An ionospheric oblique back ionogram interference suppression method based on eigen decomposition
By suppressing co-channel interference in ionospheric oblique return detection using the eigenvalue decomposition method, the problem of high hardware and software requirements was solved, achieving efficient ionospheric oblique return detection and improving signal quality and detection efficiency.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- WUHAN UNIV
- Filing Date
- 2022-08-09
- Publication Date
- 2026-07-31
AI Technical Summary
Existing technologies are unable to effectively suppress co-frequency interference in ionospheric oblique return detection, leading to difficulties in interpreting oblique return ionization maps. Furthermore, they have high hardware and software requirements, making them difficult to apply to routine engineering detection.
The feature decomposition method is adopted. The narrow-bandwidth coded modulation signal is transmitted by the oblique return detector. The feature decomposition is performed by the strong correlation of the distance dimension of the echo signal to divide the interference subspace and the signal-noise subspace, thereby suppressing co-frequency interference.
It achieves improved quality of oblique return ionization maps without increasing hardware costs, provides a continuous and clear leading edge, simplifies hardware implementation, shortens detection time, and improves signal-to-noise ratio.
Smart Images
Figure CN116184321B_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of ionospheric detection technology, and particularly relates to a method for suppressing interference in ionospheric oblique return ionographs based on feature decomposition. Background Technology
[0002] The ionosphere reflects shortwave radio waves. When radio waves emitted by a transmitter are incident obliquely into the ionosphere, they are reflected and reach the distant Earth's surface. A portion of the radio wave energy returns along the original incident path, is reflected again by the ionosphere, and then returns to the receiver at the transmitting point. Ionospheric oblique return detectors utilize this principle. By scanning frequencies, an oblique return ionosphere map showing energy, distance, and frequency can be obtained. Intelligent interpretation and inversion of this oblique return ionosphere map has significant engineering application value for shortwave frequency management systems and skywave over-the-horizon radar coordinate registration.
[0003] Because it has undergone at least two ionospheric reflections and one ground scattering, the oblique return signal energy is very weak. At the same time, since the detection system operates in the shortwave band, which has a high utilization rate and a large number of users, the receiver is subjected to a lot of co-channel interference, making the interpretation of the oblique return ionogram extremely difficult.
[0004] Currently, methods for suppressing co-channel interference in oblique return ionograms mainly rely on the image domain. Thresholds are set according to certain criteria, and data points below the threshold are cleared. Data reconstruction is then performed by interpolating adjacent missing data. However, this method is highly dependent on experience; if the threshold is not set appropriately, it can lead to false signal cancellation or unidentified interference. Furthermore, the reconstructed signal relies solely on information from neighboring frequency points, without utilizing the frequency point's own information, inevitably resulting in errors compared to the true signal. Alternatively, adaptive spatial filtering using an array antenna is also a feasible method, but this method has high hardware and software requirements due to the long shortwave wavelength and the need for large-area antenna arrays and multi-channel receivers.
[0005] In existing technologies, the "Slow Weak Target Detection Method Based on External Radar Source" is a spatial algorithm. To ensure anti-interference effectiveness, it requires a large antenna array as the algorithm's foundation, placing very high demands on hardware. Furthermore, the criteria for dividing the feature subspace are unclear, making it difficult to apply to routine engineering detection. The "High-Resolution DBS Imaging Method Based on SPICE for Airborne Radar" is a parametric method that models ground clutter signals. This method is often unrobust, and its goal is to improve the resolution of DBS imaging, not to suppress interference. The "High-Resolution Vertical Detection Method for the Ionospheric Es Layer Based on Cross-Spectrum Analysis" utilizes the relationship between the phase of adjacent frequencies in the cross-spectrum and the echo distance, employing the Capon algorithm for spectral estimation to improve ionospheric imaging resolution. Again, its goal is to improve resolution rather than suppress interference. Secondly, interference affects the spectral phase of the cross-spectrum. High-resolution imaging based on cross-spectrum analysis of oblique return scattering detection signals with interference will result in chaotic noise in the oblique return scattering ionosphere, contradicting the initial goal of interference suppression. Summary of the Invention
[0006] The purpose of this invention is to provide an interference suppression method for ionospheric oblique return ionographs when using an ionospheric oblique return detector to perform ionospheric oblique return detection. This method utilizes the strong correlation of narrowband radio frequency interference in the distance dimension and employs a feature decomposition method to divide the interference subspace and signal-noise subspace for single or multiple detections at each detection frequency, thereby realizing interference suppression based on feature decomposition.
[0007] To achieve the above objectives, the technical solution adopted in this invention is an ionospheric oblique return ionograph interference suppression method based on eigenvalue decomposition. Using an oblique return detector, a narrow-bandwidth continuous shortwave probe signal modulated by M-sequence encoding is transmitted via frequency sweeping or fixed-frequency observation. After determining the frequency point to be processed using the mode, the invalid portion of the echo is truncated. Then, utilizing the strong correlation of co-frequency interference in the echo signal along the range dimension, eigenvalue decomposition is performed on the range-gate echo covariance matrix of the invalid echo signal. The number of interference sources is estimated from the eigenvalue sequence, and the signal-noise subspace and interference subspace are divided. The range-gate signal with oblique return echo is projected onto the signal-noise subspace, suppressing co-frequency interference in the echo, without being limited by the number of repeated probes or carrier waves.
[0008] Furthermore, the process of suppressing interference from the ionospheric oblique return ionization map is as follows:
[0009] Step S1: Set the frequency sweep range and the number of frequency point detections, and determine the signal encoding length and duty cycle;
[0010] Step S2: Configure the oblique return detector to detect the oblique return signal according to the coding modulation method and transmission parameters determined in step S1, and obtain the echo signal data;
[0011] Step S3: Calculate the cyclic cross-correlation signal between the echo data of a single carrier frequency and the M-sequence encoded and modulated by the set detection signal;
[0012] Step S4: Use the mode to determine the frequency point to be processed;
[0013] Step S5: Extract invalid echo data of the carrier frequency determined in step S4;
[0014] Step S6: After cross-correlation of the invalid echo data obtained in step S5, the signal is divided into several sub-bands that partially overlap each other and have equal spacing, and a time series matrix is constructed. The covariance matrix is obtained by calculating the covariance of the matrix.
[0015] Step S7: Utilizing the strong correlation of strong co-frequency interference in the distance dimension and its dominance in the region of invalid echo data, the feature decomposition method is used to obtain the interference subspace and the signal-noise subspace for each frequency point to be processed. The valid echo data is then projected onto the signal-noise subspace, thereby achieving interference suppression of the oblique return ionization map.
[0016] Step S8: Perform coherent accumulation and constant false alarm rate processing on the signal obtained in step S7 after suppressing radio frequency interference to obtain a slanted return ionization map with continuous leading edge.
[0017] Step S9: Perform energy comparison stretching on the ionospheric oblique return signal after step S8, which is based on constant false alarm rate.
[0018] Moreover, the maximum detection range determined by the length of the M-sequence used to encode and modulate the signal in step S1 and the transmission duty cycle is greater than 3000 km.
[0019] Furthermore, if the code order of the M sequence in step S3 is n, then it is represented as follows:
[0020]
[0021] in, These are the element values in sequence M;
[0022] The sequence is represented by p cyclic shifts.
[0023]
[0024] The relevant sequence matrix S is constructed as follows:
[0025]
[0026] If the total number of detections for the single-carrier frequency echo in step S3 is m, then the sequence of the k-th detection is represented as follows:
[0027]
[0028] in, These are the 1st, ..., 2nd pulses before pulse compression in the kth detection. n -1 range gate echo data, k = 1, 2, ..., m;
[0029] The single-carrier frequency echo matrix in step S3 is expressed as follows:
[0030] W = [w1, w2, ..., w m ]
[0031] The M-sequence cyclic cross-correlation signal in step S3 is represented as follows:
[0032] X = W H S
[0033] The superscript H indicates the conjugate transpose of the matrix.
[0034] Furthermore, step S4, which uses the mode to determine the frequency point to be processed, is implemented as follows:
[0035] Construct a vector of length q consisting entirely of 1s. as follows,
[0036]
[0037] The echo signal obtained by cyclic cross-correlation at each frequency point is convolved with a vector of all 1s as follows:
[0038]
[0039] The moving average of the q range gates of the echo after cyclic cross-correlation at a single frequency point is given. Sort the elements and take the minimum value as the noise estimate of the echo. as follows,
[0040]
[0041] Taking the logarithm of the noise estimate and rounding it down, as follows:
[0042]
[0043] The noise estimation sequence for the entire frequency range is constructed as follows:
[0044]
[0045] in, Represents the noise estimates at the 1st, 2nd, ..., gth frequency points. The frequency points to be processed are shown below.
[0046]
[0047] In the above formula, Mo represents the mode, |X(f)| is the amplitude of the signal at frequency point f after cyclic cross-correlation, ∩ represents the intersection, max() represents taking the maximum value of the sequence, min() represents taking the minimum value of the sequence, and d is an empirical value.
[0048] Furthermore, the range of invalid echo data in step S5 is the continuous signal sequence that is overwhelmed by interference and noise in the original oblique return ionization diagram. If all frequency points are invalid data starting from the Kth range gate, and there are a total of M range gates in a single detection cycle, then the invalid echo data matrix is represented as follows.
[0049] X V =[x K x K+1 , ..., x M ]
[0050] Where, x K x K+1 , ..., x M The sequence of columns K, K+1, ..., M of the cyclic cross-correlation signal matrix X obtained in step S3. Furthermore, when dividing the single-carrier frequency echo invalid signal matrix into subbands in step S6, the relationships between the time interval of each subband, the length of each subband sequence, and the covariance matrix estimation are as follows.
[0051] For the Q-th probe in the invalid data matrix of a single-carrier echo, the echo sequence is represented as follows.
[0052] X VQ =[x K,Q x K+1,Q , ..., x M,Q ]
[0053] Where, x K,Q x K+1，Q , ..., x M,Q It is the Q-th row element in the sequence of columns K, K+1, ..., M of the cyclic cross-correlation signal matrix X obtained in step S3, where the superscript T represents matrix transpose;
[0054] Then the sub-bands are divided and reconstructed into a time series matrix T. Q as follows,
[0055]
[0056] Where N is the number of sub-bands, N≤MNK, which can be adjusted according to the signal-to-noise ratio and actual effect;
[0057] Calculate the covariance matrix R XQ as follows
[0058]
[0059] If multiple probes are performed, the covariance matrix of the multiple probes needs to be smoothed to obtain R. X as follows,
[0060]
[0061] Furthermore, step S7 uses eigenvalue decomposition to obtain the interference subspace and signal-noise subspace for each frequency point, and projects the valid echo data onto the signal-noise subspace. The method for achieving interference suppression of the oblique return ionization map is as follows:
[0062] First, the eigenvalues are decomposed using the covariance matrix of the invalid echo data, as follows.
[0063] R x =UAU H
[0064] Where U represents the eigenvector matrix, Λ represents the eigenvalue matrix, and the superscript H represents the matrix conjugate transpose;
[0065] The eigenvector matrix is divided as follows:
[0066] U = [G, V]
[0067] Where G represents the interference subspace and V represents the signal-noise subspace.
[0068] G = span{u1, u2, ..., u} r}
[0069] V = span{u r+1 u r+2 ,…,u N}
[0070] Where u1, u2, ..., u N represents the 1st, 2nd, ..., Nth column vectors of the eigenvector matrix U, span{} represents the vector space spanned by the vectors, and r is the subspace partitioning boundary;
[0071] After determining the interference subspace, the valid echo data is projected onto the signal-noise subspace.
[0072] Furthermore, the coherent accumulation in step S8 is as follows:
[0073] If multiple probes are performed, then... The Fourier transform of the time-domain data at each frequency point is as follows.
[0074]
[0075] The coherent accumulation is as follows:
[0076]
[0077] in, for The model.
[0078] Furthermore, the energy comparison stretching in step S9 is as follows.
[0079]
[0080] in, This is the single-frequency echo processing result of the present invention, where n is the energy contrast stretching factor, and U... T For constant false alarm threshold, This is the estimated value of the effective data echo.
[0081] In the aforementioned ionospheric oblique return ionograph interference suppression method based on eigenvalue decomposition, the suppression effect can be adjusted by compromising the row and column dimensions of the time matrix by adjusting the number of sub-band divisions. This is because, with a fixed sampling data length, adjusting the number of sub-band divisions actually adjusts the eigenvector dimension and the accuracy of the covariance matrix, which can adjust the imaging length and the interference suppression effect.
[0082] The beneficial effects of this invention are:
[0083] 1. This invention is based on the original ionospheric oblique return probe echo data, which is processed by an algorithm to suppress co-channel interference. It does not require changes to the oblique return probe's detection timing, process, or original signal processing method, and has strong versatility.
[0084] 2. This invention utilizes the strong correlation of co-frequency interference in the range dimension and performs eigenvalue decomposition on the covariance matrix formed in the range dimension. It adaptively divides the subspace by taking advantage of the low signal-to-noise ratio of oblique return scattering signals (superior to the existing technology "Slow Weak Target Detection Method Based on External Radar Source Radar"). The range gate signal with oblique return echo is projected to the signal-noise subspace, thereby effectively suppressing co-frequency interference in the echo (superior to the existing technologies "High-Resolution DBS Imaging Method for Airborne Radar Based on SPICE" and "A High-Resolution Vertical Detection Method for the Es Layer of the Ionosphere Based on Cross-Spectrum Analysis"). This method effectively utilizes the echo information of the interfered frequency points and has a higher confidence level than interpolation.
[0085] 3. The implementation of this invention can be regarded as a time-domain algorithm, which is simpler to implement in hardware than the spatial domain algorithm (superior to the existing technology "Slow Weak Target Detection Method Based on External Radar Source Radar").
[0086] 4. This invention is not limited by the number of repeated detections or the number of carriers, and can effectively improve the detection time resolution, that is, achieve better detection results with fewer detections, thereby shortening the detection time.
[0087] 5. This invention solves the problem of large differences in signal energy at different frequency points in oblique return scattering frequency sweep detection. After processing, an oblique return scattering ionization map with uniform background and small energy difference can be obtained. Attached Figure Description
[0088] Figure 1 This is a schematic diagram illustrating the principle of a method according to an embodiment of the present invention;
[0089] Figure 2 This is a step diagram illustrating the suppression of ionospheric oblique return ionograph interference based on eigenvalue decomposition according to an embodiment of the present invention;
[0090] Figure 3 This is an embodiment of the present invention showing the ionospheric oblique return single-frequency point distance-Doppler ionization map based on eigenvalue decomposition for interference suppression.
[0091] Figure 4 This is an embodiment of the present invention showing the ionospheric oblique return single-frequency point distance-time ionization map based on eigenvalue decomposition for interference suppression.
[0092] Figure 5 This is a distance-amplitude distribution diagram of the original coherent accumulation at a single frequency point of ionospheric oblique return and the coherent accumulation after interference suppression, based on eigenvalue decomposition and interference suppression, according to an embodiment of the present invention.
[0093] Figure 6 This is an example of an ionospheric oblique return frequency sweep ionization map based on eigenvalue decomposition for interference suppression, according to an embodiment of the present invention. Detailed Implementation
[0094] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings.
[0095] This embodiment provides an ionospheric oblique return ionization map interference suppression method based on feature decomposition. Utilizing conventional oblique return detection equipment, it helps to improve the quality of oblique return ionization maps and provide a more continuous and clear front without increasing additional hardware costs.
[0096] The basic principle of this embodiment is as follows: Figure 1As shown, a shortwave probe signal with a narrow bandwidth, modulated by an M-sequence (longest linear feedback shift register sequence), is transmitted obliquely upwards to the ionosphere using a frequency sweeping observation method. After determining the frequency point to be processed using the mode, the invalid part of the echo is truncated. Then, by utilizing the strong correlation of co-frequency interference in the echo signal in the range dimension, the range gate echo covariance matrix of the invalid echo signal is eigenvalued. The number of interference sources is estimated from the eigenvalue sequence, and the signal-noise subspace and interference subspace are divided. The range gate signal with oblique return echo is projected onto the signal-noise subspace to suppress co-frequency interference in the echo. After constant false alarm rate and energy stretching, an oblique return ionosphere map with a continuous leading edge is obtained, which is not limited by the number of repeated probes or the number of carriers.
[0097] like Figure 2 As shown in the figure, an embodiment of the present invention provides a method for suppressing interference in ionospheric oblique return ionization maps based on eigenvalue decomposition, which includes the following steps:
[0098] Step S1: Set the frequency sweep range and the number of frequency point detections, and determine the signal encoding length and duty cycle;
[0099] It should be noted that the maximum detection range, determined by the length of the M-sequence (longest linear feedback shift register sequence) used to encode and modulate the signal and the transmission duty cycle, is greater than 3000 km. This is because the height of the ionosphere is often calculated, and only when the distance is greater than 3000 km can there be enough interference sampling units to meet the normal detection requirements of the algorithm.
[0100] Step S2: Configure the oblique return detector to detect the oblique return signal according to the coding modulation method and transmission parameters determined in step S1, and obtain the echo signal data;
[0101] Step S3: Calculate the cyclic cross-correlation signal between the echo data of a single carrier frequency and the M-sequence (longest linear feedback shift register sequence) encoded and modulated by the set detection signal.
[0102] It should be noted that if the code order of sequence M is n, it is represented as a column vector S:
[0103]
[0104] in, These are the element values in the M sequence. Once the code order n is determined, the length of the M sequence can be determined to be 2. n -1, in ionospheric oblique backscattering detection, n is usually taken as 9 or 10 to simultaneously satisfy the detection distance and time resolution, and the superscript T represents matrix transpose.
[0105] The sequence is cyclically shifted p times and represented as a column vector S. p :
[0106]
[0107] The superscript T represents matrix transpose.
[0108] The relevant sequence matrix S is constructed as shown in equation (3):
[0109]
[0110] Where S is cyclically shifted by 0, 1, ..., 2 respectively. n -2-order column vector The combination yields the desired result.
[0111] If the total number of detections for a single carrier frequency echo is m, then the sequence of the kth detection is expressed as equation (4):
[0112]
[0113] in, These are the 1st, ..., 2nd pulses before pulse compression in the kth detection. n -1 distance gate echo data, k = 1, 2, ..., m.
[0114] The single-carrier echo matrix is then expressed as equation (5):
[0115] W = [w1, w2, ..., w m Equation (5)
[0116] The M-sequence cyclic cross-correlation signal is expressed as equation (6):
[0117] X = W H S-type (6)
[0118] Where the superscript H denotes the matrix conjugate transpose. X is m×(2 n The signal matrix after cyclic cross-correlation of -1)
[0119] Step S4: Use the mode to determine the frequency point to be processed;
[0120] It should be noted that the method for determining the frequency point to be processed using the mode is as follows:
[0121] Construct a vector of length q consisting entirely of 1s. As in equation (7):
[0122]
[0123] Where q is the noise sampling distance gate length, which is usually taken as 20 to 30.
[0124] The echo signals obtained by cyclic cross-correlation at each frequency point as described in step S3 are convolved with an all-1 vector as shown in equation (8):
[0125]
[0126] The moving average of the q range gates of the echo after cyclic cross-correlation at a single frequency point is given. Sort the elements and take the minimum value as the noise estimate of the echo. As in equation (9):
[0127]
[0128] Take the logarithm of the noise estimate and round it down as shown in equation (10):
[0129]
[0130] The noise estimation sequence for the entire frequency range is then constructed as shown in equation (11):
[0131]
[0132] in, This represents the noise estimate at the 1st, 2nd, ..., gth frequency points.
[0133] The frequency point sequence f to be processed p As shown in equation (12):
[0134]
[0135] In the above formula, Mo represents the mode, |X(f)| is the amplitude of the signal at frequency point f after cyclic cross-correlation, ∩ represents the intersection, max() represents taking the maximum value of the sequence, min() represents taking the minimum value of the sequence, and d is an empirical value, which can generally be taken as 20 to 30. This formula compares the noise estimate value of each frequency point in the frequency sweep detection with the mode. If the noise estimate value is greater than the mode and the maximum amplitude value of the pulse after compression at that frequency point is less than d times the minimum amplitude value, then it is determined to be a frequency point to be processed.
[0136] Step S5: Extract invalid echo data of the carrier frequency determined in step S4;
[0137] It should be noted that the range of invalid echo data is the continuous signal sequence that is submerged by interference and noise in the original oblique return ionization diagram. It is usually taken as the range gate sequence corresponding to the group distance of 1500km to the detection distance. If all frequency points are invalid data starting from the Kth range gate, and there are a total of M range gates in a single detection cycle, then the invalid echo data matrix is expressed as Equation (13):
[0138] X V =[x K x K+1 , ..., xM Equation (13)
[0139] Where, x K x K+1 , ..., x M The K, K+1, ..., M-th column sequence of the cyclic cross-correlation signal matrix X obtained in step S3.
[0140] Step S6: After cross-correlation of the invalid echo data obtained in step S5, the signal is divided into several sub-bands that partially overlap each other and have equal spacing, and a time series matrix is constructed. The covariance matrix is obtained by calculating the covariance of the matrix.
[0141] It should be noted that when dividing a single-carrier frequency echo invalid signal matrix into subbands, the relationships that the time intervals between subbands and the lengths of each subband sequence should possess, as well as the method for estimating the covariance matrix, are as follows:
[0142] For the Qth probe in the invalid data matrix of a single-carrier echo, the echo sequence is expressed as Equation (14):
[0143] X VQ =[x K,Q x K+1，Q , ..., x M,Q Equation (14)
[0144] Where, x K,Q x K+1，Q , ..., x M,Q It is the Q-th row element in the sequence of columns K, K+1, ..., M of the cyclic cross-correlation signal matrix X obtained in step S3, where the superscript T represents matrix transpose.
[0145] Then the sub-bands are divided and reconstructed into a time series matrix T. Q As in equation (15):
[0146]
[0147] Among them, T Q This is a Toeplitz matrix, where N is the number of subbands, N≤MNK, and can be adjusted according to the signal-to-noise ratio and actual performance. A Toeplitz matrix is a matrix where all elements on the diagonal are equal, and all elements parallel to the diagonal are also equal.
[0148] Calculate the covariance matrix R XQ As in equation (16):
[0149]
[0150] In this context, the superscript H represents the matrix conjugate transpose.
[0151] If multiple probes are performed, the covariance matrix of the multiple probes needs to be smoothed to obtain the echo invalid data covariance matrix R. X As in equation (17):
[0152]
[0153] Step S7: Utilizing the strong correlation of strong co-frequency interference in the distance dimension and its dominance in the region of invalid echo data, the feature decomposition method is used to obtain the interference subspace and the signal-noise subspace for each frequency point to be processed. The valid echo data is then projected onto the signal-noise subspace, thereby achieving interference suppression of the oblique return ionization map.
[0154] It should be noted that the method of using eigenvalue decomposition to obtain the interference subspace and signal-noise subspace for each frequency point, and projecting the effective echo data onto the signal-noise subspace to achieve interference suppression of the oblique return ionization map is as follows:
[0155] First, the covariance matrix R of the invalid echo data is used. x The eigenvalues are decomposed as shown in equation (18):
[0156] R x =UΛU H Equation (18)
[0157] Where U represents the eigenvector matrix, Λ represents the eigenvalue matrix, and the superscript H represents the matrix conjugate transpose.
[0158] The eigenvector matrix is partitioned as shown in equation (19):
[0159] U = [G, V] Equation (19)
[0160] Where G represents the interference subspace and V represents the signal-noise subspace, their expressions are shown in equations (20) and (21):
[0161] G = span{u1, u2, ..., u} r Equation (20)
[0162] V = span{u r+1 u r+2 ,…,u N Equation (21)
[0163] Where u1, u2, ..., u N Let U represent the 1st, 2nd, ..., Nth column vectors of U, span{} represent the vector space spanned by the vectors, and r is the subspace partitioning boundary, the value of which is determined by the following method:
[0164] First, take the logarithm of the diagonal of the eigenvalue matrix and round it down as shown in equation (22):
[0165]
[0166] Where, diag() represents the sequence of diagonal elements of the matrix, λ f This represents the result of taking the logarithm of the diagonal sequence and rounding it down.
[0167] Then r is as shown in equation (23):
[0168]
[0169] Where, λ f (t) represents the sequence λ f The t-th element in the matrix, Mo(λ) f ) represents the sequence λ f The mode.
[0170] After determining the interference subspace, the valid echo data is projected onto the signal-noise subspace as shown in equation (24):
[0171]
[0172] in, I represents the projection of the cyclic cross-correlation signal matrix X, where I is the identity matrix and the superscript H is the matrix conjugate transpose.
[0173] Step S8: Perform coherent accumulation and constant false alarm rate processing on the signal obtained in step S7 after suppressing radio frequency interference to obtain a slanted return ionization map with continuous leading edge.
[0174] It should be noted that the method for processing coherent accumulation is as follows:
[0175] If multiple probes are performed, then... The Fourier transform of the time-domain data at each frequency point is shown in equation (25):
[0176]
[0177] in, for The frequency domain representation of f is the frequency, and fft() represents performing a Fourier transform on the sequence.
[0178] Coherent accumulation is as shown in equation (26):
[0179]
[0180] in, for The model.
[0181] In addition, it should be noted that the handling method for constant false alarms is as follows:
[0182] Construct a vector of length q consisting entirely of 1s as shown in equation (27):
[0183]
[0184] The echo obtained by suppressing radio frequency interference at each frequency point, obtained by the method in step S4, is convolved with the all-1 vector as shown in equation (28):
[0185]
[0186] To suppress the sliding average of q range gates of the echo after radio frequency interference, for The elements are sorted, and the minimum value is taken as the noise estimate of the echo as shown in equation (29):
[0187]
[0188] Set the false alarm rate P fa The value is typically taken as 0.001 to 0.1. The constant false alarm threshold U is calculated. T As in equation (30):
[0189]
[0190] Extracting valid data echo estimates As in equation (31):
[0191]
[0192] Step S9: Perform energy comparison stretching on the ionospheric oblique return signal after step S8 (constant false alarm rate).
[0193] It should be noted that the energy contrast stretching is as shown in equation (32):
[0194]
[0195] in, The result of single-frequency echo processing in this invention is shown, where n is the energy contrast stretching factor, typically ranging from 2 to 4, which can be adjusted according to the actual effect. Rearranging all frequencies yields a slanted return ionization map with a uniform background and improved signal-to-noise ratio.
[0196] In embodiments of the present invention, such as Figure 3As shown, a practical single-frequency observation example is that in the 13MHz range-Doppler distribution map, before interference suppression, there is a strong narrowband interference at the Doppler frequency of 11Hz. This interference also raises the noise floor of adjacent frequencies, causing the oblique return signal to be submerged in noise. The oblique return signal cannot be observed in the range-Doppler distribution map. After feature decomposition in step S7 to obtain the interference subspace and the signal-noise subspace, the signal segment is projected into the signal-noise subspace, and the oblique return signal with a certain broadening near the Doppler frequency of 0Hz can be obtained. At the same time, the strong interference signal at the Doppler frequency of 11Hz is greatly attenuated, with an attenuation of more than 25dB, and the signal-to-interference ratio is greatly improved.
[0197] In embodiments of the present invention, such as Figure 4 As shown, a practical example of single-frequency observation is in the distance-time distribution map of 13MHz. Before interference suppression, the observed interference is short-duration, with each interference segment lasting approximately 2 seconds. s Similarly, the noise level rises throughout the entire detection period, and the M-sequence signal after cyclic correlation cannot be observed. However, by performing feature decomposition in step S7, the interference subspace and signal-noise subspace are obtained. Projecting this signal segment into the signal-noise subspace yields a clear oblique return signal. The leading edge of this signal appears at approximately 900 km. However, it can be observed from the figure that the signal is not stable within this distance. Conversely, at a distance of approximately 930 km, the 13 MHz signal is more stable.
[0198] In embodiments of the present invention, such as Figure 5 As shown, a practical example of single-frequency observation is that even under the prior condition that the Doppler frequencies of the ionospheric oblique return echo and sea clutter are known to be between -2Hz and 2Hz, coherent accumulation is performed as described in step S8. Due to strong interference, the rise in background noise cannot be filtered out by this prior condition, resulting in multiple false peaks in the accumulated distance-amplitude distribution. Moreover, it is difficult to distinguish between signal and noise. After constant false alarm rate (CFAR) as described in step S8, all of these false peaks are mistakenly identified as noise and eventually suppressed. However, after interference suppression as described in step S7, coherent accumulation as described in step S8 clearly shows the signal leading edge, and the signal-to-noise ratio reaches nearly 20dB.
[0199] In embodiments of the present invention, such as Figure 6As shown, a practical example of frequency sweep observation is as follows: For a conventional ionospheric oblique return ionogram, the oblique return echo data undergoes coherent accumulation, constant false alarm rate (CFAR), and energy stretching as described in steps S8-S9, resulting in an oblique return ionogram with a uniform background and significant contrast. However, due to co-frequency interference during coherent accumulation, many frequency points in the oblique return ionogram are suppressed, leading to signal loss. After adding the co-frequency interference suppression method described in step S7 through eigenvalue decomposition, the signal-to-interference ratio (SNR) is greatly improved, and the suppressed signals are recovered. From the actual processing results, the oblique return frequency sweep ionogram after interference suppression has a higher SNR and a significantly reduced number of missing frequency points, providing a reliable basis for frontier fitting and achieving the expected results.
[0200] In specific implementation, the method proposed in the technical solution of this invention can be automatically executed by those skilled in the art using computer software technology. System devices for implementing the method, such as computer-readable storage media storing the corresponding computer program of the technical solution of this invention and computer equipment including the computer program running the corresponding computer program, should also be within the protection scope of this invention.
[0201] In some possible embodiments, an ionospheric oblique return ionograph interference suppression system based on feature decomposition is provided, including a processor and a memory. The memory is used to store program instructions, and the processor is used to call the stored instructions in the memory to execute an ionospheric oblique return ionograph interference suppression method based on feature decomposition as described above.
[0202] In some possible embodiments, an ionospheric oblique return ionograph interference suppression system based on feature decomposition is provided, including a readable storage medium storing a computer program, which, when executed, implements the ionospheric oblique return ionograph interference suppression method based on feature decomposition as described above.
[0203] The specific embodiments described herein are merely illustrative of the spirit of the invention. Those skilled in the art to which this invention pertains may make various modifications or additions to the described specific embodiments or use similar methods to substitute them, without departing from the spirit of the invention or exceeding the scope defined by the appended claims.
Claims
1. A method of ionospheric oblique back ionogram interference suppression based on eigen decomposition, characterized by: Using an oblique return detector, a narrow-bandwidth, M-sequence encoded continuous shortwave probe signal is transmitted using frequency sweeping or fixed-frequency observation. After determining the frequency point to be processed using the mode, the invalid part of the echo is truncated. Then, by utilizing the strong correlation of co-frequency interference in the echo signal in the range dimension, the range gate echo covariance matrix of the invalid echo signal is eigenvalued. The number of interference sources is estimated from the eigenvalue sequence, and the signal-noise subspace and interference subspace are divided. The range gate signal with oblique return echo is projected onto the signal-noise subspace to suppress co-frequency interference in the echo, without being limited by the number of repeated probes or the number of carriers. Specifically, the method of utilizing the strong correlation of co-frequency interference in the echo signal in the distance dimension to perform eigenvalue decomposition on the range gate echo covariance matrix of the invalid echo signal includes utilizing the strong correlation of strong co-frequency interference in the distance dimension and its dominant component in the invalid echo data region to perform eigenvalue decomposition on each frequency point to be processed to obtain the interference subspace and the signal noise subspace.
2. The eigen-based ionospheric oblique back ionogram interference suppression method of claim 1, wherein: The process of suppressing interference from the ionosphere oblique return ionogram is as follows. Step S1: Set the frequency sweep range and the number of frequency point detections, and determine the signal encoding length and duty cycle; Step S2: Configure the oblique return detector to detect the oblique return signal according to the coding modulation method and transmission parameters determined in step S1, and obtain the echo signal data; Step S3: Calculate the cyclic cross-correlation signal between the echo data of a single carrier frequency and the M-sequence encoded and modulated by the set detection signal; Step S4: Use the mode to determine the frequency point to be processed; Step S5: Extract invalid echo data of the carrier frequency determined in step S4; Step S6: After cross-correlation of the invalid echo data obtained in step S5, the signal is divided into several sub-bands that partially overlap each other and have equal spacing, and a time series matrix is constructed. The covariance matrix is obtained by calculating the covariance of the matrix. Step S7: Utilizing the strong correlation of strong co-frequency interference in the distance dimension and its dominance in the region of invalid echo data, the feature decomposition method is used to obtain the interference subspace and the signal-noise subspace for each frequency point to be processed. The valid echo data is then projected onto the signal-noise subspace, thereby achieving interference suppression of the oblique return ionization map. Step S8: Perform coherent accumulation and constant false alarm rate processing on the signal obtained in step S7 after suppressing radio frequency interference to obtain a slanted return ionization map with continuous leading edge. Step S9: Perform energy comparison stretching on the ionospheric oblique return signal after step S8, which is based on constant false alarm rate.
3. The eigen-decomposition based ionogram slantwise return ionogram interference suppression method of claim 2, wherein: The maximum detection range determined by the length of the M-sequence used to encode and modulate the signal in step S1 and the transmission duty cycle is greater than 3000 km.
4. The ionospheric oblique return ionograph interference suppression method based on eigenvalue decomposition as described in claim 2 or 3, characterized in that: If the code order of the M sequence in step S3 is n, then it is represented as follows: in, These are the element values in sequence M; The sequence is represented by p cyclic shifts. Construct the correlation sequence matrix Represented as, If the total number of detections for the single-carrier frequency echo in step S3 is m, then the sequence of the k-th detection is represented as follows: in, These are the pulses before pulse compression in the k-th detection. Echo data from each distance gate, k=1,2,…,m; The single-carrier frequency echo matrix in step S3 is expressed as follows: The M-sequence cyclic cross-correlation signal in step S3 is represented as follows: Among them, superscript This represents the conjugate transpose of a matrix.
5. The ionospheric oblique return ionization map interference suppression method based on eigenvalue decomposition as described in claim 4, characterized in that: Step S4, which uses the mode to determine the frequency point to be processed, is implemented as follows: Construct a vector of length q consisting entirely of 1s. as follows, The echo signal obtained by cyclic cross-correlation at each frequency point is convolved with a vector of all 1s as follows: The moving average of the q range gates of the echo after cyclic cross-correlation at a single frequency point is given. Sort the elements and take the minimum value as the noise estimate of the echo. as follows, Taking the logarithm of the noise estimate and rounding it down, as follows: The noise estimation sequence for the entire frequency range is constructed as follows: in, Representing the Noise estimation at each frequency point ; The frequency points to be processed are shown below. In the above formula, Represents the mode. Frequency points after cyclic cross-correlation Signal amplitude, Indicates intersection, This indicates taking the maximum value in the sequence. This indicates taking the minimum value of the sequence. These are experience points.
6. The ionospheric oblique return ionization map interference suppression method based on eigenvalue decomposition as described in claim 2 or 3, characterized in that: The range of invalid echo data in step S5 is the continuous signal sequence that is overwhelmed by interference and noise in the original oblique return ionization diagram. If all frequency points are invalid data starting from the Kth range gate, and there are M range gates in a single detection cycle, then the invalid echo data matrix is represented as follows. in, The cyclic cross-correlation signal matrix obtained in step S3 The Column sequence.
7. The ionospheric oblique return ionograph interference suppression method based on eigenvalue decomposition as described in claim 6, characterized in that: Step S6, when dividing the single-carrier frequency echo invalid signal matrix into subbands, requires the following relationships between the time interval of each subband, the sequence length of each subband, and the estimation of the covariance matrix: For the Q-th probe in the invalid data matrix of a single-carrier echo, the echo sequence is represented as follows. in, The cyclic cross-correlation signal matrix obtained in step S3 The The first in the column sequence Line element, superscript Represents the transpose of a matrix; Then the sub-bands are divided and reconstructed into a time series matrix. as follows, in For the number of sub-bands, It can be adjusted according to the signal-to-noise ratio and actual effect; Calculate the covariance matrix as follows If multiple probes are performed, the covariance matrix of the multiple probes needs to be smoothed to obtain the desired result. as follows, 8. The ionospheric oblique return ionization map interference suppression method based on eigenvalue decomposition as described in claim 7, characterized in that: Step S7 uses the eigenvalue decomposition method to obtain the interference subspace and signal-noise subspace for each frequency point, and projects the valid echo data onto the signal-noise subspace. The method for achieving interference suppression of the oblique return ionization map is as follows: First, the eigenvalues are decomposed using the covariance matrix of the invalid echo data, as follows. Where U represents the eigenvector matrix. Represents the eigenvalue matrix, with superscripts Represents the conjugate transpose of a matrix; The eigenvector matrix is divided as follows: Where G represents the interference subspace and V represents the signal-noise subspace. in, Represents the eigenvector matrix The Column vectors This represents the vector space spanned by vectors. Define the boundaries of the subspace; After determining the interference subspace, the valid echo data is projected onto the signal-noise subspace.
9. The ionospheric oblique return ionograph interference suppression method based on eigenvalue decomposition as described in claim 8, characterized in that: The coherent accumulation in step S8 is as follows: If multiple probes are performed, then... The Fourier transform of the time-domain data at each frequency point is as follows. The coherent accumulation is as follows: in, for The model.
10. The ionospheric oblique return ionization map interference suppression method based on eigenvalue decomposition as described in claim 9, characterized in that: The energy comparison stretching in step S9 is as follows. in, This is the result of single-frequency echo processing. Energy contrast stretching factor, For constant false alarm threshold, This is the estimated value of the effective data echo.