JP2013183294A - Radio station database creation device, radio wave monitoring device, and method and program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To automatically create a radio station identification database in which feature vector waveforms are grouped by quantitative index on the basis of their similarity.SOLUTION: A radio station database creation device comprises: tree diagram generation means 101 which executes hierarchical cluster analysis on a feature vector waveform which is a frequency change waveform at high-going transition of a signal by using variables, etc. calculated from the feature vector waveform, to generate a tree diagram; grouping means 102 which classifies feature vector waveforms into groups on the basis of a difference in distance between merged clusters which is an amount of change in distance between merged clusters in one session of cluster merge indicated by the tree diagram; and database registration means 103 which registers feature vector waveforms classified into the same group to a database as the ones belonging to the same radio station.

Description

The present invention relates to a radio station database creation apparatus, a radio station database creation method, and a radio station database creation program that automatically create a database for identifying radio stations. The present invention also relates to a radio wave monitoring apparatus, a radio wave monitoring method, and a radio wave monitoring program for automatically detecting illegal radio stations.
About.

  When registering feature vector waveforms measured from received signals of unknown radio stations in a database as one of the methods for identifying radio stations, high-similarity feature vector waveforms are grouped together in the same group, and the same radio There is a method of registering in a database as a waveform from a station. The feature vector waveforms obtained from the same solid-state radio station fluctuate according to the transmission conditions of the radio station and other environmental conditions, and it has been confirmed by experiments that the feature vector waveforms are divided into multiple patterns. This is because if the stations are grouped and registered in the database, improvement in identification performance can be expected.

  For example, Patent Document 1 describes that a frequency variation generated at the time of rising is extracted as a feature vector, a candidate vector waveform and a feature vector waveform are compared, a similarity is calculated, and a radio station is identified. .

  Further, regarding a technique for performing grouping based on the similarity of the obtained feature quantities, for example, Patent Document 2 discloses a tree diagram by performing hierarchical cluster analysis on a face image classification result by an image classification unit. Group the face images detected from the video based on the created tree diagram, and identify the person with the highest appearance frequency from the number of face images classified into the same group It is described. In Patent Document 2, “Pseudo F static”, which is a statistic indicating the degree of separation between all clusters at each joining stage, can be used as a criterion for determining the cutting height of the tree diagram. such as cut at a height, it is described that may be predetermined where to cutting.

JP 2011-044937 A JP 2006-244279 A

  Whether a plurality of feature vector waveforms are to be the same group or different groups needs to be determined according to the degree of variation in the feature vector waveforms. However, in the method in which the operator makes such a determination qualitatively, there is a problem that a grouping standard becomes ambiguous and a database having different identification performance is generated depending on the skill of the operator.

  Further, visually determining the grouping of each feature vector waveform has a problem that the time and work load required for database creation increase.

  Note that Patent Document 2 describes that hierarchical cluster analysis is used as a method of grouping based on the similarity of the obtained feature values, but it is only used for facial image classification and is wireless. It cannot be applied directly to station classification.

  For example, the tree diagram created by hierarchical cluster analysis itself is not a classification result, so where to cut the tree diagram varies greatly depending on the characteristics of the classification target and the purpose of grouping. The classification example of the face image cannot be applied as it is. That is, in Patent Document 2, when a hierarchical cluster analysis method is used in grouping for identifying radio stations, what feature value is used and what criteria is used to determine a final group Nothing is disclosed as to whether it is okay.

  Accordingly, the present invention provides a radio station database creation apparatus, radio station database creation method, and radio station database that automatically creates a database for radio station identification in which feature vector waveforms are grouped by a quantitative index based on their similarity. The purpose is to provide a creation program. Another object of the present invention is to provide a radio wave monitoring apparatus, a radio wave monitoring method, and a radio wave monitoring program that automatically detect illegal radio stations.

  The radio station database creation device according to the present invention performs a hierarchical cluster analysis on a feature vector waveform that is a frequency fluctuation waveform at the time of signal rise, using a variable calculated from the feature vector waveform, and generates a tree diagram. Based on the difference in distance between the merged clusters, which is the amount of change in the distance between the merged clusters in one cluster merge, indicated by the generated tree diagram generating means and the tree diagram generated by the tree diagram generating means, Grouping means for grouping feature vector waveforms, and database registration means for registering in the database the feature vector waveforms classified into the same group by the grouping means as belonging to the same radio station .

  In addition, a radio wave monitoring apparatus according to the present invention receives the radio signal transmitted from the radio station database creation apparatus and the radio station and extracts a feature vector waveform which is a frequency fluctuation waveform at the time of signal rise at least. And a monitoring means for detecting an illegal radio station based on information registered in the database and indicating the identification result of the radio station registered by the database registration means.

  Further, the radio station database creation method according to the present invention performs a hierarchical cluster analysis on a feature vector waveform, which is a frequency fluctuation waveform at the time of signal rise, using a variable calculated from the feature vector waveform to form a tree shape. Based on the difference between the intercluster cluster distances, which is the amount of change in the intercluster cluster distance for one cluster merge, indicated by the tree diagram generation step for generating the diagram and the tree diagram generated in the tree diagram generation step. And a grouping step for grouping the feature vector waveforms, and a database registration step for registering the feature vector waveforms classified into the same group by the grouping step in the database as belonging to the same radio station. To do.

  In addition, the radio wave monitoring method according to the present invention includes a radio wave detection step of receiving a radio signal transmitted from a radio station and extracting a feature vector waveform which is a frequency fluctuation waveform at the time of signal rise, and the extracted feature vector waveform On the other hand, a tree diagram generation step for generating a tree diagram by performing a hierarchical cluster analysis using a variable calculated from the feature vector waveform, and a tree diagram generated in the tree diagram generation step The grouping step for grouping feature vector waveforms and the grouping step grouped into the same group based on the intercluster cluster distance difference, which is the amount of change in the intercluster cluster distance for one cluster merge shown. A database registration step of registering the feature vector waveform in the database as belonging to the same radio station; Registered in the database, based on the information indicating the identification result of the radio station that is registered by the database registration unit, characterized in that it comprises a monitoring step of detecting the illegal radio station.

  The radio station database creation program according to the present invention performs hierarchical cluster analysis on a feature vector waveform that is a frequency fluctuation waveform at the time of signal rise, using a variable calculated from the feature vector waveform. Tree diagram generation process to generate a tree diagram, and the inter-cluster distance difference, which is the amount of change in the inter-cluster distance for one cluster merge, indicated by the tree diagram generated by the tree diagram generation process Grouping processing for grouping feature vector waveforms, and database registration processing for registering the feature vector waveforms classified into the same group by grouping processing in the database as belonging to the same radio station. Features.

  In addition, the radio wave monitoring program according to the present invention receives a radio signal transmitted from a radio station to a computer and extracts a feature vector waveform that is a frequency fluctuation waveform at the time of signal rise. Tree diagram generated by tree diagram generation processing that generates a tree diagram by performing hierarchical cluster analysis on the vector waveform using variables calculated from the feature vector waveform Based on the inter-cluster distance difference, which is the amount of change in the inter-cluster distance for a single cluster merge, the grouping process that groups the feature vector waveforms, and the grouping process grouped them into the same group Database registration processing for registering feature vector waveforms in the database as belonging to the same radio station, and data Base is registered, based on the information indicating the identification result of the radio station that is registered by the database registration unit, characterized in that to perform the monitoring processing for detecting illegal radio station.

  ADVANTAGE OF THE INVENTION According to this invention, the database for radio station identification which grouped the feature vector waveform by the quantitative parameter | index based on the similarity can be created automatically. Further, according to the present invention, it is possible to automatically detect illegal radio stations.

It is a block diagram which shows the structural example of the radio station database production apparatus of 1st Embodiment. It is a flowchart which shows an example of operation | movement of the radio station database production apparatus of 1st Embodiment. It is explanatory drawing which shows the example of the dendrogram which a radio station database production apparatus produces | generates. It is a block diagram which shows the structural example of the electromagnetic wave monitoring apparatus of 2nd Embodiment. It is a block diagram which shows the outline | summary of the radio station database preparation apparatus of this invention. It is a block diagram which shows the outline | summary of the electromagnetic wave monitoring apparatus of this invention.

Embodiment 1. FIG.
Hereinafter, embodiments of the present invention will be described with reference to the drawings. FIG. 1 is a block diagram illustrating a configuration example of a radio station database creation device according to the first embodiment. A radio station database creation device 1 shown in FIG. 1 includes an arithmetic processing unit 2, a dendrogram generation unit 3, a threshold determination unit 4, and a database 5.

  The arithmetic processing unit 2 calculates feature amount data for identifying a wireless station from the acquired feature vector waveform (here, a frequency fluctuation waveform at the time of signal rise). Here, the feature amount data is one or more variables representing the degree of frequency fluctuation inherent to the acquired feature vector waveform.

  In addition to the variable obtained from the feature vector waveform, the arithmetic processing unit 2 may obtain, for example, a variable calculated from the position information, voice, modulation method, and spectrum as the feature amount data.

  The dendrogram generation unit 3 accumulates the feature amount data calculated by the arithmetic processing unit 2 every time a feature vector waveform is acquired, and performs a hierarchical cluster analysis when the number of data exceeds a certain number, Generate a dendrogram.

  Specifically, the dendrogram generation unit 3 uses a classification state with each feature vector waveform as one cluster as an initial cluster, and based on a variable calculated from each feature vector waveform, a cluster or a feature vector belonging to the cluster Find the distance between waveforms and repeat the cluster merging process of merging the pair of the closest distances into one cluster until the number of clusters becomes 1, and the merged state of the resulting cluster is merged What is necessary is just to produce | generate the information shown with the distance between clusters. Note that the intercluster distance in the dendrogram represents the degree of similarity between the clusters to be merged or feature vector waveforms belonging to the clusters.

  The Euclidean distance is widely used as a method for calculating the similarity between data in the hierarchical cluster analysis, but is not limited thereto. For example, a scaled Euclidian distance, a Minkowski distance, a Manhattan distance, a Mahalanobis general distance, a Calhoun distance, or the like may be used. Further, as a definition of similarity, in addition to using a correlation coefficient, it is also possible to use an angle between vectors, the number of linkages, and the like.

  The threshold determination unit 4 calculates a change amount of the inter-cluster distance to be merged between the step of merging a certain cluster and the step of merging the next cluster. Then, it is determined whether or not the calculated change amount of the inter-cluster distance (hereinafter referred to as the inter-cluster distance difference) is equal to or greater than a preset threshold value. When the threshold value is exceeded, the minimum number of steps to reach the step is obtained, and the state of the cluster merged immediately before that step is determined as the final cluster. That is, the minimum number of steps at which the inter-cluster distance difference is equal to or greater than the threshold value is obtained, and the feature vector waveform group classified immediately before that step is registered in the database as a result of grouping the radio stations.

  In hierarchical cluster analysis, cluster merging starts in descending order of similarity, and finally becomes one cluster. Here, where to stop cluster merging, that is, how many clusters to stop, is how to specify the number in advance, or the number of clusters, such as the cluster just before the similarity falls below a certain threshold, as the final cluster number. However, according to the present invention, the distance between merged clusters (that is, the degree of similarity) after cluster merger compared to the previous merger is determined as suitable for grouping radio stations regardless of human judgment. A method is adopted in which a greatly changing location is used as the cutting position of the dendrogram.

  Further, the threshold value determination unit 4 registers the feature vector waveforms classified into the same group as a result of grouping by hierarchical cluster analysis in the database 5 as belonging to the same radio station. When registering the grouping result in the database 5, the number of groups, that is, the number of identified radio stations, and the feature vector waveform belonging to each group, that is, the information on the feature vector waveform regarding that radio station as the transmission source are included. Although it is preferable to register so that it may be understood, a specific data structure is not ask | required. For example, when a data number is assigned to each acquired feature vector waveform, the data number and the group number to which the data number belongs can be associated and registered in the database.

  The database 5 stores information indicating the identification result of the radio station for at least the acquired feature vector waveform. Specifically, information indicating the result of grouping by the threshold determination unit 4 may be stored. The database 5 is registered so that it is possible to know how many radio stations have been identified and which feature vector waveform is regarded as the transmission source for each radio station. Although it is preferable, a specific data structure is not limited. For example, when a data number is assigned to each acquired feature vector waveform, the data number and the group number to which the data number belongs can be associated and registered in the database.

  In the present embodiment, the arithmetic processing unit 2, the dendrogram generation unit 3, and the threshold determination unit 4 are realized by an information processing apparatus that operates according to a program such as a CPU, for example. The database 5 is realized by a storage device, for example.

  Next, the operation of this embodiment will be described. FIG. 2 is a flowchart showing an example of the operation of the radio station database creation device of this embodiment.

  In the example shown in FIG. 2, when a feature vector waveform is first extracted (Yes in step S1), the arithmetic processing unit 2 calculates feature amount data from the feature vector waveform, and generates the calculated feature amount data as a dendrogram. It transmits to the part 3 (step S2). Here, the arithmetic processing unit 2 may calculate a plurality of variables representing the degree of frequency fluctuation as the feature amount data.

  The dendrogram generation unit 3 accumulates feature amount data (in this example, two types of variables representing the degree of frequency fluctuation) sent from the arithmetic processing unit 2 in a predetermined storage device as one set. The storage device may use the database 5. When a certain number or more of data is accumulated (Yes in step S3), the dendrogram generation unit 3 uses the feature vector waveform acquired so far as an analysis target and hierarchically analyzes the accumulated feature data. A cluster analysis method is performed to generate a dendrogram (step S4).

  The threshold determination unit 4 sequentially calculates the amount of change in the intercluster cluster distance for one cluster merge indicated by the generated dendrogram, that is, the merged cluster distance difference, and calculates the merged cluster distance difference. Is determined to be equal to or greater than a threshold value, and the cut position of the dendrogram is determined (step S5). Specifically, the threshold determination unit 4 obtains the minimum number of steps at which the inter-cluster distance difference is equal to or greater than the threshold by merging clusters in the hierarchical cluster analysis method, and performs the merge performed at the obtained number of steps. The merged state of the clusters immediately before processing may be used as the grouping classification result. Then, the classification result is registered in the database 5 (step S6).

  When the dendrogram generation unit 3 performs cluster merging in the dendrogram generation process, the threshold determination unit 4 stores the inter-cluster distance at that time, and is merged with the inter-cluster distance at the previous merge this time. The inter-cluster distance difference may be calculated by obtaining the difference between the inter-cluster distances. In such a case, the dendrogram cutting position can be determined at the time when the distance difference between the merged clusters becomes equal to or greater than the threshold value, and the subsequent dendrogram generation process can be terminated.

  FIG. 3 is an explanatory diagram showing a generation example of a dendrogram generated by the dendrogram generation unit 3. In FIG. 3, data for 90 feature vector waveforms obtained by generating 10 feature vector waveforms at 10 wireless stations of different models are analyzed, and two variables per data are feature data. The dendrogram obtained as a result of performing hierarchical cluster analysis on 90 data using the calculated feature value data is shown. In this example, the Ward method, which is a kind of hierarchical cluster analysis method, is applied.

  The vertical axis in FIG. 3 represents the data number (1 to 90), and the horizontal axis represents the intercluster distance. From left to right, 90 pieces of data are merged one cluster at a time in descending order of similarity, and finally one cluster is collected at the right end. Assuming that one step is to merge two clusters into one, 90 data are merged into one cluster in 89 steps. Here, it is assumed that one cluster is assigned to each data number as an initial cluster. In the following, for the sake of simplicity, the notation of a specific example of a cluster takes the form of “cluster [data number belonging to]”.

  As an example of the number of steps, for example, FIG. 3 shows an example in which the clusters [1], [9], and [6] are merged into one cluster above the vertical axis. Cluster [1] and Cluster [9] are merged into Cluster [1,9] and then Cluster [1,9] and Cluster [6] are merged, or Cluster [9] and Cluster [9] [6] is merged into cluster [9,6], then cluster [9,6] and cluster [1] are merged. Two steps are counted. Similarly, when clusters [2], [3], [4], [8], [10], [5], [7] are merged into one cluster, 6 steps are counted. One step is counted when the cluster [1, 9, 6] and the cluster [2, 3, 4, 8, 10, 5, 7] are integrated.

  In the example shown in FIG. 3, the threshold value is set to 0.5. In the process of merging clusters from left to right in the figure, a point (about 0.9) where the difference in distance between merged clusters first exceeds the threshold is indicated by a dotted line. Further, the cutting position determined by the threshold determination is indicated by a one-dot chain line. According to this cutting position, 90 pieces of data are classified into nine groups (G1 to G9), and a total of nine wireless stations are registered in the database with one group as one wireless station. From the data number of each group, it can be seen that 10 data is correctly classified into the transmission source radio station.

  In addition, since an appropriate threshold value differs depending on the degree of similarity between the elements of the target set, it is preferable that the threshold value corresponding to the environment is determined in advance by, for example, prior evaluation in an assumed environment. .

  Further, when the clusters are finally merged into one without the merged cluster distance difference being equal to or greater than the threshold value, all the data may belong to one group.

  As described above, in the present embodiment, feature vector waveforms with high similarity are grouped and registered as the same radio station, so that recognition errors due to variations in feature vector waveforms can be suppressed, and the accuracy rate of radio station identification Can be improved.

  In this embodiment, since a database for identifying radio stations is automatically created (without human intervention), it is not necessary for the operator to perform grouping by judging the similarity by visual observation or the like. Can reduce the work burden of creating a database.

  In this embodiment, the hierarchical cluster analysis method is applied to the feature amount data calculated from the feature vector waveform, and based on the quantitative index of the change amount of the intercluster distance at the time of cluster merging in the hierarchical cluster analysis. Since the grouping is performed based on the similarity between the feature vector waveforms, it is possible to automatically create a database in which wireless stations are identified with a certain identification performance without the need to determine the number of groups to be classified in advance.

  In the above example, hierarchical cluster analysis is performed using two variables as feature quantity data for one feature vector waveform. However, the type and number of variables used per waveform are appropriate according to the embodiment. It is possible to select anything.

  Hierarchical cluster analysis methods include the Ward method and the group average method, and an appropriate method can be selected according to the application.

  Extraction of the feature vector waveform is described in, for example, a method using FFT as described in Japanese Patent Application Laid-Open Nos. 09-211039 and 2011-44937, and Japanese Patent Application Laid-Open No. 2006-211250. A method using the FM detection processing can be used.

  More specifically, an antenna, a digital spectrum analyzer, and calculation means are provided. When the digital spectrum analyzer receives a radio signal via the antenna, the frequency spectrum of the received radio signal is obtained, and the calculation means calculates the frequency spectrum. A change in the center frequency may be obtained from the data as a feature vector waveform. In addition, for example, an antenna, a receiver, an A / D converter, and a feature vector waveform extraction unit are provided. When the receiver receives a radio signal from a radio station via the antenna, the reception frequency is detected and received. The signal is converted into an intermediate frequency signal, the A / D converter converts the intermediate frequency signal sent from the receiver into a digital complex envelope signal, and the feature vector waveform extraction unit sends the digital complex signal sent from the A / D converter. The envelope signal may be subjected to FM detection processing, a rising edge in the obtained FM detection signal may be detected, and a frequency fluctuation seen at the rising edge may be extracted as a feature vector waveform.

  Further, the radio station to be classified may be a radio base station.

Embodiment 2. FIG.
Next, a second embodiment of the present invention will be described. FIG. 4 is a block diagram illustrating a configuration example of the radio wave monitoring apparatus according to the second embodiment. The radio wave monitoring apparatus 20 shown in FIG. 4 has a configuration (specifically, an arithmetic processing unit 2, a dendrogram generation unit 3, a threshold value determination unit 4, a database 5) included in the radio station database creation device 1 shown in FIG. In addition, a radio wave detection unit 21 and a monitoring unit 22 are provided.

  The arithmetic processing unit 2, dendrogram generation unit 3, threshold value determination unit 4, and database 5 are basically the same as those in the first embodiment. Note that the arithmetic processing unit 2 of the present embodiment performs arithmetic processing on the feature vector waveform detected and extracted by the radio wave detection unit 21. In addition, the threshold determination unit 4 of the present embodiment may notify the monitoring unit 6 when the database 5 is updated.

  In addition, the database 5 stores information indicating the grouping result by the threshold determination unit 4 as a plurality of times as an analysis result history in chronological order in order to see a temporal transition of the identification result of the radio station. Also good. When the grouping is performed cumulatively while increasing the feature vector waveforms, a plurality of grouping results may be stored for one feature vector waveform.

  The radio wave detection unit 21 includes a receiver that receives a radio signal transmitted from a radio station. When the receiver receives a radio signal, the rising waveform at that time, the position information of the transmission source, and signal contents (sound, modulation) Method). Although the position information may be the direction of arrival, it is also possible to use position information calculated based on information detected from a plurality of location servers or the like.

  The monitoring unit 22 detects an illegal radio station based on information (including the analysis result history) indicating the identification result of the radio station stored in the database 5 and issues a warning or the like as necessary.

  As a method for detecting illegal radio stations by the monitoring unit 22, for example, the feature vector waveform of each radio station identified by the grouping by the threshold determination unit 4 is scanned, and a feature vector waveform that causes illegal radio wave emission is predetermined. A method of detecting the radio station as an illegal radio station when more than one is included is used.

  Further, for example, as a result of comparing the feature vector waveform of the regular radio station registered in advance with the feature vector waveform of each radio station identified by the grouping by the threshold determination unit 4, When dissimilar radio stations are identified, a warning that there is a possibility of illegal radio stations may be issued.

  Also, when the feature vector waveform of the regular radio station cannot be registered in advance, such as when there are an unspecified number of regular radio stations, the identification result of the radio station and the content of the signal transmitted by each radio station are displayed. You may verify together. For example, the signals identified as different radio stations include a signal to which information (ID or the like) indicating the same radio station is added, or conversely, the signals identified as the same radio station are different. If a signal to which information (ID, etc.) indicating a wireless station is added is included, there is a possibility that there is an illegal wireless station that transmits a signal by impersonating a legitimate wireless station, and this is warned. May be.

  For example, if at least one feature vector waveform classified into a different group in the previous analysis is included in the next grouping target for each group, the classification result follows the past classification result and the latest feature vector waveform. Can be obtained. It should be noted that whether or not to make the previous analysis result follow the next analysis and which group information to carry over next in the case of making it follow may be switched by setting. It is also possible to perform two types of analysis at the same time, with or without following. In addition, long-term analysis (for example, analysis using data for the past month) and short-term analysis (analysis using data for one day) are performed at a predetermined timing, and detection methods to be applied respectively. It may be used properly, for example, by changing the long-term analysis results or following the short-term analysis results.

  As described above, in this embodiment, information acquisition is automatically performed and analysis is performed, and as a result, if there is a possibility of an illegal radio station, a warning can be given to that effect. You can respond to the station 24 hours a day.

  Next, the outline of the present invention will be described. FIG. 5 is a block diagram showing an outline of the radio station database creation apparatus of the present invention. The radio station database creation apparatus shown in FIG. 5 includes a tree diagram generation unit 101, a grouping unit 102, and a database registration unit 103.

  The tree diagram generation unit 101 performs a hierarchical cluster analysis on the acquired feature vector waveform, which is a frequency fluctuation waveform at the time of signal rise, using a variable calculated from the feature vector waveform to generate a tree diagram. Generate. The dendrogram generation unit 101 is shown as the dendrogram generation unit 3 in the above embodiment.

  The grouping means 102 is a feature vector based on the difference in distance between merged clusters, which is the amount of change in the distance between merged clusters related to one cluster merge, indicated by the tree diagram generated by the tree diagram generating means 101. Group waveforms together. The grouping unit 102 is shown as the threshold value determination unit 4 in the above embodiment.

  The database registration unit 103 registers the feature vector waveforms classified into the same group by the grouping unit 102 in the database as belonging to the same radio station. The database registration unit 103 is shown as the threshold value determination unit 4 in the above embodiment.

  The grouping means 102 may group the feature vector waveforms with a point where the inter-cluster distance difference exceeds a predetermined threshold as a cutting position of the tree diagram.

  The tree diagram generation unit 101 may perform hierarchical cluster analysis using a plurality of variables representing the degree of frequency variation calculated from the feature vector waveform.

  In addition, the tree diagram generation unit 101 uses a classification state in which each feature vector waveform is one cluster as an initial cluster, and based on a variable calculated from each feature vector waveform, between the feature vector waveforms belonging to the cluster or cluster. The tree diagram may be generated by sequentially repeating the cluster merging process of obtaining the distance and merging the pair of clusters having the closest distance into one cluster.

  Further, the grouping means 102 stores the intercluster cluster distance at the time when the cluster diagram generation means 101 performs the cluster merging process in the process of generating the tree diagram, and the intercluster distance at the time of the previous merging and this time The difference between the intercluster clusters to be merged is calculated, the intercluster distance difference is calculated, it is determined whether the calculated intercluster distance difference is equal to or greater than a predetermined threshold, and the intercluster distance difference is calculated. The cutting position of the dendrogram may be determined when the threshold value is exceeded. According to such a configuration, subsequent tree diagram generation processing can be omitted.

  FIG. 6 is a block diagram showing an outline of the radio wave monitoring apparatus of the present invention. The radio wave monitoring apparatus shown in FIG. 6 includes radio wave detection means 105 in addition to each means (that is, tree diagram generation means 101, grouping means 102, database registration means 103) provided in the above-described radio station database creation apparatus. The monitoring means 106 is further provided.

  The radio wave detection means 105 receives a radio signal transmitted from a radio station and extracts a feature vector waveform that is a frequency fluctuation waveform at least when the signal rises. The radio wave detecting means 105 is shown as the radio wave detecting unit 21 in the above embodiment. In the configuration shown in FIG. 6, the tree diagram generation unit 101 performs a hierarchical cluster analysis on the feature vector waveform extracted by the radio wave detection unit 105 using a variable calculated from the feature vector waveform. A dendrogram may be generated.

  The monitoring unit 106 detects an illegal radio station based on information registered in the database and indicating the identification result of the radio station registered by the database registration unit 103. Note that the monitoring means 106 is shown as the monitoring unit 22 in the above embodiment.

  By providing such a configuration, an illegal radio station can be automatically detected.

  The present invention is not limited to the use for identifying a radio station, but can be suitably applied to, for example, a use for grasping the number of nearby radio stations and a use for detecting illegal radio stations.

DESCRIPTION OF SYMBOLS 1 Radio station database preparation apparatus 2 Arithmetic processing part 3 Dendrogram production | generation part 4 Threshold determination part 5 Database 20 Radio wave monitoring apparatus 21 Radio wave detection part 22 Monitoring part 101 Tree diagram production | generation means 102 Grouping means 103 Database registration means 104 Database 105 Radio wave detection means 106 Monitoring means

Claims (10)

Tree diagram generation means for generating a tree diagram by performing hierarchical cluster analysis on a feature vector waveform that is a frequency fluctuation waveform at the time of signal rise, using a variable calculated from the feature vector waveform;
Grouping means for grouping feature vector waveforms based on the intercluster cluster distance difference, which is the amount of change in the intercluster cluster distance for one cluster merge shown by the tree diagram;
A radio station database creation device comprising: database registration means for registering the feature vector waveforms classified into the same group by the grouping means in a database as belonging to the same radio station. The radio station database creation device according to claim 1, wherein the grouping means groups the feature vector waveforms with a point where the difference in distance between the merged clusters exceeds a predetermined threshold as a cutting position of the tree diagram. The radio station database creation device according to claim 1 or 2, wherein the tree diagram generation means performs a hierarchical cluster analysis using a plurality of variables representing a degree of frequency fluctuation calculated from a feature vector waveform. The tree diagram generation means obtains the distance between the feature vectors belonging to the cluster or the cluster based on the variable calculated from each feature vector waveform, with the classification state where each feature vector waveform is one cluster as the initial cluster. 4. The radio station database creation according to claim 1, wherein a tree diagram is generated by sequentially repeating a cluster merging process of merging a pair of clusters having the closest distance into one cluster. 5. apparatus. The grouping means memorizes the inter-cluster distance at the time of cluster merging in the process of generating the dendrogram by the dendrogram generating means, and the inter-cluster distance at the previous merging and the cluster to be merged this time Calculate the inter-cluster distance difference by calculating the inter-cluster distance difference, determine whether the calculated inter-cluster distance difference is greater than or equal to a predetermined threshold, and the inter-cluster distance difference is greater than or equal to the threshold The radio station database creation device according to claim 4, wherein a cutting position of the tree diagram is determined at a point of time. A radio station database creation device according to any one of claims 1 to 5,
Radio wave detection means for receiving a radio signal transmitted from a radio station and extracting a feature vector waveform which is a frequency fluctuation waveform at the time of signal rise;
A radio wave monitoring apparatus comprising: monitoring means for detecting an illegal radio station based on information registered in the database and indicating an identification result of the radio station registered by the database registration means. A tree diagram generation step for generating a tree diagram by performing a hierarchical cluster analysis on the obtained feature vector waveform, which is a frequency fluctuation waveform at the time of signal rise, using a variable calculated from the feature vector waveform; ,
A grouping step for grouping feature vector waveforms based on the inter-cluster distance difference, which is the amount of change in the inter-cluster distance for one cluster merge shown by the tree diagram;
A database registration step of registering in the database the feature vector waveforms classified into the same group by the grouping step as belonging to the same radio station. A radio wave detection step of receiving a radio signal transmitted from a radio station and extracting a feature vector waveform which is a frequency fluctuation waveform at the time of signal rise;
A tree diagram generation step of generating a tree diagram by performing a hierarchical cluster analysis on the extracted feature vector waveform using a variable calculated from the feature vector waveform;
A group for grouping feature vector waveforms based on a difference in distance between merged clusters, which is a change in distance between merged clusters for one cluster merge, indicated by the tree diagram generated in the tree diagram generation step. Step,
A database registration step of registering the feature vector waveforms classified into the same group in the grouping step in the database as belonging to the same radio station;
And a monitoring step of detecting an illegal radio station based on information indicating the identification result of the radio station registered by the database registration means registered in the database. On the computer,
A tree diagram generation process for generating a tree diagram by performing a hierarchical cluster analysis on a feature vector waveform that is a frequency fluctuation waveform at the time of signal rise, using a variable calculated from the feature vector waveform,
A grouping process for grouping feature vector waveforms based on a difference in distance between merged clusters, which is a change amount of a distance between merged clusters for one cluster merge, shown by the tree diagram, and by the grouping process A radio station database creation program for executing database registration processing for registering feature vector waveforms classified into the same group into a database as belonging to the same radio station. On the computer,
Radio wave detection processing that receives a radio signal transmitted from a radio station and extracts a feature vector waveform that is a frequency fluctuation waveform at the time of signal rise;
A tree diagram generation process for generating a tree diagram by performing a hierarchical cluster analysis on the extracted feature vector waveform using a variable calculated from the feature vector waveform;
A group that groups feature vector waveforms based on a difference in distance between merged clusters, which is a change in distance between merged clusters for one cluster merge, indicated by the tree diagram generated in the tree diagram generation process. Processing,
A database registration process for registering the feature vector waveforms classified into the same group by the grouping process in a database as belonging to the same radio station, and a radio station registered by the database registration means registered in the database. A radio wave monitoring program for executing monitoring processing for detecting illegal radio stations based on information indicating the identification result.

Images