JP4403042B2 - Signal classification device - Google Patents

Description

  The present invention relates to a signal classification device that receives a signal emitted from an unknown number of transmission sources and classifies the received signal for each transmission source.

Conventionally, a signal is classified by repeating the following three steps each time a signal is received (see, for example, Non-Patent Document 1). Hereinafter, an example in which signals are classified for each transmission source is referred to as a hypothesis.
Step 1: The newly received signal and the selected hypothesis are combined to generate a new hypothesis.
Step 2: The evaluation value of the generated hypothesis is calculated.
Step 3: Select hypotheses based on the calculated evaluation value.

Ishikawa, Harazawa, Iwamoto, “Proposal of Radio Classification Using Multiple Hypothesis Correlation”, IEICE SANE2003-6, April 2003, P29-34

  In the conventional method described above, in the hypothesis generation in Step 1, each transmission source always thinks that radio wave emission is continued, and a new hypothesis is generated. However, in practice, the transmission source may stop radio wave emission, and in such a case, there is a problem of misclassification after the radio wave emission of the transmission source is stopped.

  The present invention has been made to solve the above-described problems, and an object of the present invention is to provide a signal classification device that can prevent a classification error occurring after radio wave transmission of a transmission source is stopped.

  A signal classification apparatus according to the present invention is a signal classification apparatus that receives a signal radiated from an unknown number of transmission sources and classifies the received signal for each transmission source, and detects the received signal and includes a plurality of types of feature values thereof. For each transmission source that is estimated to have emitted a signal by combining a signal detection means that extracts and outputs the signals in the order of appearance of the signals, a hypothesis that classifies the signals from the signal detection means, and a hypothesis from the hypothesis integration means A hypothesis generation means for generating a hypothesis in which the signals are classified, an evaluation value calculation means for calculating an evaluation value of the hypothesis from the hypothesis generation means and outputting the hypothesis to the hypothesis, and a hypothesis evaluation value from the evaluation value calculation means Based on the hypothesis number reducing means for selecting hypotheses based on the hypothesis and reducing the number of hypotheses, and when the output request flag is input, one hypothesis having the highest evaluation value among the hypotheses from the hypothesis number reducing means is selected. Select and sort When the output request flag is not input, a hypothesis selection unit that outputs all hypotheses from the hypothesis number reduction unit, a hypothesis from the hypothesis selection unit, and the signal detection unit newly output It is determined whether the transmission source of each hypothesis is a radio wave emission stop based on the signal time information, and the hypothesis including the transmission source determined to be the radio wave emission stop based on the determination result and other corrections are unnecessary. A transmission stop determination means for outputting a hypothesis, and a determination as to whether or not a signal classified into each transmission source determined to stop radio wave emission satisfies a predetermined condition by each hypothesis from the transmission stop determination means. Noise determining means for outputting a hypothesis including a transmission source determined as noise based on the determination result and other correction-needed hypotheses, and a signal determined as noise by the noise determining means as noise. Correction of the hypothesis to be reclassified and a correction means for recalculating the hypothesis evaluation value according to the correction content and adding it to the hypothesis and outputting it as a corrected hypothesis; a hypothesis having the same content as the corrected hypothesis from the correction means If there is a modified hypothesis with the same content, the remaining hypothesis is discarded, leaving one modified hypothesis out of the corrected hypotheses with the same content And when there is no corrected hypothesis having the same content, the corrected hypothesis integrating means for outputting the corrected hypothesis from the correcting means as it is, the corrected hypothesis from the corrected hypothesis integrating means and the transmission stop judgment If the corrected hypothesis having the same content as the correction unnecessary hypothesis exists, the corrected hypothesis having the same content is discarded, and the corrected hypothesis remaining without being discarded is compared. Hypotheses and all corrections An unnecessary hypothesis is output to the hypothesis generation means, and when there is no corrected hypothesis having the same content as the correction unnecessary hypothesis, all hypotheses and all correction unnecessary hypotheses are output to the hypothesis generation means. And an integration means.

  According to the present invention, it is possible to determine the stop of radio wave emission of the transmission source and reflect it in hypothesis generation, and therefore it is possible to prevent a classification error that occurs after radio wave emission of the transmission source. Moreover, when the signal classified into the transmission source determined to stop radio wave emission does not satisfy a predetermined condition, they classify noise incorrectly. For this reason, a hypothesis having a wrong classification as described above is detected and the hypothesis is corrected. As a result of hypothesis correction, even if hypotheses having the same contents are generated, they can be combined into one, so that many hypotheses can be handled, and as a result, performance improvement can be expected.

  First, the present invention will be schematically described. In the present invention, it is determined whether or not each transmission source of each hypothesis continues to emit radio waves by comparing the feature quantities of the newly classified signal and the signal classified into each transmission source. When it is determined that the radio wave emission is stopped, a process in which the signal is not classified is added to the transmission source. Further, if the number of signals classified as transmission sources determined to stop radio wave emission is extremely small, it is considered that they have classified noise incorrectly. Although a hypothesis in which these signals are correctly classified as noise is also generated, there is a possibility that a hypothesis correctly classified in step 3 in the conventional example is already discarded. Therefore, when an error in the hypothesis as described above is found, the erroneously classified signal is classified as noise to correct the hypothesis. Furthermore, since there is a possibility that a hypothesis having the same content as the revised hypothesis exists, search for the existence of a hypothesis having the same content, and if there is a hypothesis having the same content, leave one of the hypotheses having the same content. And discard the rest. As a result, even if a hypothesis having an error in classification is generated, it is detected and corrected. Even when a plurality of hypotheses having the same contents exist as a result of the correction, they are retrieved and integrated into one hypothesis.

As described above, the following effects can be obtained and an improvement in performance can be expected.
(A) It is possible to prevent a classification error that occurs after radio wave emission of the transmission source is stopped.
(B) It is possible to correct an error in a hypothesis in which noise is erroneously classified as a transmission source.
(C) By combining hypotheses having the same contents into one, it is possible to handle many types of hypotheses.

  In the following description, a signal classified as a transmission source is determined to be noise, and a hypothesis corrected accordingly is referred to as a corrected hypothesis, and a hypothesis that is originally classified correctly and does not need to be corrected is referred to as a correction unnecessary hypothesis.

Next, the basic concept of the present invention will be described.
The signal classification device according to the present invention is intended to classify a signal received by an antenna for each transmission source, and for this purpose, the signal property such as an attribute value calculated from the physical property of the signal or the physical property is used. The signal is classified by paying attention to each feature amount. This is because signals radiated from the same transmission source are considered to have similar feature quantities.

  Therefore, signals can be classified for each transmission source by utilizing the similarity of these feature amounts. For example, if the position of the transmission source that radiates the signal and the position of the reception source that receives the signal are both fixed, the azimuth (reception azimuth angle) for receiving the signal radiated by one transmission source is equal, and so on. The signal can be classified for each transmission source by using the property.

  In addition, since the signal (frequency fixed signal) from the frequency fixed transmission source that emits a signal having a fixed frequency is similar to the carrier frequency of the signal, it can be classified using the similarity of the carrier frequency. Furthermore, a signal (frequency variation signal) from a frequency variation transmission source that periodically changes the frequency has a similar duration from the appearance of the signal to the disappearance at one frequency, and the signal appearance time is also periodic. Therefore, it is possible to classify using these.

  Specifically, a plurality of hypotheses in which signals are classified are generated, and an evaluation value of the hypothesis is calculated for the generated hypotheses using the above-described feature amount similarity. Then, the hypothesis having the highest evaluation value is set as the classification result. However, in this method, the number of hypotheses generated increases exponentially with the number of signals to be classified. For this reason, if a plurality of signals are processed at once, the number of hypotheses generated increases explosively, which places a heavy load on the computer.

  Therefore, a hypothesis is generated for each signal, an evaluation value is calculated for the generated hypothesis, and a plurality of hypotheses are selected based on the calculated evaluation value. Then, the selected hypothesis and the signal to be classified next are combined to generate a new hypothesis, the evaluation value of the generated hypothesis is calculated, and the process of selecting a plurality of hypotheses based on the evaluation value is repeated. Classify. In the case of this method, since the number of signals to be classified at one time is one, the number of generated hypotheses does not become enormous, and the signals can be classified while reducing the load on the computer.

  In the generation of hypotheses by the above method, the signal to be classified is classified as either a transmission source of each hypothesis or noise. In each hypothesis, the hypothesis that the transmission source stopped emitting radio waves was not generated, and as a result, the difference between the disappearance time of the signal last classified in the transmission source and the appearance time of the signal to be classified Even if it is large, hypotheses classified into the same transmission source have been generated. However, in reality, a transmission source that cannot receive a signal for a long time is likely to have already stopped radio wave emission. Therefore, if the difference between the extinction time of the signal last classified as a transmission source and the appearance time of the signal to be classified satisfies a predetermined condition, the transmission source generates a hypothesis that radio wave emission has already stopped. Then, the transmission source determined to stop radio wave emission is not classified thereafter.

  Furthermore, when the number of classified signals is extremely small at the transmission source determined to stop radio wave emission, these signals are inherently noise and are likely to be classified incorrectly. Therefore, in such a case, the hypothesis is corrected by reclassifying the signal classified as the transmission source determined to be the radio wave radiation stop as noise. However, depending on the correction contents of the hypothesis, (1) a plurality of corrected hypotheses having the same contents may be generated, or (2) the corrected hypothesis may have the same contents as the correction unnecessary hypothesis. Since it is completely useless to handle a plurality of hypotheses having the same content, the presence or absence of a hypothesis having the same content is detected as a result of correction of the hypothesis, and the remaining one of the hypotheses having the same content is discarded.

Next, signals to be classified and feature quantities will be described. The signals are distinguished in the order in which they are received. Therefore, the nth received signal is referred to as an “nth signal”. Then, among the plurality of feature amounts extracted from the “nth signal”, the value of the jth feature amount is represented as f _{n, j} . Further, all types of feature quantities of the “nth signal” are assumed to be feature quantity vectors F _n . These relationships are shown in the following formula (1).

  Next, hypotheses used in the present invention will be described. First, a hypothesis obtained by classifying “1st to n-th signals” is referred to as an “n-signal hypothesis”. When “n-signal hypotheses” are arranged in an arbitrary order, the h-th “n-signal hypothesis” Is called “n-signal hypothesis h”. A hypothesis that classifies the “nth signal” is referred to as an “nth signal hypothesis”. Since there are a plurality of “nth signal hypotheses”, the rth “nth signal hypothesis” is referred to as “nth signal hypothesis r”. The “n-signal hypothesis” is a hypothesis that classifies “1st to n-th signals”, whereas the “n-th signal hypothesis” classifies one signal (“n-th signal”). It is a hypothesis.

  Next, a hypothesis generation method will be described. FIG. 1 shows two “three-signal hypotheses” that tentatively classify “first to third signals”. As shown in FIG. 1, “3 signal hypothesis 1” is that the received signals 301 and 303 are regarded as signals from one transmission source, classified as the first transmission source, and one received signal 302 is generated. It represents a hypothesis that is considered to be a signal from a transmission source different from the transmission source of the eye and is classified as a second transmission source. Further, “3 signal hypothesis 2” represents a hypothesis in which the signals 301, 302, and 303 are regarded as signals from one transmission source and are classified as the first transmission source.

  Here, consider a case where the “fourth signal” is newly classified. For “3 signal hypothesis 1” in FIG. 1, “4th signal” is classified as the first transmission source, “4th signal hypothesis 1”, and it is classified as the second transmission source. “Fourth signal hypothesis 2”, “fourth signal hypothesis 3” or the like classified into the third transmission source can be considered. A plurality of “four-signal hypotheses” are generated by combining “three-signal hypothesis 1” and a plurality of “fourth signal hypotheses”. Similarly, since there are multiple “fourth signal hypotheses” for “three-signal hypothesis 2”, combine these “fourth signal hypotheses” with “three-signal hypothesis 2”. As a result, a plurality of “four-signal hypotheses” are generated. The above is the explanation of the hypothesis that classifies the signal and the generation method thereof. Hereinafter, an embodiment of a signal classification apparatus using the hypothesis will be described.

Embodiment 1 FIG.
FIG. 2 is a block diagram showing the configuration of the signal classification apparatus according to Embodiment 1 of the present invention. The signal classification apparatus shown in FIG. 2 is a signal classification apparatus that receives a signal radiated from an unknown number of transmission sources, classifies the received signal for each transmission source, detects a received signal, and extracts its feature amount. The signal detection means 10 for outputting the signals in the order of appearance of the signals, the hypothesis obtained by classifying the signals from the signal detection means 10 and the hypothesis output from the hypothesis integration means 100 are combined to classify the signals for each transmission source from which the signals are emitted. Calculated by the hypothesis generation means 20 for generating a new hypothesis, the evaluation value calculation means 30 for calculating the evaluation value of the hypothesis generated by the hypothesis generation means 20 and adding it to the hypothesis and outputting it. If a hypothesis number reduction means 40 for selecting hypotheses based on the evaluation value to be received and reducing the number of hypotheses and an output request flag are input, one hypothesis having the highest evaluation value is selected and classified. Results and Outputs Te, the output request flag if not been input, and a hypothesis selecting unit 50 for outputting a hypothesis transmission stop determination unit 60.

  Further, based on the hypothesis from the hypothesis selection means 50 and the feature amount of the signal from the signal detection means 10, it is determined whether or not each transmission source of the hypothesis has stopped radio wave emission, and radio wave emission is stopped. The hypothesis including the transmission source determined to be present is output to the noise determination unit 70, and the other hypothesis is output to the hypothesis integration unit 100 as a hypothesis that does not require correction. Based on the feature amount of the signal classified into the transmission source of the radio wave radiation stop, it is determined whether or not the signal is noise. If it is determined that the signal is noise, the correction unit 80 If not, a noise determination unit 70 that outputs to the hypothesis integration unit 100 as a correction-necessary hypothesis, correction of a hypothesis that reclassifies a signal determined to be noise as noise, and within the correction Was added to the hypothesis calculates the evaluation value of the hypothesis again according, and a correcting means 80 to output as the modified hypothesis.

  Further, the corrected hypothesis from the correcting means 80 is searched for whether there is the same content. If there is a corrected hypothesis having the same content, one of the hypotheses having the same content is left. The remaining hypothesis is discarded, and when there is no correction, the input corrected hypothesis is output to the hypothesis integration unit 100 as it is, the corrected hypothesis integration unit 90, the corrected hypothesis from the correction hypothesis integration unit 90, the transmission stop determination unit 60, and the noise The correction hypothesis from the judgment means 70 is compared, and if there is a corrected hypothesis having the same content as the correction unnecessary hypothesis, the corrected hypothesis having the same content is discarded, and the remaining corrected hypotheses and all the correction unnecessary hypotheses are hypothesized. Hypothesis integration means 100 that outputs to generation means 20 is provided.

  Next, FIG. 3 is a block diagram showing a detailed configuration of the signal detection means 10. As shown in FIG. 3, the signal detection means 10 includes an analog / digital conversion means 11 that converts a received analog signal into a digital signal, and a fast Fourier transform means 12 that performs a fast Fourier transform on the digitally converted data. The signal component detecting means 13 for detecting the signal component from the spectrum output from the fast Fourier transform means 12, the feature quantity extracting means 14 for extracting the feature quantity of the detected signal component, and the feature quantity detecting means 15. And a signal sorting means 15 for outputting the feature quantity vectors of the signals in the order of appearance of the signals.

  Next, FIG. 4 is a block diagram illustrating a detailed configuration of the transmission stop determination unit 60. As shown in FIG. 4, the transmission stop determination unit 60 includes the disappearance time of the signal finally classified into each transmission source of the hypothesis from the hypothesis selection unit 50 and the feature vector of the new signal from the signal detection unit 10. A threshold is set for the difference from the appearance time of, and it is determined that transmission sources that do not satisfy the threshold continue to emit radio waves, and transmission sources that exceed the threshold are determined to have stopped radio emission, and radio wave emission is determined to be stopped. The hypothesis including the transmitted source is included in the noise determination unit 70, and the other hypotheses are output to the hypothesis integration unit 100 as a correction unnecessary hypothesis.

  FIG. 5 is a block diagram showing a detailed configuration of the noise determination means 70. As shown in FIG. 5, the noise determination unit 70 sets a threshold for the total duration time of the signals classified as the transmission sources determined as the radio wave emission stop by the hypothesis from the transmission stop determination unit 60, and transmission that does not satisfy the threshold The source signal is determined to be a misclassification of noise, hypotheses including these transmission sources are output to the correction means 80, and other hypotheses are output to the hypothesis integration means 100 as correction-necessary hypotheses. 71.

  FIG. 6 is a block diagram showing a detailed configuration of the correction means 80. As shown in FIG. 6, the correction means 80 is a hypothesis correction means 81 that corrects a hypothesis by reclassifying a signal determined to be noise to a hypothesis from the noise determination means 70, and a hypothesis correction content. And an evaluation value correcting means 82 for re-calculating the hypothesis evaluation value, adding it to the hypothesis, and outputting it as a corrected hypothesis to the corrected hypothesis integrating means 90.

  FIG. 7 is a block diagram showing a detailed configuration of the modified hypothesis integrating means 90. As shown in FIG. 7, the corrected hypothesis integrating unit 90 searches the corrected hypothesis from the correcting unit 80 for the presence or absence of the same content hypothesis. If there is a hypothesis with the same hypothesis search unit 91 that outputs all hypotheses to the hypothesis integration unit 100, and if there is a hypothesis with the same content, one of the hypotheses with the same content is discarded and the rest is discarded, and the remaining corrected hypothesis is integrated with the hypothesis And a modified hypothesis discarding unit 92 that outputs to the unit 100.

  FIG. 8 is a block diagram showing a detailed configuration of the hypothesis integration unit 100. As shown in FIG. 8, the hypothesis integration unit 100 compares the correction unnecessary hypothesis output from the transmission stop determination unit 60 and the noise determination unit 70 with the corrected hypothesis output from the correction hypothesis integration unit 90, and corrects the correction. A hypothesis comparison unit 101 that discards a corrected hypothesis having the same content as that of the corrected hypothesis and outputs the remaining corrected hypotheses and all of the unnecessary hypotheses to the hypothesis generation unit 20 ing.

  Next, processing of the signal classification device 1 will be described. The analog / digital conversion means 11 in the signal detection means 10 samples an input analog reception signal at a constant interval and outputs digitally converted data. The fast Fourier transform unit 12 performs a fast Fourier transform process on the digitally converted data and outputs a spectral component of a signal included in the data. Thereafter, the signal component detection unit 13 detects the signal component from the spectrum component of the signal output from the fast Fourier transform unit 12. As the method, there are a method of comparing with a preset threshold value, a method of detecting a value exceeding the threshold value by using a value obtained by adding or multiplying a preset signal detection parameter to an average value for each fast Fourier transform process, etc. is there. Then, the feature quantity extraction unit 14 extracts the feature quantity (appearance time, disappearance time, duration, carrier frequency, reception azimuth angle, frequency bandwidth, peak power, etc.) of the signal component detected by the signal detection unit 13. The feature vector Fn is output. The signal sorting unit 15 outputs the feature amount vectors to the hypothesis generation unit 20 one by one in the order of appearance times of the feature amount vectors Fn extracted from the plurality of signals.

  Hereinafter, as an example of a signal to be classified, a case where a fixed frequency signal and a frequency variation signal are included will be described as an example. In the following process, the process will be described in three cases: when an initialization flag is input, when an output request flag is input, and when none is input.

(When the initialization flag is output)
When the initialization flag is output, the hypothesis generation means 20 initializes the hypothesis, generates the “first signal hypothesis” for the “first signal”, and sets the “first signal hypothesis”. It is output to the evaluation value calculation means 30 as the “first signal”. Since “noise hypothesis of the first signal” may be detected, the hypothesis generation means 20 generates the following three hypotheses (1), (2), and (3). .
(1) A hypothesis for classifying a signal into a signal radiated by the first frequency variation transmission source.
(2) A hypothesis for classifying a signal into a signal radiated by the first fixed frequency transmission source.
(3) A hypothesis for classifying a signal as a signal in which noise is erroneously detected.

  The evaluation value calculation means 30 initializes the hypothesis evaluation value added to each hypothesis, obtains a hypothesis evaluation value, adds it to each hypothesis, and outputs it to the hypothesis number reduction means 40. The method for calculating the hypothesis evaluation value will be described in the description of the processing (when neither the initialization flag nor the output request flag is output).

  The hypothesis number reduction means 40 selects hypotheses based on the hypothesis evaluation value added to the hypothesis, and reduces the number of hypotheses. This method is the same as the processing (when neither the initialization flag nor the output request flag is output), and will be described later.

  The hypothesis selection unit 50 outputs the hypothesis from the hypothesis number reduction unit 50 to the transmission stop determination unit 60.

  The transmission stop determination unit 60 compares the hypothesis output from the hypothesis selection unit 50 with the feature quantity vector of the signal output from the signal detection unit 10, and determines whether the transmission source of each hypothesis stops radio wave emission. Determine whether. Then, the hypothesis including the transmission source determined to stop radio wave emission is output to the noise determination unit 70, and the hypothesis other than that is output to the hypothesis integration unit 100 as a correction unnecessary hypothesis. This method is also the same as the processing (when neither the initialization flag nor the output request flag is output), and will be described later.

  The noise determination unit 70 determines whether or not the signals classified as the transmission sources determined to stop radio wave emission are the noises classified incorrectly in each hypothesis output from the transmission stop determination unit 60. The hypothesis including the signal determined and judged as noise is output to the correction means 80, and the hypothesis other than that is output to the hypothesis integration means 100 as a correction unnecessary hypothesis. This method is also the same as the processing (when neither the initialization flag nor the output request flag is output), and will be described later.

  The correction means 80 performs hypothesis correction on the hypotheses output from the noise determination means 70 to reclassify the signal determined to be noise into noise, and recalculates hypothesis evaluation values in accordance with the correction. Are added to the hypothesis and output to the corrected hypothesis integrating means 90 as a corrected hypothesis. This method is also the same as the processing (when neither the initialization flag nor the output request flag is output), and will be described later.

  The corrected hypothesis integration unit 90 searches for a hypothesis having the same content in the corrected hypothesis from the correction unit 80, and if it is included, leaves one of the hypotheses having the same content and discards the rest. , Put together hypotheses with the same content. This method is also the same as the processing (when neither the initialization flag nor the output request flag is output), and will be described later.

  The hypothesis integration unit 100 compares the hypothesis content of the corrected hypothesis output from the correction hypothesis integration unit 90 with the correction unnecessary hypothesis output from the transmission stop determination unit 60 and the noise determination unit 70, and the same content as the corrected hypothesis. If there is an unnecessary correction hypothesis, the corrected hypothesis having the same content is discarded. This method is also the same as the processing (when neither the initialization flag nor the output request flag is output), and will be described later.

(When neither the initialization flag nor the output request flag is output)
When the initialization flag is not input, that is, when the n-th (where n> 1) signal is input, the hypothesis generation unit 20 classifies the feature vector signal newly output by the signal detection unit 10. The generated hypothesis is combined with the hypothesis output from the hypothesis integration unit 100 to generate a new hypothesis. Here, the hypothesis output by the hypothesis integration unit 100 is a hypothesis in which “1 to (n−1) -th signals” are classified. Therefore, the hypothesis output by the hypothesis integration unit 100 is “(n−1) signal hypothesis”. The hypothesis generation means 20 classifies the “(n−1) signal hypothesis” and the feature vector signal (“nth signal”) newly output by the signal detection means 10 into the “nth signal hypothesis”. ”To generate an“ n-signal hypothesis ”.

Here, a case where the “nth signal” is classified based on “(n−1) signal hypothesis h” will be described. The “(n−1) signal hypothesis h” is a hypothesis having M frequency variation transmission sources and K frequency fixed transmission sources. It is assumed that both the M frequency variation transmission sources and the K frequency fixed transmission sources continue to emit radio waves. In this case, the hypothesis generation means 11 generates the “nth signal hypothesis” in the following (4) to (9), and combines the “(n−1) signal hypothesis h” with a new “n signal hypothesis”. Generate multiple hypotheses.
(4) A plurality of “nth signal hypotheses” that classify the “nth signal” into the mth frequency variation transmission source (where 1 ≦ m ≦ M).
(5) After detecting one or more signals radiated from the m-th frequency variation transmission source, a plurality of “n-th signal” is a signal radiated from the m-th frequency variation transmission source. nth signal hypothesis ".
(6) “nth signal” is assumed to be a signal radiated from a new frequency fluctuation transmission source ((M + 1) th frequency fluctuation transmission source) that has not been classified so far. Hypothesis ".
(7) A plurality of “nth signal hypotheses” that classify the “nth signal” into k-th frequency fixed transmission sources (where 1 ≦ k ≦ K).
(8) “nth signal” is a signal radiated from a new frequency fixed transmission source ((K + 1) th frequency fixed transmission source) that has no previously classified signal. Hypothesis ".
(9) “nth signal hypothesis” in which “nth signal” is a signal in which noise is erroneously detected.
When the m ′ frequency variation transmission source among the M frequency variation transmission sources is determined to stop radio wave emission, the above (4) becomes (4 ′) below.
(4 ′) A plurality of “n-th signals” for classifying the “n-th signal” into the (m′−1) -th frequency variation transmission source and the m′-M-th frequency variation transmission source. hypothesis".
When the k'th frequency fixed transmission source among the K frequency fixed transmission sources is determined to stop radio wave emission, the above (7) becomes the following (7 ').
(7 ′) A plurality of “nth signals” for classifying the “nth signal” into 1st to (k′−1) th frequency fixed transmission sources and k ′ to Kth frequency fixed transmission sources. hypothesis".

Now, when the total number of “(n−1) signal hypotheses” is H, the process of generating the “nth signal hypothesis” in (4) to (9) is performed on the “(n−1) signal hypothesis”. Hypothesis h (1 ≦ h ≦ H) ”is performed to generate“ n signal hypothesis ”. Furthermore, if the "n-th signal hypothesis" for the "(n-1) signal hypothesis h of" is considered number R _h, hypothesis generation means 11, the equation (2), the total number G of " The n-signal hypothesis ”is output to the evaluation value calculation means 30.

The evaluation value calculation means 30 calculates hypothesis evaluation values for the G “n-signal hypotheses” output from the hypothesis generation means 20, adds the results to each hypothesis, and outputs the hypotheses. Now, it is assumed that the “n-signal hypothesis h” is a hypothesis generated by combining “(n−1) signal hypothesis h ′” and “n-th signal hypothesis r”. In this case, the hypothesis evaluation value of “(n−1) signal hypothesis h ′” is L _p . Also, let P _t be the probability that the “nth signal hypothesis r” indicating the transmission source to which the “nth signal” is classified is correct. Furthermore, when the classification of the “nth signal” indicated by the “nth signal hypothesis r” is correct, the value of the jth feature amount of the “nth signal” is f _{n, j.} The probability P _{f, j is} assumed. In this case, the hypothesis evaluation value L _j of the “n signal hypothesis h” can be obtained by the equation (3).

  Here, j = 1 of the feature quantity indicates the duration, and the “nth signal hypothesis r” is a hypothesis for classifying the “nth signal” into the mth frequency variation transmission source. And the expected value of the duration of the signal classified as the m-th frequency variation transmission source is

, Its standard deviation is

It is assumed that In this case, P _{f, j} (j = 1) is calculated by equation (4).

The evaluation value calculation means 30 calculates a hypothesis evaluation value as described above. When n = 1, that is, when the initialization flag is output, the “(n−1) signal hypothesis” does not exist. In this case (when the initialization flag is output), the feature value evaluation value L _{P, j} of “(n−1) signal hypothesis” is set to 1 (the feature value evaluation value L _{P, j is set} to 1). Means that the feature value evaluation value is initialized).

  The hypothesis number reduction means 40 selects hypotheses according to the hypothesis evaluation values calculated by the evaluation value calculation means 30 and decreases the number of hypotheses. As the method, the number of hypotheses to be selected hs is determined in advance and hs hypotheses having a high hypothesis evaluation value are selected, or a hypothesis evaluation value is provided with a threshold and a hypothesis exceeding the threshold is selected. Can be considered. The threshold value may be a sum of all hypotheses multiplied by a constant, or a hypothesis evaluation value having the highest evaluation value among all hypotheses multiplied by a constant.

  The hypothesis selection unit 50 outputs all hypotheses input from the hypothesis number reduction unit 40 to the transmission stop determination unit 60.

  As shown in FIG. 4, the transmission stop determination means 60 includes a time difference determination means 61, and the signal detection means 10 for each of a plurality of “n-signal hypotheses” transmission sources input from the hypothesis selection means 50. Whether or not radio wave radiation continues is determined using the appearance time of the feature vector of the “(n + 1) -th signal” input from. As a method therefor, the time difference determination means 61 sets a threshold value for the difference between the disappearance time of the signal last classified for each transmission source of each “n signal hypothesis” and the appearance time of the “(n + 1) th signal”. It is determined that the transmission source that has been provided and has exceeded the threshold has already stopped radio wave emission. Then, a hypothesis including a transmission source determined to stop radio wave emission is output to the noise determination unit 70, and a hypothesis other than that is output to the hypothesis integration unit 100 as a correction unnecessary hypothesis.

  Further, as shown in FIG. 9, the transmission stop determination unit 60 generates the appearance time of the signal last classified as the hypothesis transmission source from the hypothesis selection unit 50 and the appearance time of the new signal from the signal detection unit 10. A threshold is set for the number of missed detections estimated based on the difference between the transmission source and the transmission source that does not satisfy the threshold is determined to be continuously radiating, and the transmission source that exceeds the threshold is determined to have stopped radiating. Hypotheses that include detection frequency determination means 62 and that include a transmission source determined to stop radio wave emission are output to noise determination means 70, and hypotheses that are not output to hypothesis integration means 100 as correction-necessary hypotheses. good. Here, the loss detection means that the received signal cannot be detected due to some influence. As shown in FIG. 10, there are originally three signals 401 to 403, of which the signal 402 indicated by the dotted line. Cannot be detected, and only two signals 401 and 403 can be recognized.

  Further, the transmission stop determination unit 60 may include a time difference determination unit 61 and a loss detection number determination unit 62. That is, among the transmission sources of each “n-signal hypothesis” input from the hypothesis selection unit 50, the time difference determination unit 61 determines that the fixed frequency transmission source is stopped, and the frequency variation transmission source is lost. The radio wave radiation stop may be determined by the detection frequency determination means 62, the time difference determination means 61, or both. The number of missed detections judgment means 62 estimates the number of missed detections that occurred between the signal last classified into each frequency fixed transmission source and the “(n + 1) th signal”. If the estimated value exceeds the threshold for the number of missed detections, the frequency variation transmission source determines that the radio wave emission has already been stopped, and outputs it to the noise determination means 70.

Now, it is assumed that the appearance time interval of the signal of the m-th frequency fluctuation transmission source of the “n signal hypothesis h1” is estimated as t ^(m) _SRI . The appearance time of the signal finally classified as the m-th frequency variation transmission source is f ^(m) _{Nm 2} , and the appearance time of the “(n + 1) -th signal” is f _{(n + 1) 2.} . In this case, an estimated value of the number of missed detections occurring between the signal last classified as the mth frequency variation transmission source and the “(n + 1) th signal” is obtained by the following equation (5).

Here, ROUND (*) is a function that rounds * to the nearest integer. If Z exceeds the preset maximum number of missed detections, Z _MAX , it is determined that the m-th frequency variation transmission source has already stopped radio wave emission. The hypothesis including the transmission source determined to be the radio wave emission stop by the time difference determining unit 61 or the number of missed detections determining unit 62 is output to the noise determining unit 70, and the other hypothesis is input to the hypothesis integrating unit 100 as a correction unnecessary hypothesis. Is output.

As shown in FIG. 5, the noise determination unit 70 includes a time noise determination unit 71, and whether or not the signal classified as the transmission source determined as the radio wave emission stop by the transmission stop determination unit 60 is noise. Make a decision. As a method thereof, the time noise determination means 71 provides a threshold for the total duration of signals classified as transmission sources determined to stop radio wave emission, and determines that the total time is less than the threshold as noise. Assume that the k-th frequency fixed transmission source of “n signal hypothesis h2” is determined to stop radio wave emission. It is assumed that the number of signals classified as the kth frequency fixed transmission source is Nk, and the duration thereof is f ^(k) _{n, 1} (1 ≦ n ≦ Nk). Assuming that the threshold for the total duration of noise determination is th _CCT , the condition for determining that the signal classified as the k-th frequency fixed transmission source is noise is the following equation (6).

  Then, a hypothesis including a signal determined to be noise is output to the correction unit 80, and a hypothesis other than that is output to the hypothesis integration unit 100 as a correction unnecessary hypothesis.

Further, as shown in FIG. 11, the noise determination unit 70 may include a signal number noise determination unit 72. In the signal number noise determination means 72, a threshold is provided for the number of signals classified as the transmission source determined as the radio wave emission stop by the transmission stop determination means 60, and a signal whose total number of signals is less than the threshold is determined as noise. . Now, suppose that the kth frequency fixed transmission source of “n signal hypothesis h2” is determined as radio wave radiation stop, and the number of signals classified as the kth frequency fixed transmission source is Nk. Assuming that the threshold value of the number of signals for noise determination is th _nom , the condition for determining that a signal classified as the kth frequency specific transmission source is noise is the following equation (7).

  Further, the noise determination means 70 includes a time noise determination means 71 and a signal number noise determination means 72. The fixed frequency transmission source is processed by the time noise determination means 71, and the frequency fluctuation transmission source is processed by the signal number noise determination means 72. It is also possible to adopt a configuration that Further, a plurality of configurations are conceivable, such as a configuration in which the fixed frequency transmission source and the frequency variation transmission source are processed by the time noise determination unit 71 and the frequency variation transmission source is further processed by the signal number noise determination unit 72.

  As shown in FIG. 6, the correction means 80 includes a hypothesis correction means 81 and an evaluation value correction means 82, corrects the hypothesis by reclassifying each signal determined to be noise by the noise determination means 70, and Recalculate the evaluation value that matches the correction. As a method thereof, the hypothesis correcting unit 81 corrects a hypothesis that classifies a signal determined to be noise as noise with respect to the hypothesis output from the noise determining unit 70. Then, the evaluation value correcting means 82 recalculates the hypothesis evaluation value in accordance with the hypothesis correction classified as the noise. The recalculated evaluation value is added to the corrected hypothesis, and the corrected hypothesis is output to the corrected hypothesis integrating means 90.

  As shown in FIG. 7, the modified hypothesis integrating unit 90 includes an identical hypothesis searching unit 91 and a modified hypothesis discarding unit 92. For each modified hypothesis output from the modifying unit 80, the presence / absence of a modified hypothesis having the same content is determined. Search, and if there are, output them together as one hypothesis. As the method, first, the same hypothesis search unit 91 searches the corrected hypothesis output from the correction unit 90 for the presence or absence of the same content. When there are corrected hypotheses having the same contents, the corrected hypotheses are output to the corrected hypothesis discarding unit 92, and when there are no hypotheses having the same content, they are output to the hypothesis integrating unit 100 as corrected hypotheses. The corrected hypothesis discarding unit 92 discards the remaining one of the corrected hypotheses having the same contents, discards the rest, outputs the corrected hypothesis to the hypothesis integrating unit 100, and discards the remaining hypotheses.

  As shown in FIG. 8, the hypothesis integration unit 100 compares the correction-needed hypotheses output from the transmission stop determination unit 60 and the noise determination unit 70 with the corrected hypothesis output from the correction hypothesis integration unit 90, and has the same contents. If there is a hypothesis, the hypotheses are combined into one hypothesis and output to the hypothesis generation means 20, and if not, the hypotheses output from each means are output to the hypothesis generation means 20 as they are. As a method, the hypothesis comparison unit 101 compares the corrected hypothesis output from the correction hypothesis integration unit 90 with the correction unnecessary hypothesis output from the transmission stop determination unit 60 and the noise determination unit 70. When there is a correction-needed hypothesis having the same content as the corrected hypothesis, the same corrected hypothesis is discarded. All the corrected hypotheses are compared with all the correction unnecessary hypotheses, and the remaining corrected hypotheses and all the correction unnecessary hypotheses are output to the hypothesis generation means 20.

(When the output request flag is output)
In this case, the signal detection means 10, hypothesis generation means 20, evaluation value calculation means 30, hypothesis number reduction means 40, transmission stop judgment means 60, noise judgment means 70, correction means 80, correction hypothesis integration means 90, hypothesis integration means 100 This process is the same as the above-described process (when neither the initialization flag nor the output request flag is output), and thus the description thereof is omitted. On the other hand, the hypothesis selection means 50 selects one hypothesis having the highest hypothesis evaluation value from the hypotheses input from the hypothesis number reduction means 40 and outputs it as a signal classification result. The above is an outline of the present invention.

  In the signal classification device of the present invention, a hypothesis that the transmission source has stopped radio wave emission is also generated, so that classification errors that occur after the transmission source has stopped radio wave emission can be prevented. It also detects and corrects hypotheses that misclassify noise. Furthermore, even if hypotheses having the same contents are generated as a result of correction of the hypotheses, they are integrated, so that unnecessary hypotheses can be conceived, and as many types of hypotheses can be handled accordingly. From the above effects, improvement in classification performance can be expected.

Embodiment 2. FIG.
Since the configuration of the signal classification apparatus according to the second embodiment is the same as that of the first embodiment shown in FIG. 2, FIG. 2 is used instead. Further, the processing of each means is also the signal detection means 10, hypothesis generation means 20, evaluation value calculation means 30, hypothesis number reduction means 40, hypothesis selection means 50, transmission stop judgment means 60, noise judgment means 70, hypothesis integration means 100. Are the same and will not be described. Only the correction means 80 and the correction hypothesis integration means 90 are different.

  As shown in FIG. 12, the correction means 80 includes a hypothesis correction means 81, and corrects the hypothesis by reclassifying the signal determined as noise in the hypothesis output from the noise determination means 70 into noise. The result is output to the corrected hypothesis integration means 90 as a hypothesis. A difference from the correction unit 80 of the first embodiment shown in FIG. 6 is that the evaluation value correction unit 82 is not provided, and the hypothesis evaluation value is not corrected and is output to the corrected hypothesis integration unit 90.

  As shown in FIG. 13, the corrected hypothesis integrating means 90 includes the same hypothesis searching means 91 and the maximum evaluation value hypothesis integrating means 93, and the corrected hypotheses output from the correcting means 80 include hypotheses having the same contents. If they are present, they are integrated into one and output to the hypothesis integrating means 100. As the method, first, the same hypothesis search unit 91 searches the corrected hypothesis output from the correction unit 90 for the presence or absence of a corrected hypothesis having the same content. Then, if there is no hypothesis having the same content, it is output to the hypothesis integrating means 100, and the one having the same content is output to the maximum evaluation value integrating means 93. The maximum evaluation value integration unit 93 selects one hypothesis having the highest hypothesis evaluation value from the inputted hypotheses having the same content, outputs it to the hypothesis integration unit 100, and discards the remaining hypotheses. Except for the two means, the second embodiment is the same as the first embodiment, and a description thereof will be omitted.

  Also in the signal classification device according to the second embodiment, the same effect as in the first embodiment can be obtained. However, since the evaluation value is not corrected, performance may be deteriorated. However, the amount of calculation associated with the correction of the evaluation value can be reduced accordingly.

It is a figure explaining the generation method of a hypothesis in this invention. It is a block diagram which shows the structure of the signal classification apparatus based on this invention. It is a block diagram which shows the detailed structure of the signal detection means 10 which concerns on Embodiment 1 of this invention. It is a block diagram which shows the detailed structure of the transmission stop determination means 60 which concerns on Embodiment 1 of this invention. It is a block diagram which shows the detailed structure of the noise judgment means 70 concerning Embodiment 1 of this invention. It is a block diagram which shows the detailed structure of the correction means 80 which concerns on Embodiment 1 of this invention. It is a block diagram which shows the detailed structure of the correction hypothesis integration means 90 which concerns on Embodiment 1 of this invention. It is a block diagram which shows the detailed structure of the hypothesis integration means 100 which concerns on Embodiment 1 of this invention. It is a block diagram which shows the detailed structure of the transmission stop determination means 60 which concerns on Embodiment 1 of this invention. It is a figure explaining the concept of a loss detection. It is a block diagram which shows the detailed structure of the other example of the noise judgment means 70 concerning Embodiment 1 of this invention. It is a block diagram which shows the detailed structure of the correction means 80 which concerns on Embodiment 2 of this invention. It is a block diagram which shows the detailed structure of the correction hypothesis integration means 90 which concerns on Embodiment 2 of this invention.

Explanation of symbols

  10 signal detection means, 20 hypothesis generation means, 30 evaluation value calculation means, 40 hypothesis number reduction means, 50 hypothesis selection means, 60 transmission stop judgment means, 61 time difference judgment means, 62 loss detection number judgment means, 70 noise judgment means 71 Time noise judgment means 72 Signal number noise judgment means 80 Correction means 81 Hypothesis correction means 82 Evaluation value correction means 90 Correction hypothesis integration means 91 Same hypothesis search means 92 Correction hypothesis discarding means 93 Maximum evaluation Value hypothesis integration means, 100 hypothesis integration means, 101 hypothesis comparison means.

Claims (10)

A signal classification device that receives a signal emitted from an unknown number of transmission sources and classifies the received signal for each transmission source,
A signal detection means for detecting a received signal, extracting a plurality of types of feature values, and outputting the signals in order of appearance time;
A hypothesis generating means for generating a hypothesis in which a signal is classified for each transmission source in which the signal is estimated to be combined by combining a hypothesis obtained by classifying the signal from the signal detecting means and a hypothesis from the hypothesis integrating means;
An evaluation value calculating means for calculating an evaluation value of a hypothesis from the hypothesis generation means, adding the hypothesis to an output,
Hypothesis number reduction means for selecting hypotheses based on hypothesis evaluation values from the evaluation value calculation means and reducing the number of hypotheses;
When the output request flag is input, when one hypothesis having the highest evaluation value is selected from the hypotheses from the hypothesis number reducing means and output as a classification result, and when the output request flag is not input Is a hypothesis selection means for outputting all hypotheses from the hypothesis number reduction means;
Based on the hypothesis from the hypothesis selection means and the time information of the signal newly output by the signal detection means, it is determined whether or not the transmission source of each hypothesis is a radio wave emission stop. A transmission stop determining means for outputting a hypothesis including a transmission source determined to be radiation stopped and other correction-necessary hypotheses;
Based on the hypotheses from the transmission stop determination means, it is determined whether or not the signal classified as each transmission source determined to stop radio wave emission satisfies a predetermined condition, and determined as noise based on the determination result. Noise judging means for outputting a hypothesis including the transmitted source and other correction-needed hypotheses;
Correction means for re-classifying the signal determined to be noise by the noise determination means as noise, and recalculating the hypothesis evaluation value according to the correction content and adding it to the hypothesis and outputting it as a corrected hypothesis,
It is searched whether or not a hypothesis having the same content exists with respect to the corrected hypothesis from the correcting means. If a corrected hypothesis having the same content exists, one corrected hypothesis having the same content is corrected. The remaining hypothesis is discarded, the remaining hypothesis is output without being discarded, and when there is no corrected hypothesis having the same content, a corrected hypothesis integrating unit that outputs the corrected hypothesis from the correcting unit as it is,
The corrected hypothesis from the corrected hypothesis integrating means is compared with the correction unnecessary hypotheses from the transmission stop determining means and the noise determining means. The revised hypotheses are discarded, and the corrected hypotheses that have not been discarded and all the correction-needed hypotheses are output to the hypothesis generation means. A hypothesis integration means for outputting the corrected hypotheses and all the hypotheses that do not require correction to the hypothesis generation means. The signal classification device according to claim 1,
The transmission stop judging means sets a threshold value for a difference between an extinction time of a signal last classified as a hypothesis transmission source from the hypothesis selection means and an appearance time of a new signal from the signal detection means. A signal classifying apparatus comprising: a time difference determining unit that determines that a transmission source that is less than 1 continues to radiate radio waves, and that a transmission source that exceeds a threshold value determines that radio wave radiation has been stopped. The signal classification device according to claim 1,
The transmission stop judging means is estimated based on a difference between an appearance time of a signal last classified as a hypothesis transmission source from the hypothesis selection means and an appearance time of a new signal from the signal detection means. A threshold is set for the number of missed detections, and a transmission source that does not satisfy the threshold is judged to continue radio emission, and a transmission source that exceeds the threshold has a means for judging the number of missed detections to judge that radio emission has been stopped. A signal classification device. The signal classification device according to claim 1,
The transmission stop judging means is
A threshold is provided for the difference between the disappearance time of the signal last classified as a hypothesis transmission source from the hypothesis selection means and the appearance time of a new signal from the signal detection means, and a transmission source that does not satisfy the threshold is a radio wave. A time difference judging means for judging that radiation is continued and a transmission source exceeding the threshold is judged to have stopped radio emission;
A threshold is set for the number of missed detections estimated based on the difference between the appearance time of the signal last classified as a hypothesis transmission source from the hypothesis selection means and the appearance time of a new signal from the signal detection means. A transmission source that does not satisfy the threshold value is determined to continue radio wave emission, and a transmission source that exceeds the threshold value includes a means for determining the number of missed detections to determine that radio wave emission has been stopped.
When the transmission source is a frequency variation transmission source, the radio wave emission stop of the transmission source is determined by the missed detection frequency determination means and / or the time difference determination means, or both, and the transmission source is a fixed frequency transmission source In the signal classification apparatus, the time difference determining means determines whether radio wave emission of the transmission source is stopped. In the signal classification device according to any one of claims 1 to 4,
The noise determination means sets a threshold for the number of signals classified into each transmission source determined to stop radio wave emission based on a hypothesis including the transmission source from the transmission stop determination means, and signals from transmission sources that do not satisfy the threshold A signal classification device, comprising: a signal number noise judging means for judging a noise as a noise, outputting a hypothesis including these transmission sources to the correcting means, and outputting other hypotheses as a correction unnecessary hypothesis to the hypothesis integrating means . In the signal classification device according to any one of claims 1 to 4,
The noise determination means sets a threshold for the total duration time of signals classified into each transmission source determined as a radio wave emission stop based on a hypothesis including the transmission source from the transmission stop determination means, and a transmission source that does not satisfy the threshold A signal noise classification means, comprising: a time noise judging means for judging the signal of the signal as noise and outputting a hypothesis including these transmission sources to the correcting means and outputting other hypotheses as a correction unnecessary hypothesis to the hypothesis integrating means. apparatus. In the signal classification device according to any one of claims 1 to 4,
The noise judgment means is
A threshold is provided for the number of signals classified into each transmission source determined to stop radio wave emission based on a hypothesis including a transmission source from the transmission stop determination means, and signals from transmission sources that do not satisfy the threshold are determined as noise. A signal number noise judging means for outputting a hypothesis including a transmission source to the correcting means and outputting other hypotheses as a correction unnecessary hypothesis to the hypothesis integrating means;
A threshold is provided for the total duration of signals classified as each transmission source determined as a radio wave emission stop based on a hypothesis including a transmission source from the transmission stop determination means, and a signal from a transmission source that does not satisfy the threshold is determined as noise. And a temporal noise judging means for outputting hypotheses including these transmission sources to the correcting means and outputting other hypotheses to the hypothesis integrating means as correction unnecessary hypotheses,
When the transmission source is a frequency variation transmission source, the signal number noise determination means determines that the noise is present. When the transmission source is a frequency fixed transmission source, the time noise determination means determines that the noise is determined. A characteristic signal classification device. In the signal classification device according to any one of claims 1 to 7,
The corrected hypothesis integration means searches for a hypothesis having the same content with respect to the corrected hypothesis from the correction means, and if there is no hypothesis with the same content, the corrected hypothesis is directly used as the hypothesis integration means. The same hypothesis search means to output and the search result of the same hypothesis search means, if there is a hypothesis of the same content, select one hypothesis having the highest evaluation value from the hypotheses of the same content, the hypothesis integration means And a maximum evaluation value hypothesis integration means for outputting to the signal classification device. In the signal classification device according to any one of claims 1 to 7,
The correction means includes a hypothesis correction means for correcting a hypothesis for a hypothesis including a signal determined to be noise by the noise determination means, a recalculation of a hypothesis evaluation value according to the correction content of the hypothesis, and a correction to the hypothesis. An evaluation value correcting means for outputting as a hypothesis. The signal classification device according to claim 9,
The corrected hypothesis integrating means searches for the presence or absence of hypotheses having the same contents with respect to the corrected hypotheses output from the correcting means, and when there is no identical hypothesis, the same hypothesis searching means for outputting all of the input corrected hypotheses And a modified hypothesis discarding unit that selects and outputs one of the hypotheses having the same content when a hypothesis having the same content exists.

Images