Progressive change point detection method, device, equipment and medium

Through sliding window algorithm and FIR filter processing of Mann-Whitney statistics, the accuracy problem of estimation of asymptotic point position under high-frequency noise is solved, the detection accuracy is improved, and it is suitable for non-parametric offline detection.

CN118193943BActive Publication Date: 2024-11-26GUANGDONG UNIV OF TECH
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410196748.7
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-02-22
Publication Date
2024-11-26
Estimated Expiration
2044-02-22

AI Technical Summary

Technical Problem

In the prior art, it is difficult to accurately estimate the position of the asymptotic change point under high frequency noise, especially non-Gaussian noise, such as Cauchy noise.

Method used

The Mann-Whitney statistics are calculated using the sliding window algorithm and input it into the FIR filter for low-pass filtering. Combined with the preset variable point position detection algorithm, the signal delay amount is determined to estimate the progressive variable point in the time series data.

Benefits of technology

Accurate progressive change point detection under high-frequency noise and impulse noise is realized, and the detection accuracy is improved without knowing the digital characteristics or noise distribution of the signal in advance.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118193943B_ABST
    Figure CN118193943B_ABST
Patent Text Reader

Abstract

The present application relates to a method, device, equipment and medium for detecting a gradual change point, and the method includes: obtaining a signal with high-frequency noise, sliding forward along a signal sequence corresponding to the signal with high-frequency noise according to a sliding window of preset number and width based on a sliding window algorithm, and calculating a Man-Whitney statistic sequence once for each sliding; inputting the Man-Whitney statistic sequence into an FIR filter for low-pass filtering to obtain a filtered Man-Whitney statistic sequence; determining the signal delay of the FIR filter, using a preset change point position detection algorithm to calculate and determine the time point at which a gradual change point occurs in the time series data and the value corresponding to the time point according to the filtered Man-Whitney statistic sequence and the signal delay; and performing a change point position alarm on the time series data based on the time point at which a gradual change point occurs and the value corresponding to the time point. The present application can improve the accuracy of gradual change point detection under high-frequency noise.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of change point detection, and in particular to a gradual change point detection method, a corresponding device, an electronic device and a computer-readable storage medium. Background Art

[0002] With the development of science and technology, the research on gradient point detection has been extended from the initial quality control and management to application scenarios such as weather forecasting, changes in port receiving traffic in network traffic monitoring, and collecting people's behavioral data in life through sensors to reflect changes in human behavior and mental health. This patent studies data based on time series, so the definition of gradient point is the location where the given time series data changes and the moment when the change ends, including the time point and the value corresponding to the time point.

[0003] Change point detection research first appeared in the field of quality monitoring and management, and statistical process control plays an important role in quality monitoring and management fields such as industrial production lines, network security, medical systems, financial and economic systems. SPC control charts are direct tools for reflecting statistical process control, and the earliest control chart is the Shewhart control chart. When SPC control charts reflect changes in statistical process control, change point detection is required. The initial change point detection research also started from these control charts, and many control charts based on change point detection have been proposed one after another. Page (1954) proposed the CUSUM control chart based on the log-likelihood ratio statistic, and proposed different versions of the CUSUM control chart according to different scenarios. Champ CW. and Wooda WH. supplemented the operating rules of the Shewhart control chart and the CUSUM control chart based on their predecessors. CroS1er BR introduced the CUSUM control chart under multivariate variables and compared and analyzed the robustness of the combination of the Shewhart control chart and the CUSUM control chart. Roberts SV proposed the geometric moving average (GMA) control chart, Crowder SY designed the EWMA control chart based on the exponential moving average concept, and Lowry CA et al. proposed the EWMA control chart for multivariate variables and compared it with the CUSUM control chart for multivariate variables. Early control chart research basically believed that the sequence data used for the experiment obeyed a certain distribution, such as Gaussian distribution, T-distribution, exponential distribution, etc. The change point detection control chart based on this concept is called the parameter-based change detection method, or parameter method for short. When the specific distribution obeyed by the experimental sequence data and a series of parameters corresponding to the distribution are complete, the change point detection research based on the parameter method can complete the task of accurately determining whether there is a change point in the experimental data and giving the location of the change point in time.

[0004] At present, people can obtain more and more types of data samples. The change point detection research based on parameter methods is gradually not suitable for the sequence data that needs to be studied at this stage. The change point detection research of non-parametric methods that are not restricted by the distribution and parameters of sequence data has gradually attracted the attention of scholars. Pettit AN proposed a non-parametric detection algorithm for the mean mutation model based on the Mann-Whitney statistic. In addition, a fast algorithm for calculating the Mann-Whitney statistic was given according to the definition of rank. Hawkins DM et al. proposed a new non-parametric statistic for the t-test of two samples and designed a change point detection algorithm for statistical process control. They also proposed to use Monte Carlo experiments to fit the distribution of new statistics to obtain the control threshold for determining the change point. Zhou C.Zou c.ZhangY proposed the ZEWMA statistic for mean mutation by combining the EWMA control chart and the Mann-Whitney statistic. Hawkins DM and Qiqi Deng proposed a nonparametric HQK statistic based on the Pettitt version of the Mann-Whitney statistic and applied it to the online change point detection algorithm. ROSS proposed an online nonparametric change point detection control chart based on the Cramer-van-Mises and Kolmogormov-Smirov (KS) tests. From the nonparametric algorithms listed above, we can see that many algorithms are based on a statistic, such as the Mann-Whitney statistic, the CvM statistic and the KS statistic in the Cramer-van-Mises and Kolmogormov-Smirov tests, etc., and although they are nonparametric methods, some of them are still improved on the ideas of the previous classic parametric methods.

[0005] In addition, in mathematical statistics, the Mann-Whitney statistic is equivalent to AUC, because the mathematical expectation of the Mann-Whitney statistic is AUC, so AUC will be used to refer to the Mann-Whitney statistic below. For AUC, it refers to the area of ​​the curve enclosed by the true positive rate (TP) and the false positive rate (FP) in the confusion matrix. It represents the separation of random variables in the binary classification problem, and its value range is 0.5 to 1.

[0006] Although the above method can estimate the position of the asymptotic change point in non-parametric, that is, digital characteristics of unknown data, such as mathematical expectation, variance, etc., it cannot make a relatively accurate estimate of the position of the asymptotic change point under high-frequency noise, especially under non-Gaussian noise, such as Cauchy noise.

[0007] To sum up, in order to adapt to the problem that the prior art cannot accurately estimate the position of the asymptotic change point under high-frequency noise, especially non-Gaussian noise, such as Cauchy noise, the applicant has made corresponding explorations to solve this problem. Summary of the invention

[0008] The purpose of the present application is to solve the above-mentioned problem and to provide a gradual change point detection method, a corresponding device, an electronic device and a computer-readable storage medium.

[0009] In order to meet the various objectives of this application, this application adopts the following technical solutions:

[0010] A method for detecting a gradual change point is proposed to meet one of the purposes of the present application, comprising:

[0011] Acquire a signal with high-frequency noise, and forwardly slide a sliding window of a preset number and width along a signal sequence corresponding to the signal with high-frequency noise based on a sliding window algorithm, and calculate a Mann-Whitney statistic sequence once each sliding;

[0012] Inputting the Mann-Whitney statistic sequence into a FIR filter for low-pass filtering to obtain a filtered Mann-Whitney statistic sequence;

[0013] Determine the signal delay of the FIR filter, and use a preset change point position detection algorithm to calculate the time point at which the gradual change point occurs in the time series data and the value corresponding to the time point according to the filtered Mann-Whitney statistic sequence and the signal delay;

[0014] Based on the time point at which the gradual change point occurs and the value corresponding to the time point, a change point position alarm is performed on the time series data to complete the detection of the gradual change point in the time series data.

[0015] Optionally, based on a sliding window algorithm, sliding forward along a signal sequence corresponding to the signal with high-frequency noise according to a sliding window of preset number and width, and calculating a Mann-Whitney statistic sequence once each sliding, comprises:

[0016] The calculation formula of the Mann-Whitney statistic series is:

[0017]

[0018] Among them, X and Y represent two sliding windows, θ represents the Mann-Whitney statistic sequence, m and n are the window widths, m = n, I(·) is the indicator function, and we have: i >X j , then I(·)=1, if X i <X j , I(·)=0. Optionally, the step of inputting the Mann-Whitney statistic sequence into a FIR filter for low-pass filtering to obtain a filtered Mann-Whitney statistic sequence includes:

[0019] The key parameters of the FIR filter include one or more of the cutoff frequency, the filter order and the window function type.

[0020] Optionally, the step of determining the cutoff frequency, the filter order, and the window function includes:

[0021] Determine a Mann-Whitney statistic sequence expression in a noise-free state, and calculate and determine its power spectrum expression according to the Mann-Whitney statistic sequence expression;

[0022] Based on the power spectrum expression, calculate and determine the noise-free change point and the quantity accumulation diagram corresponding to the noise-free change point to confirm the range of the selected cutoff frequency;

[0023] The filter order ranges from 10 to 50, and the window function is a rectangular window.

[0024] Optionally, determining a Mann-Whitney statistic sequence expression in the absence of noise, and calculating and determining its power spectrum expression according to the Mann-Whitney statistic sequence expression, comprises:

[0025] Perform discrete Fourier transform on the Mann-Whitney statistic sequence expression, let X(ω) be the spectrum, S p (ω) is the power spectrum. According to the definition of discrete Fourier transform, we have:

[0026]

[0027] Among them, AUC represents the Mann-Whitney statistic sequence;

[0028] The power spectrum expression is:

[0029] S p (ω)=||X(ω)|| 2 .

[0030] Optionally, the step of determining the signal delay of the FIR filter, using a preset change point position detection algorithm to calculate and determine the time point at which the gradual change point occurs in the time series data and the value corresponding to the time point according to the filtered Mann-Whitney statistic sequence and the signal delay, includes:

[0031] The penalty item comparison result is determined by using the penalty item comparison algorithm to calculate the filtered Mann-Whitney statistic sequence;

[0032] The difference between the penalty item comparison result and the delay amount caused by the FIR filter is calculated and determined, and the estimated position of the asymptotic change point is determined according to the difference between the penalty item comparison result and the delay amount caused by the FIR filter, the compensation amount and the offset coefficient.

[0033] Optionally, the step of calculating and determining a difference between the penalty item comparison result and the delay amount caused by the FIR filter, and determining an estimated position of the asymptotic change point according to the difference between the penalty item comparison result and the delay amount caused by the FIR filter, the compensation amount, and the offset coefficient, comprises:

[0034] The estimated location of the asymptotic change point is calculated as:

[0035]

[0036] Among them, P h , P e represents the estimated position of the gradual change point, idx1, idx2 represent the penalty comparison results, v delay represents the delay caused by the FIR filter, λ represents the compensation amount, and v offset Indicates the offset factor.

[0037] A device for detecting a gradual change point provided for another purpose of the present application includes:

[0038] A signal acquisition module, configured to acquire a signal with high-frequency noise, and forwardly slide along a signal sequence corresponding to the signal with high-frequency noise according to a sliding window of preset number and width based on a sliding window algorithm, and calculate a Mann-Whitney statistic sequence once each sliding;

[0039] A low-pass filtering module is configured to input the Mann-Whitney statistic sequence into a FIR filter for low-pass filtering to obtain a filtered Mann-Whitney statistic sequence;

[0040] A gradual change point detection module is configured to determine the signal delay of the FIR filter, and to calculate and determine the time point at which the gradual change point occurs in the time series data and the value corresponding to the time point based on the filtered Mann-Whitney statistic sequence and the signal delay using a preset change point position detection algorithm;

[0041] The change point position detection module is configured to perform a change point position alarm on the time series data based on the time point at which the gradual change point occurs and the value corresponding to the time point, so as to complete the detection of the gradual change point in the time series data.

[0042] An electronic device provided to meet another purpose of the present application includes a central processing unit and a memory, wherein the central processing unit is used to call and run a computer program stored in the memory to execute the steps of the gradual change point detection method described in the present application.

[0043] A computer-readable storage medium is provided to meet another purpose of the present application, which stores a computer program implemented according to the gradual change point detection method in the form of computer-readable instructions. When the computer program is called and executed by a computer, the steps included in the corresponding method are executed.

[0044] Compared with the prior art, the present application aims to solve the problem that the prior art cannot accurately estimate the position of the asymptotic change point under high-frequency noise, especially under non-Gaussian noise, such as Cauchy noise. The present application includes but is not limited to the following beneficial effects:

[0045] First, the gradual change point detection method of the present application is a non-parametric offline gradual change point detection method. The user does not need to know the digital characteristics of the change point in advance, such as the mean, variance, covariance, etc., nor does it need to know in advance what distribution the noise of the signal obeys. The user only needs to take the one-dimensional signal as input, and the system will perform data processing and calculation as described above, and estimate the change point position as the output result;

[0046] Secondly, the gradual change point detection method of the present application can use the Mann-Whitney statistic of the original signal sequence as the input signal under high-frequency noise and impulse noise, and design an FIR filter to filter the noise, so as to better estimate the location where the gradual change point occurs, thereby improving the accuracy of the gradual change point detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

[0048] Figure 1 Schematic diagram of the process of the progressive change point detection method in the embodiment of the present application;

[0049] Figure 2 This is a schematic diagram of dividing the original signal without noise into three segments in an embodiment of the present application;

[0050] Figure 3 Schematic diagram of the AUC image in the embodiment of the present application;

[0051] Figure 4 A schematic diagram of determining the total spectrum in an embodiment of the present application;

[0052] Figure 5 This is a schematic diagram for verifying the derivation results in the embodiments of the present application;

[0053] Figure 6 It is a schematic diagram of an energy accumulation diagram in the case of a change point without noise and a change point with noise;

[0054] Figures 7 to 12A schematic diagram of determining a suitable cutoff frequency and window function type in an embodiment of the present application;

[0055] Fig.13 This is a schematic diagram of the gradual change point detection after adding a filter in the embodiment of the present application;

[0056] Fig.14 Schematic diagram of progressive change point detection under high frequency noise in an embodiment of the present application;

[0057] Fig.15 This is a principle block diagram of a progressive change point detection device in an embodiment of the present application;

[0058] Fig.16 It is a schematic diagram of the structure of the computer device in the embodiment of the present application. DETAILED DESCRIPTION

[0059] The embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present application, and cannot be interpreted as limiting the present application.

[0060] It will be understood by those skilled in the art that, unless expressly stated, the singular forms "one", "said", and "the" used herein may also include plural forms. It should be further understood that the term "comprising" used in the specification of the present application refers to the presence of the features, integers, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or groups thereof. It should be understood that when we refer to an element as being "connected" or "coupled" to another element, it may be directly connected or coupled to the other element, or there may be an intermediate element. In addition, the "connection" or "coupling" used herein may include wireless connection or wireless coupling. The term "and / or" used herein includes all or any unit and all combinations of one or more associated listed items.

[0061] It will be understood by those skilled in the art that, unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as those generally understood by those skilled in the art to which this application belongs. It should also be understood that terms such as those defined in general dictionaries should be understood to have meanings consistent with those in the context of the prior art, and will not be interpreted with idealized or overly formal meanings unless specifically defined as here.

[0062] It will be understood by those skilled in the art that the "client", "terminal" and "terminal device" used herein include both devices with wireless signal receivers, which are devices with only wireless signal receivers without transmission capabilities, and devices with receiving and transmitting hardware, which are devices with receiving and transmitting hardware capable of two-way communication on a two-way communication link. Such devices may include: cellular or other communication devices such as personal computers, tablet computers, which have single-line displays or multi-line displays or cellular or other communication devices without multi-line displays; PCS (Personal Communications Service, personal communication system), which can combine voice, data processing, fax and / or data communication capabilities; PDA (Personal Digital Assistant, personal digital assistant), which may include a radio frequency receiver, pager, Internet / intranet access, web browser, notepad, calendar and / or GPS (Global Positioning System, global positioning system) receiver; conventional laptop and / or palmtop computers or other devices, which have and / or include a conventional laptop and / or palmtop computer or other device with and / or including a radio frequency receiver. The "client", "terminal" and "terminal device" used herein may be portable, transportable, installed in a vehicle (air, sea and / or land), or suitable for and / or configured to run locally, and / or in a distributed form, at any other location on the earth and / or in space. The "client", "terminal" and "terminal device" used herein may also be a communication terminal, an Internet terminal, a music / video playing terminal, for example, a PDA, a MID (Mobile Internet Device) and / or a mobile phone with a music / video playing function, or a smart TV, a set-top box and other devices.

[0063] The hardware referred to by the names such as "server", "client", and "service node" in this application is essentially an electronic device with the equivalent capabilities of a personal computer. It is a hardware device with the necessary components revealed by the von Neumann principle, such as a central processing unit (including an arithmetic unit and a controller), a memory, an input device, and an output device. The computer program is stored in its memory, and the central processing unit calls the program stored in the external memory into the internal memory for execution, executes the instructions in the program, and interacts with the input and output devices to complete specific functions.

[0064] It should be pointed out that the concept of "server" referred to in this application can also be extended to the case of server clusters. According to the network deployment principle understood by those skilled in the art, the servers should be logically divided. In physical space, these servers can be independent of each other but can be called through interfaces, or integrated into a physical computer or a set of computer clusters. Those skilled in the art should understand this flexibility, and should not use it to restrict the implementation of the network deployment method of this application.

[0065] Unless expressly specified, one or more technical features of the present application can be deployed on a server for implementation and accessed by a client through a remote call to obtain an online service interface provided by the server, or can be directly deployed and run on a client for access.

[0066] The neural network models referenced or may be referenced in this application, unless expressly specified, can be deployed on a remote server and remotely called on the client, or can be deployed and directly called on a client with sufficient device capabilities. In some embodiments, when it runs on the client, its corresponding intelligence can be obtained through transfer learning to reduce the requirements for the client's hardware operating resources and avoid excessive occupation of the client's hardware operating resources.

[0067] Unless explicitly specified, the various data involved in this application can be stored remotely on a server or on a local terminal device, as long as it is suitable for being called by the technical solution of this application.

[0068] Those skilled in the art should be aware that, although the various methods of the present application are described based on the same concept and thus present commonality to each other, unless otherwise specified, these methods can be independently executed. Similarly, for each embodiment disclosed in the present application, they are all proposed based on the same inventive concept, therefore, concepts with the same expression, and concepts that are appropriately changed for convenience despite different expressions, should be understood as equivalent.

[0069] Unless the mutually exclusive relationship between the embodiments to be disclosed in this application is explicitly stated, the relevant technical features involved in each embodiment can be cross-combined to flexibly construct a new embodiment, as long as such combination does not deviate from the creative spirit of this application and can meet the needs of the prior art or solve certain deficiencies in the prior art. Those skilled in the art should be aware of this flexibility.

[0070] In mathematical statistics, the Mann-Whitney statistic is equivalent to AUC, because the mathematical expectation of the Mann-Whitney statistic is AUC, and AUC represents the Mann-Whitney statistic sequence, so AUC will be used to refer to the Mann-Whitney statistic below. For AUC, it refers to the area of ​​the curve enclosed by the true positive rate (TP) and the false positive rate (FP) in the confusion matrix. It represents the separation of random variables in the binary classification problem, and its value range is 0.5 to 1.

[0071] Based on the above exemplary scenarios, please refer to Figure 1 In one embodiment, the gradual change point detection method of the present application includes:

[0072] Step S10, obtaining a signal with high-frequency noise, sliding a sliding window with a preset number and width along a signal sequence corresponding to the signal with high-frequency noise based on a sliding window algorithm, and calculating a Mann-Whitney statistic sequence once each sliding;

[0073] The computer terminal device can obtain a signal with high-frequency noise, and based on the sliding window algorithm, slide the signal sequence corresponding to the signal with high-frequency noise forward according to the sliding window of preset number and width, calculate the Mann-Whitney statistic sequence once for each sliding, and calculate the AUC of the signal with high-frequency noise. The AUC represents the degree of separability between two random variables X and Y, and the relationship is as follows:

[0074] AUC=Pr(X>Y),

[0075] The specific calculation uses the idea of ​​sliding window calculation. There are two windows X and Y, which slide forward along the signal sequence. The AUC value is calculated once each time it slides. If θ is used to represent AUC, its definition is as follows:

[0076]

[0077] Among them, X and Y represent two sliding windows, θ represents the Mann-Whitney statistic sequence, m and n are the window widths, m = n, I(·) is the indicator function, and we have: i >X j , then I(·)=1, if X i <X j , I(·)=0.

[0078] Then, the calculated AUC is passed to the FIR filter for filtering. This step involves the design of the FIR filter. There are three key parameters, namely the cutoff frequency, the filter order, and the window function type.

[0079] Step S20, inputting the Mann-Whitney statistic sequence into a FIR filter for low-pass filtering to obtain a filtered Mann-Whitney statistic sequence;

[0080] After the Mann-Whitney statistic sequence is calculated, the Mann-Whitney statistic sequence is input into a FIR filter for low-pass filtering to obtain a filtered Mann-Whitney statistic sequence.

[0081] The step of inputting the Mann-Whitney statistic sequence into a FIR filter for low-pass filtering to obtain a filtered Mann-Whitney statistic sequence comprises:

[0082] The key parameters of the FIR filter include one or more of the cutoff frequency, the filter order and the window function type.

[0083] Further, the step of determining the cut-off frequency, the filter order and the window function includes:

[0084] Step S201, determining a Mann-Whitney statistic sequence expression in a noise-free state, and calculating and determining a power spectrum expression thereof according to the Mann-Whitney statistic sequence expression;

[0085] Step S203, calculating and determining the noise-free change point and the quantity accumulation diagram corresponding to the noise-free change point based on the power spectrum expression, so as to confirm the range of the selected cutoff frequency;

[0086] The filter order ranges from 10 to 50, and the window function is a rectangular window.

[0087] Furthermore, determining a Mann-Whitney statistic sequence expression in the absence of noise, and calculating and determining its power spectrum expression according to the Mann-Whitney statistic sequence expression, comprises:

[0088] Perform discrete Fourier transform on the Mann-Whitney statistic sequence expression, let X(ω) be the spectrum, S p (ω) is the power spectrum. According to the definition of discrete Fourier transform, we have:

[0089]

[0090] Among them, AUC represents the Mann-Whitney statistic sequence;

[0091] The power spectrum expression is:

[0092] S p (ω)=||X(ω)|| 2 .

[0093] Here is the derivation process, divided into six steps:

[0094] For the first step, see Figure 2 , first derive the mathematical expression of AUC in the noise-free state, divide the original signal in the noise-free state into three segments, with lengths of L1, L2 and L3, and L1, L2, L3∈Z + ,like Figure 2 shown.

[0095] See also Figure 3 , the calculated AUC image is an isosceles trapezoid. For the convenience of calculation, the image corresponding to the calculated AUC value is translated to the left so that the image is symmetrical about the y-axis, as shown in Figure 3 As shown:

[0096] According to the symmetry of the graph, the coordinates of special points can be obtained, and the expression of AUC can be derived using the two-point equation (as shown below):

[0097]

[0098] The derived AUC expression is expressed as a piecewise function as follows:

[0099]

[0100] In the second step, based on the AUC expression derived in the first step, its power spectrum expression is derived. This application regards AUC as the signal input to the FIR filter, and first derives its spectrum expression. On this basis, the power spectrum can be obtained by taking the square of the spectrum modulus. Then, the AUC expression is first subjected to discrete Fourier transform, and X(ω) is set as the spectrum, S p (ω) is the power spectrum. According to the definition of discrete Fourier transform, we have:

[0101]

[0102] Then the power spectrum is:

[0103] S p (ω)=||X(ω)|| 2 ;

[0104] See also Figure 4 , because the AUC image is an isosceles trapezoid, using its graphic characteristics, first extend the two waistlines of the trapezoid and intersect them on the y-axis, that is, the total spectrum is equal to the two waistlines plus the spectrum of their respective extension lines minus the spectrum of their respective extension lines, and finally add the spectrum corresponding to the upper base,

[0105] The following is a detailed derivation process, which divides the AUC expression into several parts for derivation, that is,

[0106]

[0107]

[0108] The spectrum is:

[0109]

[0110] make

[0111]

[0112]

[0113]

[0114] So X(ω) is:

[0115] X(ω)=T1(ω)-T2(ω)+T3(ω),

[0116] Then derive T1(ω), T2(ω) and T3(ω) respectively,

[0117] make n3=n2, we have:

[0118]

[0119] make

[0120]

[0121]

[0122] because

[0123]

[0124]

[0125] Therefore

[0126]

[0127]

[0128] Similarly, for the derivation of the expression of T2(ω), let

[0129]

[0130]

[0131] have:

[0132]

[0133]

[0134] For T3(ω), we have:

[0135]

[0136] Then the expression of X(ω) is as follows, back-substituting for all n1, n2 and n3,

[0137]

[0138] Therefore, we have X(ω), which is expressed as follows:

[0139]

[0140] When ω→0, taking the above formula and taking the limit, we get:

[0141]

[0142] Calculate the above items separately, we have:

[0143]

[0144]

[0145]

[0146] Therefore

[0147]

[0148] The expression of X(ω) is expressed by piecewise function as follows:

[0149]

[0150] See also Figure 5 On the basis of the above formula, the correctness of the result derived from the formula is verified in the simulation experiment, that is, the AUC fast algorithm is used to calculate the AUC corresponding to the noise-free sequence with gradual change points, and then the spectrum is calculated using Fourier transform, and then the power spectrum is obtained by taking the square of the modulus, which is recorded as V(ω). On the other hand, based on the derived X(ω), the derived power spectrum can be obtained by taking the square of the modulus.

[0151] Let E(ω) be the error, we have:

[0152] E(ω)=V(ω)-X(ω);

[0153] The third step is to calculate the corresponding quantity accumulation diagrams for the two cases of noise-free change points and noise-free change points based on the power spectrum, and confirm the selection of the cutoff frequency ω c The energy accumulation of the power spectrum is defined as follows:

[0154]

[0155] The experimental results can be found in Figure 6 , the blue line is the energy accumulation diagram in the case of noise-free and change points, and the red line is the energy accumulation diagram in the case of noise-free and change points.

[0156] As shown by the green line, when the frequency ω is 0.1, the energy accumulation in the case of no noise and change point converges to 1, while the energy accumulation in the case of no noise and change point has not yet converged, that is, the cutoff frequency ω c Can be selected between 0 and 0.1;

[0157] Step 5: Based on step 4, select the filter order, which ranges from 10 to 50, and the window function type, and find the appropriate cutoff frequency W by calculating the corresponding signal-to-noise ratio. c and window function type;

[0158] The signal-to-noise ratio is defined as follows:

[0159]

[0160] Among them, E s is the energy of the signal, E N is the energy of the noise, calculated as follows:

[0161]

[0162] Where n is the signal length, e i is each signal value;

[0163] See also Figures 7 to 12 , it can be seen that when the cutoff frequency ω c In the range of 0 to 0.01, the signal-to-noise ratio decays very slowly, so ω can be arbitrarily selected within this range. c On the other hand, when the filter order is arbitrarily selected, the signal-to-noise ratio of the rectangular window is higher than that of the other window functions, so the rectangular window should be selected as the window function of the FIR filter.

[0164] Step S30, determining the signal delay of the FIR filter, using a preset change point position detection algorithm to calculate and determine the time point at which the gradual change point occurs in the time series data and the value corresponding to the time point according to the filtered Mann-Whitney statistic sequence and the signal delay;

[0165] After inputting the Mann-Whitney statistic sequence into an FIR filter for low-pass filtering, determining the signal delay of the FIR filter, and using a preset change point position detection algorithm to calculate and determine the time point at which the gradual change point in the time series data occurs and the value corresponding to the time point according to the filtered Mann-Whitney statistic sequence and the signal delay;

[0166] Further, the signal delay of the FIR filter is determined, and a preset change point position detection algorithm is used to calculate the time point at which the gradual change point in the time series data occurs and the value corresponding to the time point according to the filtered Mann-Whitney statistic sequence and the signal delay, including:

[0167] Step S301, using a penalty item comparison algorithm to calculate and determine a penalty item comparison result for the filtered Mann-Whitney statistic sequence;

[0168] Step S303, calculate and determine the difference between the penalty item comparison result and the delay amount caused by the FIR filter, and determine the estimated position of the asymptotic change point according to the difference between the penalty item comparison result and the delay amount caused by the FIR filter, the compensation amount and the offset coefficient.

[0169] Further, the step of calculating and determining the difference between the penalty item comparison result and the delay amount caused by the FIR filter, and determining the estimated position of the asymptotic change point according to the difference between the penalty item comparison result and the delay amount caused by the FIR filter, the compensation amount and the offset coefficient, comprises:

[0170] The estimated location of the asymptotic change point is calculated as:

[0171]

[0172] Among them, P h , P e represents the estimated position of the gradual change point, idx1, idx2 represent the penalty comparison results, v delay represents the delay caused by the FIR filter, λ represents the compensation amount, and v offset Indicates the offset factor.

[0173] Specifically, the preset change point position detection algorithm is a penalty term comparison algorithm, which selects the filter order and estimates the change point position. The higher the filter order, the better the filtering effect, but the longer the operation time, so it should be selected in the range of 10 to 40; the penalty term comparison algorithm is used on the filtered AUC, and the result is recorded as idx1, idx2 minus the delay caused by the FIR filter, recorded as v delay , then add the compensation amount, denoted as λ, and the offset coefficient as v offset, according to the definition of the asymptotic change point, the estimated position of the asymptotic change point is obtained, as shown in the following formula:

[0174]

[0175] It is not difficult to understand that after obtaining the estimated position of the asymptotic change point, the time point at which the asymptotic change point occurs in the time series data and the value corresponding to the time point are also determined.

[0176] Among them, the delay caused by the FIR filter is half of the filter order, and the compensation amount is 0 to half the width of the sliding window, that is,

[0177]

[0178]

[0179] Step S40: Based on the time point at which the gradual change point occurs and the value corresponding to the time point, a change point position alarm is performed on the time series data to complete the detection of the gradual change point in the time series data.

[0180] After calculating and determining the time point at which a gradual change point occurs in the time series data and the value corresponding to the time point, a change point position alarm is performed on the time series data based on the time point at which the gradual change point occurs and the value corresponding to the time point to complete the detection of the gradual change point in the time series data.

[0181] In some embodiments, see Fig.13 ,After adding the filter, the detection accuracy is greatly improved compared to the case without ,the filter;

[0182] See also Fig.14 , under high-frequency noise, such as a signal sequence containing Gaussian noise, the higher the noise intensity, the technical solution of the present application still has a higher detection rate. The experimental results are as follows Fig.14 As shown, the visible offset coefficient v offset The optimum is obtained at 0.1.

[0183] It can be seen from the above embodiments that, compared with the prior art, the present application aims at the problem that the prior art cannot accurately estimate the position of the asymptotic change point under high-frequency noise, especially under non-Gaussian noise, such as Cauchy noise, and the like. The present application includes but is not limited to the following beneficial effects:

[0184] First, the gradual change point detection method of the present application is a non-parametric offline gradual change point detection method. The user does not need to know the digital characteristics of the change point in advance, such as the mean, variance, covariance, etc., nor does it need to know in advance what distribution the noise of the signal obeys. The user only needs to take the one-dimensional signal as input, and the system will perform data processing and calculation as described above, and estimate the change point position as the output result;

[0185] Secondly, the gradual change point detection method of the present application can use the Mann-Whitney statistic of the original signal sequence as the input signal under high-frequency noise and impulse noise, and design an FIR filter to filter the noise, so as to better estimate the location where the gradual change point occurs, thereby improving the accuracy of the gradual change point detection.

[0186] See also Fig.15 A gradual change point detection device provided to meet one of the purposes of the present application includes a signal acquisition module 1100, a low-pass filtering module 1200, a gradual change point detection module 1300 and a change point position detection module 1400. Among them, the signal acquisition module 1100 is configured to acquire a signal with high-frequency noise, and based on the sliding window algorithm, slides forward along the signal sequence corresponding to the signal with high-frequency noise according to a sliding window of preset number and width, and calculates a Mann-Whitney statistic sequence once each sliding; the low-pass filtering module 1200 is configured to input the Mann-Whitney statistic sequence into an FIR filter for low-pass filtering to obtain a filtered Mann-Whitney statistic sequence; the gradual change point detection module 1300 is configured to determine the signal delay of the FIR filter, and use a preset change point position detection algorithm to calculate and determine the time point at which the gradual change point occurs in the time series data and the value corresponding to the time point according to the filtered Mann-Whitney statistic sequence and the signal delay; the change point position detection module 1400 is configured to perform a change point position alarm on the time series data based on the time point at which the gradual change point occurs and the value corresponding to the time point, so as to complete the detection of the gradual change point in the time series data.

[0187] Based on any embodiment of this application, please refer to Fig.16 Another embodiment of the present application further provides an electronic device, which can be implemented by a computer device, such as Fig.16As shown, a schematic diagram of the internal structure of a computer device. The computer device includes a processor, a computer-readable storage medium, a memory, and a network interface connected via a system bus. Among them, the computer-readable storage medium of the computer device stores an operating system, a database, and computer-readable instructions. The database may store a control information sequence. When the computer-readable instructions are executed by the processor, the processor can implement a gradual change point detection method. The processor of the computer device is used to provide computing and control capabilities to support the operation of the entire computer device. The memory of the computer device may store computer-readable instructions. When the computer-readable instructions are executed by the processor, the processor can execute the gradual change point detection method of the present application. The network interface of the computer device is used to connect and communicate with a terminal. Those skilled in the art can understand that Fig.16 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0188] In this embodiment, the processor is used to execute Fig.15 The memory stores the program code and various data required to execute the above modules or submodules. The network interface is used to transmit data between user terminals or servers. The memory in this embodiment stores the program code and data required to execute all modules / submodules in the progressive change point detection device of the present application, and the server can call the program code and data of the server to execute the functions of all submodules.

[0189] The present application also provides a storage medium storing computer-readable instructions. When the computer-readable instructions are executed by one or more processors, the one or more processors execute the steps of the gradual change point detection method described in any embodiment of the present application.

[0190] The present application also provides a computer program product, including a computer program / instruction, which, when executed by one or more processors, implements the steps of the gradual change point detection method described in any embodiment of the present application.

[0191] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments of the present application can be implemented by instructing the relevant hardware through a computer program, and the computer program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, the aforementioned storage medium can be a computer-readable storage medium such as a disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).

[0192] The above description is only a partial implementation method of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

[0193] In summary, the asymptotic change point detection method of the present application can use the Mann-Whitney statistic of the original signal sequence as the input signal under high-frequency noise and impulse noise, and design an FIR filter to filter the noise, so as to better estimate the location where the asymptotic change point occurs, thereby improving the accuracy of asymptotic change point detection.

Claims

1. A method for detecting a gradual change point, characterized in that: include: Acquire a signal with high-frequency noise, and forwardly slide a sliding window of a preset number and width along a signal sequence corresponding to the signal with high-frequency noise based on a sliding window algorithm, and calculate a Mann-Whitney statistic sequence once each sliding; Inputting the Mann-Whitney statistic sequence into a FIR filter for low-pass filtering to obtain a filtered Mann-Whitney statistic sequence; Determine the signal delay of the FIR filter, and use a preset change point position detection algorithm to calculate the time point at which the gradual change point occurs in the time series data and the value corresponding to the time point according to the filtered Mann-Whitney statistic sequence and the signal delay, wherein the change point position detection algorithm is a penalty term comparison algorithm; The step of determining the signal delay of the FIR filter, using a preset change point position detection algorithm to calculate and determine the time point at which the gradual change point occurs in the time series data and the value corresponding to the time point according to the filtered Mann-Whitney statistic sequence and the signal delay, includes: The penalty item comparison result is determined by using the penalty item comparison algorithm to calculate the filtered Mann-Whitney statistic sequence; Calculate and determine the difference between the penalty item comparison result and the delay amount caused by the FIR filter, and determine the estimated position of the asymptotic change point according to the difference between the penalty item comparison result and the delay amount caused by the FIR filter, the compensation amount, and the offset coefficient; The step of calculating and determining the difference between the penalty item comparison result and the delay amount caused by the FIR filter, and determining the estimated position of the asymptotic change point according to the difference between the penalty item comparison result and the delay amount caused by the FIR filter, the compensation amount and the offset coefficient, comprises: The estimated location of the asymptotic change point is calculated as: p h =1-v delay +λ p e =idx2-v delay -l, Among them, P h , P e represents the estimated position of the gradual change point, idx1, idx2 represent the penalty comparison results, v delay represents the delay caused by the FIR filter, λ represents the compensation amount, and v offset represents the offset coefficient; Among them, the delay caused by the FIR filter is half of the filter order, and the expression of the delay caused by the FIR filter is: Wherein, N is the filter order, and the value range of N is between 10 and 40; The compensation amount is 0 to half the width of the sliding window, and the expression of the compensation amount is: λ=v offset *m, Where m is the width of the sliding window, v offset The value range of is 0 to one-half; Based on the time point at which the gradual change point occurs and the value corresponding to the time point, a change point position alarm is performed on the time series data to complete the detection of the gradual change point in the time series data.

2. The method for detecting gradual change points according to claim 1, characterized in that: The step of sliding forward along the signal sequence corresponding to the signal with high-frequency noise according to a preset number of sliding windows based on a sliding window algorithm, and calculating the Mann-Whitney statistic sequence once each sliding, comprises: The calculation formula of the Mann-Whitney statistic series is: Among them, X and Y represent two sliding windows, θ represents the Mann-Whitney statistic sequence, m and n are the window widths, m = n, I(·) is the indicator function, and we have: i >X j , then I(·)=1, if X i <X j , I(·)=0.

3. The method for detecting gradual change points according to claim 1, characterized in that: The step of inputting the Mann-Whitney statistic sequence into a FIR filter for low-pass filtering to obtain a filtered Mann-Whitney statistic sequence comprises: The key parameters of the FIR filter include one or more of the cutoff frequency, the filter order and the window function type.

4. The method for detecting gradual change points according to claim 3, characterized in that: The steps to determine the cutoff frequency, filter order, and window function include: Determine a Mann-Whitney statistic sequence expression in a noise-free state, and calculate and determine its power spectrum expression according to the Mann-Whitney statistic sequence expression; Based on the power spectrum expression, calculate and determine the noise-free change point and the quantity accumulation diagram corresponding to the noise-free change point to confirm the range of the selected cutoff frequency; The filter order ranges from 10 to 50, and the window function is a rectangular window.

5. The method for detecting gradual change points according to claim 4, characterized in that: The step of determining a Mann-Whitney statistic sequence expression in a noise-free state and calculating and determining its power spectrum expression according to the Mann-Whitney statistic sequence expression comprises: Perform discrete Fourier transform on the Mann-Whitney statistic sequence expression, let X(ω) be the spectrum, S p (ω) is the power spectrum. According to the definition of discrete Fourier transform, we have: Among them, AUC represents the Mann-Whitney statistic sequence; The power spectrum expression is: S p (ω)=||X(ω)|| 2 。 6. A device for detecting a gradual change point, characterized in that: include: A signal acquisition module, configured to acquire a signal with high-frequency noise, and forwardly slide along a signal sequence corresponding to the signal with high-frequency noise according to a sliding window of preset number and width based on a sliding window algorithm, and calculate a Mann-Whitney statistic sequence once each sliding; A low-pass filtering module is configured to input the Mann-Whitney statistic sequence into a FIR filter for low-pass filtering to obtain a filtered Mann-Whitney statistic sequence; a gradual change point detection module, configured to determine the signal delay of the FIR filter, and to calculate and determine the time point at which the gradual change point occurs in the time series data and the value corresponding to the time point according to the filtered Mann-Whitney statistic sequence and the signal delay using a preset change point position detection algorithm, wherein the change point position detection algorithm is a penalty term comparison algorithm; The step of determining the signal delay of the FIR filter, using a preset change point position detection algorithm to calculate and determine the time point at which the gradual change point occurs in the time series data and the value corresponding to the time point according to the filtered Mann-Whitney statistic sequence and the signal delay, includes: The penalty item comparison result is determined by using the penalty item comparison algorithm to calculate the filtered Mann-Whitney statistic sequence; Calculate and determine the difference between the penalty item comparison result and the delay amount caused by the FIR filter, and determine the estimated position of the asymptotic change point according to the difference between the penalty item comparison result and the delay amount caused by the FIR filter, the compensation amount, and the offset coefficient; The step of calculating and determining the difference between the penalty item comparison result and the delay amount caused by the FIR filter, and determining the estimated position of the asymptotic change point according to the difference between the penalty item comparison result and the delay amount caused by the FIR filter, the compensation amount and the offset coefficient, comprises: The estimated location of the asymptotic change point is calculated as: p h =1-v delay +λ p e =idx2-v delay -l, Among them, P h , P e represents the estimated position of the gradual change point, idx1 and idx2 represent the penalty comparison results, and v delay represents the delay caused by the FIR filter, λ represents the compensation amount, and v offset represents the offset coefficient; Among them, the delay caused by the FIR filter is half of the filter order, and the expression of the delay caused by the FIR filter is: Wherein, N is the filter order, and the value range of N is between 10 and 40; The compensation amount is 0 to half the width of the sliding window, and the expression of the compensation amount is: λ=v offset *m, Where m is the width of the sliding window, v offset The value range of is 0 to one-half; The change point position detection module is configured to perform a change point position alarm on the time series data based on the time point at which the gradual change point occurs and the value corresponding to the time point, so as to complete the detection of the gradual change point in the time series data.

7. An electronic device, comprising a central processing unit and a memory, characterized in that: The central processing unit is used to call and run the computer program stored in the memory to execute the steps of the method according to any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that: It stores a computer program implemented according to the method described in any one of claims 1 to 5 in the form of computer-readable instructions, and when the computer program is called and executed by a computer, the steps included in the corresponding method are executed.