Dynamic threshold signal detection method based on morphological filtering and eigenvalue decomposition

Abstract

The application provides a dynamic threshold signal detection method based on morphological filtering and eigenvalue decomposition, which comprises the following steps: receiving a signal and sampling, calculating the original power spectrum of the signal according to the average periodogram method, sequentially performing minimum filtering and morphological filtering open operation on the original power spectrum, interpolating the original power spectrum to the length of the original power spectrum to obtain a noise floor estimation sequence, subtracting the noise floor estimation sequence from the original power spectrum to construct a Hankel matrix and performing eigenvalue decomposition to obtain each order component of the power spectrum after removing the noise floor, taking one of the order components as a reconstructed power spectrum, calculating a signal detection threshold according to an iterative approximation algorithm, performing spectrum monitoring according to the reconstructed power spectrum and the signal detection threshold, and judging whether the signal exists and analyzing the frequency and bandwidth of the signal. The application improves the accuracy in the case that the carrier frequency distance is too close, the noise floor fluctuates greatly, and the weak signal is weak.

Description

Technical Field

[0001] This invention relates to the field of signal detection, and more particularly to a dynamic threshold signal detection method based on morphological filtering and eigenvalue decomposition. Background Technology

[0002] Multi-signal detection technology under broadband reception conditions is widely used in civilian and military fields such as electromagnetic spectrum monitoring and electronic warfare. Research on related technologies has significant theoretical and practical value. A wide frequency band typically contains multiple signals, each with different carrier frequencies and bandwidths, and may even exhibit time-frequency aliasing. Due to the complex electromagnetic environment and impedance mismatch in the analog front-end of the broadband receiver, the noise floor of the broadband signal obtained by the broadband receiver has significant fluctuations.

[0003] Traditional fixed-threshold signal detection methods are prone to false alarms and missed detections when the carrier frequencies of the signals are too close together or when the noise floor fluctuates significantly. Typically, power spectrum smoothing is required before signal detection. Signals with closely spaced carrier frequencies are easily smoothed into a single signal, and weak signals are treated as noise, leading to detection errors. Based on this background and similar problems encountered in practical engineering, we aim to propose a method to improve signal detection performance. Summary of the Invention

[0004] The technical problem to be solved by this invention is: in view of the technical problems existing in the prior art, this invention provides a dynamic threshold signal detection method based on morphological filtering and eigenvalue decomposition, which solves the problem of signal detection difficulties caused by in-band unevenness, weak signals, and too small signal carrier frequency intervals.

[0005] To solve the above-mentioned technical problems, the technical solution proposed by this invention is as follows:

[0006] A dynamic threshold signal detection method based on morphological filtering and eigenvalue decomposition includes the following steps:

[0007] S1) Receive the signal and sample it. Calculate the original power spectrum of the signal from the sampled sequence of the received signal using the average periodogram method.

[0008] S2) Divide the original power spectrum sequence into uniform segments, extract the minimum value of each segment, and use the extracted sequence as the minimum value to filter the extracted power spectrum.

[0009] S3) Perform morphological filtering on the power spectrum after minimum value filtering to obtain the morphologically filtered power spectrum.

[0010] S4) Interpolate the power spectrum after morphological filtering to the length of the original power spectrum to obtain the noise basis estimation sequence;

[0011] S5) Subtract the noise basis estimation sequence from the original power spectrum to obtain the power spectrum after noise basis removal;

[0012] S6) Construct a Hankel matrix based on the power spectrum after the noise reduction basis, perform eigenvalue decomposition on the Hankel matrix to obtain each order component of the power spectrum after the noise reduction basis, and take the first order component as the reconstructed power spectrum.

[0013] S7) The signal detection threshold is calculated by the reconstructed power spectrum using an iterative approximation algorithm;

[0014] S8) Perform spectrum monitoring based on the reconstructed power spectrum and signal detection threshold to determine the presence of the signal and analyze its frequency and bandwidth.

[0015] Furthermore, in step S3), the structuring element for enabling morphological filtering is an element in the power spectrum sequence that is less than a preset threshold.

[0016] Furthermore, the iterative approximation algorithm in step S7) specifically includes the following steps:

[0017] S71) Obtain the power spectrum and threshold reference value;

[0018] S72) Calculate the maximum and minimum values ​​of the power spectrum, and initialize the count of data points near the threshold and the threshold value;

[0019] S73) Calculate the number of data points within the first preset value range of the maximum power spectrum;

[0020] S74) If the number of data points is greater than the data point count, the number of data points is used as the new data point count, and the maximum power spectrum value is used as the new threshold value.

[0021] S75) The maximum power spectrum value is decremented by the second preset value, and the calculation result is used as the new maximum power spectrum value;

[0022] S76) Return to step S73) until the maximum power spectrum value is less than the minimum power spectrum value;

[0023] S77) The threshold value and the threshold reference value are added together and then used as the signal detection threshold output.

[0024] Furthermore, the first preset value is 0.5, and the second preset value is 0.1.

[0025] Furthermore, in step S1), 262,144 signal sample points are sampled, and the power spectrum of the signal is estimated by the average periodogram method with 65,536 points of single frame and 50% overlap.

[0026] Furthermore, in step S2), the original power spectrum sequence is evenly divided into 1024 segments.

[0027] Furthermore, the Hankel matrix mentioned in step S6) is a 100*262045 order matrix.

[0028] The present invention also proposes a dynamic threshold signal detection device based on morphological filtering and eigenvalue decomposition, including a computer device, the computer device being programmed or configured to execute the dynamic threshold signal detection method based on morphological filtering and eigenvalue decomposition.

[0029] The present invention also proposes a computer-readable storage medium storing a computer program programmed or configured to perform the described dynamic threshold signal detection method based on morphological filtering and eigenvalue decomposition.

[0030] Compared with the prior art, the advantages of the present invention are as follows:

[0031] The method of this invention performs minimum extraction and morphological filtering on the power spectrum of the signal to remove the noise floor. Based on the removal of the noise floor, the power spectrum is decomposed into eigenvalues, and then the power spectrum is reconstructed. The signal detection threshold is estimated by using an iterative approximation method, thereby improving the signal detection accuracy in cases where the carrier frequency distance is too close, the noise floor fluctuation is large, or the signal is weak. Attached Figure Description

[0032] Figure 1 This is a schematic diagram of a specific process block of an embodiment of the present invention.

[0033] Figure 2 This is a block diagram illustrating the noise floor estimation process according to an embodiment of the present invention. Detailed Implementation

[0034] The present invention will be further described below with reference to the accompanying drawings and specific preferred embodiments, but this does not limit the scope of protection of the present invention.

[0035] like Figure 1 As shown, this invention proposes a dynamic threshold signal detection method based on morphological filtering and eigenvalue decomposition, comprising the following steps:

[0036] S1) Receive the signal and sample it. Calculate the original power spectrum of the signal from the sampled sequence of the received signal using the average periodogram method.

[0037] S2) Divide the original power spectrum sequence into uniform segments, extract the minimum value of each segment, and use the extracted sequence as the minimum value to filter the extracted power spectrum.

[0038] S3) Perform morphological filtering on the power spectrum after minimum value filtering to obtain the morphologically filtered power spectrum.

[0039] S4) Interpolate the power spectrum after morphological filtering to the length of the original power spectrum to obtain the noise basis estimation sequence;

[0040] S5) Subtract the noise basis estimation sequence from the original power spectrum to obtain the power spectrum after noise basis removal;

[0041] S6) Construct a Hankel matrix based on the power spectrum after noise reduction, perform eigenvalue decomposition on the Hankel matrix to obtain each order component of the power spectrum after noise reduction, and use the first order component as the reconstructed power spectrum.

[0042] S7) The signal detection threshold is calculated by the reconstructed power spectrum using an iterative approximation algorithm;

[0043] S8) Perform spectrum monitoring based on the reconstructed power spectrum and signal detection threshold to determine the presence of the signal and analyze its frequency and bandwidth.

[0044] In this embodiment, morphological filtering is used in steps S3) to S5) to estimate the noise floor and flatten it. Morphological operations are a nonlinear transformation theory based on shape, which has the advantage of changing the local characteristics of a signal. Let signal f be a discrete function defined on F = {0,1,…,N}, and structuring element g be a function defined on G = {0,1,…,M} where N>M. Then the four basic morphological operations can be defined as follows:

[0045] The expansion operation is

[0046] Erosion operation is

[0047] Enable operation as

[0048] Closure operation is

[0049] Dilation reduces signal valleys and expands peaks, while erosion reduces signal peaks and widens valleys. Opening and closing operations are combinations of dilation and erosion; opening eliminates signal peaks while closing suppresses signal valleys.

[0050] While morphological filtering directly on the original power spectrum can estimate the noise floor, it requires a large structuring element when the signal bandwidth is wide, resulting in significant computational overhead. Furthermore, a larger structuring element makes it easier to eliminate the peaks and valleys at the bottom of the noise spectrum. To reduce computation time and improve estimation accuracy, such as... Figure 1 and Figure 2 As shown in Figure 2 , in this embodiment, before performing morphological filtering in step S2), minimum filtering extraction is also performed on the original power spectrum.

[0051] In step S3) of this embodiment, the structuring element for the opening operation of morphological filtering is the element in the power spectrum sequence that is less than the preset threshold. That is, in this embodiment, a smaller structuring element is selected for the power spectrum sequence to perform the morphological opening operation to improve the estimation accuracy.

[0052] In this embodiment, the power spectrum is reconstructed through step S6), and the specific implementation process is as follows:

[0053] For the power spectrum X = [X(1), X(2),..., X(N)], the Hankel matrix D can be constructed as follows:

[0054]

[0055] where 1 < n < N, let m = N - n + 1, then D ∈ R i , i ,

[0061] , i , i , i , i , T , i , i , R m*n is the set of Hankel matrices that can be constructed from the signal power spectrum.

[0056] Perform singular value decomposition on D:

[0057]

[0058] In the above formula, u i ∈ R m*1 (u i is a subset of the column vectors of matrix D, R m*1 ), v i ∈ R n*1 (v i is a subset of the row vectors of matrix D, R n*1 ), i = 1, 2,..., r, r = min(m, n).

[0059] Let (σ i is the i-th singular value), that is, there is:

[0060] <无限符号>If we let O i = [X i (1), X i (2),..., X i (n)], E i = [X i (n + 1), X i (n + 2),..., X i (N)] TThus, the i-th order component of the component singular values ​​can be constructed:

[0062]

[0063] For the Hankel matrix D, let:

[0064] S=[X(1),X(2),…,X(N)] (9)

[0065] Y = [X(n+1),X(n+2),…,X(n)] T (10)

[0066] According to equation (6), we can obtain:

[0067] S = O1 + O2 + ... + O R (11)

[0068]

[0069] From the Hankel matrix construction process, we know that the power spectrum X = [S,Y] T Combining equations (8), (11), and (12), we can obtain:

[0070] X = P1 + P2 + ... + P r (13)

[0071] The above formula is the decomposition formula of the power spectrum X after singular value decomposition using the Hankel matrix. We observe that the first-order component P1 is the main component of the power spectrum X, which can well reflect the envelope of the power spectrum and can also distinguish weak signals and adjacent signals well. Therefore, the first-order component P1 is used as the power spectrum reconstructed after decomposition.

[0072] After preprocessing in steps S3) to S6), when there is no signal in the frequency band, the noise floor is relatively flat, and the energy in the signal band is significantly higher than that in the noise region. A reasonable single threshold can be used to detect the signal and roughly estimate its center frequency and bandwidth for subsequent down-conversion and other signal processing. When the power spectrum of a signal changes, a fixed threshold is often insufficient to correctly distinguish between signal and noise. We observed that the power spectrum has the highest number of energy points in the energy range at the signal-noise boundary. Based on this, step S7) proposes an iterative approximation algorithm to estimate the noise floor and signal boundary of the power spectrum to achieve threshold adaptation. The iterative approximation algorithm specifically includes the following steps:

[0073] S71) Obtain the power spectrum and threshold reference value;

[0074] S72) Calculate the maximum and minimum values ​​of the power spectrum, and initialize the count of data points near the threshold and the threshold value;

[0075] S73) Calculate the number of data points within the first preset value range of the maximum power spectrum. In this embodiment, the first preset value is 0.5.

[0076] S74) If the number of data points is greater than the data point count, the number of data points is used as the new data point count, and the maximum power spectrum value is used as the new threshold value.

[0077] S75) The maximum power spectrum value is decremented by the second preset value, and the calculation result is used as the new maximum power spectrum value. In this embodiment, the second preset value is 0.1.

[0078] S76) Return to step S73) until the maximum power spectrum value is less than the minimum power spectrum value;

[0079] S77) The threshold value and the threshold reference value are added together and then used as the signal detection threshold output.

[0080] The specific algorithm flow is as follows:

[0081]

[0082]

[0083] The following is combined with Figure 1 Further explanation of the preferred embodiments:

[0084] Step 1, execute step S1), receive the signal and sample to obtain 262,144 signal sample points, perform power spectrum estimation of the 262,144 signal sample points using the average periodogram method with 65,536 points of single frame and 50% overlap to obtain the original power spectrum of the signal. Using the above parameters for calculation can obtain high calculation accuracy with less computation.

[0085] Step 2, execute step S2), divide the original power spectrum sequence into 1024 segments evenly, extract the minimum value of each segment, and use the extracted sequence as the minimum value to filter the extracted power spectrum.

[0086] Step 3, execute step S3), perform morphological filtering on the power spectrum after minimum value filtering extraction to obtain the morphologically filtered power spectrum;

[0087] Step 4, execute step S4), interpolate the power spectrum after morphological filtering to the length of the original power spectrum to obtain the noise basis estimation sequence;

[0088] Step 5, execute step S5), subtract the noise basis estimation sequence from the original power spectrum to obtain the power spectrum after noise basis removal;

[0089] Step 6, execute step S6), construct a 100*262045 order Hankel matrix based on the power spectrum after noise reduction, perform eigenvalue decomposition on the Hankel matrix to obtain each order component of the power spectrum after noise reduction, and take the first order component as the reconstructed power spectrum.

[0090] Step 7, execute step S7), calculate the signal detection threshold based on the reconstructed power spectrum using an iterative approximation algorithm;

[0091] Step 8, execute step S8), perform spectrum monitoring based on the reconstructed power spectrum and signal detection threshold, determine the presence of the signal, and analyze the signal's frequency and bandwidth.

[0092] The invention also proposes a dynamic threshold signal detection device based on morphological filtering and eigenvalue decomposition, including a computer device programmed or configured to perform the dynamic threshold signal detection method based on morphological filtering and eigenvalue decomposition.

[0093] The present invention also proposes a computer-readable storage medium storing a computer program programmed or configured to perform the described dynamic threshold signal detection method based on morphological filtering and eigenvalue decomposition.

[0094] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the invention. Therefore, any simple modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention should fall within the protection scope of the present invention.

Claims

1. A dynamic threshold signal detection method based on morphological filtering and eigenvalue decomposition, characterized in that, Includes the following steps: S1) Receive the signal and sample it. Calculate the original power spectrum of the signal from the sampled sequence of the received signal using the average periodogram method. S2) Divide the original power spectrum sequence into uniform segments, extract the minimum value of each segment, and use the extracted sequence as the minimum value to filter the extracted power spectrum. S3) Perform morphological filtering on the power spectrum after minimum value filtering to obtain the morphologically filtered power spectrum. S4) Interpolate the power spectrum after morphological filtering to the length of the original power spectrum to obtain the noise basis estimation sequence; S5) Subtract the noise basis estimation sequence from the original power spectrum to obtain the power spectrum after noise basis removal; S6) Construct a Hankel matrix based on the power spectrum after the noise reduction basis, perform eigenvalue decomposition on the Hankel matrix to obtain each order component of the power spectrum after the noise reduction basis, and take the first order component as the reconstructed power spectrum. S7) The signal detection threshold is calculated from the reconstructed power spectrum using an iterative approximation algorithm. The iterative approximation algorithm specifically includes the following steps: S71) Obtain the power spectrum and threshold reference value; S72) Calculate the maximum and minimum values ​​of the power spectrum, and initialize the count of data points near the threshold and the threshold value; S73) Calculate the number of data points within the first preset value range of the maximum power spectrum; S74) If the number of data points is greater than the data point count, the number of data points is used as the new data point count, and the maximum power spectrum value is used as the new threshold value; S75) The maximum power spectrum value is decremented by the second preset value, and the calculation result is used as the new maximum power spectrum value; S76) Return to step S73) until the maximum power spectrum value is less than the minimum power spectrum value; S77) The threshold value and the threshold reference value are added together and then used as the signal detection threshold output; S8) Perform spectrum monitoring based on the reconstructed power spectrum and signal detection threshold to determine the presence of the signal and analyze its frequency and bandwidth.

2. The dynamic threshold signal detection method based on morphological filtering and eigenvalue decomposition according to claim 1, characterized in that, In step S3), the structuring element for enabling morphological filtering is an element in the power spectrum sequence that is less than a preset threshold.

3. The dynamic threshold signal detection method based on morphological filtering and eigenvalue decomposition according to claim 1, characterized in that, The first preset value is 0.5, and the second preset value is 0.

1.

4. The dynamic threshold signal detection method based on morphological filtering and eigenvalue decomposition according to claim 1, characterized in that, In step S1), 262,144 signal sample points are obtained by sampling. The power spectrum of the signal is estimated by the average periodogram method with 65,536 points of single frame and 50% overlap.

5. The dynamic threshold signal detection method based on morphological filtering and eigenvalue decomposition according to claim 4, characterized in that, In step S2), the original power spectrum sequence is evenly divided into 1024 segments.

6. The dynamic threshold signal detection method based on morphological filtering and eigenvalue decomposition according to claim 5, characterized in that, The Hankel matrix mentioned in step S6) is a 100*262045 order matrix.

7. A dynamic threshold signal detection device based on morphological filtering and eigenvalue decomposition, characterized in that, Includes a computer device, which is programmed or configured to perform the dynamic threshold signal detection method based on morphological filtering and eigenvalue decomposition as described in any one of claims 1 to 6.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that is programmed or configured to perform the dynamic threshold signal detection method based on morphological filtering and eigenvalue decomposition as described in any one of claims 1 to 6.

Images