JP2020173172A - Radio wave specification learning device and target identifying device - Google Patents

Abstract

To obtain a radio wave specification learning device for learning to be able to determine that an unknown target (radio wave source) is unknown, and a target identifying device using the same.SOLUTION: The radio wave specification learning device comprises: a radio wave specification input unit to which a plurality of parameters derived by A/D converting a radio wave having arrived from a target or a radio wave specification simulating the target and having the plurality of parameters are inputted for each target; a matching degree information input unit to which matching degree information in which the parameter range of identification object specification of an identification object for determining as matching the radio wave specification is inputted for each radio wave specification; and a learning unit for learning the range of identification object specification for determining as matching the radio wave specification on the basis of the radio wave specification inputted from the radio wave specification input unit and the matching degree information inputted from the matching degree information input unit, and determining that the identification object specification being off the learned range is an unknown specification as the radio wave specification is unknown.SELECTED DRAWING: Figure 1

Description

The present invention relates to radio wave specification learning for learning radio wave specifications of a target (radio wave source), and a target identification device using the same.

Conventionally, as a target identification device for identifying a target (radio wave source), radio wave specifications such as one using a received pulse signal and a known pulse signal (see, for example, Patent Document 1) and PDW (Pulse Description Word). (See, for example, Patent Documents 2 and 3). There is also a device that separates a pulse train from an electromagnetic wave arriving from a target in order to obtain radio wave specifications (see, for example, Patent Document 4).

On the other hand, there is one that identifies a target (radio wave source) from an image by using a learning model by machine learning using AI (Artificial Intelligence) or the like. For example, Patent Document 5 discloses one that identifies an unknown target included in an image or identifies a plurality of candidates to which the unknown target belongs by using the similarity.

Japanese Unexamined Patent Publication No. 2003-329765 Japanese Unexamined Patent Publication No. 2003-270326 Japanese Unexamined Patent Publication No. 2011-107138 Japanese Unexamined Patent Publication No. 2014-138252 Japanese Unexamined Patent Publication No. 2012-128558

However, the conventional learning model by machine learning has a problem that it does not use radio wave specifications, and even if it is a really unknown target (radio wave source), it identifies what it is, and multiple candidates. There was also the problem of identifying.

The present invention has been made to solve the above-mentioned problems, and is a radio wave specification learning that learns an unknown target (radio wave source) so that it can be determined to be unknown, and a target using the same. The purpose is to obtain an identification device.

In the radio wave specification learning according to the present invention, a plurality of parameters derived by A / D conversion of an electromagnetic wave arriving from a target, or a radio wave specification having the plurality of parameters simulating the target, is used for each target. Matching degree information input that defines the input radio wave specification input unit and the parameter range of the identification target specification of the identification target that is determined to match the radio wave specification for each radio wave specification. Based on the unit, the radio wave specifications input from the radio wave specification input unit, and the matching degree information input from the matching degree information input unit, it is determined that the radio wave specifications match. It is characterized by having a learning unit that learns the range of the identification target specifications and learns to determine the identification target specifications that are out of the learned range as unknown specifications whose radio wave specifications are unknown. Is to be.

In the target identification device according to the present invention, a plurality of parameters derived by A / D conversion of an electromagnetic wave arriving from the target, or radio wave specifications having the plurality of parameters simulating the target are input for each target. The radio wave specification input unit and the matching degree information input unit that defines the parameter range of the identification target specification of the identification target to be determined to match the radio wave specification for each radio wave specification are input. Based on the radio wave specifications input from the radio wave specification input unit and the matching degree information input from the matching degree information input unit, the identification is determined to match the radio wave specifications. A radio wave specification learning device having a learning unit that learns the range of the target specifications and learns to determine that the identification target specifications outside the learned range are unknown specifications whose radio wave specifications are unknown. It is a target identification device that identifies the target by using the learning result of the above, and the identification target is based on the identification target specification input unit in which the identification target specifications are input and the learning result learned by the learning unit. It is characterized by including a radio wave specification determination unit for determining the radio wave specifications corresponding to the specifications or determining that the identification target specifications are the unknown specifications.

According to the present invention, the range of identification target specifications to be determined to match the radio wave specifications is learned, and the identification target specifications outside the learned range are determined to be unknown specifications whose radio wave specifications are unknown. By doing so, it is possible to obtain radio wave specification learning that has a special effect of being able to perform learning with a low risk of identifying an unknown target as known, and a target identification device using the radio wave specification learning.

It is a functional block diagram of the radio wave specification learning apparatus which concerns on Embodiment 1 of this invention. It is a flowchart explaining operation (radio wave specification learning method) of the radio wave specification learning apparatus which concerns on Embodiment 1 of this invention. It is a functional block diagram of the radio wave specification learning device and the target identification device which concerns on Embodiment 1 of this invention. It is a flowchart explaining the operation (target identification method) of the radio wave specification learning apparatus and the target identification apparatus which concerns on Embodiment 1 of this invention. It is a functional block diagram of the radio wave specification learning device and the target identification device which concerns on Embodiment 1 of this invention. It is a flowchart explaining the operation (target identification method) of the radio wave specification learning apparatus and the target identification apparatus which concerns on Embodiment 1 of this invention. It is a functional block diagram of the radio wave specification learning apparatus which concerns on Embodiment 2 of this invention. It is a flowchart explaining operation (radio wave specification learning method) of the radio wave specification learning apparatus which concerns on Embodiment 2 of this invention. It is a functional block diagram of the radio wave specification learning device and the target identification device which concerns on Embodiment 2 of this invention. It is a functional block diagram of the radio wave specification learning apparatus which concerns on Embodiment 3 of this invention. It is a flowchart explaining operation (radio wave specification learning method) of the radio wave specification learning apparatus which concerns on Embodiment 3 of this invention. It is a functional block diagram of the radio wave specification learning device and the target identification device which concerns on Embodiment 3 of this invention.

Embodiment 1.
Hereinafter, the radio wave specification learning device and the target identification device according to the first embodiment of the present invention will be described with reference to FIGS. 1 to 6. In the figure, the same reference numerals indicate the same or corresponding parts, and detailed description thereof will be omitted. In FIGS. 1, 3, and 5, target 1 (radio wave source 1) is an arrival source of electromagnetic waves. The radio wave specification learning device 2 learns radio wave specifications such as PDW (Pulse Description Word) and IQ data. The target identification device 3 identifies a specific target 1 from a plurality of targets 1 or identifies an unknown target as unknown by using the learning result (learning model) of the radio wave specification learning device 2. Is. In the present application, the association between information is referred to as association.

In FIGS. 1, 3, and 5, the radio wave specification learning device 2 has the following configuration. In the radio wave specification input unit 4, a plurality of parameters derived by A / D conversion of the electromagnetic wave arriving from the target 1 or a radio wave specification having a plurality of parameters simulating the target 1 is input for each target 1. It is a thing. The matching degree information input unit 5 inputs the matching degree information that defines the parameter range of the identification target specification of the identification target to be determined to match the radio wave specification for each radio wave specification. The radio wave specifications and the corresponding target 1 are associated with each other. Similarly, the degree of agreement information and the corresponding goal 1 are associated with each other. The associated target 1 information (target information) may include at least the model name or model number. The model name or model number may be a code name. The learning unit 6 determines that the radio specifications match with the radio specifications based on the radio specifications input from the radio specification input unit 4 and the matching degree information input from the matching degree information input unit 5. It learns the range of specifications and learns to determine the identification target specifications outside the learned range as unknown specifications whose radio wave specifications are unknown.

Next, the operation of the radio wave specification learning device according to the first embodiment (the radio wave specification learning method according to the first embodiment) will be described with reference to FIG. In FIG. 2, in step 1, the radio wave specification input unit 4 has a plurality of parameters derived by A / D conversion of the electromagnetic wave arriving from the target 1, or a plurality of parameters simulating the target 1. Is a processing step input for each target 1. The step 2 is a processing step in which the matching degree information is input to the matching degree information input unit 5 for each radio wave specification, which defines the parameter range of the identification target specification of the identification target to be determined to match the radio wave specification. .. In step 1 and step 2, the order of processing does not matter. It may be at the same time. Step 3 determines that the learning unit 6 matches the radio wave specifications based on the radio wave specifications input from the radio wave specification input unit 4 and the matching degree information input from the matching degree information input unit 5. This is a processing step of learning the range of identification target specifications to be performed and learning to determine that the identification target specifications outside the learned range are unknown specifications whose radio wave specifications are unknown.

In the radio wave specification learning device according to the first embodiment (the radio wave specification learning method according to the first embodiment), the radio wave specification input unit 4 (step 1) is a PDW or a radio wave specification which is a radio wave specification. At least one of the IQ data may be input, or at least one of the frequency, amplitude, pulse width, and signal strength of the pulse train data among the PDWs that are radio wave specifications may be input as parameters. Further, the matching degree information input unit 5 (step 2) may input matching degree information in which a parameter range is defined for each parameter constituting a plurality of parameters.

In FIGS. 1, 3, and 5, the target identification device 3 has the following configuration. The identification target specification input unit 7 inputs the identification target specifications (radio wave specifications of the target 1 (radio wave source 1) to be newly identified). The radio wave specification determination unit 8 determines the radio wave specifications corresponding to the identification target specifications based on the learning result (learning model) learned by the learning unit 6, or determines that the identification target specifications are unknown specifications. Is what you do. When the radio wave specifications corresponding to the identification target specifications are determined, the matching target 1 (radio wave source 1) can be identified, so that it can be said that the target 1 (radio wave source 1) has been identified.

In FIG. 5, the radio wave specification determination unit 8 performs the first additional input of the identification target specifications determined to be unknown specifications to the radio wave specification input unit 4 of the radio wave specification learning device 2 as radio wave specifications. It is a thing. Further, the radio wave specification determination unit 8 sets the matching degree information having a parameter range corresponding only to the identification target specifications determined to be unknown specifications, and performs a second additional input to the matching degree information input unit 5. To do. With these two input operations, the radio wave specification determination unit 8 (target identification device 3) can relearn the learning unit 5.

In FIGS. 3 and 5, the receiving device 9 inputs the identification target specification input unit 7 (target identification device 3) to the identification target specifications (radio wave specifications of the target 1 (radio wave source 1) to be newly identified). ) Is input. The receiving device 9 has the following configuration. The antenna unit 10 receives the electromagnetic wave arriving from the target 1 (radio wave source 1). The A / D conversion unit 11 A / D-converts the electromagnetic wave received by the antenna unit 10 to generate digital data. The specification detection unit 12 detects the identification target specifications (radio wave specifications) of the digital data generated by the A / D conversion unit 11. The pulse train separation unit 13 sends the identification target specifications (radio wave specifications) that have been subjected to the pulse train separation processing to the identification target specification input unit 7 (target identification device 3).

Next, the target identification device (the target identification method according to the first embodiment) of the target identification device according to the first embodiment will be described with reference to FIG. In FIG. 4, step 11 is a processing step in which the identification target specifications, which are the radio wave specifications of the target 1 (radio wave source 1) to be newly identified, are input to the identification target specification input unit 7. Step 12 is a processing step in which the identification target specifications are input from the identification target specification input unit 7 to the learning unit 6 and the learning model is used. In step 13, the radio wave specification determination unit 8 determines the radio wave specifications corresponding to the identification target specifications based on the learning result (learning model) learned by the learning unit 6, or the identification target specifications are unknown. This is a processing step for determining the original. When the radio wave specifications corresponding to the identification target specifications are determined in step 13, the matching target 1 (radio wave source 1) can be identified, so that it can be said that the target 1 (radio wave source 1) has been identified.

Next, a case where the determination result of the target identification device according to the first embodiment is fed back to the radio wave specification learning device according to the first embodiment and re-learned by the learning unit 6 (reinforcement learning) will be described. .. FIG. 6 describes re-learning (reinforcement learning) in the operation of the radio wave specification learning device according to the first embodiment (the radio wave specification learning method according to the first embodiment). Since steps 11 to 13 in FIG. 6 are the same as those described with reference to FIG. 4, the description thereof will be omitted. In FIG. 6, step 14 is a step of determining whether or not the identification target specifications are determined to be unknown specifications based on the learning result (learning model) learned by the learning unit 6 in step 13. If it is determined in step 14 that there are unknown specifications, the process proceeds to step 21 and step 22. In steps 21 and 22, the order of processing does not matter. It may be at the same time.

In FIG. 6, step 11 is a processing step in which the identification target specifications determined to be unknown specifications are additionally input to the radio wave specification input unit 4 as radio wave specifications. In step 22, the matching degree information input unit 5 is set with the matching degree information having a parameter range corresponding only to the identification target specifications determined to be unknown specifications, and the matching degree information input unit is additionally input with the second. This is a processing step. In step 23, the radio wave specifications (identification target specifications) input from the radio wave specification input unit 4 and the matching degree information (matching newly set according to the identification target specifications) input from the matching degree information input unit 5 are performed. Based on the degree information), the learning unit 6 is made to relearn the range of the identification target specifications to be determined to match the radio wave specifications, and the radio wave specifications are unknown for the identification target specifications outside the learned range. This is a processing step for re-learning to determine that the unknown specifications are.

The following can be considered as the trigger for the execution of step 14. That is, the target identification device (radio wave specification determination unit 8) according to the first embodiment is a trigger for re-learning (reinforcement learning) the radio wave specification learning device according to the first embodiment. For example, when the identification target specification input unit 7 inputs the same identification target specifications determined by the radio wave specification determination unit 8 to be unknown specifications over a preset number of times, or the identification target. It is conceivable that the specification input unit 7 records the identification target specifications determined by the radio wave specification determination unit 8 to be unknown specifications, and a re-learning execution command is input from the outside. In such a case, the radio wave specification determination unit 8 may perform the first additional input and the second additional input (step 21 and step 22) to relearn the learning unit 6 (step 23). .. It should be noted that the preset number of times may be set according to the situation, such as the number of times that the unknown specifications are determined to be unknown and the number of times that the unknown specifications cannot be ignored. Further, if the learning unit 6 is in the process of learning, the preset number of times may be one.

As described above, in the radio wave specification learning device and the target identification device (radio wave specification learning method and target identification method) according to the first embodiment, the identification target specifications are two or more of the radio wave specifications due to the overlap of all the parameter ranges. When there is an input (duplicate matching degree information) that causes a state to be judged to match the original, it is not assumed. The duplication of the agreement degree information will be described in the radio wave specification learning device and the target identification device (radio wave specification learning method and target identification method) according to the second and third embodiments.

Embodiment 2.
Hereinafter, the radio wave specification learning device according to the second embodiment of the present invention and the target identification device using the radio wave specification learning device will be described with reference to FIGS. 7, 8 and 9. In the drawings, the same reference numerals indicate the same or corresponding parts, including the parts common to the first embodiment, and detailed description thereof will be omitted. In FIGS. 7 and 9, the learning unit 6 has a duplication detection unit 14 and a notification unit 15. The duplication detection unit 14 determines that the matching degree information input unit 5 has an input that causes a state in which it is determined that the identification target specifications match two or more radio wave specifications due to the duplication of the parameter range. When the duplication detection unit 14 determines that there is duplication of the parameter range, the notification unit 15 notifies the outside.

In this way, when there is duplication in the parameter range, the duplication can be eliminated by reviewing the matching degree information. Further, in the radio wave specification learning device according to the second embodiment, learning can proceed as follows even when the parameter ranges overlap. When there is an input in which the learning unit 6 determines that the identification target specifications match two or more radio wave specifications due to the overlapping of the parameter ranges, the learning unit 6 identifies each of the radio wave specifications having the overlapping parameter ranges and the identification target. It is possible to learn to calculate the similarity with the specifications and determine that the radio wave specifications having the highest similarity match the identification target specifications.

Next, the operation of the radio wave specification learning device according to the second embodiment (the radio wave specification learning method according to the second embodiment) will be described with reference to FIG. In FIG. 8, since steps 1 and 2 are the same as those described with reference to FIG. 2, the description thereof will be omitted. Of course, in step 1 and step 2, the order of processing does not matter. It may be at the same time.

In FIG. 8, steps 31 and 32 correspond to the duplication of the parameter range by subdividing step 3 (FIG. 2). In step 31, the duplication detection unit 14 determines that the matching degree information input unit 5 has an input that causes a state in which the identification target specifications are determined to match two or more radio wave specifications due to the duplication of the parameter range, and notifies the user. When the duplication detection unit 14 determines that there is duplication of the parameter range, the unit 15 is a processing step of notifying the outside. Step 32 determines that the learning unit 6 matches the radio wave specifications based on the radio wave specifications input from the radio wave specification input unit 4 and the matching degree information input from the matching degree information input unit 5. This is a processing step of learning the range of identification target specifications to be performed and learning to determine that the identification target specifications outside the learned range are unknown specifications whose radio wave specifications are unknown. Further, as described above, the similarity between each of the radio wave specifications having overlapping parameter ranges and the identification target specification is calculated, and it is determined that the radio wave specification having the highest similarity matches with the identification target specification. You may learn that.

When the determination result of the target identification device according to the second embodiment is fed back to the radio wave specification learning device according to the second embodiment and relearned by the learning unit 6 (reinforcement learning), the first embodiment 1 Since it is the same as the above, the description is omitted.

Embodiment 3.
Hereinafter, the radio wave specification learning device according to the third embodiment of the present invention and the target identification device using the radio wave specification learning device will be described with reference to FIGS. 10, 11, and 12. In the drawings, the same reference numerals indicate the same or corresponding parts, including the parts common to the first and second embodiments, and detailed description thereof will be omitted. In the third embodiment, when there is an input in which the identification target specifications are determined to match two or more radio wave specifications due to the overlap of the parameter ranges, the learning unit 6 corrects the matching degree information and parameters. Learning is done by eliminating overlapping parts from the range. In FIGS. 10 and 12, the learning unit 6 has a duplication detection unit 14 and a duplication elimination unit 16. When the duplication detection unit 14 determines that there is duplication of the parameter range, the duplication elimination unit 16 corrects the matching degree information and eliminates the duplication portion from the parameter range.

The duplication elimination unit 16 corrects the matching degree information by deleting the overlapping portion of the parameter range from at least one of the parameter ranges to eliminate the duplication, or by deleting the overlapping portion of the parameter range and eliminating the duplication. It is conceivable to do. Further, it is conceivable that the duplication elimination unit 16 divides the overlapping portion of the parameter range and incorporates the divided portion into at least one of the parameter ranges to eliminate the duplication. The duplication elimination unit 16 and the notification unit 15 may be used together. For example, it is conceivable to operate the duplication elimination unit 16 after confirming the notification of the notification unit 15. Depending on the content of the notification (overlapping content) of the notification unit 15, it is possible that the duplication elimination unit 16 is not operated.

Next, the operation of the radio wave specification learning device according to the third embodiment (the radio wave specification learning method according to the third embodiment) will be described with reference to FIG. In FIG. 11, since steps 1 and 2 are the same as those described with reference to FIG. 2, the description thereof will be omitted. Of course, in step 1 and step 2, the order of processing does not matter. It may be at the same time. Further, in FIG. 11, step 32 is the same as that described with reference to FIG. 8, but the overlapping state of the parameter range is eliminated.

In FIG. 11, steps 31 and 32 are subdivided steps 3 (FIG. 2) to correspond to overlapping parameter ranges. In step 31, the duplication detection unit 14 determines that there is an input in the matching degree information input unit 5 that causes a state in which the identification target specifications are determined to match two or more radio wave specifications due to the duplication of the parameter range, and the duplication. The elimination unit 16 is a processing step in which the duplication detection unit 14 eliminates the duplication of the parameter range. In step 32, after the duplication of the parameter range is eliminated, the learning unit 6 is based on the radio wave specifications input from the radio wave specification input unit 4 and the matching degree information input from the matching degree information input unit 5. Learns the range of identification target specifications that are determined to match the radio wave specifications, and learns to determine that the identification target specifications that are out of the learned range are unknown specifications whose radio wave specifications are unknown. It is a processing step.

When the determination result of the target identification device according to the third embodiment is fed back to the radio wave specification learning device according to the third embodiment and relearned by the learning unit 6 (reinforcement learning), the first embodiment 1 Since it is the same as 2 and 2, the description thereof will be omitted.

As described above, in the radio wave specification learning device and the target identification device (radio wave specification learning method and target identification method) according to the first to third embodiments of the present invention, the learning unit 6 uses the radio wave specifications and the matching degree information. Based on the above, the range of identification target specifications to be determined to match the radio wave specifications is learned, and the identification target specifications outside the learned range are determined to be unknown specifications whose radio wave specifications are unknown. Therefore, even if the target 1 is really unknown (radio wave source 1), there is little possibility of identifying what it is or identifying a plurality of candidates.

1 target (radio wave source), 2 radio wave specification learning device, 3 target identification device,
4 Radio wave specification input unit, 5 Matching degree information input unit, 6 Learning unit, 7 Identification target specification input unit,
8 Radio specification judgment unit, 9 Receiver, 10 Antenna unit, 11 A / D converter,
12 Specifications detection unit, 13 Pulse train separation unit, 14 Duplicate detection unit, 15 Notification unit,
16 Duplicate elimination section.

Claims (13)

A radio wave specification input unit in which a plurality of parameters derived by A / D conversion of an electromagnetic wave arriving from a target, or a radio wave specification having the plurality of parameters simulating the target, is input for each target. From the matching degree information input unit for inputting the matching degree information that defines the parameter range of the identification target specification of the identification target to be determined to match the radio wave specification for each radio wave specification, and the radio wave specification input unit. Based on the input radio wave specifications and the matching degree information input from the matching degree information input unit, the range of the identification target specifications to be determined to match the radio wave specifications is learned. A radio wave specification learning device including a learning unit that learns to determine that the radio wave specifications are unknown specifications whose radio wave specifications are unknown, for the identification target specifications that are out of the learned range. The radio wave specification according to claim 1, wherein the radio wave specification input unit inputs at least one of PDW (Pulse Description Word) which is the radio wave specification or IQ data which is the radio wave specification. Learning device. The radio wave specification input unit is characterized in that at least one of the frequency, amplitude, pulse width, and signal strength of the pulse train data is input as the parameter among the PDW (Pulse Description Word) which is the radio wave specification. The radio wave specification learning device according to claim 1. The agreement degree information input unit is any one of claims 1 to 3, wherein the agreement degree information in which the parameter range is defined for each parameter constituting the plurality of parameters is input. Described radio specification learning device. 4. The learning unit is characterized in that when there is an input in which it is determined that the identification target specifications match two or more radio wave specifications due to the overlap of the parameter ranges, the learning unit notifies the learning unit. Radio wave specification learning device described in. When there is an input in which the learning unit determines that the identification target specifications match two or more of the radio wave specifications due to the duplication of the parameter ranges, the learning unit has the radio wave specifications with the overlapping parameter ranges. The claim is characterized in that it learns to calculate the similarity between each of the above and the identification target specifications and determine that the radio wave specifications having the highest similarity are the same as the identification target specifications. The radio wave specification learning device according to 4. When there is an input in which the learning unit determines that the identification target specifications match two or more radio wave specifications due to the overlap of the parameter ranges, the learning unit corrects the matching degree information and the parameters. The radio wave specification learning device according to claim 4, wherein the learning is performed by eliminating the overlapping portion from the range. The learning unit corrects the matching degree information by deleting the overlapping portion from at least one of the parameter ranges to eliminate the duplication, or deleting the overlapping portion and eliminating the duplication. The radio wave specification learning device according to claim 7. The radio wave specification learning device according to claim 7, wherein the learning unit divides the overlapping portion and incorporates the divided portion into at least one of the parameter ranges to eliminate the duplication. A target identification device that identifies the target using the learning result of the radio wave specification learning device according to any one of claims 1 to 9.
The radio wave specifications corresponding to the identification target specifications are determined based on the identification target specification input unit into which the identification target specifications are input and the learning result learned by the learning unit, or the identification target is determined. A target identification device including a radio wave specification determination unit that determines that the specifications are unknown specifications. The radio wave specification determination unit performs the first additional input of the identification target specifications determined to be the unknown specifications to the radio wave specification input unit as the radio wave specifications.
The radio wave specification determination unit sets the matching degree information having the parameter range corresponding only to the identification target specification determined to be the unknown specification, and inputs the second additional input to the matching degree information input unit. And carry out
The target identification device according to claim 10, wherein the learning unit is relearned. When the same identification target specifications determined by the radio wave specification determination unit to be the unknown specifications are input over a preset number of times, the identification target specification input unit is used.
The target identification device according to claim 11, wherein the radio wave specification determination unit executes the first additional input and the second additional input to relearn the learning unit. The identification target specification input unit records the identification target specification determined by the radio wave specification determination unit to be the unknown specification, and when a re-learning execution command is input from the outside, the identification target specification input unit records the identification target specification.
The target identification device according to claim 11, wherein the radio wave specification determination unit executes the first additional input and the second additional input to relearn the learning unit.

Images