Tag Reader

Abstract

To improve possibility of correctly decoding tag data even if a coding speed as to a wireless tag is fluctuated.SOLUTION: A tag reader apparatus according to an embodiment of the present invention has an extracting unit, an acquiring unit, and a machine learning unit. The extracting unit extracts a data signal obtained by encoding data from a wireless signal. The acquiring unit acquires tag data transmitted by a wireless tag as a learning target without obtaining transmission from the wireless tag as the learning target. The machine learning unit carries out machine learning using the data signal extracted by the extracting unit from the wireless signal transmitted from the wireless tag as the learning target with setting the tag data acquired by the acquiring unit as a correct data, and determines, based on relationship obtained by the machine learning, a response data included in the data signal extracted by the extracting unit from the wireless signal transmitted from the wireless tag as a reading target.SELECTED DRAWING: Figure 5

Description

[Technical field]

[0001] SUMMARY OF THE DISCLOSURE The present invention relates to a tag reader. [Background technology]

[0002] The wireless tag backscatters and transmits a data signal that is generated by encoding tag data at a predetermined encoding speed. A preamble section is provided at the beginning of the tag data to generate a clock signal with a frequency corresponding to the encoding speed. The tag reader then decodes the tag data from the data signal at a timing corresponding to the clock signal generated in synchronization with the preamble section included in the received data signal. However, due to a decrease in accuracy accompanying the fall in the cost of wireless tags, the encoding speed of the wireless tag may vary during one data transmission period. If this variation becomes large, there is a risk that the tag data cannot be correctly decoded by the clock signal generated according to the preamble portion. In view of these circumstances, it has been desirable to increase the likelihood that tag data can be correctly decoded even if the encoding speed of wireless tags varies. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] JP 2008-35407 A Summary of the Invention [Problem to be solved by the invention]

[0004] An object of the present invention is to provide a tag reader that can increase the possibility of correctly decoding tag data even if the encoding speed varies among wireless tags. [Means for solving the problem]

[0005] A tag reader according to an embodiment includes an extraction unit, an acquisition unit, and a machine learning unit. The extraction unit extracts a data signal obtained by encoding data from the wireless signal. The acquisition unit acquires tag data transmitted by a wireless tag to be learned without being transmitted from the wireless tag to be learned. The machine learning unit processes the data signal extracted by the extraction unit from the wireless signal transmitted from the wireless tag to be learned, and the tag data acquired by the acquisition unit as correct answer data. , associated with the feature of the fluctuation of one bit period in the data signal extracted by the extraction unit. Machine learning is performed, and response data contained in a data signal extracted by an extraction unit from a wireless signal transmitted from a wireless tag to be read is determined based on the machine-learned relationship. [Brief description of the drawings]

[0006] [Figure 1] FIG. 2 is a block diagram showing the main circuit configuration of a tag reading device according to an embodiment. [Diagram 2] FIG. 4 is a diagram showing an example of a structure of tag data. [Diagram 3] FIG. 4 is a diagram illustrating a data structure of one of the data records included in the tag data table. [Figure 4] FIG. 13 is a diagram showing an example of the waveform of a data signal when the encoding speed is gradually slowed down. [Diagram 5] 4 is a flowchart of a learning control process performed by a control unit shown in FIG. [Figure 6] 4 is a flowchart of a reading control process performed by a control unit shown in FIG. 1; DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0007] Hereinafter, an embodiment will be described with reference to the drawings. In the following, a tag reader that reads data from a wireless tag will be described as an example. In addition, a wireless tag is also widely called an RFID (radio frequency identification) tag, and therefore, in the following, it will be referred to as an RFID tag. FIG. 1 is a block diagram showing a main circuit configuration of a tag reader 100 according to an embodiment.

[0008] The tag reading device 100 includes an oscillator 1, a DA (digital to analog) converter 2, a modulator 3, a power amplifier 4, an antenna duplexer 5, an antenna 6, a quadrature detector 7, an amplifier 8, an AD (analog to digital) converter 9, an operation unit 10, and a baseband processing unit 11. The antenna 6 may not be included in the tag reading device 100, but may be externally attached to the tag reading device 100.

[0009] An oscillator 1 generates a sine wave and a cosine wave of a predetermined frequency as a carrier wave. The DA converter 2 converts the transmission data signal outputted in a digital state from the baseband processing unit 11 into an analog signal. The modulator 3 converts the data signal converted into an analog signal by the DA converter 2 into a modulated wave, and modulates the carrier wave generated by the oscillator 1 to obtain a transmission signal. Power amplifier 4 amplifies the power of the transmission signal output from modulator 3 to a level suitable for wireless transmission.

[0010] The antenna duplexer 5 has an input end connected to the power amplifier 4. The antenna duplexer 5 has an input / output end connected to the antenna 6. The antenna duplexer 5 has an output end connected to a quadrature detector 7. When a transmission signal amplified by the power amplifier 4 is input to the input end of the antenna duplexer 5, the antenna duplexer 5 supplies this transmission signal from the input / output end to the antenna 6. The antenna duplexer 5 receives a received signal output from the antenna 6 from the input / output end and outputs the received signal from the output end.

[0011] The antenna 6 emits radio waves corresponding to the transmission signal supplied from the antenna duplexer 5. The antenna 6 generates an electric signal as a reception signal corresponding to the radio waves transmitted from and arriving at the RFID tag 200. The reception signal generated at the antenna 6 in response to the radio waves transmitted from and arriving at the RFID tag 200 is a signal obtained by backscattering the radio waves radiated from the antenna 6 by the RFID tag 200 and modulating it with a data signal obtained by encoding the tag data. For example, an RZ (return to zero) code is used for the above encoding.

[0012] The quadrature detector 7 performs quadrature detection on the received signal output from the output terminal of the antenna duplexer 5, using the carrier wave generated by the oscillator 1. The quadrature detector 7 outputs an analog data signal obtained by detection. Amplifier 8 amplifies the data signal output from quadrature detector 7 to a level suitable for digitization by AD converter 9. To prevent amplifier 8 from being saturated due to the influence of DC components, amplifier 8 blocks the DC components through a capacitor prior to amplification. The AD converter 9 digitizes the data signal amplified by the amplifier 8 .

[0013] The operation unit 10 provides a user interface for an operator to operate the tag reading device 100. The operation unit 10 includes at least an operation device for inputting various instructions by the operator. The operation device may be a well-known device such as a touch panel. The operation unit 10 may include, for example, a display device that displays various information to notify the operator. The display device may be a well-known device such as a liquid crystal display. However, the user interface for operating the tag reading device 100 may be provided by a host computer 300 (described later), and the tag reading device 100 may not necessarily include the operation unit 10 .

[0014] The baseband processing unit 11 includes a coding unit 111, a demodulation unit 112, a machine learning unit 113, and a control unit 114. The encoding unit 111 encodes the transmission data output from the control unit 114 to be sent to the RFID tag 200 using a predetermined encoding method to obtain a data signal for transmission, and outputs the data signal to the DA converter 2 . The demodulation unit 112 generates a data signal having a waveform conforming to the original RZ code from the data signal, which has a differentiated waveform due to passing through a capacitor in the amplifier 8 . Thus, through a series of processes by the quadrature detector 7, amplifier 8, AD converter 9 and demodulator 112, the data signal is extracted from the radio signal received by the antenna 6, and the function of the extraction unit is realized by the cooperation of these quadrature detector 7, amplifier 8, AD converter 9 and demodulator 112.

[0015] The machine learning unit 113 has a learning mode and a decoding mode as operation modes. In the decoding mode, the machine learning unit 113 decodes tag data from the data signal provided by the demodulation unit 112 based on the learning result in the learning mode. When communicating with the RFID tag 200, the control unit 114 provides transmission data to the encoding unit 111 according to a predetermined sequence. The control unit 114 imports tag data output by the machine learning unit 113 as tag data transmitted from the RFID tag 200 according to a predetermined sequence. The control unit 114 performs a predetermined process on the imported tag data. For example, the process is a process for displaying a screen showing a reading result based on the tag data on the operation unit 10. Alternatively, the process is a process for notifying the host computer 300 of the imported tag data or a part of the data.

[0016] The host computer 300 is an information processing device connected to the tag reading device 100 and used as a user interface for controlling the tag reading device 100 or for checking the reading results by the tag reading device 100. In this embodiment, the tag data table TAA is stored in a storage unit 301 provided in the host computer 300.

[0017] The tag data table TAA is a data table that the control unit 114 refers to for tag data transmitted by a plurality of RFID tags 200 used for learning by the machine learning unit 113. FIG. 2 is a diagram showing an example of a structure of tag data. The tag data includes a preamble section, a start bit, a data section, and a checksum section. The preamble section and the start bit are common to all RFID tags 200. The data stored in each of the RFID tags 200 is set in the data section. The data set in this data section is the data to be read by the tag reader 100. Hereinafter, the data set in the data section is referred to as real data. The real data includes at least an identification code as an identifier for identifying each of the RFID tags 200. Note that the real data is often only the identification code. However, the real data may include any data different from the identification code in addition to the identification code. An error detection code calculated for the real data according to a predetermined rule is set in the checksum section.

[0018] FIG. 3 is a diagram showing a schematic data structure of one of the data records REA contained in the tag data table TAA. The tag data table TAA is a collection of a plurality of data records REA associated with each of a plurality of RFID tags 200. The data record REA includes fields FAA and FAB. The identification code of the associated RFID tag 200 is set in the field FAA. The actual data transmitted by the associated RFID tag 200 is set in the field FAB. Note that the field FAB may include at least one of the data set in the preamble portion, the start bit, and the checksum portion in addition to the actual data.

[0019] Next, the operation of the tag reading device 100 configured as above will be described. Note that most of the operations for reading the RFID tag 200 are similar to already known operations. Here, the description will focus on operations that are characteristic of the tag reading device 100. Also, the contents of the various processes described below are merely examples, and it is possible to change the order of some of the processes, omit some of the processes, or add other processes as appropriate.

[0020] Before commencing the explanation of the operation of tag reader 100, the nature of the data signal generated by demodulator 112 will be explained. A data signal is originally a signal obtained by encoding data at a predetermined encoding speed. Therefore, in the case of an RZ code, the data signal should be a signal that appears every period either in a pattern where high and low levels appear every 1 / 2 period for bits that are "1", or in a pattern where low levels appear for one period for bits that are "0". However, if the encoding speed varies due to a decrease in the operational accuracy accompanying a drop in the cost of the RFID tag 200, the period during which the data signal changes will vary during one reception period of the tag data.

[0021] FIG. 4 is a diagram showing an example of the waveform of a data signal when the encoding speed is gradually slowed down. In the data signal shown in Fig. 4, the period PB of the 12th bit of the data section is longer than the period PA of the 1st bit of the preamble section, which makes it difficult to correctly decode the data section and checksum section using a clock signal generated based on the preamble section. In addition, even if the encoding speed during one reception period of tag data is constant, it is possible that the encoding speed varies greatly for each tag response or for each tag. If the encoding speed varies greatly, the preamble portion will not be recognized as a preamble portion and will not be able to be decoded.

[0022] For example, a maintenance worker such as an administrator or serviceman of the tag reading device 100 issues an instruction to start learning via the operation unit 10 or the host computer 300 when preparing to start using the tag reading device 100 or during maintenance. Note that when starting learning, the maintenance worker prepares multiple RFID tags 200 for learning and adds data records REA associated with each of the multiple learning RFID tags 200 to the tag data table TAA. Alternatively, the maintenance worker prepares multiple RFID tags 200 associated with data records REA already included in the tag data table TAA as learning RFID tags 200.

[0023] When a learning start instruction is input via the operation unit 10 or when a notification is received from the host computer 300, the control unit 114 executes a learning control process. FIG. 5 is a flowchart of the learning control process by the control unit 114.

[0024] In ACT1, the control unit 114 checks whether or not an identification code has been input by operating the operation unit 10. If the control unit 114 cannot check the relevant event, it judges NO and proceeds to ACT2. In ACT2, the control unit 114 checks whether or not an identification code has been notified from the host computer 300. If the control unit 114 cannot check the relevant event, it judges NO and returns to ACT1. Thus, the control unit 114 waits for the identification code to be input or notified as ACT1 and ACT2.

[0025] The maintenance worker operates the operation unit 10 to select one of the learning RFID tags 200 and input its identification code. Alternatively, the maintenance worker selects one of the learning RFID tags 200 and designates the learning RFID tag 200 on the host computer 300. This designation may be performed, for example, by inputting the identification code, or may be performed by designation on a screen showing a list of the RFID tags 200 associated with the data records REA included in the tag data table TAA.

[0026] When the above-mentioned operation of inputting the identification code is performed on the operation unit 10, the control unit 114 judges YES in ACT1 and proceeds to ACT3. As ACT3, the control unit 114 inputs an identification code based on the operation on the operation unit 10. On the other hand, when the identification code is designated as described above, the host computer 300 notifies the tag reader 100 of the designated identification code. In response to this, the control unit 114 judges YES in ACT2 and proceeds to ACT4. As ACT 4 , the control unit 114 acquires the identification code notified by the host computer 300 . Upon completing ACT3 or ACT4, the control unit 114 proceeds to ACT5 in either case. The control unit 114 may omit ACT1 and ACT3, or ACT2 and ACT4, and receive the identification code designation only from either the operation unit 10 or the host computer 300.

[0027] In ACT5, the control unit 114 acquires data associated in the tag data table TAA with the identification code input in ACT3 or the identification code acquired in ACT4. The control unit 114 sends request data including the corresponding identification code, for example, to the host computer 300. The host computer 300 then selects from the tag data table TAA a data record REA in which the identification code included in the request data is set in the field FAA, and sends response data including the actual data set in the field FAB of the corresponding data record REA to the tag reader 100. The control unit 114 then acquires the actual data included in this response data. Note that the host computer 300 may also notify the actual data associated with the identification code in the tag data table TAA in addition to notifying the identification code as described above. In this case, the control unit 114 acquires the actual data thus notified in ACT4, and proceeds to ACT6 by skipping ACT5. Also, when the actual data includes only the identification code, the identification code input in ACT3 or acquired in ACT4 becomes the actual data as it is, so the control unit 114 skips ACT5 and proceeds to ACT6.

[0028] In ACT6, the control unit 114 generates tag data including the data acquired in ACT5 as the correct answer data. For example, the control unit 114 calculates a checksum for the actual data acquired in ACT5, and generates correct answer data as data including a prescribed preamble section and start bit, and setting the actual data acquired in ACT5 in the data section and the calculated checksum in the checksum section. The control unit 114 reads out the preamble section and start bit stored in advance in a storage device, for example, and includes them in the correct answer data. The preamble section and start bit may be stored in any storage device, such as a memory built in the control unit 114 or a memory separately provided in the baseband processing unit 11. Note that if at least a part of the preamble section, start bit, and checksum section is included in the data acquired in ACT5, the control unit 114 may include the data as it is in the correct answer data. In addition, if the field FAB of the tag data table TAA contains all of the preamble portion, start bit, actual data, and checksum, the control unit 114 may adopt the data as the correct data as is and omit generating the correct data as described above. Thus, the control unit 114 has a function as an acquisition unit that acquires correct answer data corresponding to the tag data transmitted by the learning target RFID tag 200 without relying on transmission from the RFID tag 200 in question.

[0029] As ACT7, control unit 114 commands machine learning unit 113 to start learning. Simultaneously with this command, or before or after this command, control unit 114 notifies machine learning unit 113 of the supervised answer data generated as described above. When the command to start learning is received, machine learning unit 113 starts operating in the learning mode.

[0030] As ACT8, the control unit 114 starts read transmission to read tag data from the RFID tag 200 that is in a state where communication with the tag reader 100 is possible. At this time, the control unit 114 outputs transmission data in a well-known sequence so that the RFID tag 200 backscatters the radio waves emitted from the antenna 6 and generates reflected waves modulated with the tag data.

[0031] The maintenance worker brings the selected learning RFID tag 200 close to the antenna 6. Then, the RFID tag 200 backscatters the radio waves emitted from the antenna 6 by the read transmission started as described above, and generates a reflected wave modulated by a data signal corresponding to the stored tag data. The tag data transmitted from the learning RFID tag 200 in this way is the same as the correct answer data.

[0032] A data signal generated by the quadrature detector 7, amplifier 8, AD converter 9, and demodulator 112 from a received signal generated at the antenna 6 in response to a reflected wave from the RFID tag 200 is input to the machine learning unit 113. In the learning mode, the machine learning unit 113 extracts feature amounts such as fluctuations in the period of each bit of the input data signal, and learns by associating them with correct answer data. For example, the machine learning unit 113 uses a well-known machine learning model for the process of this learning.

[0033] The maintenance worker sequentially changes the selected learning RFIDs, and issues an instruction to change the RFIDs by a predetermined operation on the operation unit 10 or the host computer 300.

[0034] Upon completing ACT8, the control unit 114 proceeds to ACT9. In ACT 9, the control unit 114 checks whether a change has been instructed. If the control unit 114 cannot check the event, it judges the result as NO and proceeds to ACT 10. In ACT 10, the control unit 114 checks whether or not an end command has been issued. If the control unit 114 cannot check the event, it determines NO and returns to ACT 9. Thus, the control unit 114 waits for an instruction to change or end in ACT9 and ACT10.

[0035] If a change has been instructed by the maintenance worker as described above, the control unit 114 judges YES in ACT9 and proceeds to ACT11. In ACT11, the control unit 114 stops reading and transmitting. In ACT12, the control unit 114 instructs the machine learning unit 113 to end learning. Thereafter, the control unit 114 returns to the standby state for ACT1 and ACT2.

[0036] After issuing the instruction to change as described above, the maintenance worker inputs or specifies the identification code of the newly selected learning RFID tag 200 in the same manner as described above. In response to this, the tag reading device 100 performs learning in the machine learning unit 113 in the same manner as described above.

[0037] When the maintenance worker has finished learning for each of the plurality of learning RFID tags 200, he or she issues an instruction to end the learning by performing a predetermined operation on the operation unit 10 or the host computer 300.

[0038] In response to such an end instruction, the control unit 114 judges YES in ACT10 and proceeds to ACT13. In ACT13, the control unit 114 stops reading and transmitting. In ACT14, the control unit 114 instructs the machine learning unit 113 to end learning. Then, the control unit 114 ends the learning control process.

[0039] A reader who uses the tag reader 100 to read the RFID tag 200 issues an instruction to start reading using the operation unit 10 or the host computer 300 .

[0040] When a command to start reading is input via the operation unit 10 or when a notification is received from the host computer 300, the control unit 114 executes a reading control process. FIG. 6 is a flowchart of the reading control process by the control unit 114.

[0041] As ACT21, the control unit 114 commands the start of decoding to the machine learning unit 113. When the start of decoding is commanded, the machine learning unit 113 starts operation in the decoding mode. As ACT22, the control unit 114 starts reading and transmitting. At this time, the control unit 114 outputs transmission data in a known sequence so that the unknown RFID tag 200 in a communicable state backscatters the radio waves emitted from the antenna 6 and generates a reflected wave modulated by a data signal.

[0042] The reader brings the RFID tag 200 to be read close to the antenna 6. Then, the RFID tag 200 backscatters the radio waves emitted from the antenna 6 by the read transmission started as described above, and generates a reflected wave modulated by a data signal corresponding to the stored tag data.

[0043] A data signal generated by the quadrature detector 7, amplifier 8, AD converter 9, and demodulator 112 from a received signal generated at the antenna 6 in response to a reflected wave from the RFID tag 200 is input to the machine learning unit 113. In the decoding mode, the machine learning unit 113 determines the tag data contained in the input data signal based on the learning result in the learning mode, and notifies the control unit 114 of the tag data as the decoding result.

[0044] After starting reading and sending in ACT22, the control unit 114 proceeds to ACT23. In ACT23, the control unit 114 checks whether or not tag data has been notified from the machine learning unit 113. If the control unit 114 cannot check the relevant event, it determines NO and proceeds to ACT24. In ACT 24, the control unit 114 checks whether or not a timeout has occurred. If the control unit 114 cannot check the event, it determines NO and proceeds to ACT 25. In ACT 25, the control unit 114 checks whether or not an end instruction has been issued. If the control unit 114 cannot check the event, it determines NO and returns to ACT 23. Thus, in ACT23 to ACT25, the control unit 114 waits for notification of tag data, a timeout, or an end instruction to occur. Then, if the control unit 114 receives notification of tag data from the machine learning unit 113 as described above, the control unit 114 judges YES in ACT23 and proceeds to ACT26.

[0045] In ACT26, the control unit 114 stops reading and transmitting. As ACT27, the control unit 114 calculates a checksum for the data portion of the notified tag data. In ACT28, the control unit 114 checks whether there is an error in the data portion of the notified tag data. For example, the control unit 114 compares the checksum calculated in ACT27 with the checksum portion of the notified tag data, and if they match, it is determined that there is no error and the result is NO, and the process proceeds to ACT29. In ACT 29, the control unit 114 outputs the notified tag data as a reading result to the host computer 300. The control unit 114 may directly notify the reading operator of the reading result by displaying a reading result screen based on the tag data on the operation unit 10. Then, the control unit 114 returns to ACT 22 and prepares to read the next RFID tag 200. If the checksum calculated in ACT27 does not match the checksum portion of the notified tag data, the control unit 114 determines that there is an error and answers YES in ACT28, and returns to ACT22 without executing ACT29.

[0046] On the other hand, if a predetermined condition is met, such as a prescribed waiting time having elapsed since the start of reading and transmission in ACT22, the control unit 114 determines YES in ACT24 as a timeout, and proceeds to ACT30. The control unit 114 stops reading and transmitting in ACT 30. Then, the control unit 114 returns to ACT 22, restarts reading and transmitting from the beginning, and attempts a new reading.

[0047] When the reader operator has finished reading the RFID tag 200, he or she issues an instruction to end reading via the operation unit 10 or the host computer 300.

[0048] When an instruction to end reading is input via the operation unit 10 or is notified by the host computer 300, the control unit 114 judges YES in ACT25 and proceeds to ACT31. In ACT31, the control unit 114 stops reading and transmitting. As ACT32, the control unit 114 instructs the machine learning unit 113 to end the decoding. Then, the control unit 114 ends the reading control process. When the end of decoding is commanded, the machine learning unit 113 ends operation in the decoding mode.

[0049] As described above, the tag reader 100 determines the tag data contained in the data signal by machine learning, thereby increasing the possibility that the tag data can be correctly decoded even if the encoding speed varies among wireless tags.

[0050] This embodiment can be modified in various ways as follows. The tag data table TAA may be stored in a storage device provided in the tag reader 100, and the control unit 114 may obtain the actual data by reading from the storage device.

[0051] Although some embodiments of the present invention have been described, these embodiments are presented as examples and are not intended to limit the scope of the invention. These novel embodiments can be implemented in various other forms, and various omissions, substitutions, and modifications can be made without departing from the spirit of the invention. These embodiments and their modifications are included in the scope and spirit of the invention, and are included in the scope of the invention and its equivalents described in the claims. The invention as originally claimed in the present application is set forth below. [Supplementary Note 1] An extractor that extracts a data signal obtained by encoding data from a wireless signal; an acquisition unit that acquires tag data transmitted by a wireless tag to be learned without relying on transmission from the wireless tag to be learned; a machine learning unit that performs machine learning on a data signal extracted by the extraction unit from a wireless signal transmitted from a wireless tag to be learned, using the tag data acquired by the acquisition unit as correct answer data, and determines response data included in the data signal extracted by the extraction unit from the wireless signal transmitted from the wireless tag to be read, based on the machine-learned relationship; A tag reading device comprising: [Additional Note 2] The acquiring unit acquires data unique to each wireless tag included in the tag data transmitted by the wireless tag to be learned, and generates tag data by adding data common to the tag data transmitted by the plurality of wireless tags and data for error detection to the acquired data. 2. A tag reader as defined in claim 1. [Supplementary Note 3] The acquiring unit acquires data unique to each wireless tag included in the tag data transmitted by the wireless tag to be learned from a storage device external to the tag reading device. 3. A tag reader as defined in claim 2. [Appendix 4] The machine learning unit learns characteristics of fluctuations in a 1-bit period in the data signal extracted by the extraction unit from the wireless signal transmitted from the wireless tag to be read. 2. A tag reader as defined in claim 1. [Explanation of symbols]

[0052] 1...oscillator, 2...DA converter, 3...modulator, 4...power amplifier, 5...antenna duplexer, 6...antenna, 7...quadrature detector, 8...amplifier, 9...AD converter, 10...operation unit, 11...baseband processing unit, 100...tag reader, 111...encoding unit, 112...demodulation unit, 113...machine learning unit, 114...control unit, 200...RFID tag, 300...host computer, 301...storage unit.

Claims

1. an extracting unit that extracts a data signal obtained by encoding data from the wireless signal; an acquisition unit that acquires tag data transmitted by a wireless tag to be learned without relying on transmission from the wireless tag to be learned; a machine learning unit that performs machine learning on a data signal extracted by the extraction unit from a wireless signal transmitted from a wireless tag to be learned, by associating the data signal extracted by the extraction unit with a feature of a fluctuation in a 1-bit period in the data signal extracted by the extraction unit, using the tag data acquired by the acquisition unit as correct answer data, and determines response data included in the data signal extracted by the extraction unit from the wireless signal transmitted from the wireless tag to be read, based on the machine-learned relationship; A tag reading device comprising:

2. the acquiring unit acquires data unique to each wireless tag included in the tag data transmitted by the wireless tag to be learned, and generates tag data by adding data common to the tag data transmitted by a plurality of wireless tags and data for error detection to the acquired data; The tag reader of claim 1 .

3. the acquiring unit acquires data unique to each wireless tag included in tag data transmitted by the wireless tag to be learned from a storage device external to the tag reading device; The tag reader according to claim 2 .

Images