RFID tag identification method based on radio frequency fingerprints
By extracting the fingerprint features of RFID tags and establishing fingerprint libraries, and using machine learning to quickly match, the problem of slow RFID tag recognition speed in the prior art is solved, and the recognition efficiency and accuracy are improved.
Patent Information
- Application Number
- CN202411133079.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-19
- Publication Date
- 2024-12-06
AI Technical Summary
The existing RFID tag recognition methods are slow to identify, especially when multiple tags respond simultaneously, channel blocking and data collisions are prone to occur, resulting in some tags not being read correctly.
By obtaining the complete timing information of RFID tags, fingerprint features are extracted and fingerprint libraries are established, and similarity is calculated using machine learning to quickly match, reducing channel occupation and improving identification efficiency.
It realizes rapid identification of RFID tags, reduces the frequency of channel blockage, and improves the speed and accuracy of tag identification.
Smart Images

Figure CN119089917A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of radio frequency identification technology, and in particular to a method for identifying an RFID tag based on radio frequency fingerprint. Background Art
[0002] In the typical RFID (Radio Frequency Identification) tag recognition process, the reader sends a radio frequency signal to the tag through the antenna. After the tag receives the signal, it activates the internal circuit according to the energy provided by the received signal. First, the data it carries is modulated and then sent to the outside. A total of two rounds of information need to be sent, the first round sends RN16 (16-bit random number) data, and the second round sends EPC (Electronic Product Code) data. After the reader receives this data, it performs decoding operations according to the corresponding rules and finally feeds back the results to the user.
[0003] In a general RFID system, wireless communication is used between the RFID reader and the RFID electronic tag. Since the channel used for data information transmission between the reader and the tag is the same, when multiple tags respond at the same time, more than one data signal will be transmitted in the channel, which may cause channel blockage, resulting in data collision, causing some tags to not be read correctly. The most commonly used solution at present is to use the ALOHA algorithm to reduce collisions.
[0004] However, in the existing method, the tag recognition method used for tag recognition has a slow recognition speed. Summary of the invention
[0005] The purpose of the present invention is to provide a method for RFID tag identification based on radio frequency fingerprint, aiming to solve the technical problem of slow identification speed of tag identification methods used in the prior art for tag identification.
[0006] To achieve the above object, the present invention adopts a method for RFID tag identification based on radio frequency fingerprint, comprising:
[0007] Obtain complete timing information of all RFID tags that need to be matched;
[0008] Extract fingerprint features of known tags;
[0009] Using fingerprint features to build a fingerprint database;
[0010] Read in the tag to be matched;
[0011] Get the fingerprint information of the read tag;
[0012] Perform fingerprint matching;
[0013] If a matching tag is found, the matching tag information is returned, if not found, empty is returned.
[0014] The specific method of extracting the fingerprint features of known tags is as follows:
[0015] For the tag S needed to build the fingerprint library i (i=0,1,2…n), find its RN16 data segment and EPC data segment;
[0016] Then split it into an array of 0 and 1 code elements in the time domain; the specific steps are as follows:
[0017] First, the data needs to be normalized:
[0018] α 0 (n) = α 0 (n)-μ α0
[0019] a 1 (n) = a 1 (n)-μ a1 ;
[0020] where μ α is the average amplitude, a 0 is an array of code element 0, a 1 is an array of code element 0;
[0021] After normalization, the average amplitude μ is re-averaged α As one of the fingerprint features;
[0022] The variance σ^2, skewness coefficient γ, kurtosis coefficient κ, and difference h of code element 0 and 1 are calculated respectively. The calculation formula is as follows:
[0023] Variance σ 2 :
[0024]
[0025] Skewness coefficient γ:
[0026]
[0027] Kurtosis coefficient κ:
[0028]
[0029] Difference h:
[0030] h 0 =max(a 0 )-min(a 0 )
[0031] h 1 =max(a 1 )-min(a 1 ).
[0032] The specific method of establishing a fingerprint database is as follows:
[0033] Collect data through a collection device and extract the interest interval of the collected data;
[0034] The acquired signal is further processed and split into a set of 0 and 1 code elements;
[0035] Perform feature extraction on each set of code elements;
[0036] Through machine learning, the feature weights required to establish the fingerprint library are trained;
[0037] Read in known tags, calculate the fingerprint data of each tag, and store it in the fingerprint library.
[0038] The specific method of fingerprint matching is as follows:
[0039] Read in data;
[0040] If the data can be read completely, the fingerprint feature is calculated for the whole data, put into the feature library for matching, and the one with the highest similarity is selected as the confirmed label. If the similarity is lower than the set threshold, it is considered that the library does not contain the label and the label is an unknown label or a cloned label.
[0041] If the data cannot be read completely, the fingerprint feature is calculated for each code element of the valid data. The feature library returns the five most similar tags each time, and finally takes the tag with the highest similarity as the confirmed tag. If the similarity is lower than the set threshold, it is considered that the library does not contain the tag and the tag is an unknown tag or a cloned tag.
[0042] The formula for calculating label similarity is as follows:
[0043]
[0044] where x 1 …x 10 Represents the weight of the corresponding feature on the result. The specific value is obtained through machine learning.
[0045] Among them, the feature extraction for each set of code elements should include: average amplitude, variance, maximum and minimum difference, kurtosis coefficient, and skewness coefficient.
[0046] Among them, the specific method of training the feature weights required to establish the fingerprint library through machine learning is as follows:
[0047] Extract the features of RFID data and build a data set.
[0048] Set 5 learnable parameters, namely, the average amplitude, variance, maximum and minimum difference, kurtosis coefficient, and skewness coefficient weights; when calculating similarity, each feature has a different degree of influence on the result, so machine learning is used to learn these weights;
[0049] Obtain the optimal values of parameters through machine learning;
[0050] The formula for constructing the features of the machine learning dataset is as follows:
[0051]
[0052] The tags are defined as follows:
[0053] X Label =[Featrures A , Featrures B ];
[0054] If A and B point to the same tag ID, then Y Label =0, otherwise Y Label =1;
[0055] The network design is as follows:
[0056] Y=Linear(X a -X b );
[0057] Right now:
[0058]
[0059] Where a(i) and b(i) are the fingerprint features of the two labels, B is the bias term, the training result is 0, so it is omitted in the fingerprint matching process, and xi is a learnable parameter.
[0060] The present invention discloses a method for RFID tag identification based on radio frequency fingerprint. First, feature extraction is performed on the complete timing information of all tags to be identified, and fingerprint features are further extracted. Then, a fingerprint library is established with the extracted fingerprint features, and the fingerprint library is used for subsequent rapid identification. After the fingerprint library is successfully established, it is no longer necessary to read the complete signal and then decode it to read the known tag. Instead, the fingerprint features of the acquired signal are calculated, and then the signal is matched in the fingerprint library. If the fingerprint library contains fingerprint information with extremely high similarity, the match is considered successful, and the information of the tag corresponding to the fingerprint is found. Finally, the tag information is returned, and the effect of rapid matching is achieved. In this way, the technical problem of slow recognition speed of the tag recognition method adopted for tag recognition in the prior art is solved. BRIEF DESCRIPTION OF THE DRAWINGS
[0061] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0062] Figure 1 It is a schematic diagram of the fingerprint database establishment process of the present invention.
[0063] Figure 2 It is a schematic diagram of the data reading process of the present invention.
[0064] Figure 3 It is a schematic diagram of the loss function of the machine learning model training process of the present invention. DETAILED DESCRIPTION
[0065] Embodiments of the present invention are described in detail below. Examples of the embodiments are shown in the accompanying drawings. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present invention, but should not be construed as limiting the present invention.
[0066] See also Figures 1 to 3 , Figure 1 It is a schematic diagram of the fingerprint database establishment process of the present invention. Figure 2 It is a schematic diagram of the data reading process of the present invention. Figure 3 Schematic diagram of the loss function of the machine learning model training process of the present invention. The present invention provides a method for RFID tag recognition based on radio frequency fingerprint, comprising the following steps:
[0067] S101, obtaining complete timing information of all RFID tags that need to be matched;
[0068] S102, extracting fingerprint features of known tags;
[0069] For this specific implementation, for the tag S needed to establish the fingerprint library i (i=0,1,2…n), find its RN16 data segment and EPC data segment;
[0070] Then split it into an array of 0 and 1 code elements in the time domain; the specific steps are as follows:
[0071] First, the data needs to be normalized:
[0072] α 0 (n) = α 0 (n)-μ α0
[0073] a 1 (n) = a 1 (n)-μ a1 ;
[0074] where μ α is the average amplitude, a 0 is an array of code element 0, a 1 is an array of code element 0;
[0075] After normalization, the average amplitude μ is re-averaged α As one of the fingerprint features;
[0076] The variance σ^2, skewness coefficient γ, kurtosis coefficient κ, and difference h of code element 0 and 1 are calculated respectively. The calculation formula is as follows:
[0077] Variance σ 2 :
[0078]
[0079] Skewness coefficient γ:
[0080]
[0081] Kurtosis coefficient κ:
[0082]
[0083] Difference h:
[0084] h 0 =max(a 0 )-min(a 0 )
[0085] h 1 =max(a 1 )-min(a 1 ).
[0086] S103, using the fingerprint features to establish a fingerprint database;
[0087] According to this specific implementation method, data is collected by a collection device, and the interest interval of the collected data is extracted; the acquired signal is further processed and split into a set of 0 and 1 code elements; feature extraction is performed on each set of code elements, and the feature extraction performed should include: average amplitude, variance, maximum and minimum difference, kurtosis coefficient, and skewness coefficient; through machine learning means, the feature weights required to establish a fingerprint library are trained; known labels are read in, the fingerprint data of each label is calculated, and stored in the fingerprint library;
[0088] In addition, the fingerprint library mainly includes two modules: fingerprint extraction module and analysis module; the main function of the fingerprint extraction module is to input the complete time sequence information of the RFID, extract the fingerprint features and store them; the main function of the analysis module is to input the RFID information to be matched, compare it with the fingerprint features in the library one by one, and return the most similar RFID information;
[0089] The specific method of training the feature weights required to establish the fingerprint library through machine learning is as follows:
[0090] Extract the features of RFID data and build a data set.
[0091] Set 5 learnable parameters, namely, the average amplitude, variance, maximum and minimum difference, kurtosis coefficient, and skewness coefficient weights; when calculating similarity, each feature has a different degree of influence on the result, so machine learning is used to learn these weights;
[0092] Obtain the optimal values of parameters through machine learning;
[0093] The formula for constructing the features of the machine learning dataset is as follows:
[0094]
[0095] The tags are defined as follows:
[0096] X Label =[Featrures A , Featrures B ];
[0097] If A and B point to the same tag ID, then Y Label =0, otherwise Y Label =1;
[0098] The network design is as follows:
[0099] Y=Linear(X a -X b );
[0100] Right now:
[0101]
[0102] Where a(i) and b(i) are the fingerprint features of the two labels, B is the bias term, the training result is 0, so it is omitted in the fingerprint matching process, and xi is a learnable parameter.
[0103] S104, read in the tag to be matched;
[0104] S105, obtaining the fingerprint information of the read-in tag;
[0105] S106, performing fingerprint matching;
[0106] For this specific implementation, read in data;
[0107] If the data can be read completely, the fingerprint feature is calculated for the whole data, put into the feature library for matching, and the one with the highest similarity is selected as the confirmed label. If the similarity is lower than the set threshold, it is considered that the library does not contain the label and the label is an unknown label or a cloned label.
[0108] If the data cannot be read completely, the fingerprint feature is calculated for each code element of the valid data. The feature library returns the five most similar tags each time, and finally takes the tag with the highest similarity as the confirmed tag. If the similarity is lower than the set threshold, it is considered that the library does not contain the tag and the tag is an unknown tag or a cloned tag.
[0109] The formula for calculating label similarity is as follows:
[0110]
[0111] where x 1 …x 10 Represents the weight of the corresponding feature on the result. The specific value is obtained through machine learning.
[0112] S107. If a matching tag is found, the matching tag information is returned; if not found, empty is returned.
[0113] When using the RFID tag identification method based on radio frequency fingerprint of the present embodiment, firstly, feature extraction is performed on the complete timing information of all tags to be identified, and fingerprint features are further extracted, and then a fingerprint library is established with the extracted fingerprint features, and the fingerprint library is used for subsequent rapid identification; after the fingerprint library is successfully established, reading a known tag does not need to be done by reading the complete signal and then decoding it, but the fingerprint features of the acquired signal are calculated, and then matched in the fingerprint library, if the fingerprint library contains fingerprint information with extremely high similarity, it is considered that the match is successful, the information of the tag corresponding to the fingerprint is found, and finally the tag information is returned, so as to achieve a rapid matching effect, thereby solving the technical problem of slow recognition speed of the tag recognition method adopted for tag recognition in the prior art.
[0114] What is disclosed above is only a preferred embodiment of the present invention, and it certainly cannot be used to limit the scope of rights of the present invention. Ordinary technicians in this field can understand that all or part of the processes of the above embodiment and equivalent changes made according to the claims of the present invention still fall within the scope of the invention.
Claims
1. A method for RFID tag identification based on radio frequency fingerprint, It is characterized in that include: Obtain complete timing information of all RFID tags that need to be matched; Extract fingerprint features of known tags; Using fingerprint features to build a fingerprint database; Read in the tag to be matched; Get the fingerprint information of the read tag; Conduct fingerprint matching; If a matching tag is found, the matching tag information is returned, if not found, empty is returned.
2. The method for RFID tag identification based on radio frequency fingerprint as claimed in claim 1, It is characterized in that The specific method of extracting the fingerprint features of known tags is as follows: For the tag S needed to build the fingerprint library i (i==0,1,2....n), find its RN16 data segment and EPC data segment; Then split it into an array of 0 and 1 code elements in the time domain; the specific steps are as follows: First, the data needs to be normalized: a 0 (n)=a 0 (n)-m a0 a 1 (n)=a 1 (n)-m a1 ; where μ a is the average amplitude, a 0 is the array of code element 0, α 1 is an array of code element 0; After normalization, the average amplitude μ is re-averaged a As one of the fingerprint features; The variance σ^2, skewness coefficient γ, kurtosis coefficient K, and difference h of code element 0 and 1 are calculated respectively. The calculation formula is as follows: Variance σ 2 : Skewness coefficient γ: Kurtosis coefficient κ: Difference h: h 0 =max(a 0 )-min(a 0 ) h 1 =max(a 1 )-min(a 1 )。 3. The method for RFID tag identification based on radio frequency fingerprint as claimed in claim 2, It is characterized in that The specific method of establishing a fingerprint database is as follows: Collect data through a collection device and extract the interest interval of the collected data; The acquired signal is further processed and split into a set of 0 and 1 code elements; Perform feature extraction on each set of code elements; Through machine learning, the feature weights required to establish the fingerprint library are trained; Read in known tags, calculate the fingerprint data of each tag, and store it in the fingerprint library.
4. The method for RFID tag identification based on radio frequency fingerprint as claimed in claim 3, It is characterized in that The specific method of fingerprint matching is as follows: Read in data; If the data can be read completely, the fingerprint feature is calculated for the whole data, put into the feature library for matching, and the one with the highest similarity is selected as the confirmed label. If the similarity is lower than the set threshold, it is considered that the library does not contain the label and the label is an unknown label or a cloned label. If the data cannot be read completely, the fingerprint feature is calculated for each code element of the valid data. The feature library returns the five most similar tags each time, and finally takes the tag with the highest similarity as the confirmed tag. If the similarity is lower than the set threshold, it is considered that the library does not contain the tag and the tag is an unknown tag or a cloned tag. The formula for calculating label similarity is as follows: where x 1 ...x 10 Represents the weight of the corresponding feature on the result. The specific value is obtained through machine learning.
5. The method for RFID tag identification based on radio frequency fingerprint as claimed in claim 4, It is characterized in that The feature extraction for each set of code elements should include: average amplitude, variance, maximum and minimum difference, kurtosis coefficient, and skewness coefficient.
6. The method for RFID tag identification based on radio frequency fingerprint as claimed in claim 5, It is characterized in that The specific method of training the feature weights required to establish the fingerprint library through machine learning is as follows: Extract the features of RFID data and build a data set. Set 5 learnable parameters, namely, the average amplitude, variance, maximum and minimum difference, kurtosis coefficient, and skewness coefficient weights; Obtain the optimal values of parameters through machine learning; The formula for constructing the features of the machine learning dataset is as follows: The tags are defined as follows: X Label =[Featrures A ,Featrures B ]; If A and B point to the same tag ID, then Y Label =0, otherwise Y Label =1; The network design is as follows: Y=Linear(X a -X b ); Right now: Where a(i) and b(i) are the fingerprint features of the two labels, B is the bias term, the training result is 0, so it is omitted in the fingerprint matching process, and xi is a learnable parameter.
Citation Information
Cited By
Full-chain tracing method and system for agricultural products based on tracing labels
CN119647499A
