A segmented SVD adaptive wavelet denoising method for OTDR signals

Abstract

The present invention discloses a segmented SVD adaptive wavelet denoising method for OTDR signals. When segmenting the signals, a certain percentage of overlapping segments are used to reduce the errors introduced by sudden changes. A series of singular values arranged from large to small are obtained according to the SVD decomposition. The singular value feature mean method is used to select the reconstruction thresholds for each group, and each signal segment is reconstructed. Combining wavelet threshold denoising, compared with the signals obtained by traditional singular value decomposition denoising, the signal-to-noise ratio has been significantly improved. It has better adaptability compared with wavelet soft threshold and hard threshold denoising, retains the useful information of the OTDR signal, significantly reduces noise, can effectively improve the denoising effect of the signal, obtains high-quality denoised signals, highlights event features, is more conducive to the detection and analysis of event points, and solves the technical problems of poor denoising effects of the existing cumulative average denoising method, wavelet threshold denoising method, and singular value decomposition denoising method for OTDR signal denoising.

Description

Technical Field

[0001] The present invention relates to the technical field of OTDR signal denoising, and particularly to a segmented SVD adaptive wavelet denoising method for OTDR signals. Background Art

[0002] An optical time domain reflectometer (OTDR) is an instrument that uses the backward Rayleigh scattering optical signal generated when pulsed light travels in an optical fiber to characterize the transmission characteristics of the optical fiber. As a non-destructive optical fiber measurement technology, OTDR can measure the optical fiber length, the transmission attenuation of the optical fiber, and fault location, etc., and is widely used in the production, construction, maintenance, etc. of optical fiber cables. It is an indispensable test instrument in the field of optical fiber communication. The OTDR curve can reflect the loss distribution of the backward scattered light along the optical fiber. In the OTDR test curve, the main event types included are non-reflection events, reflection events, and the end of the optical fiber, such as events like fiber joints, fusions, bends, breaks, etc. These events correspond to the losses at various locations in the optical fiber. For reflection events, under normal circumstances, the OTDR curve has a relatively high signal-to-noise ratio and is easily detected and identified. However, for smaller event points such as non-reflection events, the signal is easily submerged by noise. When the curve is severely contaminated by noise, the events in the curve are difficult to identify. Therefore, it is a necessary operation to perform denoising processing on the signal during the signal analysis process.

[0003] When a mutation event occurs, there is usually some aliased strong noise in the high-frequency part of the OTDR detection signal. Traditional OTDRs generally use the cumulative average denoising method, wavelet threshold denoising method or singular value decomposition denoising method for denoising. Although the cumulative average denoising method can improve the denoising effect through multiple cumulative averages, its efficiency is low, and the denoising effect will not improve with the increase of the cumulative number after a certain number of accumulations. There are two types of traditional wavelet threshold denoising methods: hard threshold function and soft threshold function. There is distortion in the wavelet soft threshold denoising method, and the denoising effect is not good. In the wavelet hard threshold denoising method, the signal is discontinuous at the threshold, which easily leads to signal oscillation. Moreover, both the wavelet soft threshold denoising method and the wavelet hard threshold denoising method need to select appropriate wavelet bases, decomposition levels, thresholds and threshold functions according to the characteristics of the signal, and their adaptive ability is poor. For the singular value decomposition (SVD) denoising method, the original OTDR signal is a non-stationary signal affected by noise. According to the distribution characteristics of the noise in the signal, it can be analyzed that the intensity of the noise gradually increases with the increase of the distance. This means that the signal-to-noise ratio of signals at different length ends will gradually decrease, and the ratio of the useful signal to the noise signal will also gradually change. If a single SVD denoising is performed on all sequences of the OTDR signal, and a unified critical point is selected as the reconstruction threshold during the signal reconstruction process, it will cause signal distortion or poor denoising effect.

[0004] To solve the problem of poor denoising effect existing in the above cumulative average denoising method, wavelet threshold denoising method and singular value decomposition denoising method, the present invention provides a segmented SVD adaptive wavelet denoising method for OTDR signals. Compared with the signals obtained by traditional singular value decomposition denoising, the signal-to-noise ratio is significantly improved. It has better adaptability compared with wavelet soft threshold and hard threshold denoising, retains the useful information of the OTDR signal, significantly reduces noise, can effectively improve the denoising effect of the signal, obtains high-quality denoised signals, highlights event characteristics, and is more conducive to the detection and analysis of event points. Summary of the Invention

[0005] An embodiment of the present invention provides a segmented SVD adaptive wavelet denoising method for OTDR signals, which is used to solve the technical problem of poor denoising effect of the existing cumulative average denoising method, wavelet threshold denoising method and singular value decomposition denoising method used for OTDR signal denoising.

[0006] In view of this, the first aspect of the present invention provides a segmented SVD adaptive wavelet denoising method for OTDR signals, including:

[0007] Perform signal overlapping segmentation processing on the to-be-processed OTDR original signal, and divide the to-be-processed OTDR original signal into several segmented signals with preset percentage overlapping segments;

[0008] Construct trajectory matrices for all segmented signals respectively, perform singular value decomposition on each trajectory matrix respectively, and obtain the singular values corresponding to each trajectory matrix;

[0009] Sort the singular values corresponding to each trajectory matrix in descending order respectively, and use the singular value feature mean method to select the reconstruction threshold for each group of sorted singular values;

[0010] Perform SVD reconstruction on the segmented signals according to the selected reconstruction threshold, and obtain the corresponding segmented reconstructed signals;

[0011] According to the determined wavelet decomposition level, wavelet threshold, and adaptive threshold function, perform wavelet threshold denoising on each segmented reconstructed signal respectively, and obtain the denoised segmented reconstructed signals;

[0012] Perform signal sequence restoration on the denoised segmented reconstructed signals to obtain the denoised OTDR signal.

[0013] Optionally, sorting the singular values corresponding to each trajectory matrix in descending order respectively, and using the singular value feature mean method to select the reconstruction threshold for each group of sorted singular values includes:

[0014] Sort the singular values corresponding to each trajectory matrix in descending order respectively to obtain each group of sorted singular value sequences;

[0015] Determine the reconstruction threshold of each group of sorted singular value sequences through the mean value of the eigenvalues of matrix AA T , where A is the trajectory matrix, and A T is the transpose of A.

[0016] Optionally, performing SVD reconstruction on the segmented signals according to the selected reconstruction threshold to obtain the corresponding segmented reconstructed signals includes:

[0017] For each group of sorted singular value sequences, set all singular values after the reconstruction threshold to zero, retain all singular values before the reconstruction threshold, and construct a new singular value matrix;

[0018] Perform SVD reconstruction on the segmented signals according to the new singular value matrix to obtain the corresponding segmented reconstructed signals.

[0019] Optionally, the calculation formula for the wavelet decomposition level is:

[0020] J = [log2N]

[0021] where J is the wavelet decomposition level and N is the length of the segmented signal.

[0022] Optionally, the process of determining the wavelet threshold is:

[0023] Take the absolute value of all values of the target signal x(t) of the wavelet transform, and arrange all the absolute values in ascending order;

[0024] Calculate the square value of each arranged element to obtain the recombined sequence f(k), where k = 0, 1,..., N - 1;

[0025] Perform risk assessment on the recombined sequence f(k) to obtain the risk assessment value R(k) of each data point. The risk assessment formula is:

[0026]

[0027] Determine the wavelet threshold according to the risk assessment value R(k). The calculation formula is:

[0028]

[0029] Among them, γ j is the wavelet threshold of the j-th layer, and f(k min ) is the sequence point corresponding to the minimum risk assessment value R(k).

[0030] Optionally, the adaptive threshold function is:

[0031]

[0032] Among them, j is the decomposition layer number, W j,k is the estimated wavelet coefficient of w j,k , w j,k is the k-th wavelet coefficient at the j-th scale of the decomposition, and h is the adjustment factor.

[0033] Optionally, perform signal sequence recovery on the denoised piecewise reconstructed signal. The signal sequence recovery formula for the denoised OTDR signal is:

[0034]

[0035] Among them, x′(n) is the recovered signal, n is the n-th point of x′(n), x d (t) is the t-th data point in the d-th piecewise reconstructed signal, x d is the average sequence of the overlapping part of two adjacent segments, l is the length of each segment signal during continuous segmentation, and p is the length of the overlapping part of the segment signal.

[0036] The second aspect of the present invention provides a wavelet denoising system for a piecewise SVD adaptive function of an OTDR signal, including:

[0037] A segmentation module, which is used to perform signal overlapping segmentation processing on the to-be-processed OTDR raw signal, and divide the to-be-processed OTDR raw signal into several segmented signals with preset percentage overlapping segments;

[0038] A trajectory matrix establishment module, which is used to construct trajectory matrices for all segmented signals respectively, perform singular value decomposition on each trajectory matrix respectively, and obtain the singular values corresponding to each trajectory matrix;

[0039] A reconstruction threshold determination module, which is used to sort the singular values corresponding to each trajectory matrix in descending order respectively, and use the singular value feature mean method to select the reconstruction threshold for each group of sorted singular values;

[0040] An SVD reconstruction module, which is used to perform SVD reconstruction on the segmented signal according to the selected reconstruction threshold to obtain the corresponding segmented reconstruction signal;

[0041] A wavelet denoising module, which is used to perform wavelet threshold denoising on each segmented reconstruction signal respectively according to the determined wavelet decomposition level, wavelet threshold and adaptive threshold function to obtain the denoised segmented reconstruction signal;

[0042] A signal recovery module, which is used to perform signal sequence recovery on the denoised segmented reconstruction signal to obtain the denoised OTDR signal.

[0043] Optionally, the reconstruction threshold determination module is specifically used for:

[0044] Sort the singular values corresponding to each trajectory matrix in descending order respectively to obtain each group of descending sorted singular value sequences;

[0045] Determine the reconstruction threshold of each group of descending sorted singular value sequences through the mean value of the eigenvalues of matrix AA T where A is the trajectory matrix, and A T is the transpose of A.

[0046] Optionally, the SVD reconstruction module is specifically used for:

[0047] For each group of descending sorted singular value sequences, set all singular values after the reconstruction threshold to zero, retain all singular values before the reconstruction threshold, and construct a new singular value matrix;

[0048] Perform SVD reconstruction on the segmented signal according to the new singular value matrix to obtain the corresponding segmented reconstruction signal.

[0049] It can be seen from the above technical solutions that the segmented SVD adaptive wavelet denoising method and system for OTDR signals provided by the present invention have the following advantages:

[0050] In the embodiment of the present invention, for the segmented SVD adaptive wavelet denoising method of OTDR signals, when segmenting the signals, a certain percentage of overlapping segments are used to reduce the errors introduced by sudden changes. A series of singular values arranged from large to small are obtained according to the SVD decomposition. The singular value feature mean method is used to select the reconstruction threshold for each group, and each signal segment is reconstructed. Combining wavelet threshold denoising, compared with the signals obtained by traditional singular value decomposition denoising, the signal-to-noise ratio is significantly improved. Compared with wavelet soft threshold and hard threshold denoising, it has better adaptability, retains the useful information of the OTDR signal, significantly reduces the noise, can effectively improve the denoising effect of the signal, obtain high-quality denoised signals, highlight the event characteristics, and is more conducive to the detection and analysis of event points, solving the technical problem that the existing OTDR signal denoising methods such as cumulative average denoising method, wavelet threshold denoising method, and singular value decomposition denoising method have poor denoising effects.

[0051] The segmented SVD adaptive wavelet denoising system for OTDR signals provided by the present invention is used to execute the segmented SVD adaptive wavelet denoising method for OTDR signals provided by the present invention. Its principle and the achieved technical effects are the same as those of the segmented SVD adaptive wavelet denoising method for OTDR signals provided by the present invention, and will not be elaborated here. Brief Description of the Drawings

[0052] Figure 1 It is a schematic flowchart of a segmented SVD adaptive wavelet denoising method for OTDR signals provided in an embodiment of the present invention;

[0053] Figure 2 It is a schematic diagram of signal overlapping segmentation of a segmented SVD adaptive wavelet denoising method for OTDR signals provided in an embodiment of the present invention;

[0054] Figure 3 It is a comparison diagram of soft threshold function, hard threshold function, and adaptive threshold function provided in an embodiment of the present invention;

[0055] Figure 4 It is a schematic structural diagram of a segmented SVD adaptive wavelet denoising system for OTDR signals provided in an embodiment of the present invention. Detailed Embodiment

[0056] In order to enable those skilled in the art to better understand the solution of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0057] For ease of understanding, please refer to Figure 1 , the segmented SVD adaptive wavelet denoising method for OTDR signals provided in the embodiments of the present invention includes:

[0058] Step 101: Perform signal overlapping segmentation processing on the to-be-processed OTDR raw signal, and divide the to-be-processed OTDR raw signal into several segmented signals with a preset percentage overlapping segment.

[0059] It should be noted that if the existing continuous segmentation method is used to segment the to-be-processed OTDR raw signal, after the adjacent segments are subjected to singular value decomposition processing and the signals are restored, mutations will occur between the continuous segments, resulting in signal distortion and affecting the denoising effect. In the embodiments of the present invention, a preset percentage of overlapping segments are used to reduce the error introduced by the mutation. The signal segmentation schematic diagram in the embodiments of the present invention is as Figure 2 shown Figure 2 In, x1 to x4 are 4 segmented signals obtained after overlapping segmentation, x(l) is the midpoint of the overlapping segment between x1 and x2, x(2l) is the midpoint of the overlapping segment between x2 and x3, x(3l) is the midpoint of the overlapping segment between x3 and x4, x(4l) is the last data point of the original signal, the length of the overlapping segment is p, and x(1) is the first data point of the original signal.

[0060] The to-be-processed OTDR raw signal is overlapped and divided into m segmented signals, and the overlapping length of the adjacent upper segment and lower segment is 2p. Therefore, the lengths of the first segment and the last segment of the segmented signal are l + p, and the lengths of the other intermediate segments are l + 2p, where l is the length of each segmented signal during continuous segmentation (if the signal length is n and it is continuously segmented into m segments, then l = n / m).

[0061] Step 102: Construct trajectory matrices for all the segmented signals respectively, and perform singular value decomposition on each trajectory matrix to obtain the singular values corresponding to each trajectory matrix.

[0062] It should be noted that the m segmented signals obtained after overlapping segmentation are used to construct trajectory matrices, that is, m Hankel matrices are constructed. The i-th (i ≤ m) matrix is:

[0063]

[0064] Among them, A iLet \(P_{i}\) be the trajectory matrix corresponding to the \(i\)-th segmented signal, \(P = N - L+1\), where \(N\) is the data length of each segmented signal, and the step size for constructing the matrix is 1, that is, the data points in each row lag behind those in the previous row by one position. Different values of the number of rows \(L\) and the number of columns \(P\) result in different denoising effects on the signal. To separate the useful signal and the noise signal in the noisy signal more fully, the product of the selected number of rows \(L\) and the number of columns \(P\) should be as large as possible. When \(N\) is even, \(L = N / 2\); when \(N\) is odd, \(L=(N + 1) / 2\).

[0065] According to the SVD principle, the constructed trajectory matrix can be expressed as:

[0066] \(A = USV\ T

[0067] where \(A\) is the trajectory matrix, \(U\) and \(V\) are orthogonal matrices, \(U\in R\ g×g \), \(V\in R\ h×h \), \(S\in R\ g×h \), \(UU\ T =I\), \(VV\ T =I\), \(U = [u_1,u_2,u_3,\cdots,u\ g \) is the left singular matrix, \(V = [v_1,v_2,v_3,\cdots,v\ g \) is the right singular matrix, \(S\) is a diagonal matrix, the values on the diagonal are singular values, and other elements are 0, and \(S=\text{diag}(\sigma_1,\sigma_2,\sigma_3,\cdots,\sigma\ r )\) satisfies \(\sigma_1\geq\sigma_2\geq\sigma_3\geq\cdots\geq\sigma\ r .

[0068] Perform SVD processing on each trajectory matrix through the above formula to obtain \(m\) groups of left singular matrices \(U\), right singular matrices \(V\), and diagonal matrices \(S\). The values on the diagonal of the diagonal matrix \(S\) are the decomposed singular values.

[0069] Step 103: Sort the singular values corresponding to each trajectory matrix in descending order, and use the singular value characteristic mean method to select the reconstruction threshold for each group of sorted singular values.

[0070] It should be noted that the singular values corresponding to each decomposed trajectory matrix are arranged from large to small respectively to obtain the descending sorted singular value sequences for each group. The singular value characteristic mean method is used to select the reconstruction threshold for each group of descending sorted singular value sequences. The reconstruction threshold of each group of descending sorted singular value sequences is determined by the mean of the eigenvalues of the matrix \(A^TA\ T . The matrix \(A^T\ T \) is the transpose of the matrix \(A\). Among them:

[0071]

[0072] where \(\sigma\ i \) is the \(i\)-th singular value of the trajectory matrix, \(\lambda\i is the matrix AA T 's i-th eigenvalue. The eigenvalues of the matrix AA T correspond one-to-one with the singular values of the trajectory matrix.

[0073] Step 104: Perform SVD reconstruction on the segmented signal according to the selected reconstruction threshold to obtain the corresponding segmented reconstructed signal.

[0074] It should be noted that for each group of singular value sequences sorted in descending order, all singular values after the reconstruction threshold are discarded, and all singular values before the reconstruction threshold are used for reconstructing the signal. That is, all singular values after the reconstruction threshold are set to zero, and all singular values before the reconstruction threshold are retained to construct a new singular value matrix.

[0075] Step 105: Perform wavelet threshold denoising on each segmented reconstructed signal according to the determined wavelet decomposition level, wavelet threshold, and adaptive threshold function to obtain the denoised segmented reconstructed signal.

[0076] It should be noted that after obtaining the segmented reconstructed signal, wavelet threshold denoising is performed on the segmented reconstructed signal. The first step of wavelet threshold denoising is to perform wavelet decomposition on the noise signal. For a noise signal with a length of N, the wavelet decomposition level is:

[0077] J = [log2N]

[0078] where J is the wavelet decomposition level and N is the length of the segmented signal.

[0079] The wavelet threshold is selected using the unbiased risk estimation threshold method. Let the noisy signal (target signal) to be wavelet-transformed be x(t). Take the absolute value of all values of this signal, then sort them from smallest to largest according to the absolute value, and finally calculate the square value of each arranged element to obtain the recombined sequence f(k):

[0080] f(k) = (sort(|x(t)|)) 2

[0081] where k = 0, 1,..., N - 1 and sort is the sorting function.

[0082] Perform risk assessment on the recombined sequence f(k):

[0083]

[0084] where R(k) is the risk assessment value of the recombined sequence f(k) at the k-th data point.

[0085] Determine the wavelet threshold according to the risk assessment value R(k). The calculation formula is:

[0086]

[0087] Among them, γ j is the wavelet threshold of the j-th layer, and f(k min ) is the sequence point corresponding to the minimum risk assessment value R(k).

[0088] After determining the decomposition level and the wavelet threshold, it is also necessary to select an appropriate threshold function for processing. Since the noise conditions in the reconstructed signals of each component segment are different, in order to adapt to the noise levels of each segment of the signal, the adaptive threshold function adopted in the embodiments of the present invention is:

[0089]

[0090] Among them, j is the decomposition level, W j,k is the estimated wavelet coefficient of w j,k , w j,k is the k-th wavelet coefficient at the j-th scale of the decomposition, h is an adjustment factor, and specific parameters can be determined according to the noise level. Generally, h = (0.5, 1, 2, 3, 4). The comparison diagrams of the existing soft threshold function, hard threshold function and the adaptive threshold function provided by the embodiments of the present invention are as Figure 3 shown. The adaptive threshold function provided by the embodiments of the present invention overcomes the defects of the existing soft threshold function and hard threshold function, improves the calculation method at the threshold, makes the function curve continuous, solves the problem of insufficient denoising ability of the soft threshold denoising for wavelet coefficients in the threshold segment. Along with the increase of the wavelet coefficient amplitude, when the wavelet coefficient is greater than the set threshold, the wavelet coefficient after the improvement of the threshold function approaches the wavelet coefficient before processing infinitely, reducing the constant deviation brought by the hard threshold function.

[0091] Step 106: Recover the signal sequence of the denoised segmented reconstructed signal to obtain the denoised OTDR signal.

[0092] It should be noted that through the above steps, the denoised segmented reconstructed signals have been obtained. Among these signals, there are overlapping parts between adjacent signals. The overlapping signal segments are averaged to obtain a new sequence, and the non-overlapping parts keep the corresponding data values to restore the data length of the original signal. The calculation process is shown in the following formula:

[0093]

[0094] Among them, x′(n) is the restored signal, n is the n-th point of x′(n), x d (t) is the t-th data point in the d-th segmented reconstructed signal, x d is the average sequence of the overlapping part between two adjacent segments, l is the length of each segmented signal during continuous segmentation, and p is the length of the overlapping part of the segmented signal.

[0095] In the embodiment of the present invention, for the segmented SVD adaptive wavelet denoising method of the OTDR signal, when segmenting the signal, a certain percentage of overlapping segments are used to reduce the error introduced by sudden changes. According to the SVD decomposition, a series of singular values arranged from large to small are obtained. The singular value feature mean method is used to select the reconstruction threshold for each group, and each signal segment is reconstructed. Combining wavelet threshold denoising, compared with the signal obtained by traditional singular value decomposition denoising, the signal-to-noise ratio is significantly improved. Compared with wavelet soft threshold and hard threshold denoising, it has better adaptability, retains the useful information of the OTDR signal, significantly reduces noise, can effectively improve the denoising effect of the signal, obtains a high-quality denoised signal, highlights the event characteristics, is more conducive to the detection and analysis of event points, and solves the technical problem that the existing OTDR signal denoising methods such as cumulative average denoising method, wavelet threshold denoising method and singular value decomposition denoising method have poor denoising effects.

[0096] For ease of understanding, please refer to Figure 4 , and an embodiment of a segmented SVD adaptive wavelet denoising system for OTDR signals is also provided in the present invention, including:

[0097] A segmentation module, configured to perform signal overlapping segmentation processing on the to-be-processed OTDR original signal, and divide the to-be-processed OTDR original signal into several segmented signals with a preset percentage of overlapping segments;

[0098] A trajectory matrix establishment module, configured to respectively construct a trajectory matrix for all the segmented signals, and perform singular value decomposition on each trajectory matrix to obtain the singular values corresponding to each trajectory matrix;

[0099] A reconstruction threshold determination module, configured to respectively sort the singular values corresponding to each trajectory matrix in descending order, and use the singular value feature mean method to select the reconstruction threshold for each group of sorted singular values;

[0100] An SVD reconstruction module, configured to perform SVD reconstruction on the segmented signals according to the selected reconstruction threshold to obtain corresponding segmented reconstruction signals;

[0101] A wavelet denoising module, configured to perform wavelet threshold denoising on each segmented reconstruction signal respectively according to the determined wavelet decomposition level, wavelet threshold and adaptive threshold function to obtain the denoised segmented reconstruction signals;

[0102] A signal restoration module, configured to perform signal sequence restoration on the denoised segmented reconstruction signals to obtain the denoised OTDR signal.

[0103] The reconstruction threshold determination module is specifically configured to:

[0104] Sort the singular values corresponding to each trajectory matrix in descending order to obtain a sequence of singular values sorted in descending order for each group;

[0105] Determine the reconstruction threshold of the sequence of singular values sorted in descending order for each group through the mean value of the eigenvalues of matrix AA T , where A is the trajectory matrix, and A T is the transpose of A.

[0106] The SVD reconstruction module is specifically used for:

[0107] For each sequence of singular values sorted in descending order, set all singular values after the reconstruction threshold to zero, retain all singular values before the reconstruction threshold, and construct a new singular value matrix;

[0108] Perform SVD reconstruction on the segmented signal according to the new singular value matrix to obtain the corresponding segmented reconstructed signal.

[0109] The calculation formula for the number of wavelet decomposition levels is:

[0110] J = [log2N]

[0111] where J is the number of wavelet decomposition levels and N is the length of the segmented signal.

[0112] The process of determining the wavelet threshold is as follows:

[0113] Take the absolute value of all values of the target signal x(t) to be wavelet-transformed, and arrange all absolute values in ascending order;

[0114] Calculate the square value of each arranged element to obtain the recombined sequence f(k), where k = 0, 1,..., N - 1;

[0115] Perform risk assessment on the recombined sequence f(k) to obtain the risk assessment value R(k) for each data point. The risk assessment formula is:

[0116]

[0117] Determine the wavelet threshold according to the risk assessment value R(k). The calculation formula is:

[0118]

[0119] where γ j is the wavelet threshold of the j-th layer, and f(k min ) is the sequence point corresponding to the minimum risk assessment value R(k).

[0120] The adaptive threshold function is:

[0121]

[0122] where j is the decomposition level, and W j,k is the estimated wavelet coefficient of w j,k , w j,k is the k-th wavelet coefficient at the j-th scale of the decomposition, and h is the adjustment factor.

[0123] Performing signal sequence recovery on the segmented reconstructed signal after denoising, the signal sequence recovery formula for the OTDR signal after denoising is as follows:

[0124]

[0125] where x′(n) is the recovered signal, n is the n-th point of x′(n), and x d (t) is the t-th data point in the d-th segmented reconstructed signal, and x d is the average sequence of the overlapping part of two adjacent segments, l is the length of each segmented signal during continuous segmentation, and p is the length of the overlapping part of the segmented signals.

[0126] In the segmented SVD adaptive wavelet denoising system for OTDR signals provided in the embodiments of the present invention, when performing signal segmentation, a certain percentage of overlapping segments are used to reduce the error introduced by sudden changes. A series of singular values arranged from large to small are obtained according to the SVD decomposition. The singular value characteristic mean method is used to select the reconstruction threshold for each group and reconstruct each signal segment. Combining wavelet threshold denoising, compared with the signal obtained by traditional singular value decomposition denoising, the signal-to-noise ratio has a significant improvement. Compared with wavelet soft threshold and hard threshold denoising, it has better adaptability, retains the useful information of the OTDR signal, significantly reduces noise, can effectively improve the denoising effect of the signal, obtains a high-quality denoised signal, highlights the event characteristics, is more conducive to the detection and analysis of event points, and solves the technical problem of poor denoising effect of the existing cumulative average denoising method, wavelet threshold denoising method, and singular value decomposition denoising method for OTDR signal denoising.

[0127] The segmented SVD adaptive wavelet denoising system for OTDR signals provided in the embodiments of the present invention is used to execute the segmented SVD adaptive wavelet denoising method for OTDR signals in the foregoing embodiments of the segmented SVD adaptive wavelet denoising method for OTDR signals, and can achieve the same technical effects as the foregoing embodiments of the segmented SVD adaptive wavelet denoising method for OTDR signals. Its principle is the same as that of the foregoing embodiments of the segmented SVD adaptive wavelet denoising method for OTDR signals, and will not be elaborated herein.

[0128] Those skilled in the art can clearly understand that for the convenience and conciseness of description, the specific working processes of the above-described systems, devices, and units can refer to the corresponding processes in the foregoing method embodiments, and will not be elaborated herein.

[0129] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A segmented SVD adaptive wavelet denoising method for OTDR signals, characterized in that, Including: Perform signal overlapping segmentation processing on the OTDR raw signal to be processed, and divide the OTDR raw signal to be processed into several segmented signals with a preset percentage overlapping segment; Construct a trajectory matrix for all segmented signals respectively, perform singular value decomposition on each trajectory matrix respectively, and obtain the singular values corresponding to each trajectory matrix; Sort the singular values corresponding to each trajectory matrix in descending order respectively, and use the singular value feature mean method to select the reconstruction threshold for each group of sorted singular values; Perform SVD reconstruction on the segmented signal according to the selected reconstruction threshold to obtain the corresponding segmented reconstructed signal; According to the determined wavelet decomposition level, wavelet threshold and adaptive threshold function, perform wavelet threshold denoising on each segmented reconstructed signal respectively to obtain the denoised segmented reconstructed signal; Perform signal sequence restoration on the denoised segmented reconstructed signal to obtain the denoised OTDR signal.

2. The segmented SVD adaptive wavelet denoising method for OTDR signals according to claim 1, wherein Sort the singular values corresponding to each trajectory matrix in descending order respectively, and use the singular value feature mean method to select the reconstruction threshold for each group of sorted singular values, including: Sort the singular values corresponding to each trajectory matrix in descending order respectively to obtain each group of descending sorted singular value sequences; Determine the reconstruction threshold of the singular value sequence sorted in descending order for each group through the mean value of the eigenvalues of matrix AA T , where A is the trajectory matrix, and A T is the transpose of A.

3. The segmented SVD adaptive wavelet denoising method for OTDR signals according to claim 2, wherein Perform SVD reconstruction on the segmented signal according to the selected reconstruction threshold to obtain the corresponding segmented reconstructed signal, including: For each group of descending sorted singular value sequences, set all singular values after the reconstruction threshold to zero, retain all singular values before the reconstruction threshold, and construct a new singular value matrix; Perform SVD reconstruction on the segmented signal according to the new singular value matrix to obtain the corresponding segmented reconstructed signal.

4. The segmented SVD adaptive wavelet denoising method for OTDR signals according to claim 1, wherein, The calculation formula for the wavelet decomposition level is: J = [log2N] Where J is the wavelet decomposition level and N is the length of the segmented signal.

5. The segmented SVD adaptive wavelet denoising method for OTDR signals according to claim 4, characterized in that, The determination process of the wavelet threshold is: Take the absolute value of all values of the target signal x(t) to be wavelet-transformed, and arrange all absolute values in ascending order; Calculate the square value for each arranged element to obtain the recombined sequence f(k), where k = 0, 1,..., N - 1; Perform risk assessment on the recombined sequence f(k) to obtain the risk assessment value R(k) for each data point. The risk assessment formula is: Determine the wavelet threshold according to the risk assessment value R(k). The calculation formula is: Among them, γ j is the wavelet threshold of the j-th layer, and f(k min ) is the sequence point corresponding to the minimum risk assessment value R(k).

6. The segmented SVD adaptive wavelet denoising method for OTDR signals according to claim 5, wherein The adaptive threshold function is: where j is the decomposition level, and W j,k is the estimated wavelet coefficient of w j,k , w j,k is the k-th wavelet coefficient at the j-th scale of the decomposition, and h is the adjustment factor.

7. The segmented SVD adaptive wavelet denoising method for OTDR signals according to claim 6, characterized in that The signal sequence restoration formula for performing signal sequence restoration on the denoised segmented reconstructed signal to obtain the denoised OTDR signal is: Among them, x′(n) is the restored signal, n is the nth point of x′(n), and x d (t) is the tth data point in the dth segmented reconstruction signal, and x d is the average sequence of the overlapping part of two adjacent segments, l is the length of each segmented signal during continuous segmentation, p is the length of the overlapping part of the segmented signals, and m is the number of signal segments.

8. A segmented SVD adaptive wavelet denoising system for OTDR signals, characterized in that, Including: A segmentation module for performing signal overlapping segmentation processing on the OTDR raw signal to be processed, and dividing the OTDR raw signal to be processed into several segmented signals with a preset percentage overlapping segment; A trajectory matrix establishment module for constructing a trajectory matrix for all segmented signals respectively, performing singular value decomposition on each trajectory matrix respectively, and obtaining the singular values corresponding to each trajectory matrix; A reconstruction threshold determination module for sorting the singular values corresponding to each trajectory matrix in descending order respectively, and using the singular value feature mean method to select the reconstruction threshold for each group of sorted singular values; An SVD reconstruction module for performing SVD reconstruction on the segmented signal according to the selected reconstruction threshold to obtain the corresponding segmented reconstructed signal; Wavelet denoising module, which is used to perform wavelet threshold denoising on each segmented reconstructed signal respectively according to the determined wavelet decomposition level, wavelet threshold and adaptive threshold function, so as to obtain the denoised segmented reconstructed signal; Signal recovery module, which is used to recover the signal sequence of the denoised segmented reconstructed signal to obtain the denoised OTDR signal.

9. The segmented SVD adaptive wavelet denoising system for OTDR signals according to claim 8, wherein The reconstruction threshold determination module is specifically used for: Sort the singular values corresponding to each trajectory matrix in descending order respectively to obtain each group of singular value sequences sorted in descending order; Determine the reconstruction threshold of the singular value sequence sorted in descending order for each group through the mean value of the eigenvalues of matrix AA T where A is the trajectory matrix, and A T is the transpose of A.

10. The segmented SVD adaptive wavelet denoising system for OTDR signals according to claim 9, wherein, The SVD reconstruction module is specifically used for: For each group of singular value sequences sorted in descending order, set all singular values after the reconstruction threshold to zero, retain all singular values before the reconstruction threshold, and construct a new singular value matrix; Perform SVD reconstruction on the segmented signal according to the new singular value matrix to obtain the corresponding segmented reconstructed signal.

Images