprises by estimation number of tags fast.
4. label counting method efficiently in the described radio frequency identification system of claim 1, it is characterized in that: after cutting apart and calculating the quantity of each subclass, if the number of tags that comprises in certain subclass is greater than a threshold value, then this subclass need divide once more, splitting method is with estimation is similar fast, this time division is referred to as the two dimension division, if still there is the subclass greater than threshold value, then proceeds three-dimensional division, four-dimensional division ... till the subclass that does not exist greater than threshold value.
CN2008100292061A 2008-07-04 2008-07-04 High-efficiency label counting method in radio frequency recognition system Active CN101620661B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN2008100292061A CN101620661B (en) 2008-07-04 2008-07-04 High-efficiency label counting method in radio frequency recognition system

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN2008100292061A CN101620661B (en) 2008-07-04 2008-07-04 High-efficiency label counting method in radio frequency recognition system

Publications (2)

Publication Number Publication Date
CN101620661A CN101620661A (en) 2010-01-06
CN101620661B true CN101620661B (en) 2011-11-30

Family

ID=41513892

Family Applications (1)

Application Number Title Priority Date Filing Date
CN2008100292061A Active CN101620661B (en) 2008-07-04 2008-07-04 High-efficiency label counting method in radio frequency recognition system

Country Status (1)

Country Link
CN (1) CN101620661B (en)

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103544459B (en) * 2013-10-25 2016-09-28 上海交通大学 The evaluation method of a kind of RFID tag sum and system
CN103605946B (en) * 2013-12-03 2017-01-18 无锡儒安科技有限公司 Scanning method and scanning device of radio frequency tags
CN105095806B (en) * 2014-05-19 2017-10-10 郑静晨 RFID tag quantity survey method in Emergency Logistics management system
CN113836961B (en) * 2021-11-26 2022-02-11 广州优刻谷科技有限公司 Method and system for estimating quantity of passive ultrahigh frequency RFID tags

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1881271A (en) * 2005-06-17 2006-12-20 东芝泰格有限公司 Rfid tag recognition apparatus and method and article sales data processing apparatus
CN1975755A (en) * 2006-12-01 2007-06-06 华南理工大学 Intelligent RFID reading system anticonflict scheduling method
CN1983305A (en) * 2005-12-16 2007-06-20 王强 Method for automatically batch distinguishing non-contact electronic lable
CN101145181A (en) * 2006-10-10 2008-03-19 中兴通讯股份有限公司 Radio frequency identification multiple tag anti-collision method

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1881271A (en) * 2005-06-17 2006-12-20 东芝泰格有限公司 Rfid tag recognition apparatus and method and article sales data processing apparatus
CN1983305A (en) * 2005-12-16 2007-06-20 王强 Method for automatically batch distinguishing non-contact electronic lable
CN101145181A (en) * 2006-10-10 2008-03-19 中兴通讯股份有限公司 Radio frequency identification multiple tag anti-collision method
CN1975755A (en) * 2006-12-01 2007-06-06 华南理工大学 Intelligent RFID reading system anticonflict scheduling method

Also Published As

Publication number Publication date
CN101620661A (en) 2010-01-06

Similar Documents

Publication Publication Date Title
Zheng et al. PET: Probabilistic estimating tree for large-scale RFID estimation
CN101554015B (en) Anonymous tracking using a set of wireles devices
CN106384067B (en) A kind of RFID Unknown Label identifying system and recognition methods
CN101620661B (en) High-efficiency label counting method in radio frequency recognition system
CN106203212B (en) A kind of binary tree RFID anti-collision method based on dynamic frame slot
CN103473524B (en) RFID multi-tag identification method, reader and tags
CN103020569B (en) A kind of radio-frequency identification multi-label anti-collision method
Liu et al. Generic composite counting in RFID systems
CN107038398B (en) The loss label uncertainty parallel detecting method of anonymous multigroup radio-frequency recognition system
Liu et al. Fast counting the key tags in anonymous RFID systems
CN103679099A (en) Anti-collision method applied to mass-RFID-label system
CN104573593B (en) A kind of deficient based on frame slot determines blind separation RFID anti-collision method
Luo et al. An efficient protocol for RFID multigroup threshold-based classification based on sampling and logical bitmap
CN103218635B (en) A kind of RFID anti-collision method based on mixing multiway tree search
Qiao et al. RFID as an Infrastructure
CN104166867B (en) A kind of many HASH functions multiframe coupled mode RFID anti-collision algorithms (MHMFG)
Yan et al. BFSearch: Bloom filter based tag searching for large-scale RFID systems
Li et al. Towards time-efficient localized polling for large-scale RFID systems
CN104063627B (en) A kind of optimal Q determination method for parameter and device
Wang A Fast Hybrid Strategy‐Based RFID Tag Identification Protocol
Bonuccelli et al. A very fast tags polling protocol for single and multiple readers RFID systems, and its applications
CN105718968A (en) RFID adaptive N tree anti-collision method based on bitmap construction
CN105205511B (en) A kind of binary system splitting method eliminated based on free timeslot
Xie et al. A tag-correlation-based approach to fast identification of group tags
CN110852129B (en) Dynamic label estimation method

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
ASS Succession or assignment of patent right

Owner name: HKUST FOK YING TUNG GRADUATE SCHOOL

Free format text: FORMER OWNER: GUANGZHOU HKUST R+D CORPORATION LTD.

Effective date: 20100806

C41 Transfer of patent application or patent right or utility model
COR Change of bibliographic data

Free format text: CORRECT: ADDRESS; FROM: 511458 302A, SOUTH ZONE, SOFTWARE BUILDING, INFORMATION SCIENCE AND TECHNOLOGY PARK, NANSHA DISTRICT, GUANGZHOU CITY, GUANGDONG PROVINCE TO: 511458 ROOM N301, SOFTWARE BUILDING, INFORMATION SCIENCE AND TECHNOLOGY PARK, NANSHA, GUANGZHOU CITY

TA01 Transfer of patent application right

Effective date of registration: 20100806

Address after: 511458 N301 room, software building, Nansha Guangzhou Information Technology Park

Applicant after: HKUST Fok Ying Tung Graduate School

Address before: 511458 302A, South Building, Nansha District Information Technology Park, Guangdong, Guangzhou

Applicant before: Guangzhou HKUST R&D Corporation Ltd.

C14 Grant of patent or utility model
GR01 Patent grant