A rough sorting method of UAV signals based on variable cycle power spectrum

Abstract

The present invention discloses a method for rough sorting of UAV signals based on variable cycle power spectrum, comprising acquiring a signal; determining an expected signal S according to the type of signal to be detected; E , intercept or accumulate the signal to be processed S P ; The signal to be processed S P and the expected signal sequence to obtain the result data sequence R E , and then obtain the discrete Fourier transform to obtain the power spectral density estimate P F The value is compared with the feature discrimination threshold, and the first parameter feature sequence F is recorded; the first parameter feature sequence F is compared with the modulation parameter feature of each expected signal to obtain a valid comparison result. After the above operation, the second parameter feature sequence F' is obtained and compared with a separate expected signal data sequence. When the discrimination result exceeds the set decision threshold, the discrimination result is pushed to the signal sorting module; the variable cyclic power spectrum detection method is adopted, and the detection of multiple target signals can be processed in real time in a pipeline manner to realize the recognition and sorting of multiple signals.

Description

Technical Field

[0001] The present invention relates to the field of signals, and in particular to coarse sorting of UAV signals based on variable cycle power spectrum. Background Art

[0002] When using radio spectrum detection technology to detect illegal drones, we often face the complex electromagnetic environment in cities. The detected signals are difficult to distinguish from the same-frequency industrial and civil communication signals. Accurate identification of illegal drone signals, especially accurate identification of non-cooperative target signals, is the basic prerequisite for subsequent effective sorting and targeted interference and deception. In the field of civil communication signal processing, it involves spectrum resource optimization design and cognitive radio spectrum perception; the existing technology has the disadvantages of weak real-time performance in engineering implementation, inability to process the detection of multiple target signals in real time, low hardware resource utilization, and low detection probability of simultaneously identifying multiple signals. Summary of the invention

[0003] Purpose of the invention: In order to overcome the deficiencies in the prior art, the present invention provides a method for coarse sorting of UAV signals based on variable cyclic power spectrum, and proposes a variable cyclic power spectrum detection method that can be implemented in a distributed computing chip, which can process the detection of multiple target signals in real-time pipeline processing.

[0004] Technical solution: To achieve the above purpose, a method for rough sorting of UAV signals based on variable cycle power spectrum of the present invention comprises the following steps:

[0005] a1: Get signal;

[0006] a2: Determine the expected signal S according to the type of signal to be detected E , intercept or accumulate the signal to be processed S P ;

[0007] a3: The signal to be processed S P Perform correlation operation with each set of expected signal sequences to obtain the result data sequence R of the correlation operation E ;

[0008] a4: For each result data sequence R E , obtain the discrete Fourier transform and obtain the corresponding power spectrum density estimate, denoted as P F ;

[0009] a5: Estimation of the power spectral density P obtained each time F The value of is compared with the set feature discrimination threshold ΔP, and the size of the data point exceeding the feature discrimination threshold ΔP and the position of the data point in the sequence are recorded as the first parameter feature sequence F;

[0010] a6: compare the first parameter feature sequence F with the modulation parameter features of each expected signal. When no valid comparison result is obtained, repeat steps a3 to a5 continuously. When a valid comparison result is obtained and a certain number of times are accumulated and the set valid number threshold ΔK is triggered, the data signal to be processed is compared with only a single expected signal data sequence corresponding to the expected signal sequence with a valid comparison result, and steps a3 to a5 are performed. The obtained second parameter feature sequence is recorded as F'.

[0011] a7: The second parameter feature sequence F' is compared with a single expected signal data sequence to obtain a recognition result, and when the recognition result exceeds the set decision threshold ΔN, the recognition result is pushed to the signal sorting module.

[0012] Furthermore, in step a2, the expected signal refers to a known signal of the type to be detected; the expected signal S E Use superscripts to distinguish different types of expected signals S according to different signal types E k , when determining the data sequence of the expected signal, the expected signal S E According to the different signal types, the expected signals with similar modulation types and modulation parameters are accumulated to form a new sequence, S E (k-1) and S E k Combined into an expected signal sequence, denoted as S E (k-1)k .

[0013] Furthermore, the signal to be processed S P It refers to the sampling sequence of the input signal, the sampling rate of the signal is set to Fs, the length of the analyzed data segment is τ and the inherent delay τ′ of the pipeline processing; S P The sequence length of the signal to be processed is:

[0014] M=[Fs*(τ+τ′)]

[0015] And the sequence length M is rounded; each data element in the sequence length of the signal to be processed is recorded as:

[0016] S P =[S P1 , S P2 , S P3 .....S PM ]

[0017] When performing pipeline processing, the inherent delay τ of each pipeline processing ′ Time, the data elements in the signal to be processed are shifted and rounded to Fs*τ', and so on.

[0018] Furthermore, in step a3, the signal to be processed S p When the accumulated sequence size reaches the preset number N, the signal to be processed S p Perform correlation operation with a set of expected signal sequences to obtain the result data sequence R of the correlation operation E ; The definition of the correlation operation between two sequences x and y is:

[0019]

[0020] The signal to be processed S p and a set of expected signals S E x and y to be related to the operation, and perform the related operation;

[0021] The corresponding expected signal data sequence is marked with a superscript, and the subscript number is marked with the number of loop operations; the signal data sequence to be processed and the expected signal S E (k-1) and S E k The result data sequence obtained by the first cycle of the expected data group composed of E1 (k-1)k ; After τ′ time, the translated signal sequence to be processed is again correlated with the expected signal sequence, which is recorded as the result data sequence of the second cycle correlation operation, and so on.

[0022] Furthermore, in step a4, the result data sequence R of each cyclic correlation operation is obtained. E Obtain the discrete Fourier transform and obtain the corresponding power spectrum density estimate in the form of fast Fourier transform FFT, denoted as P F ; Use the superscript of the corresponding expected signal data sequence as a mark, and use the subscript number to mark the number of loop operations; the result data sequence R E1 (k-1)k Obtain the power spectral density estimate P by taking the discrete Fourier transform F1 (k-1)k .

[0023] Furthermore, in the step a5, the value of the power spectrum density estimation obtained each time is compared with the set feature discrimination threshold ΔP, and the size of the data point exceeding the feature discrimination threshold ΔP and the position of the data point in the sequence are recorded as the first parameter feature sequence F, with the superscript of the corresponding expected signal data sequence as the mark, and the number of cyclic operations marked with the subscript number, and the specific comparison is as follows:

[0024] P F1 (k-1)k >ΔP

[0025] Then the power spectral density estimate PF1 (k-1)k The size of the data point whose value exceeds the feature discrimination threshold ΔP and the position of the data point in the sequence are recorded as the first parameter feature sequence

[0026] P F1 (k-1)k ≤ΔP

[0027] Then the power spectral density estimate P F1 (k-1)k If the value of does not exceed the feature recognition threshold ΔP, the first parameter feature sequence F is not recorded.

[0028] Further, the first parameter feature sequence F is compared with the modulation parameter feature of each expected signal. When no valid result is obtained, steps a3 to a5 are repeated continuously. When a valid comparison result is obtained, a certain number of times is accumulated and a set valid number threshold ΔK is triggered, and the data signal to be processed is compared with only a single expected signal data sequence S corresponding to the expected signal sequence with a valid comparison result. E k , perform steps a3 to a5; at this time, the delay time τ′ becomes a single clock cycle, that is, pipeline processing, and no longer delays the τ′ time to perform each operation from step a3 to step a5; the obtained second parameter characteristic sequence is recorded as F'.

[0029] Furthermore, the first parameter feature sequence F obtained is compared with a set of expected signal data sequences to be effective, and the second parameter feature sequence F' is compared with a single expected signal data sequence to obtain a recognition result, and the recognition result of the comparison is set as F E , and the identification result F E Compared with the set decision threshold ΔN, the specific comparison is as follows:

[0030] F E >ΔN

[0031] Then the identification result F E Push to the signal sorting module;

[0032] F E ≤ΔN

[0033] Then the identification result F E Do not push to the signal sorting module, return to step a1 to obtain the signal.

[0034] Beneficial effects: The present invention provides a method for coarse sorting of UAV signals based on variable cyclic power spectrum, proposes a variable cyclic power spectrum detection method that can be implemented in a distributed computing chip, and can process the detection of multiple target signals in real-time pipeline processing. This method has low requirements for hardware storage resources, has a high probability of detecting multiple signals, and realizes the identification and sorting of multiple signals; at the same time, it optimizes the design of expected signals and processing procedures, thereby improving detection efficiency and resource utilization; it can be applied to the coarse sorting of other related and similar communication signals, improve the success rate of radio detection of non-cooperative target signals, and can also effectively improve the real-time effect of processing. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Attached Figure 1 Flowchart of coarse sorting of UAV signals based on variable cycle power spectrum. DETAILED DESCRIPTION

[0036] The present invention will be further described below in conjunction with the accompanying drawings.

[0037] As attached Figure 1 As shown, a method for rough sorting of UAV signals based on variable cycle power spectrum includes the following steps:

[0038] a1: Get signal;

[0039] a2: Determine the expected signal S according to the type of signal to be detected E , intercept or accumulate the signal to be processed S P ;

[0040] a3: The signal to be processed S P Perform correlation operation with each set of expected signal sequences to obtain the result data sequence R of the correlation operation E ;

[0041] a4: For each result data sequence R E , obtain the discrete Fourier transform and obtain the corresponding power spectrum density estimate, denoted as P F ;

[0042] a5: Estimation of the power spectral density P obtained each time F The value of is compared with the set feature discrimination threshold ΔP, and the size of the data point exceeding the feature discrimination threshold ΔP and the position of the data point in the sequence are recorded as the first parameter feature sequence F:

[0043] a6: compare the first parameter feature sequence F with the modulation parameter features of each expected signal. When no valid comparison result is obtained, repeat steps a3 to a5 continuously. When a valid comparison result is obtained and a certain number of times are accumulated and the set valid number threshold ΔK is triggered, the data signal to be processed is compared with only a single expected signal data sequence corresponding to the expected signal sequence with a valid comparison result, and steps a3 to a5 are performed. The obtained second parameter feature sequence is recorded as F'.

[0044] a7: The second parameter feature sequence F' is compared with a single expected signal data sequence to obtain a recognition result, and when the recognition result exceeds the set decision threshold ΔN, the recognition result is pushed to the signal sorting module.

[0045] In step a2, the expected signal refers to a known signal of the type to be detected; the expected signal S E Use superscripts to distinguish different types of expected signals S according to different signal types E k , when determining the data sequence of the expected signal, the expected signal S E According to the different signal types, the expected signals with similar modulation types and modulation parameters are accumulated to form a new sequence, S E (k-1) and S E k Combined into an expected signal sequence, denoted as S E (k-1)k .

[0046] The signal to be processed S p It refers to the sampling sequence of the input signal, the sampling rate of the signal is set to Fs, the length of the analyzed data segment is τ and the inherent delay τ′ of the pipeline processing; S p The sequence length of the signal to be processed is:

[0047] M=[Fs*(τ+τ′)]

[0048] And the sequence length M is rounded; each data element in the sequence length of the signal to be processed is recorded as:

[0049] S P =[S P1 , S P2 , S P3 .....S PM ]

[0050] When pipeline processing is performed, every time the inherent delay τ′ of the pipeline processing passes, the data elements in the signal to be processed are shifted and rounded to (Fs*τ′), and so on.

[0051] In step a3, the signal to be processed Sp When the accumulated sequence size reaches the preset number N, the signal to be processed S p Perform correlation operation with a set of expected signal sequences to obtain the result data sequence R of the correlation operation E ; The definition of the correlation operation between two sequences x and y is:

[0052]

[0053] The signal to be processed S p and a set of expected signals S E x and y to be related to the operation, and perform the related operation;

[0054] The corresponding expected signal data sequence is marked with a superscript, and the subscript number is marked with the number of loop operations; the signal data sequence to be processed and the expected signal S E (k-1) and S E k The result data sequence obtained by the first cycle of the expected data group composed of E1 (k-1)k ; After τ′ time, the translated signal sequence to be processed is again correlated with the expected signal sequence, which is recorded as the result data sequence of the second cycle correlation operation, and so on.

[0055] In the step a4, the result data sequence R of each cyclic correlation operation is obtained. E Obtain the discrete Fourier transform and obtain the corresponding power spectrum density estimate in the form of fast Fourier transform FFT, denoted as P F ; Use the superscript of the corresponding expected signal data sequence as a mark, and use the subscript number to mark the number of loop operations; the result data sequence R E1 (k-1)k Obtain the power spectral density estimate P by taking the discrete Fourier transform F1 (k-1)k .

[0056] In the step a5, the value of the power spectrum density estimation obtained each time is compared with the set feature discrimination threshold ΔP, and the size of the data point exceeding the feature discrimination threshold ΔP and the position of the data point in the sequence are recorded as the first parameter feature sequence F, with the superscript of the corresponding expected signal data sequence as the mark, and the number of cyclic operations is marked with the subscript number. The specific comparison is as follows:

[0057] P F1 (k-1)k >ΔP

[0058] Then the power spectral density estimate P F1 (k-1)kThe size of the data point whose value exceeds the feature discrimination threshold ΔP and the position of the data point in the sequence are recorded as the first parameter feature sequence

[0059] P F1 (k-1)k ≤ΔP

[0060] Then the power spectral density estimate P F1 (k-1)k If the value of does not exceed the feature recognition threshold ΔP, the first parameter feature sequence F is not recorded.

[0061] The first parameter feature sequence F is compared with the modulation parameter feature of each expected signal. When no valid result is obtained, steps a3 to a5 are repeated continuously. When a valid comparison result is obtained, a certain number of times is accumulated and the set valid number threshold ΔK is triggered, and the data signal to be processed is compared with only the single expected signal data sequence S corresponding to the expected signal sequence with the valid comparison result. E k , and perform steps a3 to a5; at this time, the delay time τ′ becomes a single clock cycle, that is, pipeline processing, and the operation of steps a3 to a5 is no longer delayed by τ′; the obtained second parameter characteristic sequence is recorded as F'. When the hardware storage or operation limit is reached, it can be paused for a certain time, and then the above-mentioned pipeline operation of steps a3 to a5 is continued. The above-mentioned transformation process is one of the characteristics of the variable cycle power spectrum.

[0062] The obtained parameter feature sequence F is compared with a set of expected signal data sequences to be effective, and the second parameter feature sequence F' is compared with a single expected signal data sequence to obtain a recognition result, and the recognition result of the comparison is set as F E , and the identification result F E Compared with the set decision threshold ΔN, the specific comparison is as follows:

[0063] F E >ΔN

[0064] Then the identification result F E Push to the signal sorting module;

[0065] F E ≤ΔN

[0066] Then the identification result F E Do not push to the signal sorting module, return to step a1 to obtain the signal;

[0067] When the identification result is pushed to the signal sorting module, the next step of operation can be carried out, and the required signal is sorted out and then the next step of signal processing is carried out; when the identification result is not pushed to the signal sorting module, the signal is acquired again.

[0068] The feature identification threshold ΔP, the effective number threshold ΔK and the decision threshold ΔN can be set as a fixed threshold or a floating threshold; the feature identification threshold ΔP is set according to the type of signal to be selected, modulation parameters, modulation parameters and other characteristics.

[0069] Taking the above operation inside FPGA as an example, under the parameters of working clock 300MHz, single channel, 4 groups of expected signals, processing sequence length 1024 points, the single processing delay τ′ is no more than 30us; secondly, it can reduce the occupancy of on-chip registers, which is mainly reflected in operations such as expected signal merging, sharing of data to be processed and pipeline processing; finally, it can effectively improve the detection probability. Compared with the expected signal detection by only using amplitude comparison, the required signal-to-noise ratio requirement can be reduced by more than 10dB.

[0070] Example

[0071] The expected signal types include image transmission signal BL, image transmission signal O1, image transmission signal O2, image transmission signal O3 and remote control signal St. Superscripts can be used to distinguish different types of expected signals, such as S E 1 Represents the image transmission signal BL, S E 2 Represents the image transmission signal O1, S E 3 Represents the image transmission signal O2, S E 4 Represents the image transmission signals O3 and S E 5 Represents the remote control signal St; the image transmission signals O1, O2, and O3 are the same type of expected signals, S E 2 , S E 3 and S E 4 can be combined into an expected signal, and the combined expected signal is S E 234 ; Therefore, it can be concluded that there are three groups of expected signal sequences S after merging in the example E 1 , S E 234 and S E 5 , which together constitute the desired signal S E Set the sampling rate of the signal to be processed Fs to 8MHz, the length of the analysis segment data segment τ to 10ms and the inherent delay τ′ of the pipeline processing to 20us; get S p The sequence length of the signal to be processed is M = 80160; each data element in it is denoted by S P =[SP1 , S P2 , S P3 ……S P80160 ]; In this example, when pipeline processing is performed, the data elements in the signal to be processed are shifted and rounded to 160 every 20us.

[0072] The result data sequence obtained in the first cycle is recorded as R E1 1 , R E1 234 and R E1 5 ; After 20us, the translated signal sequence to be processed is again correlated with the expected signal sequence, and the result data sequence obtained by the second cycle is recorded as R E2 1 , R E2 234 and R E2 5 ; The result data sequence obtained each time is subjected to Fourier transformation to obtain the power spectrum density estimation. The power spectrum density estimation result data sequence P obtained in the first cycle F1 1 , P F1 234 and P F1 5 ; The power spectrum density estimation result data sequence P obtained in the second cycle F2 1 , P F2 234 and P F2 5 ; Compare the value of the power spectrum density estimation obtained each time with the feature discrimination threshold ΔP, and record the size of the data point exceeding the feature discrimination threshold ΔP and the position of the data point in the sequence as the first parameter feature sequence F; The first parameter feature sequence obtained in the first cycle and The first parameter characteristic sequence obtained in the second cycle and

[0073] The first parameter feature sequence F is compared with the modulation parameter feature of each expected signal. When a valid comparison result appears, a certain number of times will be accumulated and the set valid number threshold ΔK will be triggered. For example, when S E 234 When a valid comparison result appears, the data stream to be processed is only compared with the single expected signal data sequence S corresponding to the valid expected signal sequence. E 2 , S E 3 and S E4 , perform step a3 and step a5; obtain the second parameter characteristic sequence F'.

[0074] The second parameter feature sequence F' is compared with a single expected signal data sequence to obtain a recognition result, and when the recognition result exceeds a set decision threshold ΔN, the recognition result is pushed to the signal sorting module.

[0075] The above are preferred embodiments of the present invention. It should be pointed out that a person skilled in the art may make several improvements and modifications without departing from the principles of the present invention. These improvements and modifications should also be regarded as within the protection scope of the present invention.

Claims

1. A method for rough sorting of UAV signals based on variable cycle power spectrum, characterized by: The following steps are involved: a1: Get signal; a2: Determine the expected signal S according to the type of signal to be detected E , intercept or accumulate the signal to be processed S P ; a3: The signal to be processed S P Perform correlation operation with each set of expected signal sequences to obtain the result data sequence R of the correlation operation E ; a4: For each result data sequence R E , obtain the discrete Fourier transform and obtain the corresponding power spectrum density estimate, denoted as P F ; a5: Estimation of the power spectral density P obtained each time F The value of is compared with the set feature discrimination threshold ΔP, and the size of the data point exceeding the feature discrimination threshold ΔP and the position of the data point in the sequence are recorded as the first parameter feature sequence F; a6: compare the first parameter feature sequence F with the modulation parameter features of each expected signal. When no valid comparison result is obtained, repeat steps a3 to a5 continuously. When a valid comparison result is obtained and a certain number of times are accumulated and the set valid number threshold ΔK is triggered, the data signal to be processed is compared with only a single expected signal data sequence corresponding to the expected signal sequence with a valid comparison result, and steps a3 to a5 are performed. The obtained second parameter feature sequence is recorded as F'. a7: The second parameter feature sequence F' is compared with a single expected signal data sequence to obtain a recognition result, and when the recognition result exceeds the set decision threshold ΔN, the recognition result is pushed to the signal sorting module.

2. The method for rough sorting of UAV signals based on variable cycle power spectrum according to claim 1 is characterized in that: In step a2, the expected signal refers to a known signal of the type to be detected; the expected signal S E Use superscripts to distinguish different types of expected signals S according to different signal types E k , when determining the data sequence of the expected signal, the expected signal S E According to the different signal types, the expected signals with similar modulation types and modulation parameters are accumulated to form a new sequence, S E (k-1) and S E k Combined into an expected signal sequence, denoted as S E (k-1)k .

3. The method for rough sorting of UAV signals based on variable cycle power spectrum according to claim 2 is characterized in that: The signal to be processed S p It refers to the sampling sequence of the input signal, the sampling rate of the signal is set to Fs, the length of the analyzed data segment is τ and the inherent delay τ′ of the pipeline processing; S p The sequence length of the signal to be processed is: M=[Fs*(τ+τ′)] And the sequence length M is rounded; each data element in the sequence length of the signal to be processed is recorded as: S P =[S P1 、S P2 、S P3 .....S PM ] When pipeline processing is performed, every time an inherent delay τ′ of the pipeline processing passes, the data elements in the signal to be processed are shifted and rounded to Fs*τ′, and so on.

4. The method for rough sorting of UAV signals based on variable cycle power spectrum according to claim 3 is characterized in that: In step a3, the signal to be processed S p When the accumulated sequence size reaches the preset number N, the signal to be processed S P Perform correlation operation with a set of expected signal sequences to obtain the result data sequence R of the correlation operation E ; The definition of the correlation operation between two sequences x and y is: The signal to be processed S P and a set of expected signals S E Substitute x and y of the relevant operation and perform the relevant operation; The corresponding expected signal data sequence is marked with a superscript, and the subscript number is marked with the number of loop operations; the signal data sequence to be processed and the expected signal S E (k-1) and S E k The result data sequence obtained by the first cycle of the expected signal sequence composed of E1 (k-1)k ; After τ′ time, the translated signal sequence to be processed is again correlated with the expected signal sequence, which is recorded as the result data sequence of the second cycle correlation operation, and so on.

5. The method for rough sorting of UAV signals based on variable cycle power spectrum according to claim 4 is characterized in that: In step a4, the result data sequence R of each correlation operation is obtained. E Obtain the discrete Fourier transform and obtain the corresponding power spectrum density estimate in the form of fast Fourier transform FFT, denoted as P F ; Use the superscript of the corresponding expected signal data sequence as a mark, and use the subscript number to mark the number of loop operations; the result data sequence R E1 (k-1))k Obtain the power spectral density estimate P by taking the discrete Fourier transform F1 (k-1)k .

6. The method for rough sorting of UAV signals based on variable cycle power spectrum according to claim 5 is characterized in that: In the step a5, the value of the power spectrum density estimation obtained each time is compared with the set feature discrimination threshold ΔP, and the size of the data point exceeding the feature discrimination threshold ΔP and the position of the data point in the sequence are recorded as the first parameter feature sequence F, with the superscript of the corresponding expected signal data sequence as the mark, and the number of cyclic operations is marked with the subscript number. The specific comparison is as follows: P F1 (k-1)k >ΔP Then the power spectral density estimate P F1 (k-1)k The size of the data point whose value exceeds the feature discrimination threshold ΔP and the position of the data point in the sequence are recorded as the first parameter feature sequence P F1 (k-1)k ≤ΔP Then the power spectral density estimate P F1 (k-1)k If the value of does not exceed the feature recognition threshold ΔP, the first parameter feature sequence F is not recorded.

7. The method for rough sorting of UAV signals based on variable cycle power spectrum according to claim 6 is characterized in that: The first parameter feature sequence F is compared with the modulation parameter feature of each expected signal. When no valid result is obtained, steps a3 to a5 are repeated continuously. When a valid comparison result is obtained, a certain number of times is accumulated and the set valid number threshold ΔK is triggered, and the data signal to be processed is compared with only the single expected signal data sequence S corresponding to the expected signal sequence with the valid comparison result. E k , perform steps a3 to a5; at this time, the delay time τ′ becomes a single clock cycle, that is, pipeline processing, and no longer delays the τ′ time to perform each operation from step a3 to step a5; the obtained second parameter characteristic sequence is recorded as F'.

8. The method for rough sorting of UAV signals based on variable cycle power spectrum according to claim 7 is characterized in that: The first parameter feature sequence F obtained is compared with a set of expected signal data sequences to be effective, and the second parameter feature sequence F' is compared with a single expected signal data sequence to obtain a recognition result, and the recognition result of the comparison is set as F E , and the identification result F E Compared with the set decision threshold ΔN, the specific comparison is as follows: F E >ΔN Then the identification result F E Push to the signal sorting module; F E ≤ΔN Then the identification result F E Do not push to the signal sorting module, return to step a1 to obtain the signal.

Images