A parallelized signal sorting method that enhances feature clustering

Abstract

The application discloses a parallel signal sorting method with enhanced feature clustering, and belongs to the field of electronic countermeasures. Firstly, pulse description words (PDW) received by detection and interception are collected, and multi-feature joint parameter clustering is performed. Then, frequency channel division is performed by using a small carrier frequency tolerance, and the small carrier frequency tolerance is 6MHz. Channel number calculation is performed according to a length threshold of the clustered pulse description words, then step-time difference histogram calculation is performed on the pulse description words in the channel, detection and judgment are performed on radar radiation sources with fixed carrier frequencies, and sorting results of the signals with fixed carrier frequencies are temporarily batched. The sorted radiation source results are tracked and fused, and sorting of carrier frequency agile signals is performed. Then, channel number calculation is performed according to the length threshold of the clustered pulse description words, step-time difference histogram calculation is performed on the pulse description words in the channel, and sorted pulses in each channel are counted, and finally, the sorting results are output.

Description

TECHNICAL FIELD

[0001] The present application relates to radar signal analysis reconnaissance technology, in particular to a parallel signal sorting method with enhanced feature clustering. BACKGROUND

[0002] In electronic reconnaissance and countermeasure, signal sorting plays an extremely important role. Only by correctly sorting the signals of each radar existing in the environment, can the target radar emitter signal be parameter estimated and type identified, so as to determine the threat nature of the radar.

[0003] With the continuous improvement of modern radar technology, in order to improve the anti-jamming capability and anti-interception capability of radar, military radars generally adopt frequency agile mode. Frequency agile radar increases the uncertainty of frequency by rapidly changing the frequency of radar signal, usually randomly changing the pulse group, thereby increasing the difficulty of receiver reconnaissance. If the traditional emitter signal sorting method is adopted, the batch will be increased, and the specific emitter cannot be correctly sorted. In the article Radar Emitter Signal Sorting Technology Research by Li Qiang, the pulse signal is diluted to reduce the pressure of batch increase, but the useful signals that can be intercepted in space are not fully utilized. Therefore, a parallel signal sorting method with enhanced feature clustering is proposed, which quickly and accurately adapts to the sorting of signal patterns such as frequency agile and the sorting of multiple radars in the same frequency band, and is one of the research hotspots in the field of current electronic reconnaissance and countermeasure. SUMMARY

[0004] The present application proposes a parallel signal sorting method with enhanced feature clustering. In the pulse description word clustering step, the parameters such as the carrier frequency feature, pulse width feature, front feature, and intra-pulse modulation feature of the radar are used to realize pulse clustering, which can improve the accuracy of radar preliminary clustering, so as to adapt to the sorting of signal patterns such as frequency agile and the sorting of multiple radars in the same frequency band. At the same time, parallel processing technology is used to greatly reduce the processing time of signal sorting algorithm, thereby greatly improving the signal sorting speed.

[0005] The technical solution for realizing the present application is as follows: a parallel signal sorting method with enhanced feature clustering, comprising the following steps:

[0006] Step 1: collect pulse signals according to fixed pulse number timing, and go to step 2.

[0007] Step 2: use the parameters of small carrier frequency tolerance, angle of arrival, pulse width, intra-pulse modulation, and pulse envelope to perform multi-parameter joint clustering on the pulse signals, send the clustered pulse signals to a fixed frequency timing sorting channel, and go to step 3.

[0008] Step 3: calculate the effective channel number of the above fixed frequency sorting channel, and go to step 4.

[0009] Step 4, K-order histogram calculation is performed on the pulse signals in each fixed frequency time sequence effective channel. If K reaches the specified order, go to Step 6. Otherwise, go to Step 5.

[0010] Step 5, the pulse signal under the K-order histogram is subjected to a repetition frequency type decision. If the pulse signal is a fixed mode radar radiation source, a radiation source description word of the pulse signal is generated and sent to Step 12 for radiation source fusion. Otherwise, return to Step 4 to calculate the next order histogram.

[0011] Step 6, collect all unsorted pulse signals in the fixed frequency time sequence sorting channels, and go to Step 7.

[0012] Step 7, multi-parameter joint clustering is performed on the above unsorted pulse signals using the large carrier frequency tolerance-arrival angle-pulse width-pulse-in modulation-pulse envelope parameters. The clustered pulse signals are put into the frequency-agile time sequence sorting channels, and go to Step 8.

[0013] Step 8, calculate the number of effective channels in the frequency-agile time sequence sorting channels, and go to Step 9.

[0014] Step 9, K-order histogram calculation is performed on each effective channel in the frequency-agile time sequence sorting channel. If K reaches the specified order, go to Step 11. Otherwise, go to Step 10.

[0015] Step 10, the pulse signal under the K-order histogram is subjected to a repetition frequency type decision. If the pulse signal is a fixed mode radar radiation source, a radiation source description word of the pulse signal is generated and sent to Step 13 for radiation source fusion. The program jumps to Step 9 to start re-computing from the first order histogram. Otherwise, return to Step 9 to calculate the next order histogram.

[0016] Step 11, mark the sorted pulse signals in each effective channel of the frequency-agile time sequence sorting channel. Collect all pulse signals that do not sort out targets in the frequency-agile time sequence sorting channels, and put the pulse signals that do not sort out targets into the radiation source temporary result table for collection next time in Step 1.

[0017] Step 12, batch the radiation source description words of the pulse signals in the fixed frequency time sequence effective channel. If the pulse signal is a new signal, immediately report the result and store it in the radiation source temporary result table. If it matches the signal in the radiation source temporary result table, update the signal parameters in the temporary result table and report the updated parameter result. Return to Step 4 to start re-computing from the first order histogram.

[0018] Step 13, the radiation source description word of the pulse signal in the agile frequency time sequence effective channel is batched, if the pulse signal is a new signal, the result is reported immediately, and stored in the radiation source temporary result table. If it matches the signal in the radiation source temporary result table, the signal parameters in the temporary result table are updated, and the updated parameter result is reported, and step 9 is returned to start from the first-order histogram to recalculate.

[0019] Compared with the prior art, the present application has the following advantages:

[0020] 1) Using parallel processing technology, the signal sorting algorithm processing time can be greatly reduced, thereby greatly improving the signal sorting speed.

[0021] 2) Pulse clustering is realized by fixed frequency and agile frequency signal channels respectively, which can improve the accuracy of radar preliminary clustering.

[0022] 3) It can adapt to the sorting of frequency agile signals and the sorting of multiple radars in the same frequency band. BRIEF DESCRIPTION OF DRAWINGS

[0023] Figure 1 The overall block diagram of the sorting of the present application.

[0024] Figure 2 The overall block diagram of the fusion of the present application.

[0025] Figure 3 The parallelization acceleration sorting time diagram of the present application.

[0026] Figure 4 The cumulative histogram sorting result diagram. DETAILED DESCRIPTION

[0027] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0028] The specific embodiments, technical difficulties and points of the present application will be further introduced below in combination with the design examples.

[0029] In combination with Figure 1 and Figure 3The application provides a parallel signal sorting method with enhanced feature clustering. In the pulse description word clustering step, the pulse clustering is realized by using the parameter features of the radar carrier frequency feature, pulse width feature, front edge feature and intra-pulse modulation feature, so that the correctness of the radar preliminary clustering is improved, so as to adapt to the sorting of signals of the frequency agility type and the sorting of multiple radars in the same frequency band, and the processing time of the signal sorting algorithm is greatly reduced by using the parallel processing technology, so that the signal sorting speed is greatly improved. Figure 1 As shown in the figure, a parallel signal sorting method with enhanced feature clustering has the specific process as follows.

[0030] Step 1: Collecting pulses according to a fixed pulse number, first collecting the pulses not sorted out in the previous time interval collected in step 11, and removing the pulses exceeding the fixed time interval; then collecting the pulses in the current time interval. The number of collected pulses can be dynamically adjusted according to the pulse density. In order to ensure the timeliness of signal sorting, a timing mechanism is added to collect pulses once every fixed time interval of 100 milliseconds when the number of pulses is not reached. In order to ensure the clustering and sorting in the subsequent steps, the pulse description word parameters include carrier frequency, arrival time, pulse width, amplitude, arrival angle and intra-pulse modulation parameters. After the pulse collection is completed, the process goes to step 2.

[0031] Step 2: Multi-parameter joint clustering is performed by using the small carrier frequency tolerance-arrival angle-pulse width-intra-pulse modulation-pulse envelope parameters. At this time, the carrier frequency tolerance depends on the carrier frequency measurement error of the frequency measurement receiver, and is generally set to 6 MHz. The clustered pulses are put into the fixed frequency time sorting channel. After clustering, the process for the fixed frequency time sorting channel enters step 3.

[0032] During the parameter clustering, the truncated kernel or the Gaussian kernel method can be used. The Gaussian kernel method can effectively prevent the hard boundary error problem of data clustering, and also increases the calculation amount.

[0033] For a data point x i , the matching probability p i of the truncated kernel χ(x) of the center point is calculated as follows:

[0034]

[0035]

[0036] The calculation method of the Gaussian kernel is as follows:

[0037]

[0038] Wherein, the distance d between the i-th data point x i and the j-th data point x j isc For the truncation distance, d c > 0.

[0039] Step 3, calculate the number of effective channels of the above fixed frequency sorting channel:

[0040] According to the total number of pulses in each fixed frequency timing sorting channel, it is judged whether it is an effective channel. When there are ≥ 6 pulses in the channel, it is judged that the channel is an effective channel. The total number of effective channels is calculated. Enter step 4 to open the parallel fixed frequency timing sorting thread with the same number of effective channels.

[0041]

[0042]

[0043] Where C in is the number of pulses of the i-th cluster channel, C iv is whether the i-th is an effective channel, N v is the total number of effective channels.

[0044] Step 4, for each effective fixed frequency timing sorting channel in step 3, first calculate the K step histogram. The initial value of K is 1. If K reaches the specified order, enter step 6, otherwise enter step 5.

[0045] Where the calculation method of the cumulative histogram is as follows:

[0046] a) Calculate the first-order histogram;

[0047] b) Calculate the second-order histogram;

[0048] c) The result of the first-order histogram plus the second-order histogram is the second-order cumulative histogram;

[0049] d) Calculate the K-order histogram in turn.

[0050] The cumulative histogram accumulation result is as shown in Figure 4 .

[0051] For the existence threshold judgment of the cumulative histogram signal, the following empirical formula can be used:

[0052] Threshold (τ) = x(E-c)e -τ / (kN)

[0053] Where E is the total number of pulses, N is the total interval number of the histogram, and c is the order of the histogram. The coefficients x and k need to be selected according to the empirical value.

[0054] Step 5, the pulse signal under the Kth histogram is subjected to a PRF type decision, if the pulse signal is a radar radiation source of fixed mode, a radiation source description word of the pulse signal is generated and sent to step 12 for radiation source fusion; otherwise, return to step 4 to calculate the next order histogram.

[0055] The PRF type decision sequence is: 5-1) carrier frequency fixed, PRF fixed signal; 5-2) carrier frequency fixed, PRF jitter signal; 5-3) carrier frequency fixed, PRF uneven signal; 5-4) carrier frequency fixed, PRF sliding signal; 5-5) continuous wave signal.

[0056] Step 6, collect all unsorted pulse signals in the fixed frequency time sequence sorting channel. Enter step 7.

[0057] Step 7, use the large carrier frequency tolerance-azimuth-pulse width-pulse modulation-pulse envelope parameters to perform multi-parameter joint clustering on the above unsorted pulse signals, and put the clustered pulse signals into the frequency-agile time sequence sorting channel:

[0058] Multi-parameter joint clustering is performed using the large carrier frequency tolerance-azimuth-pulse width-pulse modulation-pulse envelope parameters, at this time the carrier frequency tolerance depends on the frequency-agile range of each frequency band and radar for use, and the setting basis of the large carrier frequency tolerance is the frequency-agile range of the radiation source target. The frequency-agile range of a typical X-band target is 1 GHz, and the carrier frequency tolerance can be adjusted according to the characteristics of targets in each frequency band.

[0059] The clustered pulses are put into the frequency-agile time sequence sorting channel, and the processing in the frequency-agile time sequence sorting channel enters step 8.

[0060] Step 8, calculate the number of effective channels in the frequency-agile time sequence sorting channel, and go to step 9.

[0061] The calculation of the frequency-agile time sequence sorting channel uses the OpenMP technology to perform parallelization, and the number of threads is determined according to the maximum number of sorting channels and the number of processor cores.

[0062] Step 9, for each effective channel in the frequency-agile time sequence sorting channel, K order histograms are calculated, if K reaches the specified order, go to step 11, otherwise go to step 10:

[0063] According to the total number of pulses in each frequency-agile time sequence sorting channel, it is judged whether it is an effective channel, when there are ≥6 pulses in the channel, it is judged that the channel is an effective channel. The total number of effective channels is calculated. Enter step 10 to start the parallel frequency-agile time sequence sorting thread with the same number of effective channels.

[0064] Step 10, judge whether there are fixed frequency, variable frequency, jitter frequency and sliding frequency signals under the current histogram, if there are, mark the pulses belonging to the signals as sorted pulses, and generate the radiation source description word of the signals, and send to step 13 for radiation source fusion, the overall block diagram is shown in Figure 2 If the existence of the above signals is not determined under the histogram, jump to step 9 to calculate the next order histogram.

[0065] Step 11, mark the sorted pulse signals in the effective channels of each frequency agile time sequence sorting channel, collect the pulse signals not sorted in all the frequency agile time sequence sorting channels, and put the pulse signals not sorted into the radiation source temporary result table for collection in the next sorting in step 1.

[0066] Step 12, batch the radiation source description word of the pulse signals in the fixed frequency time sequence effective channel, if the pulse signal is a new signal, immediately report the result and store it in the radiation source temporary result table; if it matches the signal in the radiation source temporary result table, update the signal parameters in the temporary result table, and report the updated parameter result, and return to step 4 to start calculating from the first order histogram.

[0067] Step 13, batch the radiation source description word of the pulse signals in the frequency agile time sequence effective channel, if the pulse signal is a new signal, immediately report the result and store it in the radiation source temporary result table; if it matches the signal in the radiation source temporary result table, update the signal parameters in the temporary result table, and report the updated parameter result, and return to step 9 to start calculating from the first order histogram.

Claims

1. A parallelized signal sorting method with enhanced feature clustering, characterized in that, The steps are as follows: Step 1: Collect pulse signals at fixed intervals according to a fixed number of pulses, then proceed to Step 2; Step 2: Perform multi-parameter joint clustering of the pulse signal using parameters such as small carrier frequency tolerance, angle of arrival, pulse width, intra-pulse modulation, and pulse envelope. Then, send the clustered pulse signal into a fixed frequency timing sorting channel and proceed to Step 3. Step 3: Calculate the number of effective channels for the fixed frequency sorting channels mentioned above, and proceed to Step 4; Step 4: Calculate the K-order histogram for the pulse signal in each fixed frequency timing effective channel. If K reaches the specified order, proceed to step 6; otherwise, proceed to step 5. Step 5: Perform a repetition rate type determination on the pulse signal under the Kth order histogram. If the pulse signal is a radar radiation source with a fixed pattern, generate the radiation source descriptor of the pulse signal and send it to step 12 for radiation source fusion; otherwise, return to step 4 to calculate the next order histogram. Step 6: Collect all unsorted pulse signals in the fixed frequency timing sorting channels, then proceed to Step 7; Step 7: Perform multi-parameter joint clustering on the above unsorted pulse signals using the parameters of large carrier frequency tolerance, angle of arrival, pulse width, intra-pulse modulation, and pulse envelope. Place the clustered pulse signals into the agile frequency timing sorting channel and proceed to step 8. Step 8: Calculate the number of effective channels in the frequency agile timing sorting channel, then proceed to step 9; Step 9: For each effective channel in the frequency agile timing sorting channel, calculate the K-order histogram. If K reaches the specified order, proceed to step 11; otherwise, proceed to step 10. Step 10: Perform a repetition rate type determination on the pulse signal under the Kth order histogram. If the pulse signal is a radar radiation source with a fixed pattern, generate the radiation source description word of the pulse signal and send it to step 13 for radiation source fusion. The program jumps to step 9 to recalculate from the first order histogram. Otherwise, return to step 9 to calculate the next order histogram; Step 11: Mark the pulse signals that have been sorted in the effective channels of each frequency agile timing sorting channel, collect the pulse signals of the unsorted targets in all frequency agile timing sorting channels, and put the pulse signals of the unsorted targets into the radiation source temporary result table for collection in the next sorting in Step 1. Step 12: Batch the radiation source descriptors of the pulse signals in the effective channels of the fixed frequency timing. If the pulse signal is a new signal, report the result immediately and store it in the temporary radiation source result table. If the signal matches the signal in the provisional result table of the radiation source, update the signal parameters in the provisional result table, report the updated parameter results, and return to step 4 to recalculate from the first-order histogram. Step 13: Batch the radiation source descriptors of the pulse signals in the effective channels of the agile frequency timing. If the pulse signal is a new signal, report the result immediately and store it in the temporary radiation source result table. If the signal matches the signal in the provisional result table of the radiation source, update the signal parameters in the provisional result table, report the updated parameter results, and return to step 9 to recalculate from the first-order histogram.

2. The parallelized signal sorting method with enhanced feature clustering according to claim 1, characterized in that: In step 1, the pulse description word (PDW) of the acquired pulse signal includes pulse carrier frequency, arrival time, pulse width, amplitude, angle of arrival, intra-pulse modulation type, intra-pulse modulation parameters, and pulse leading-edge feature vector. The intra-pulse modulation type includes no intra-pulse modulation, linear frequency modulation, nonlinear frequency modulation, binary phase code, quadrature phase code, and frequency coding. The intra-pulse modulation parameters include the bandwidth of frequency modulation and the code pattern of frequency and phase coding. The pulse leading-edge feature is a vector composed of the first 6 coefficients of the amplitude and time-domain normalized pulse leading-edge fitting polynomial.

3. The parallelized signal sorting method with enhanced feature clustering according to claim 2, characterized in that: In step 2, the small carrier frequency tolerance setting is based on the receiver's frequency measurement error, which is generally set to 6MHz and can be adjusted according to the electromagnetic environment density to adapt to applications under different electromagnetic environments.

4. The parallelized signal sorting method with enhanced feature clustering according to claim 3, characterized in that: In step 3, the calculation of the fixed frequency timing sorting channel adopts OpenMP technology for parallel expansion, and the maximum number of sorting channels is determined by combining the number of threads and the number of processor cores.

5. The parallelized signal sorting method with enhanced feature clustering according to claim 4, characterized in that: In step 5, the order of frequency repetition type determination is as follows: 5-1) fixed carrier frequency and fixed repetition frequency signal; 5-2) fixed carrier frequency and jitter repetition frequency signal; 5-3) fixed carrier frequency and uneven repetition frequency signal; 5-4) fixed carrier frequency and slip repetition frequency signal; 5-5) continuous wave signal.

6. The parallelized signal sorting method with enhanced feature clustering according to claim 5, characterized in that: In step 7, the large carrier frequency tolerance is set based on the agility range of the radiation source target. The agility range of a typical X-band target is 1 GHz. The carrier frequency tolerance can be adjusted according to the characteristics of each frequency band target.

7. The parallelized signal sorting method with enhanced feature clustering according to claim 6, characterized in that: In step 8, the calculation of the agile frequency timing sorting channel adopts OpenMP technology for parallel expansion, and the number of threads is determined by combining the maximum number of sorting channels and the number of processor cores.

8. The parallelized signal sorting method with enhanced feature clustering according to claim 7, characterized in that: In step 10, the order of frequency repetition types is as follows: 10-1) fixed carrier frequency and fixed frequency repetition signal; 10-2) fixed carrier frequency and frequency jitter signal; 10-3) fixed carrier frequency and frequency uneven signal; 10-4) fixed carrier frequency and frequency slip signal; 10-5) continuous wave signal.

Images