Adaptive threshold blind signal detection method based on power spectrum gradient

Abstract

The invention discloses a self-adaptive threshold blind signal detection method based on a power spectrum gradient, and belongs to the technical field of wireless communication. The technical problem to be solved by the invention is to provide the adaptive threshold blind signal detection method based on the power spectrum gradient for an environment with a wide space-based field of view, a relatively large Doppler effect and time-frequency domain signal aliasing, so that blind signals can be detected in an undulating and complicated electromagnetic environment and under noise interference, and the blind signal detection accuracy is improved. According to the invention, signal detection and frequency estimation can be effectively realized, signal detection can be completed without any prior information, the influence of noise can be effectively overcome, signal detection is carried out under a low signal-to-noise ratio, and the accuracy of frequency spectrum estimation is high. According to the method disclosed by the invention, a self-adaption-based threshold setting method is provided, and different signal and noise conditions can be more flexibly adapted in frequency domain analysis, so that the robustness and accuracy of detection are improved. Compared with a fixed threshold value, the self-adaptive dual threshold value can adjust the threshold value according to signal characteristics and noise level observed in real time, the detection sensitivity is improved, the influence of noise on a detection result is reduced, and the performance of the system in a complex environment is improved. Compared with a single threshold value, the self-adaptive double threshold values avoid missing detection in a noise environment, and the accuracy and sensitivity of signal detection are improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of wireless communication, and particularly relates to an adaptive threshold blind signal detection method based on power spectrum gradient in this field. Background Art

[0002] In radio spectrum detection, since the signal frequency position and bandwidth range are unknown, it is necessary to search within a relatively wide frequency range in order to detect the target signal with unknown working frequency points and unknown bandwidths in a timely manner. For blind signal detection without available prior information, in a complex electromagnetic environment and noise interference factors in the real scenario, it is particularly important to overcome the influence of noise to achieve signal detection.

[0003] Traditional signal detection techniques mainly include energy detection, matched filtering, eigenvalue detection, singular value detection, cyclostationary detection, etc. These detection methods have their own advantages and disadvantages in different scenarios: Matched filtering detection has the theoretically best sensing performance, but it requires prior knowledge of the specific information of the signal to be detected and is not a blind detection method. The energy detection method is based on the difference between the noise energy and the signal energy in the environment to achieve the purpose of signal detection. It does not require any prior information, has a low computational complexity, and has been widely used. It is a blind detection method, but the algorithm performance is limited by the energy information and is easily interfered by unstable noise. Eigenvalue-based detection does not require any prior information and is less affected by unstable noise, but has a large computational complexity and requires a large number of sampling points. Among these methods, most are based on narrowband signal detection. Wideband signal detection mainly includes morphological detection methods and gradient feature detection methods, etc. The advantage of the morphological detection method is that it can accurately estimate the noise floor, but the algorithm is complex and the computational amount is large; the gradient feature detection method has high detection accuracy and good anti-noise performance, but the gradient-based detection mainly has the following two disadvantages: one is that the amplitude characteristics of the signal will be severely lost when calculating the gradient features, and the other is that improper scale selection will cause signal missed detection. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide an adaptive threshold blind signal detection method based on power spectrum gradient for an environment with a wide space-based field of view, a large Doppler effect, and time-frequency domain signal aliasing, which can detect blind signals under fluctuating and complex electromagnetic environments and noise interference, can effectively achieve signal detection and frequency estimation, can complete signal detection without any prior information, and can effectively overcome the influence of noise, and has high accuracy in signal detection and spectrum estimation at low signal-to-noise ratios.

[0005] The present invention adopts the following technical solutions:

[0006] An adaptive threshold blind signal detection method based on power spectrum gradient, the improvement of which includes the following steps:

[0007] Step 1: Read the time-domain sequence that has been converted by analog-to-digital conversion at the receiver output, and use the Welch algorithm to estimate the power spectrum of the time-domain sequence;

[0008] A method of segmenting and windowing is used to divide a data segment of length N, r(n), n = 0, 1,..., N - 1, into L segments, each segment having M data. The data of the i-th segment is expressed as:

[0009] r i (n) = r(n + iM - M), 0 ≤ n ≤ M, 1 ≤ i ≤ L

[0010] Then, the window function w(n) is added to each data segment, and the periodogram of each segment is obtained. The periodogram of the a-th segment is:

[0011]

[0012] In the above formula, U is the normalization factor:

[0013]

[0014] Approximating the periodograms of each segment as uncorrelated with each other, the power spectrum estimate value G xx (e jω ) is:

[0015]

[0016] In the above formula,

[0017] Step 2: Calculate the gradient of the power spectrum obtained in Step 1, use the gradient operator to obtain the gradient spectrum, and perform an averaging process on the gradient spectrum data to obtain the mean gradient spectrum:

[0018] Let the initial gradient value S sub (0) = 0, and the calculation method of the gradient is as follows:

[0019] S sub (k) = G(k) - G(k - 1)

[0020] In the above formula, k is a natural number, and G(k) = G xx (e jω );

[0021] Step 3: Calculate the mean and standard deviation of the power spectrum gradient obtained in Step 2, calculate the minimum upper threshold and the maximum lower threshold, calculate the adaptive window size according to the local change rate of the power spectrum gradient, determine the left and right boundaries of the window, extract the data within the window, calculate the local mean and standard deviation, calculate the adaptive k value, and obtain the adaptive upper and lower threshold arrays:

[0022] The minimum upper threshold th1 and the maximum lower threshold th2 are respectively:

[0023]

[0024]

[0025] In the above formula, is the mean of the input power spectrum gradient, and V global is the standard deviation of the input power spectrum gradient;

[0026] Traverse the input power spectrum gradient in the way of a sliding window, and calculate the adaptive window size within each window. Set the initial parameter values, MSize is the minimum window size, and BSize is the basic window size:

[0027]

[0028] In the above formula, ASize is the adaptive window size, is the local change rate of the power spectrum gradient;

[0029] For each window, set the initial parameter values, K base is the basic K value, α is the adjustment factor, extract the data within the window, and calculate its mean and the standard deviation V local , and then adjust the threshold according to the adaptive k value, and set the threshold to the mean plus k times the standard deviation:

[0030]

[0031] In the above formula, N local is the length of the local window. If the calculated threshold is lower than the minimum upper threshold th1, then use the minimum upper threshold; if it is higher than the maximum lower threshold th2, then use the maximum lower threshold. Finally, obtain the adaptive upper and lower threshold arrays:

[0032]

[0033] In the above formula, th1 A is the adaptive upper threshold, and th2 A is the adaptive lower threshold;

[0034] Step 4, set the adaptive threshold, traverse the power spectrum gradient according to the threshold monitoring criterion, judge the signal and noise, and output the framed signal interval information:

[0035] Compare the adaptive upper and lower threshold arrays obtained in Step 3 with the power spectrum gradient obtained in Step 2, find the data points where the power spectrum gradient exceeds the upper threshold and the data points below the lower threshold, set the first point exceeding the upper threshold as the starting point and the last point as the ending point, and set the first point below the lower threshold as the starting point and the last point as the ending point;

[0036] Take the ending point exceeding the upper threshold as the interval starting point, and the starting point exceeding the lower threshold as the interval ending point, perform matching, output the final successfully matched signal interval array, and obtain the final result of signal detection;

[0037] Take the frequency component corresponding to the signal starting interval index bit as the starting frequency point of the signal, fK start_idx is the index of the starting frequency point of the detected signal, fK end_idx is the index of the ending frequency point of the detected signal, C is the center frequency, and B is the bandwidth:

[0038] C = (fK start_idx + fK end_idx ) / 2

[0039] Obtain the signal interval according to the starting index to the ending index, define the frequency within the interval as fK in , the power spectrum value within the interval as G in , find the maximum power spectrum value G max_in within the interval range, subtract 3db to get the value G half_in at the half-power point, find the index G half_in of all points within the interval that are greater than or equal to G idxs :

[0040] G half_in = G max_in - 3

[0041] G idxs = find(G in >= G half_in )

[0042] B = abs(fK in (max(G idxs )) - fK in (min(G idxs ))).

[0043] Further, the threshold monitoring criterion in step 4 is as follows: judge the signal and noise from the power spectrum gradient map and output the interval information of the signal; the logic is: only one value crosses the upper threshold and the next value also crosses the upper threshold, indicating that the current area is noise; only one value crosses the lower threshold and the next value also crosses the lower threshold, indicating that the current area is also noise; when the first value crosses the upper threshold and the next value crosses the lower threshold, it indicates that the current area is a signal.

[0044] The beneficial effects of the present invention are as follows:

[0045] The present invention is proposed for an environment with a broad space-based field of view, a large Doppler effect, and time-frequency domain signal aliasing. The present invention can detect blind signals under a fluctuating and complex electromagnetic environment and noise interference, can effectively realize signal detection and frequency estimation, can complete signal detection without any prior information, and can effectively overcome the influence of noise, and has high accuracy in signal detection and spectrum estimation under low signal-to-noise ratios. The adaptive threshold setting method given by the present invention can more flexibly adapt to different signal and noise conditions in frequency domain analysis, greatly improving the robustness and accuracy of detection. Compared with a fixed threshold, the adaptive dual threshold can adjust the threshold value according to the signal characteristics and noise level observed in real time, improving the detection sensitivity, helping to reduce the influence of noise on the detection result, and improving the performance of the system in a complex environment. Compared with a single threshold, the adaptive dual threshold avoids the situation of missed detection in a noise environment, improving the accuracy and sensitivity of signal detection.

[0046] The method disclosed by the present invention has high detection accuracy, can effectively overcome the influence of noise, and can realize blind signal detection in a space-based background without any prior information.

[0047] The method disclosed by the present invention is sensitive to frequency changes. Through the gradient information of the power spectrum in frequency, it can capture the local changes in the frequency domain structure of the signal, and shows strong robustness in spectrum analysis under complex signal and noise environments.

[0048] The method disclosed by the present invention proposes an adaptive threshold setting method, which can more flexibly adapt to different signal and noise conditions in frequency domain analysis, thereby improving the robustness and accuracy of detection. Compared with a fixed threshold, the adaptive dual threshold can adjust the threshold value according to the signal characteristics and noise level observed in real time, improving the detection sensitivity, helping to reduce the influence of noise on the detection result, and improving the performance of the system in a complex environment. Compared with a single threshold, the adaptive dual threshold avoids the situation of missed detection in a noise environment, improving the accuracy and sensitivity of signal detection. Description of the Drawings

[0049] Figure 1It is a schematic flow chart of the method of the present invention;

[0050] Figure 2 It is an adaptive threshold flow chart;

[0051] Figure 3 It is a schematic diagram of the power spectrum gradient of a certain narrowband signal;

[0052] Figure 4 It is a schematic diagram of adaptive threshold detection;

[0053] Figure 5 It is a simulation result diagram of modulation signal aliasing under the condition of the change of the number;

[0054] Figure 6 It is a simulation result diagram of the acquired signal under low signal-to-noise ratio;

[0055] Figure 7 It is a simulation result diagram of a narrowband signal under a broadband noise background;

[0056] Figure 8 It is a simulation result diagram of a broadband signal under a broadband noise background. Specific embodiments

[0057] In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0058] Embodiment 1. This embodiment discloses an adaptive threshold blind signal detection method based on power spectrum gradient. For blind signal detection, a power spectrum gradient sequence of the data to be detected is generated, the mean and standard deviation of the power spectrum gradient sequence are calculated in real time, and the double thresholds for signal detection are adaptively calculated through this statistical value. As Figure 1 shown, it includes the following steps:

[0059] Step 1, read the time-domain sequence output by the receiver and already subjected to analog-to-digital conversion, and perform power spectrum estimation on the time-domain sequence using the Welch algorithm;

[0060] Using the Welch algorithm can improve the variance characteristics of the spectrum estimated by the direct periodogram method. A method of segmenting and windowing is adopted to divide a data sequence r(n) of length N, n = 0, 1, L, N - 1 into L segments, each segment having M data. The i-th segment of data is expressed as:

[0061] r i (n) = r(n + iM - M), 0 ≤ n ≤ M, 1 ≤ i ≤ L

[0062] Then the window function w(n) is added to each data segment, and the periodogram of each segment is obtained. The periodogram of the a-th segment is:

[0063]

[0064] In the above formula, U is the normalization factor:

[0065]

[0066] Regarding each periodogram as approximately uncorrelated with each other, the power spectrum estimate value G is obtained xx (e jω ) is:

[0067]

[0068] In the above formula,

[0069] According to the probability and statistics theory, it can be obtained that: the variance of the power spectrum estimated by the Welch method is approximately 1 / L times the variance of the directly estimated spectrum, and the more segments there are, the smaller the variance. However, the overlapping of data reduces the uncorrelation of each segment, making the reduction of variance not reach the degree of theoretical calculation. In addition, increasing L will increase the bias and decrease the resolution. Therefore, in actual use, the values of L and M should be appropriately selected while considering the requirements of variance and resolution.

[0070] Step 2: Calculate the gradient of the power spectrum obtained in Step 1, use the gradient operator to obtain the gradient spectrum, and perform an averaging process on the gradient spectrum data to obtain the mean gradient spectrum:

[0071] Let the initial gradient value be S sub (0) = 0, and the calculation method of the gradient is as follows:

[0072] S sub (k) = G(k) - G(k - 1)

[0073] In the above formula, k is a natural number, and G(k) = G xx (e jω );

[0074] Step 3: Calculate the mean and standard deviation of the power spectrum gradient obtained in Step 2, calculate the minimum upper threshold and the maximum lower threshold, calculate the adaptive window size according to the local change rate of the power spectrum gradient, determine the left and right boundaries of the window, extract the data within the window, calculate the local mean and standard deviation, calculate the adaptive k value, and obtain the adaptive upper and lower threshold arrays:

[0075] As Figure 2 shown, according to the adaptive threshold solving flowchart, the minimum upper threshold th1 and the maximum lower threshold th2 calculated using the power spectrum data are respectively:

[0076]

[0077] In the above formula, is the mean of the input power spectrum gradient, V global is the standard deviation of the input power spectrum gradient;

[0078] Traverse the input power spectrum gradient in the way of a sliding window, calculate the adaptive window size within each window, and set the initial parameter values. MSize is the minimum window size, and BSize is the basic window size:

[0079]

[0080] In the above formula, ASize is the adaptive window size, is the local change rate of the power spectrum gradient;

[0081] For each window, set the initial parameter values. K base is the basic K value, α is the adjustment factor. Extract the data within the window by the algorithm, and calculate its mean value and the standard deviation V local , and then adjust the threshold according to the adaptive k value, and set the threshold to the mean value plus k times the standard deviation:

[0082]

[0083] In the above formula, N local is the length of the local window. To ensure the rationality of the threshold and suppress the excessive change of the threshold to a certain extent, if the calculated threshold is lower than the minimum upper threshold th1, then use the minimum upper threshold; if it is higher than the maximum lower threshold th2, then use the maximum lower threshold, and finally obtain the adaptive upper and lower threshold arrays:

[0084]

[0085] In the above formula, th1 A is the adaptive upper threshold, and th2 A is the adaptive lower threshold;

[0086] Step 4, set the adaptive threshold, traverse the power spectrum gradient according to the threshold monitoring criterion, judge the signal and noise, and output the framed signal interval information:

[0087] As Figure 3 shown, the threshold monitoring criterion is: judge the signal and noise from the power spectrum gradient diagram, and output the interval information of the signal; the logic is: only one value crosses the upper threshold, and the next value that crosses is also the upper threshold, indicating that the current area is noise; only one value crosses the lower threshold, and the next value that crosses is also the lower threshold, indicating that the current area is also noise; when the first value crosses the upper threshold and the next value crosses the lower threshold, it indicates that the current area is a signal.

[0088] As shown Figure 4 in the figure, compare the adaptive upper and lower threshold arrays obtained in step 3 with the power spectrum gradient obtained in step 2, find the data points where the power spectrum gradient exceeds the upper threshold and the data points below the lower threshold, set the first point exceeding the upper threshold as the starting point and the last point as the ending point, and set the first point below the lower threshold as the starting point and the last point as the ending point;

[0089] Take the ending point exceeding the upper threshold as the starting point of the interval and the starting point exceeding the lower threshold as the ending point of the interval for matching, output the final successfully matched signal interval array, and obtain the final result of signal detection;

[0090] Take the frequency component corresponding to the signal starting interval index bit as the starting frequency point of the signal, which can be used to estimate the center frequency and bandwidth. fK start_idx is the index of the starting frequency point of the detected signal, fK end_idx is the index of the ending frequency point of the detected signal, C is the center frequency, and B is the bandwidth:

[0091] C = (fK start_idx + fK end_idx ) / 2

[0092] Obtain the signal interval according to the starting index to the ending index, define the frequency within the interval as fK in , and the power spectrum value within the interval as G in , find the maximum value G max_in of the power spectrum within the interval range, subtract 3 db to get the value G half_in at the half-power point, and find the index G half_in of all points within the interval that are greater than or equal to G idxs :

[0093] G half_in = G max_in - 3

[0094] G idxs = find(G in >= G half_in )

[0095] B = abs(fK in (max(G idxs )) - fK in (min(G idxs ))).

[0096] The performance of the method of the present invention is described below through simulation experiments:

[0097] Simulation parameters

[0098]

[0099] The performance simulation results are as follows Figures 5 to 8 shown. From Figure 5 it can be seen that the correct rate of the detection result of the method of the present invention is 100%. Therefore, under the Gaussian channel, the anti-noise performance of the method of the present invention is very good.

[0100] To verify the feasibility of the method of the present invention in practical engineering, the following uses the method of the present invention to perform spectrum sensing on a certain actual acquisition data. From Figure 6 it can be seen that the amplitude of the power spectral density fluctuates greatly, showing the power spectral characteristics caused by colored noise. The width of the rectangular area is the signal bandwidth found by the method of the present invention, and the starting interval of the signal can be detected. It can be seen that the estimated carrier frequency is 70014460.858896 Hz, and the average error ratio between the estimated carrier frequency and the true carrier frequency is 2.066×10 -4 .

[0101] Figure 7 , Figure 8 The detection situations of narrowband signals and broadband signals under broadband noise are compared. Under the Gaussian channel, on the premise that the signal-to-noise ratio is -20 dB, the method of the present invention can accurately detect the signal when the signal-to-noise ratio is about -20 dB, the carrier frequency estimation is relatively accurate, and the error ratio of the signal carrier frequency is at the 10 -4 magnitude.

[0102] The supplementary table content is the signal detection correct rate and carrier frequency estimation accuracy rate for verifying the method of the present invention in the simulation scenario. The modulation methods include 8 communication signals (BPSK, QPSK, 8PSK, 2FSK, 4FSK, 2ASK, 16QAM, GMSK), 2 radar signals (simple pulse and linear frequency modulation), the number of signals are 3, 5, 8 respectively, and the signal-to-noise ratios are 20 dB, 8 dB, 5 dB respectively. As comprehensive a simulation as possible to various possible situations, the signal can be accurately detected and the carrier frequency estimation is also relatively accurate.

[0103] The specific simulation results are shown in the following table, where the average error ratio of the estimated carrier frequency value and the signal detection correct rate are both the results after 1000 Monte Carlo random simulations.

[0104] Signal aliasing results

[0105]

[0106]

[0107]

[0108]

[0109]

[0110]

[0111]

[0112]

[0113]

[0114]

[0115]

[0116]

[0117]

[0118]

[0119] It can be found that for some signals in the table, the correct rate is higher when the number of signals is larger (marked with *). The reason is that the adaptive detection algorithm determines the adaptive threshold based on the mean and standard deviation of the difference in the signal power spectrum. When there are more signals, the mean and standard deviation are also higher, so the obtained adaptive threshold is higher. Under the same signal-to-noise ratio, noise is less likely to cross the threshold, and the correct rate of signal detection is higher.

[0120] Aliasing of different modulation types

[0121]

[0122]

[0123] It has been verified that the method proposed in the present invention solves the technical problems proposed in the present invention. The method of the present invention has been practically applied, verifying the practicability of the present invention and being able to achieve the technical problems proposed in the specification of this application and reach the technical effects recorded in the specification of this application. The method of the present invention has been verified by simulation experiments and practical applications, both verifying the technical effects claimed by the present invention.

[0124] The algorithm (method) proposed by the present invention is the underlying technical core of the present invention. Based on the algorithm, various products can be derived, such as developing corresponding software systems.

[0125] It should be understood that various forms of the process shown above can be used, with steps reordered, added, or deleted. For example, the steps described in this application can be executed in parallel, sequentially, or in a different order, as long as the desired results of the technical solutions disclosed in this application can be achieved, all within the protection scope of the present invention.

Claims

1. A method for detecting blind signals based on adaptive thresholds based on power spectrum gradient, characterized in that: The steps include: Step 1, read the time domain sequence output by the receiver and after analog-to-digital conversion, and use the Welch algorithm to estimate the power spectrum of the time domain sequence; The segmented windowing method is used to divide a data segment of length N, r(n), n = 0, 1, ..., N-1, into L segments, each segment has M data, and the i-th segment data is expressed as: r i (n)=r(n+iM-M),0≤n≤M,1≤i≤L Then add the window function w(n) to each data segment and find the periodogram of each segment. The periodogram of segment a is: In the above formula, U is the normalization factor: Each periodogram is approximately regarded as unrelated to each other, and the power spectrum estimate G is obtained. xx (e jω )for: In the above formula, Step 2, calculate the gradient of the power spectrum obtained in step 1, use the gradient operator to get the gradient spectrum, and average the gradient spectrum data to get the mean gradient spectrum: Assume the initial value of the gradient S sub (0)=0, the gradient is calculated as follows: S sub (k)=G(k)-G(k-1) In the above formula, k is a natural number, G(k)=G xx (e jω ); Step 3, calculate the mean and standard deviation of the power spectrum gradient obtained in step 2, calculate the minimum upper threshold and the maximum lower threshold, calculate the adaptive window size according to the local change rate of the power spectrum gradient, determine the left and right boundaries of the window, extract the data in the window, calculate the local mean and standard deviation, calculate the adaptive k value, and obtain the adaptive upper and lower threshold array: The minimum upper threshold th1 and the maximum lower threshold th2 are: In the above formula, is the mean value of the input power spectrum gradient, V global is the standard deviation of the input power spectrum gradient; Use a sliding window to traverse the input power spectrum gradient, calculate the adaptive window size in each window, set the initial parameter value, MSize is the minimum window size, and BSize is the basic window size: In the above formula, ASize is the adaptive window size, is the local change rate of power spectrum gradient; For each window, set the initial parameter value, K base is the basic K value, α is the adjustment factor, extract the data in the window and calculate its mean and standard deviation V local , and then adjust the threshold according to the adaptive k value, setting the threshold to the mean plus k times the standard deviation: In the above formula, N local is the length of the local window. If the calculated threshold is lower than the minimum upper threshold th1, the minimum upper threshold is used; if it is higher than the maximum lower threshold th2, the maximum lower threshold is used, and finally the adaptive upper and lower threshold array is obtained: In the above formula, th1 A is the adaptive upper threshold, th2 A is the adaptive lower threshold; Step 4: Set the adaptive threshold, traverse the power spectrum gradient according to the threshold monitoring criteria, judge the signal and noise, and output the framed signal interval information: Compare the adaptive upper and lower threshold arrays obtained in step 3 with the power spectrum gradient obtained in step 2, find the data points where the power spectrum gradient exceeds the upper threshold and the data points where the power spectrum gradient is lower than the lower threshold, and define the first point exceeding the upper threshold as the starting point and the last point as the end point, and define the first point below the lower threshold as the starting point and the last point as the end point; The end point exceeding the upper threshold is taken as the interval starting point, and the start point exceeding the lower threshold is taken as the interval end point, and matching is performed, and the signal interval array of the final successful match is output to obtain the final result of signal detection; The frequency component corresponding to the signal starting interval index is taken as the starting frequency point of the signal, fK start_idx fK is the index of the detected signal start frequency point, end_idx is the index of the detected signal end frequency point, C is the center frequency, and B is the bandwidth: C=(fK start_idx +fK end_idx ) / 2 Get the signal interval from the start index to the end index, and define the frequency within the interval as fK in , the power spectrum value in the interval is G in , find the maximum value of the power spectrum G in the interval range max_in , minus 3db is the half power point value G half_in , find all the values ​​greater than or equal to G in the interval half_in The index of the point G idxs : G half_in =G max_in -3 G idxs =find(G in >=G half_in ) B=abs(fKin(max(Gidxs))-fKin(min(Gidxs))).

2. The method for detecting blind signals by adaptive threshold based on power spectrum gradient according to claim 1, characterized in that: The threshold monitoring criterion of step 4 is: judge the signal and noise from the power spectrum gradient diagram, and output the interval information of the signal; the logic is: only one value crosses the upper threshold, and the next value crosses the upper threshold, indicating that the current area is noise; only one value crosses the lower threshold, and the next value crosses the lower threshold, indicating that the current area is also noise; when the first value crosses the upper threshold and the next value crosses the lower threshold, it means that the current area is signal.

3. An adaptive threshold blind signal detection system based on power spectrum gradient, characterized in that: The system has a program module corresponding to the steps of any one of claims 1-2 above, and executes the steps in the adaptive threshold blind signal detection method based on power spectrum gradient when running.

4. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and the computer program is configured to implement the steps of a power spectrum gradient-based adaptive threshold blind signal detection method according to any one of claims 1 to 2 when called by a processor.

5. A power spectrum gradient-based adaptive threshold blind signal detection device, the device comprising at least one processor and a memory connected to the at least one processor, wherein: The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the above-mentioned adaptive threshold blind signal detection method based on power spectrum gradient, so as to realize the detection of blind signals under fluctuating and complex electromagnetic environments and noise interference.

Images