Abnormal capture method for transient signal in digital oscilloscope

Abstract

The invention discloses an anomaly capture method for a transient signal in a digital oscilloscope, which comprises the following steps of: firstly extracting an abnormal signal and system noise in the transient signal, and then carrying out multi-section adaptive signal fitting on the abnormal signal to obtain a reference template matrix; and finally, calculating the Manhattan distance between each sampling point and an element between the corresponding reference template matrix based on a reference template matrix sampling sliding window comparison mode, counting a proportion meeting a distance threshold in a window by combining with a judgment threshold, and if the proportion meeting the distance threshold is met, judging that matching exists, triggering a capture signal and uploading a captured abnormal signal.

Description

Technical Field

[0001] The present invention belongs to the technical field of digital oscilloscopes. More specifically, it relates to a method for abnormally capturing transient signals in a digital oscilloscope. Background Art

[0002] With the rapid development of electronic technologies such as communication and radar, the signal frequency is continuously increasing, and the characteristics of transient signals are becoming increasingly complex. Their non-stationarity and suddenness pose great challenges to abnormal capture. Transient signals usually have characteristics such as short duration, concentrated energy, and fast frequency change. These characteristics make it difficult for traditional signal processing methods to effectively capture and identify abnormal phenomena therein. In practical applications, the abnormal capture of transient signals not only requires high-precision detection capabilities but also demands that the system can respond quickly and accurately identify the type of abnormality in order to timely evaluate its impact on system performance, quickly locate faults, and improve system reliability. However, factors such as the suddenness, diversity, and noise interference of transient signals further increase the difficulty of abnormal capture. The suddenness results in the random appearance of abnormal signals, and the uncertainty of their characteristic changes increases the complexity of capture. In addition, transient abnormal signals in different scenarios may exhibit similar characteristics, which may not only lead to misjudgment but also trigger repeated analysis, increasing the burden of data processing. Therefore, researching an efficient method for abnormally capturing transient signals is of great significance for improving the system's detection and response capabilities to transient abnormal signals.

[0003] In recent years, with the rapid development of technologies such as compressive sensing, time-frequency analysis, and deep learning, significant progress has been made in the field of transient signal processing. These technologies not only improve the accuracy of signal processing but also enhance the robustness of algorithms in complex environments. The methods for abnormally capturing transient signals can be mainly divided into methods based on time-domain analysis, frequency-domain analysis, and time-frequency joint analysis. Time-domain analysis methods capture abnormalities by directly analyzing the time-domain characteristics of signals, but they are sensitive to noise; frequency-domain analysis methods analyze the frequency-domain characteristics of signals through means such as Fourier transform, which are suitable for steady-state signals but have limited processing capabilities for non-stationary transient signals; time-frequency joint analysis methods combine the advantages of the time domain and the frequency domain and can better capture the non-stationary characteristics of transient signals, but their computational complexity is high. In addition, methods based on deep learning can automatically extract the characteristics of signals and achieve abnormal capture by training a large amount of data, but they are strongly dependent on sample data and may perform poorly when facing new types of abnormal signals. Therefore, in the research of methods for abnormally capturing transient signals, how to design a capture algorithm with high precision, strong robustness, and real-time performance in high-noise, low-sample data, and complex and changeable scenarios has become the focus of current research. By improving the accuracy and efficiency of abnormally capturing transient signals, the system's detection and processing capabilities for transient abnormal signals can be further enhanced to meet the application requirements in complex and changeable environments. Summary of the Invention

[0004] The object of the present invention is to overcome the deficiencies of the prior art and provide an abnormal capture method for transient signals in a digital oscilloscope. Based on multi-segment adaptive signal fitting and dynamic threshold determination technology, real-time capture of transient signals is achieved, solving the problems of low real-time capture rate of complex accidental signals and inaccurate classification and poor anti-noise ability caused by noise interference.

[0005] To achieve the above object of the invention, an abnormal capture method for transient signals in a digital oscilloscope according to the present invention is characterized by including the following steps:

[0006] (1) Extract abnormal signals and system noise in the transient signal;

[0007] (2) Extract the noise threshold ξ;

[0008] (3) Perform multi-segment adaptive signal fitting on the abnormal signal to obtain a reference template matrix;

[0009] (4) Based on the reference template matrix, complete the abnormal capture of the signal to be measured;

[0010] The object of the invention of the present invention is realized as follows:

[0011] An abnormal capture method for transient signals in a digital oscilloscope according to the present invention first extracts abnormal signals and system noise in the transient signal, then performs multi-segment adaptive signal fitting on the abnormal signal to obtain a reference template matrix, and finally calculates the Manhattan distance between each sampling point and the corresponding elements in the reference template matrix based on the reference template matrix sampling sliding window comparison method, and combines the decision threshold to count the proportion of the distance threshold satisfied within the window. If the proportion of the distance threshold is satisfied, it is determined as a match, triggering a capture signal and uploading the captured abnormal signal.

[0012] At the same time, an abnormal capture method for transient signals in a digital oscilloscope according to the present invention also has the following beneficial effects:

[0013] (1) The present invention adopts multi-segment adaptive signal fitting and dynamic threshold determination technology, and co-removes pseudo-extremes through extreme value optimization and noise threshold, significantly improving the robustness of transient signal feature extraction. Compared with traditional fixed threshold or uniform segmentation methods, the present invention can effectively distinguish noise interference from real signal features, and still maintain a capture accuracy of more than 90% when the signal-to-noise ratio is lower than 10dB, solving the problem of false triggering caused by noise sensitivity of traditional methods.

[0014] (2) By adaptively optimizing the order of piecewise polynomials and dynamically determining the fitting order of each segment based on the minimum criterion of the residual sum of squares (RSS), the present invention realizes high-precision signal modeling. Compared with the traditional fixed-order fitting method, while ensuring the fitting accuracy, the computational complexity of the present invention is reduced by more than 30%, which is especially suitable for the real-time processing requirements of transient signals in high-speed acquisition systems.

[0015] (3) The present invention proposes a probability threshold determination mechanism based on Manhattan distance, combined with the joint constraint of dynamic noise threshold and probability threshold, which significantly reduces the risk of mis-matching caused by accidental noise interference. Compared with the traditional single point-to-point matching method, the misjudgment rate is reduced by more than 50%, and it has stronger tolerance for local distortion of signal morphology.

[0016] (4) The present invention designs an extreme value segmentation rule and a signal trend adaptive judgment mechanism, and dynamically selects the segmentation logic by analyzing the time series distribution of extreme points to ensure that the segmentation boundary strictly matches the signal change trend. Compared with the manual experience segmentation method, the segmentation accuracy of the present invention for non-stationary transient signals is improved by 40%, significantly reducing the structural error of the fitting model.

[0017] (5) By dynamically generating the template matrix and quickly matching the sliding window, the present invention converts the fitting model into a reference template matrix, and realizes millisecond-level real-time detection by using parallel window comparison. Compared with the traditional full waveform matching technology, the memory occupancy is reduced by 60%, and the processing speed on FPGA is increased by more than 3 times, meeting the instant response requirements of high-speed oscilloscopes for transient abnormal signals. Description of the Drawings

[0018] Figure 1 is a flow chart of an abnormal capture method for transient signals in a digital oscilloscope of the present invention;

[0019] Figure 2 is a diagram of a signal to be measured and a specifically captured abnormal signal. Detailed Embodiments

[0020] The following describes the detailed embodiments of the present invention with reference to the drawings, so that those skilled in the art can better understand the present invention. It should be particularly noted that in the following description, when the detailed description of known functions and designs may dilute the main content of the present invention, these descriptions will be omitted here.

[0021] Embodiment

[0022] Figure 1 is a flow chart of an abnormal capture method for transient signals in a digital oscilloscope of the present invention.

[0023] In this embodiment, as Figure 1As shown in the figure, an abnormal capture method for transient signals in a digital oscilloscope according to the present invention includes the following steps:

[0024] (1). Extract abnormal signals and system noise in the transient signal;

[0025] (1.1). Turn on the abnormal detection mode of the digital oscilloscope, then access the transient signal for abnormal detection. When an abnormality is detected in the transient signal, extract the abnormal part of the transient signal and store it in matrix form, denoted as abnormal signal X;

[0026]

[0027] where x ij represents the value of the i-th sampling point of the abnormal signal collected at the j-th sampling moment, W is the number of sampling points collected at each sampling moment, and T represents the number of sampling moments;

[0028] In this embodiment, the accessed transient signal x(n) is a modulation signal with a carrier frequency of 1 GHz, a modulation frequency of 50 MHz, and a modulation depth of 50%, and is superimposed with a noise signal having a mean value of 0 and a standard deviation of 0.03. Then, the sequence signal is discretized at a sampling rate of 20 GSPS, and the total sampling duration is 1 s, obtaining 1D sampling data with a length of 20×10 9 .

[0029] In this embodiment, the number of rows of the abnormal signal X is 80, and the number of columns is 250×10 6 , indicating that the signal is processed at a processing clock frequency of 250 MHz.

[0030] (1.2). The digital oscilloscope collects system noise when no signal is input, and also stores it in matrix form. Let the stored system noise be

[0031]

[0032] where represents the value of the i-th sampling point of the system noise collected at the j-th sampling moment;

[0033] (2). Extract the noise threshold ξ;

[0034] Take the difference between the maximum value and the minimum value in the system noise as the noise threshold ξ;

[0035] (3). Perform multi-segment adaptive signal fitting on the abnormal signal X to obtain a reference template matrix;

[0036] (3.1) Determine the segmentation points according to the time-domain extreme value distribution characteristics of the abnormal signal;

[0037] (3.1.1) Extract the amplitudes of each sampling point in the abnormal signal X to form the amplitude matrix Y;

[0038]

[0039] where y ij is the amplitude corresponding to x ij ;

[0040] (3.1.2) Represent the amplitude matrix Y in the form of a serial vector according to the sampling time as:

[0041] Y = [y 11 , …, y W1 ; y 12 , …, y W2 ; …; y 1T , …, y WT

[0042] (3.1.3) In this embodiment, by obtaining the local maximum and minimum values in the amplitude matrix, the segmentation points are the extreme points. The specific process is as follows:

[0043] Traverse each element y ij in the amplitude matrix Y, compare the sizes of the two adjacent elements on the left and right of y ij . If y ij is greater than both elements, record the amplitude y ij as the maximum value y maxk , and record its corresponding coordinate as x maxk ;

[0044] After the elements are traversed, construct the maximum value matrix Y max ;

[0045]

[0046] where k is the maximum value number, k = 1, 2, …, K, and K is the number of maximum values;

[0047] Traverse each maximum value in the maximum value matrix Y max . If two adjacent maximum values are equal and there is a minimum value between the two adjacent maximum values, retain these two adjacent maximum values; otherwise, retain only any one of the two maximum values;

[0048] (3.1.4) Traverse each element y ij in the amplitude matrix Y, compare the sizes of the two adjacent elements on the left and right of y ij . If y ij is less than both elements, record the amplitude yij is the minimum value y minh , and its corresponding coordinate is denoted as x minh ;

[0049] After the elements are traversed, construct the minimum value matrix Y min ;

[0050]

[0051] Among them, h is the minimum value number, h = 1, 2,..., H, and H is the number of minimum values;

[0052] Traverse each minimum value in the minimum value matrix Y min If two adjacent minimum values are equal and there is no maximum value between them, then keep any one of the two minimum values; otherwise, keep both minimum values;

[0053] (3.1.5), Determine the segmentation point according to the coordinate of the first extreme point in the extreme value matrix;

[0054] (3.1.5.1), Based on min(K, H), delete the elements in the maximum value matrix Y max or the minimum value matrix Y min after min(K, H), and denote the remaining element numbers as l, l = 1, 2,..., L, and L = min(K, H);

[0055] (3.1.5.2), Compare the coordinate x max1 of the first maximum point with the coordinate x min1 of the first minimum point. When x max1 > x min1 , go to step (3.1.5.3); when x max1 < x min1 , go to step (3.1.5.4);

[0056] (3.1.5.3), Using the coordinates x max1 , x min1 as the starting points, that is, initialize l = 1, and traverse each extreme point in turn to judge the maximum and minimum values in the abnormal signal rising trend: If it satisfies y maxl+1 - y maxl > 0 and 0 < y maxl - y minl+1 < ξ, then determine that y maxl is a noise point in the abnormal signal rising trend and not a maximum value, and remove it from the maximum value matrix Y max , and at the same time remove y minl+1 from the minimum value matrix Y minRemove from it; after the judgment is completed, continue to judge the maximum and minimum values in the downward trend: If it satisfies y minl-1 -y minl > 0 and 0 < y maxl -y minl < ξ, then determine that y minl is a noise point in the downward trend of the abnormal signal, not a minimum value, and remove it from the minimum value matrix Y min At the same time, remove y maxl from the maximum value matrix Y max The maximum value matrix after removal is denoted as The minimum value matrix is

[0057] (3.1.5.4), with coordinates x max1 , x min1 as the starting point, that is, initialize l = 1, and sequentially traverse each extreme point to judge the maximum and minimum values in the upward trend of the abnormal signal: If it satisfies y maxl+1 -y maxl > 0 and 0 < y maxl -y minl < ξ, then determine that y maxl is a noise point in the upward trend of the abnormal signal, not a maximum value, and remove it from the maximum value matrix Y max At the same time, remove y minl from the minimum value matrix Y min After the judgment is completed, continue to judge the maximum and minimum values in the downward trend: If it satisfies y minl-1 -y minl > 0 and 0 < y maxl+1 -y minl < ξ, then determine that y minl is a noise point in the downward trend of the abnormal signal, not a minimum value, and remove it from the minimum value matrix Y min At the same time, remove y maxl from the maximum value matrix Y max The maximum value matrix after removal is denoted as The minimum value matrix is

[0058] (3.1.5.5), obtain the segmentation points according to the maximum value matrix and the minimum value matrix ;

[0059] In this embodiment, it is discussed in two cases. The first case is when x max1 < x min1 . If l is odd, the segmentation point is the maximum value x in the maximum value matrix maxl; If l is even, the segmentation point is the minimum value in the minimum value matrix of the minimum value x minl , so the segmentation point s l is expressed by the formula:

[0060]

[0061] When x max1 > x min1 , if l is odd, the segmentation point is the minimum value in the minimum value matrix of the minimum value x minl ; If l is even, the segmentation point is the maximum value in the maximum value matrix of the maximum value x maxl , so the segmentation point s l is expressed by the formula:

[0062]

[0063] where ρ represents the number of segmentation points

[0064] (3.2) Independently fit each segment based on the segmentation point;

[0065] In this embodiment, according to the abnormal signal matrix X, the start and end intervals [x 11 , x WT of the abnormal signal can be divided into ρ signal interval segments, which are sequentially denoted as: [x 11 , s1), [s1, s2), …, [s l , s l+1 ), …, [s ρ-1 , x WT ;

[0066] Use the least squares method to fit each signal interval segment respectively, and the fitting formula is as follows:

[0067]

[0068] where x is the fitting variable, a0, a τ are the polynomial coefficients to be fitted, and n l is the fitting order corresponding to l signal interval segments;

[0069] Substitute any sampling point x ij in the lth signal interval segment into the above fitting formula to obtain the fitting value P ij corresponding to x l (x ij );

[0070] Calculate the residual r ij of the sampling point x ij : rij = x ij - P l (x ij )

[0071] Calculate the sum of squared differences RSS of all sampling points in the l-th signal interval segment l ;

[0072]

[0073] For RSS l Take the partial derivative and set the derivative to 0 to obtain the polynomial coefficient a τ ; Then by incrementing the order n l Find the order n corresponding to the minimum of RSS l ; l ;

[0074] Traverse each signal interval segment to determine the minimum order n of the fitting for each segment l and the polynomial coefficient a τ , and finally construct the piecewise polynomial fitting model P(x) of the abnormal signal;

[0075]

[0076] (3.3) Based on the fitting values of each sampling point within each segment of the piecewise polynomial fitting model P(x), replace the sampling values of each sampling point in the abnormal signal X with the fitting values to obtain the reference template matrix X of the abnormal signal ref ;

[0077]

[0078] Among them, represents the fitting value of the sampling point x ij ;

[0079] (4) Abnormal capture of the signal to be measured;

[0080] (4.1) Turn on the abnormal capture function of the digital oscilloscope, then connect the signal to be measured for real-time acquisition, and record the matrix of the signal to be measured collected at the m-th time as:

[0081]

[0082] In this embodiment, a triangular wave is randomly superimposed on the normal signal, each abnormal length is 256 points, and the number of abnormal signals is 25, so as to form the signal to be measured;

[0083] (4.2) Calculate the Manhattan distance D m between each corresponding element of the matrix C of the signal to be measured ref and the reference template matrix X ij ;

[0084]

[0085] (4.3), Statistically calculate the Manhattan distance D ij The number of elements greater than the distance threshold χ, and the total number is denoted as N d ;

[0086] In this embodiment, the distance threshold χ can be obtained according to the sum of squared residuals of the abnormal signal fitting and the noise threshold, and the specific calculation formula is:

[0087]

[0088] (4.4), Calculate the distance probability P d ;

[0089]

[0090] (4.5), Judge the distance probability P d Whether it is greater than the preset probability threshold δ. If P d > δ, the acquisition stops and the measured signal matrix C m is the detected abnormal signal; otherwise, go to step (4.6);

[0091] In this embodiment, the probability threshold δ satisfies: 0.8 < δ < 1;

[0092] (4.6), Let m = m + 1; remove the first element c m in C m,11 , and then supplement the next sampling point c m,WT at the end of c m,1T+1 , so as to obtain the measured signal matrix C m+1 collected for the (m + 1)th time;

[0093]

[0094] (4.7), Return to step (4.2) for the next round of iteration.

[0095] In this embodiment, after obtaining the abnormal position coordinates in the measured signal, the waveform of this section is displayed, as Figure 2 shown. Among them, in the measured signal with a length of 1 s, an abnormality is detected at the position of 2.505×10 -6 s to 2.515×10 -6 s, as shown in the left picture; the abnormal signal obtained after extraction is shown in the right picture.

[0096] Although the above description of the illustrative embodiments of the present invention is provided for the understanding of those skilled in the art of the present technology, it should be clear that the present invention is not limited to the scope of the specific embodiments. For those of ordinary skill in the art of the present technology, as long as various changes are within the spirit and scope of the present invention defined and determined by the appended claims, these changes are obvious, and all inventions and creations using the concept of the present invention are within the scope of protection.

Claims

1. An abnormal capture method for transient signals in a digital oscilloscope, characterized in that, It includes the following steps: (1) Extract the abnormal signals and system noise in the transient signal; (1.1) Turn on the abnormal detection mode of the digital oscilloscope, then connect the transient signal for abnormal detection. When an abnormality is detected in the transient signal, extract the abnormal part of the transient signal and store it in matrix form, denoted as the abnormal signal X; where x ij represents the value of the i-th sampling point collected at the j-th sampling moment of the abnormal signal, W is the number of sampling points collected at each sampling moment, and T represents the number of sampling moments; (1.2) When the digital oscilloscope collects the system noise without signal input, it is also stored in matrix form. Let the stored system noise be Among them, represents the value of the i-th sampling point collected by the system noise at the j-th sampling moment; (2) Extract the noise threshold ξ; Subtract the minimum value from the maximum value in the system noise to obtain the noise threshold ξ; The difference between the maximum value and the minimum value in the system noise is used as the noise threshold ξ; (3) Perform multi-segment adaptive signal fitting on the abnormal signal X to obtain a reference template matrix; (3.1) Determine the segmentation points according to the time-domain extreme value distribution characteristics of the abnormal signal; (3.2) Perform independent fitting on each segment based on the segmentation points; Divide the start and end interval [x 11 , x WT of the abnormal signal into ρ signal interval segments, which are successively denoted as: [x 11 , s1), [s1, s2), …, [s l , s l+1 ), …, [s ρ-1 , x WT ; Use the least squares method to fit each signal interval segment respectively, and the fitting formula is as follows: where x is the fitting variable, a0, a τ are the polynomial coefficients to be fitted, and n l is the fitting order corresponding to l signal interval segments; Construct a piecewise polynomial fitting model P(x) of the abnormal signal; (3.3) Based on the fitting values of each sampling point within each segment of the piecewise polynomial fitting model P(x), replace the sampling values of each sampling point in the abnormal signal X with the fitting values to obtain the reference template matrix X of the abnormal signal ref ; Among them, represents the fitting value of the sampling point x ij ; (4) Abnormal capture of the signal to be measured; (4.1) Turn on the abnormal capture function of the digital oscilloscope, then connect the signal to be measured for real-time acquisition, and record the matrix of the signal to be measured collected at the mth time as: (4.2) Calculate the matrix C of the signal to be measured m and the reference template matrix X ref to obtain the Manhattan distance D between each corresponding element ij ; (4.3), Statistical Manhattan distance D ij Count the number of elements greater than the distance threshold χ, and the total number is denoted as N d ; (4.4), Calculate the distance probability P d ; (4.5) Determine the distance probability P d whether it is greater than the preset probability threshold δ. If P d > δ, the acquisition stops and the measured signal matrix C m is output as the detected abnormal signal; otherwise, go to step (4.6); (4.6) Let m = m + 1; Remove the first element c m from C m,11 , and then supplement the next sampling point c m,WT at the end of c m,1T+1 to obtain the matrix C m+1 of the signal to be measured collected for the (m + 1)-th time; (4.7) Return to step (4.2) for the next round of iteration.

2. The abnormal capture method of transient signals in a digital oscilloscope according to claim 1, characterized in that The specific method for determining the segmentation points is as follows: (2.1.1) Extract the amplitudes of each sampling point in the abnormal signal X to form an amplitude matrix Y; Among them, y ij is the amplitude corresponding to x ij ; (2.1.2) Represent the amplitude matrix Y in the form of a serial vector according to the sampling time as: Y = [y 11 , …, y W1 ; y 12 , …, y W2 ; …; y 1T , …, y WT ​ (2.1.3), traverse each element y in the amplitude matrix Y ij , compare y ij with the magnitudes of the two adjacent elements on its left and right. If y ij is greater than both elements, record the amplitude y ij as the maximum value y maxk , and record the corresponding coordinate as x maxk ; Construct the maximum value matrix Y after the elements are traversed max ; where k is the maximum value number, k = 1, 2, …, K, and K is the number of maximum values; Traverse the maximum value matrix Y max For each maximum value in it, if two adjacent maximum values are equal and there is a minimum value between the two adjacent maximum values, then keep these two adjacent maximum values; otherwise, keep only any one of the two maximum values; (2.1.4) Traverse each element y in the amplitude matrix Y ij and compare the magnitudes of the two adjacent elements on the left and right of y ij . If y ij is less than both elements, record the amplitude y ij as the minimum value y minh , and record the corresponding coordinate as x minh ; Construct the minimum value matrix Y after the elements are traversed min ; where h is the minimum value number, h = 1, 2, …, H, and H is the number of minimum values; Traverse the minimum value matrix Y min For each minimum value in it, if two adjacent minimum values are equal and there is no maximum value between them, then keep any one of the two minimum values; otherwise, keep both minimum values; (2.1.5) Determine the segmentation points according to the coordinates of the first extreme point in the extreme value matrix; (2.1.5.1), taking min(K, H) as the benchmark, for the maximum value matrix Y max or the minimum value matrix Y min delete the elements after min(K, H) in it, and denote the remaining element numbers as l, l = 1, 2,..., L, L = min(K, H); (2.1.5.2), compare the coordinates x of the first maximum point max1 with the coordinates x of the first minimum point min1 . When x max1 > x min1 , go to step (2.1.5.3); when x max1 < x min1 , go to step (2.1.5.4); (2.1.5.3), with the coordinate x max1 , x min1 as the starting point, that is, initialize l = 1, and traverse each extreme point in turn to judge the maximum and minimum values in the rising trend of the abnormal signal: If it satisfies y maxl+1 -y maxl > 0 and 0 < y maxl -y minl+1 < ξ, then determine that y maxl is a noise point in the rising trend of the abnormal signal, and remove it from the maximum value matrix Y max , and at the same time remove y minl+1 from the minimum value matrix Y min ; After the judgment is completed, continue to judge the maximum and minimum values in the falling trend: If it satisfies y minl-1 -y minl > 0 and 0 < y maxl -y minl < ξ, then determine that y minl is a noise point in the falling trend of the abnormal signal, and remove it from the minimum value matrix Y min , and at the same time remove y maxl from the maximum value matrix Y max , and record the maximum value matrix after removal as the minimum value matrix is (2.1.5.4), with the coordinate x max1 , x min1 as the starting point, that is, initialize l = 1, and traverse each extreme point in turn to judge the maximum and minimum values in the rising trend of the abnormal signal: If it satisfies y maxl+1 - y maxl > 0 and 0 < y maxl - y minl < ξ, then determine that y maxl is a noise point in the rising trend of the abnormal signal, and remove it from the maximum value matrix Y max , and at the same time remove y minl from the minimum value matrix Y min ; After the judgment is completed, continue to judge the maximum and minimum values in the falling trend: If it satisfies y minl-1 - y minl > 0 and 0 < y maxl+1 - y minl < ξ, then determine that y minl is a noise point in the falling trend of the abnormal signal, and remove it from the minimum value matrix Y min , and at the same time remove y maxl from the maximum value matrix Y max , and record the maximum value matrix after removal as the minimum value matrix as (2.1.5.5) Obtain the segmentation points according to the maximum value matrix and the minimum value matrix ​ When x max1 <x min1 If l is odd, the segmentation point is the maximum value in the maximum value matrix The maximum value x maxl ; if l is even, the segmentation point is the minimum value in the minimum value matrix The minimum value x minl , so the segmentation point s l is expressed by the formula as: When x max1 > x min1 If l is odd, the segmentation point is the minimum value in the minimum value matrix x minl ; if l is even, the segmentation point is the maximum value in the maximum value matrix x maxl , so the segmentation point s l is expressed by the formula: where ρ represents the number of segmentation points.

3. The abnormal capture method of transient signals in a digital oscilloscope according to claim 1, characterized in that, The fitting order n l is obtained by the following method: Substitute any sampling point \(x\) in the \(l\)-th signal interval segment ij into the above fitting formula to obtain the fitting value \(P\) of \(x\) ij (\(x\) l ) ij ; Calculate the residual r of the sampling point x ij : r ij : r ij = x ij - P l (x ij ) Calculate the sum of squared differences RSS of all sampling points in the l-th signal interval segment l ; For RSS l Take the partial derivative and set the derivative to 0 to obtain the polynomial coefficient a τ ; Then by incrementing the order n l Find the order n corresponding to the minimum of RSS l l .​ 4. The abnormal capture method of transient signals in a digital oscilloscope according to claim 1, characterized in that, The calculation method of the distance threshold χ is as follows:

Images