Real-time detection method and system for low-duty-ratio narrow pulse in broadband frequency spectrum monitoring

Abstract

The invention discloses a real-time detection method and system for a low-duty-ratio narrow pulse in broadband spectrum monitoring, and the method comprises the steps: obtaining a signal of the low-duty-ratio narrow pulse, and carrying out the real-time spectrum estimation, and obtaining a corresponding estimation spectral line; performing incoherent accumulation on the estimated spectral line to obtain a cumulative spectral line, and updating a cumulative spectral line estimated value in real time according to the cumulative spectral line at the previous moment and the current estimated spectral line; performing linear transformation on each frame of estimated spectral line according to the corresponding cumulative spectral line estimated value to obtain a linearly transformed spectral line; and according to the minimum pulse width to be detected, accumulating the spectral lines of the specified number of frames after linear transformation and carrying out constant false alarm detection, carrying out secondary accumulation on the frames corresponding to the spectral lines passing through the constant false alarm detection, and then carrying out secondary constant false alarm detection on the spectral lines of the frames subjected to secondary accumulation to obtain a pulse signal detection result. According to the invention, the hardware resource consumption can be reduced as much as possible, and the low-duty-ratio narrow pulse can be detected in real time on the broadband spectrum.

Description

Technical Field

[0001] The present invention relates to signal detection technology, and in particular to a real-time detection method and system for low duty cycle narrow pulses in broadband spectrum monitoring. Background Art

[0002] Energy detection is an algorithm with the most application scenarios. It can be performed in the time domain or frequency domain. It does not require prior knowledge of the interference signal, is simple to implement, and has low complexity. It is a commonly used method for real-time detection in spectrum monitoring. The broadband spectrum monitoring system is wide open in the frequency domain. In addition, low duty cycle narrow pulse signals are bursty in the time domain. For such signals, the broadband spectrum monitoring system is also "wide open" in the time domain. Therefore, it is necessary to detect and judge in both the time domain and the frequency domain to capture low duty cycle narrow pulses. Therefore, the existing detection algorithms are not strong in real-time detection capabilities for low duty cycle narrow pulse signals, and require a lot of hardware resource consumption.

[0003] In order to provide frequency resolution and distinguish signals arriving at the same time, the time domain energy detection method usually combines channelization technology to divide the broadband received signal into multiple narrowbands and then detect the output of each channel; or a single channel samples the ADC point by point to detect the pulse and then send it to the next stage for further analysis. If the pulse signal bandwidth is relatively narrow and the peak signal-to-noise ratio (or interference-to-signal ratio) is not high enough, it will cause the loss of signal-to-noise ratio (interference-to-signal ratio) within the channel bandwidth, and the pulse cannot be detected. In order to provide time resolution, the frequency domain energy detection method can only reduce the sample length of the spectrum estimation, that is, the FFT length is limited, so the signal-to-noise ratio of each frame of the estimated spectrum is low. If multiple frames of spectrum are accumulated before detection, when the duty cycle of the pulse signal is very low, then the accumulation of multiple frames of spectrum will only further reduce the signal-to-noise ratio (interference-to-signal ratio) of the accumulated pulse signal, making it difficult to detect. The time-frequency analysis method can obtain better resolution in both time and frequency dimensions, but it has to compromise in multiple performance aspects such as computational complexity and time-frequency cross-term interference. In addition, no matter which of the above methods is adopted, it is ultimately necessary to detect the output of each time and frequency two-dimensional resolution unit one by one, and the resource consumption of this part of the calculation will not be very low.

[0004] At present, the main technical solutions for pulse signal detection are:

[0005] 1. Pulse interference detection based on time domain energy judgment. Perform multiple energy judgments on the time domain sampling sequence or multiple sampling cumulative energy judgments to achieve the effect of detecting pulse interference. However, for high-speed ADC signals, the instantaneous bandwidth is large and the signal-to-noise ratio is low, so the detection performance will be affected, and the signal's center frequency, bandwidth and other spectral information cannot be obtained. As mentioned earlier, this method is usually combined with channelization technology to detect the channelized narrowband low-speed signal, but due to the loss of signal-to-noise ratio, only pulse signals with extremely high peak power can be detected.

[0006] 2. Pulse interference detection based on frequency domain energy judgment and periodicity of judgment results. This type of method also uses multiple judgment results of a single frame spectrum, but its purpose is to target periodic pulse signals, that is, after binary judgment is performed on the accumulation of multiple frames of spectrum lines, the observation is long enough. On the one hand, it is necessary to judge whether the signal is periodic based on whether it appears repeatedly in the same frequency resolution unit, and on the other hand, it is necessary to judge whether it is periodic based on the frequency of the signal's presence or absence during the observation time. Combining these two aspects can greatly reduce the false alarm rate. However, this method is only for periodic pulse signals and requires the signal to have a higher peak power.

[0007] In summary, the existing technical solutions are mainly aimed at pulse signals with large peak power, high duty cycle or specific regularity. The reason is that the detection of low duty cycle narrow pulses requires two-dimensional detection in the frequency domain and time domain, especially for the wide-band spectrum monitoring system with high-speed ADC architecture, and this process consumes a lot of hardware resources. Summary of the invention

[0008] The technical problem to be solved by the present invention is as follows: In view of the above-mentioned problems in the prior art, a real-time detection method and system for low duty cycle narrow pulses in broadband spectrum monitoring are provided to reduce the consumption of hardware resources as much as possible and achieve real-time detection of low duty cycle narrow pulses on a broadband spectrum.

[0009] In order to solve the above technical problems, the technical solution adopted by the present invention is:

[0010] A real-time detection method for low duty cycle narrow pulses in broadband spectrum monitoring comprises the following steps:

[0011] Acquire the signal of low duty cycle narrow pulse and perform real-time spectrum estimation to obtain the corresponding estimated spectrum line;

[0012] The estimated spectrum line is incoherently accumulated to obtain the cumulative spectrum line, and the estimated value of the cumulative spectrum line is updated in real time according to the cumulative spectrum line at the previous moment and the current estimated spectrum line;

[0013] Performing a linear transformation on each frame of estimated spectrum according to the corresponding cumulative spectrum line estimation value to obtain a spectrum line after linear transformation;

[0014] The first accumulation times and the second accumulation times are determined according to the minimum pulse width to be detected and the maximum pulse repetition period, respectively; the spectrum lines after linear transformation of a specified number of frames are accumulated according to the first accumulation times and the first constant false alarm detection is performed; the frames corresponding to the spectrum lines passing the first constant false alarm detection are secondarily accumulated according to the second accumulation times; then the spectrum lines of the second accumulated frames are secondarily accumulated to obtain the pulse signal detection result.

[0015] Furthermore, when acquiring a low duty cycle narrow pulse signal and performing real-time spectrum estimation, it includes:

[0016] If the current signal is a noise-superimposed stationary signal, assume that the real and imaginary parts of the kth spectral line obey independent and identically distributed Gaussian distributions. Then the probability function of the amplitude spectrum distribution is:

[0017]

[0018] Among them, x k represents the amplitude of the kth spectral line;

[0019] If the current signal is a noise superimposed input pulse signal, the amplitude spectrum distribution probability function is:

[0020]

[0021] Among them, I 0 [·] represents the first kind of modified Bessel function of order 0, s k is the amplitude of the kth spectral line of the deterministic signal, is the statistical characteristic of the noise component in the spectrum.

[0022] Furthermore, when the estimated spectral lines are incoherently accumulated to obtain the accumulated spectral lines, the spectral lines of the noise-superimposed stationary type signal and / or the noise-superimposed input pulse signal within the specified time window up to the current frame are incoherently accumulated so that the statistical characteristics after the incoherent accumulation obey the Gaussian distribution.

[0023] Furthermore, when the cumulative spectrum line estimation value is updated in real time according to the cumulative spectrum line at the last moment and the current estimated spectrum line, the cumulative spectrum line at the last moment and the current estimated spectrum line are weightedly summed using a smoothing coefficient.

[0024] Furthermore, the expression of the cumulative spectral line estimate is as follows:

[0025]

[0026] in, is the cumulative spectral line estimate of the current frame, is the cumulative spectrum line estimation value of the previous frame, and N(n) is the estimated spectrum line of the current frame.

[0027] Furthermore, when linear transformation is performed on each frame of estimated spectral lines according to the corresponding cumulative spectral line estimation value, the inverse of the corresponding cumulative spectral line estimation value is used as a weighting coefficient to perform linear transformation on each frame of estimated spectral lines.

[0028] Furthermore, the spectral line expression after linear transformation is as follows:

[0029]

[0030] in, is the cumulative spectral line estimation value of the current frame, and N(n) is the estimated spectral line of the current frame.

[0031] Furthermore, when the first accumulation times and the second accumulation times are determined respectively according to the minimum pulse width to be detected and the maximum pulse repetition period, the expressions are as follows:

[0032] M 1 =T 2 / T 1

[0033] M 2 =k×(T 3 ÷T 1 )

[0034] Among them, M 1 is the number of initial accumulations, M 2 is the number of secondary accumulations, T 1 is the single frame spectrum analysis duration, T 2 is the minimum pulse width to be detected, T 3 is the maximum pulse repetition period, k is an integer;

[0035] After respectively determining the number of initial accumulation times and the number of secondary accumulation times according to the minimum pulse width to be detected and the maximum pulse repetition period, the method further includes:

[0036] Obtain the Gaussian distribution characteristics of the spectrum of the frame accumulated according to the initial accumulation times and the first constant false alarm rate to determine the detection threshold of the first constant false alarm detection. The Gaussian distribution characteristics of the spectrum of the frame accumulated according to the initial accumulation times are:

[0037] Obtain the Gaussian distribution characteristics of the spectral lines of the frames accumulated according to the second accumulation times and the second constant false alarm rate to determine the detection threshold of the second constant false alarm detection, and the second constant false alarm rate is greater than the first constant false alarm rate. The Gaussian distribution characteristics of the spectral lines of the frames accumulated according to the second accumulation times are:

[0038] The present invention also proposes a real-time detection system for low duty cycle narrow pulses in broadband spectrum monitoring, comprising a microprocessor and a computer-readable storage medium connected to each other, wherein the microprocessor is programmed or configured to execute any one of the real-time detection methods for low duty cycle narrow pulses in broadband spectrum monitoring.

[0039] The present invention also proposes a computer-readable storage medium, in which a computer program is stored. When the computer program is executed by a microprocessor, the steps of any one of the methods for real-time detection of low duty cycle narrow pulses in broadband spectrum monitoring are implemented.

[0040] Compared with the prior art, the advantages of the present invention are:

[0041] The present invention estimates the spectral line statistical parameters in real time by accumulating the spectrum and updating it, and processes the spectral line amplitude in advance based on this, so that the entire spectrum obeys independent and identical distribution, which is equivalent to reducing the search of the frequency resolution unit dimension and reducing the consumption of hardware resources. On this basis, through two accumulations and two CFAR detections, the signal-to-noise ratio of the signal is greatly improved, and such signals can be effectively detected. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 is a flow chart of an embodiment of the present invention.

[0043] Figure 2 Schematic diagram of the characteristics of the signal and noise spectrum after linear changes in an embodiment of the present invention.

[0044] Figure 3 Schematic diagram for comparing detection results of low duty cycle narrow pulses in an embodiment of the present invention. DETAILED DESCRIPTION

[0045] The present invention is further described below in conjunction with the accompanying drawings and specific preferred embodiments, but the protection scope of the present invention is not limited thereby.

[0046] Embodiment 1

[0047] The average power of low duty cycle pulse signals is low, and if the peak power is also low, it is difficult to detect using existing methods. The energy detection method requires searching and detecting one by one on the resolution units of both time and frequency dimensions at the same time, which consumes a lot of hardware resources during real-time detection.

[0048] This embodiment proposes a real-time detection method for low duty cycle narrow pulses in broadband spectrum monitoring, which uses the property transformation of estimated spectrum lines to make the statistical characteristics of full-band spectrum lines tend to be independent and identically distributed, thereby achieving real-time detection and parameter estimation of such signals at a lower hardware resource cost, such as Figure 1 As shown, the following steps are included:

[0049] S1) acquiring a low duty cycle narrow pulse signal and performing real-time spectrum estimation to obtain a corresponding estimated spectrum line;

[0050] S2) performing incoherent accumulation of the estimated spectral lines to obtain cumulative spectral lines, and updating the estimated value of the cumulative spectral lines in real time according to the cumulative spectral lines at the previous moment and the current estimated spectral lines;

[0051] S3) performing a linear transformation on each frame of estimated spectral lines according to the corresponding cumulative spectral line estimation value to obtain a spectral line after linear transformation;

[0052] S4) accumulating the linearly transformed spectral lines of a specified number of frames according to the minimum pulse width to be detected and performing constant false alarm detection, performing secondary accumulation on the frames corresponding to the spectral lines that pass the constant false alarm detection, and then performing secondary constant false alarm detection on the spectral lines of the secondary accumulated frames to obtain a pulse signal detection result.

[0053] The assumption of this embodiment is that the main components of the monitored spectrum are background noise and time-domain continuous stationary signals. Under the assumption, through the above steps, the statistical parameters of the spectrum of background noise and stationary signals can be accumulated multiple times to achieve asymptotic unbiased estimation. Secondly, all spectrum lines are transformed into independent and identically distributed variables with the same parameters through linear transformation using the estimated parameters. Then the spectrum lines of both are regarded as the "background noise" of pulse detection, and the spectrum lines of several real-time spectrum estimates are designed for accumulation. Then, each frame of spectrum estimation uses a unified decision threshold to perform CFAR detection on the spectrum lines, marking whether there is a short pulse signal in the frame spectrum. Finally, by performing secondary accumulation and secondary judgment on the spectrum lines of the marked frame, the false alarm rate and missed detection rate can be reduced.

[0054] Each step is described in detail below.

[0055] In step S1 of this embodiment, when performing real-time spectrum estimation, a FFT of appropriate length is specifically selected for spectrum estimation to ensure that the single analysis duration is less than the pulse width, which can provide better time resolution on the one hand, and ensure that the bandwidth of the single spectrum line of FFT is equivalent to the bandwidth of the pulse on the other hand so as not to lose the signal-to-noise ratio within the unit bandwidth. How to use FFT for spectrum estimation is well known to those skilled in the art, and this embodiment does not involve improvements in the specific implementation process, and the specific implementation process will not be repeated here.

[0056] The results of real-time spectrum estimation for low duty cycle narrow pulse signals include the following two situations:

[0057] When the noise is superimposed on a stationary signal (a continuous signal in the time domain), the amplitude spectrum obeys the Rayleigh distribution. Assume that the real and imaginary parts of the kth spectral line obey independent and identically distributed Gaussian distributions. σ sk is the variance of the noise, then the probability function of the amplitude spectrum distribution is:

[0058]

[0059] Among them, x k represents the amplitude of the kth spectral line, and its mean The variance is

[0060] When the noise is superimposed on the input pulse signal, it can be regarded as an instantaneous stable deterministic signal within the FFT time, and its spectrum line amplitude obeys the Rician distribution. The amplitude spectrum distribution probability function is:

[0061]

[0062] Among them, I 0 [·] represents the first kind of modified Bessel function of order 0, s k is the amplitude of the kth spectral line of the deterministic signal, is the statistical characteristic of the noise component in the spectrum. Let the signal-to-noise ratio of a single spectrum line be is the Rice coefficient. The expected value and variance of the amplitude of each spectral line can be described by the Rice coefficient:

[0063]

[0064] In step S2 of this embodiment, in the initialization stage, multiple frame spectra are accumulated to obtain asymptotic unbiased estimates of the current channel continuous signal and background noise, and the accumulated spectrum is updated by using a weighted accumulation method of real-time estimated spectra.

[0065] Specifically, when the estimated spectral lines are incoherently accumulated to obtain the accumulated spectral lines, the spectral lines of multiple frames of the above-mentioned noise-superimposed stationary type signals and / or noise-superimposed input pulse signals within a specified time window up to the current frame are incoherently accumulated, so that the statistical characteristics after the incoherent accumulation obey the Gaussian distribution.

[0066] Since the amplitude of any spectral line in each frame of spectrum trace can be regarded as independent and identically distributed, according to the limit center theorem, the distribution after sufficient accumulation can be approximated by Gaussian random process, that is, the statistic Follows standard normal distribution Among them, μ n represents the expected value of the noise, x n,k represents the amplitude of the kth spectral line in the nth frame, and the statistic of the incoherent accumulation of multiple frames of spectral lines is Z n , then the statistic obey Distribution, this statistic is an asymptotically unbiased estimate. By increasing the number of accumulated frames, the statistic Z is reduced n The variance can make the output amplitude close to the expected value of the amplitude of each spectral line.

[0067] In step S3 of this embodiment, the accumulated spectrum is smoothed and used as an unbiased estimate of the spectrum mean, and its reciprocal is taken as a weighting coefficient for linear transformation to perform linear transformation on the estimated spectrum of each current frame. This transformation has little effect on the original signal-to-noise ratio (interference-to-signal ratio) of low duty cycle pulses. Figure 2 As shown, Figure 2 (a) represents the PSK real-time spectrum sequence of the estimated spectrum line. The spectrum line height is unevenly distributed and is clearly divided into the background noise area and the signal area. Figure 2 (b) represents the PSK real-time spectrum sequence of the spectrum after linear transformation. It can be seen that the spectrum after linear transformation is highly "flattened", whether it is the spectrum area with pure background noise or the spectrum area with modulated time domain continuous signal superimposed; Figure 2 (c) shows the statistical fitting of the PSK real-time spectrum sequence (mean not zeroed) after linearization. The statistical characteristics of the "flat" spectrum line amplitude obey the same distribution, which is consistent with the theoretical model in this scheme. In summary, the "flattened" spectrum line can be regarded as the "noise floor" of pulse detection, without the need to detect the spectrum line first, distinguish the signal interval and the noise floor interval, and estimate the statistical parameters of each spectrum line. There is no need to set different detection thresholds for the noise floor area and the signal area, and there is no need to compare different thresholds one by one.

[0068] Specifically, when the cumulative spectrum line estimate is updated in real time according to the cumulative spectrum line at the previous moment and the current estimated spectrum line, the cumulative spectrum line at the previous moment and the current estimated spectrum line are weighted summed using the smoothing coefficient. The expression of the cumulative spectrum line estimate is as follows:

[0069]

[0070] in, is the cumulative spectral line estimate of the current frame, is the cumulative spectrum line estimation value of the previous frame, and N(n) is the estimated spectrum line of the current frame.

[0071] Furthermore, when each frame of estimated spectral line is linearly transformed according to the corresponding cumulative spectral line estimation value, the inverse of the corresponding cumulative spectral line estimation value is used as a weighting coefficient to perform a linear transformation on each frame of estimated spectral line. The spectral line expression after linear transformation is as follows:

[0072]

[0073] in, is the cumulative spectral line estimation value of the current frame, and N(n) is the estimated spectral line of the current frame.

[0074] For the spectrum lines after linear transformation, if the spectrum estimation of the current frame does not contain pulse signals, then all spectrum lines obey the uniform Rayleigh distribution and are independent. It can be seen that the probability distribution function of the random variable y is

[0075]

[0076] That is, the distribution of the current spectrum line has nothing to do with whether the spectrum line is noise or a time-domain continuous signal, and it uniformly obeys the mean of 0 and the variance of The Rayleigh distribution of .

[0077] If the current estimated spectrum contains an impulse signal, the probability distribution function of the variable y is the following Rice distribution:

[0078]

[0079] in It can be seen that the signal amplitude is also scaled proportionally at this time, and the signal-to-noise ratio (interference-to-signal ratio) of the new random variable y remains unchanged.

[0080] At this time, the judgment of the pulse signal in each frame estimation spectrum is transformed into a signal detection problem under a background noise with constant statistical characteristics and independent and identical distribution. Therefore, the constant false alarm detection threshold does not need to be estimated and adjusted according to the background environment, and the two-dimensional search problem does not need to be searched one by one on the frequency resolution unit, which reduces the consumption of hardware resources. According to the false alarm rate and the detection rate and formula (7) and formula (8), the CFAR thresholds of the two under different signal-to-noise ratios can be obtained respectively, and the threshold interval that satisfies the false alarm rate and the detection rate at the same time is selected as the first CFAR detection threshold. When designing the threshold, the detection rate should be met as much as possible and the false alarm rate index should be appropriately relaxed. The false alarm rate can be reduced by the secondary accumulation and secondary detection of the next step S4.

[0081] In step S4 of this embodiment, the number of initial accumulations of the real-time spectrum and the length of the secondary accumulation FIFO are first determined according to the minimum pulse width to be detected and the maximum pulse repetition period. In step S3, in any frame of the initial accumulation spectrum, as long as no less than one spectrum line is detected to exceed the threshold, the current initial accumulation frame is sent to the FIFO for secondary accumulation; all the spectrum lines of the initial accumulation frames in the FIFO are accumulated along the time dimension to obtain a secondary accumulation spectrum, and the secondary accumulation spectrum is subjected to constant false alarm detection again, and the pulse signal is subjected to binary judgment, and the frequency domain parameter measurement can be obtained at the same time.

[0082] Specifically, when accumulating the linearly transformed spectrum lines of a specified number of frames according to the minimum pulse width to be detected and performing constant false alarm detection, the linearly transformed single-frame spectrum estimation is subjected to smoothing filtering of several frames according to the minimum pulse width to be detected. Assuming that the single-frame spectrum analysis time is T 1 , minimum pulse width T 2 , then take the initial accumulation times M 1 =T 2 / T 1 , for example, the minimum pulse width is twice the single real-time spectrum length, then M 1Take 2. According to the derivation in the previous article, the spectrum line of the smoothed pulse-free signal (that is, the spectrum line of the frame accumulated according to the number of initial accumulations) is approximately distributed in Gaussian distribution. The standard deviation is According to the Gaussian distribution characteristics and constant false alarm rate P fa Determine the detection threshold T fa To satisfy:

[0083]

[0084] The smoothed spectrum is tested with constant false alarm. As long as the spectrum of each frame exceeds the threshold, the frame is marked and sent to the second-level accumulation. The width of the accumulation smoothing window is M. 2 That is, the FIFO length, assuming that the maximum pulse repetition period is T 3 , then M 2 =k×(T 3 ÷T 1 ), k is an integer, which can be 3 to 8 according to experience. 2 When it is particularly large, the recursive summation method can also be used, S new =S old +x new -x old , only one adder, one subtractor and one register are needed to store the current sum, avoiding excessive hardware resource consumption. The detection threshold of the second constant false alarm detection is still calculated based on the Gaussian distribution characteristics and the constant false alarm rate. The difference is that the Replace with And the secondary detection can use a higher constant false alarm rate indicator to determine the detection threshold.

[0085] The detection performance of the entire algorithm is approximately determined by the above two CFAR detections: the initial accumulation and detection determines the signal detection rate; the secondary accumulation and detection determines the false alarm rate. Therefore, the initial detection should ensure the signal detection rate as much as possible, and the false alarm rate index should be decomposed into the secondary accumulation and detection.

[0086] In order to verify the effectiveness of this embodiment, a pulse detection process with a duty cycle of 2‰, an in-band signal-to-noise ratio of -12dB, and a pulse width of 3 times the real-time spectrum processing time was simulated. Figure 3 As shown, Figure 3 (a) represents the time domain envelope of a signal with a low duty cycle narrow pulse, Figure 3 (b) represents the accumulated spectrum obtained by incoherently integrating the estimated spectrum. Figure 3 (c) represents the real-time spectrum sequence of the estimated spectrum line, Figure 3 (d) represents the real-time spectrum sequence of the spectral lines after linear transformation, Figure 3(e) shows the result of smoothing the real-time spectrum sequence of the spectrum line after linear transformation three times. Figure 3 (f) represents the result of the detection after the second accumulation of the real-time spectrum sequence of the spectrum line after the linear transformation. From the simulation point of view, the method of simply detecting the energy in the time domain or the energy detection in the frequency domain accumulation spectrum cannot detect the existence of the pulse signal. According to this method, the pulse is difficult to detect directly from the frequency domain energy until the third smoothing process. Through the process of detection, second accumulation, and re-detection, the energy of the signal unit where the pulse signal is located is highlighted, and the signal can be easily detected.

[0087] Embodiment 2

[0088] This embodiment proposes a real-time detection system for low duty cycle narrow pulses in broadband spectrum monitoring, including a microprocessor and a computer-readable storage medium connected to each other, and the microprocessor is programmed or configured to execute the real-time detection method for low duty cycle narrow pulses in broadband spectrum monitoring described in any one of Embodiment 1 and Embodiment 2.

[0089] This embodiment further proposes a computer-readable storage medium, in which a computer program is stored. When the computer program is executed by a microprocessor, the steps of the real-time detection method of low duty cycle narrow pulses in broadband spectrum monitoring described in any one of Embodiment 1 and Embodiment 2 are implemented.

[0090] The above is only a preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technicians in this technical field, some improvements and modifications without departing from the principle of the present invention should also be regarded as the protection scope of the present invention.

Claims

1. A real-time detection method for low duty cycle narrow pulses in broadband spectrum monitoring, characterized in that: The following steps are involved: Acquire the signal of low duty cycle narrow pulse and perform real-time spectrum estimation to obtain the corresponding estimated spectrum line; The estimated spectrum line is incoherently accumulated to obtain the cumulative spectrum line, and the estimated value of the cumulative spectrum line is updated in real time according to the cumulative spectrum line at the previous moment and the current estimated spectrum line; Performing a linear transformation on each frame of estimated spectrum according to the corresponding cumulative spectrum line estimation value to obtain a spectrum line after linear transformation; The first accumulation times and the second accumulation times are determined according to the minimum pulse width to be detected and the maximum pulse repetition period, respectively; the spectrum lines after linear transformation of a specified number of frames are accumulated according to the first accumulation times and the first constant false alarm detection is performed; the frames corresponding to the spectrum lines passing the first constant false alarm detection are secondarily accumulated according to the second accumulation times; then the spectrum lines of the second accumulated frames are secondarily accumulated to obtain the pulse signal detection result.

2. The real-time detection method of low duty cycle narrow pulses in broadband spectrum monitoring according to claim 1 is characterized in that: When acquiring a low duty cycle narrow pulse signal and performing real-time spectrum estimation, it includes: If the current signal is a noise-superimposed stationary signal, assume that the real and imaginary parts of the kth spectral line obey independent and identically distributed Gaussian distributions. Then the probability function of the amplitude spectrum distribution is: Among them, x k represents the amplitude of the kth spectral line; If the current signal is a noise superimposed input pulse signal, the amplitude spectrum distribution probability function is: where I0[·] represents the first kind of modified Bessel function of order 0, s k is the amplitude of the kth spectral line of the deterministic signal, is the statistical characteristic of the noise component in the spectrum.

3. The real-time detection method of low duty cycle narrow pulses in broadband spectrum monitoring according to claim 2 is characterized in that: When the estimated spectral lines are incoherently accumulated to obtain the accumulated spectral lines, specifically, the spectral lines of the noise-superimposed stationary type signal and / or the noise-superimposed input pulse signal within the specified time window up to the current frame are incoherently accumulated so that the statistical characteristics after the incoherent accumulation obey the Gaussian distribution.

4. The real-time detection method of low duty cycle narrow pulses in broadband spectrum monitoring according to claim 1 is characterized in that: When the cumulative spectrum line estimation value is updated in real time according to the cumulative spectrum line at the last moment and the current estimated spectrum line, the cumulative spectrum line at the last moment and the current estimated spectrum line are weightedly summed using a smoothing coefficient.

5. The real-time detection method of low duty cycle narrow pulses in broadband spectrum monitoring according to claim 4 is characterized in that: The expression of the cumulative spectrum line estimate is as follows: in, is the cumulative spectral line estimate of the current frame, is the cumulative spectrum line estimation value of the previous frame, and N(n) is the estimated spectrum line of the current frame.

6. The real-time detection method of low duty cycle narrow pulses in broadband spectrum monitoring according to claim 1 is characterized in that: When linear transformation is performed on each frame of estimated spectral lines according to the corresponding cumulative spectral line estimation value, specifically, the inverse of the corresponding cumulative spectral line estimation value is used as a weighting coefficient to perform linear transformation on each frame of estimated spectral lines.

7. The real-time detection method of low duty cycle narrow pulses in broadband spectrum monitoring according to claim 6 is characterized in that: The spectral line expression after linear transformation is as follows: in, is the cumulative spectral line estimation value of the current frame, and N(n) is the estimated spectral line of the current frame.

8. The real-time detection method of low duty cycle narrow pulses in broadband spectrum monitoring according to claim 1 is characterized in that: When the first accumulation times and the second accumulation times are determined according to the minimum pulse width to be detected and the maximum pulse repetition period, the expressions are as follows: M1=T2 / T1 M2=k×(T3÷T1) Wherein, M1 is the number of initial accumulations, M2 is the number of secondary accumulations, T1 is the duration of single-frame spectrum analysis, T2 is the minimum pulse width to be detected, T3 is the maximum pulse repetition period, and k is an integer; After respectively determining the number of initial accumulations and the number of secondary accumulations according to the minimum pulse width to be detected and the maximum pulse repetition period, the method further includes: Obtain the Gaussian distribution characteristics of the spectrum of the frame accumulated according to the initial accumulation times and the first constant false alarm rate to determine the detection threshold of the first constant false alarm detection. The Gaussian distribution characteristics of the spectrum of the frame accumulated according to the initial accumulation times are: Obtain the Gaussian distribution characteristics of the spectral lines of the frames accumulated according to the second accumulation times and the second constant false alarm rate to determine the detection threshold of the second constant false alarm detection, and the second constant false alarm rate is greater than the first constant false alarm rate. The Gaussian distribution characteristics of the spectral lines of the frames accumulated according to the second accumulation times are:

9. A real-time detection system for low duty cycle narrow pulses in broadband spectrum monitoring, characterized in that: The invention comprises a microprocessor and a computer-readable storage medium connected to each other, wherein the microprocessor is programmed or configured to execute the real-time detection method of low duty cycle narrow pulses in broadband spectrum monitoring as claimed in any one of claims 1 to 8.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, which, when executed by a microprocessor, implements the steps of the real-time detection method for low duty cycle narrow pulses in broadband spectrum monitoring according to any one of claims 1 to 8.

Images