A method, system, device, and storage medium for waveform comparison of signals from a deep fusion device.

Abstract

This invention discloses a waveform comparison method, system, device, and storage medium for signals from a deep fusion device. The method employs the DTW distance algorithm, calculating the minimum similarity distance between the standard waveform of the waveform recorder and the output waveform of the deep fusion device. Hierarchical clustering is then used to obtain a similarity score, resulting in three waveform similarity states: similar, suspected similar, and dissimilar. Furthermore, multiple waveform similarity evaluation indicators are preset for various influencing factors, and a judgment matrix is ​​constructed using the analytic hierarchy process to obtain the relative weights of each indicator. The relative weights are combined with the clustering results for deep fusion device waveform comparison, improving the matching accuracy of waveform similarity results and providing real-time and accurate reflection of the operational health status of the deep fusion device.

Description

Technical Field

[0001] This invention relates to a waveform comparison method, system, device, and storage medium for deep fusion equipment signals, belonging to the field of power distribution network technology. Background Technology

[0002] Currently, there are many methods for evaluating the health of distribution terminals. The literature (Gong Fangliang, Mou Longhua, Guo Wenming, Zhu Jiran, Leng Hua, Tang Haiguo, Gong Hanyang, Zhang Zhidan. Distribution Terminal Status Diagnosis Based on DS Evidence Theory [J]. Power System Protection and Control, 2018, 46(01):30-36.) uses highly accurate membership functions to determine the degree of membership of health indicators of distribution terminals, but does not consider the randomness of complex systems. The literature (Liang Dong, Qin Chaoyun, Yang Jing, Zhu Xiaoling, Zong Jin. Research on Multi-Dimensional Health Assessment Model of Secondary Equipment in Distribution Automation System [J]. Electrical Appliances and Energy Efficiency Management Technology, 2018(20):68-72.) addresses the health factor assessment problem of secondary equipment in distribution automation by selecting the equipment operation status, environmental conditions, and inspection status that affect the secondary equipment as assessment objects, but it cannot reflect the real-time health status of the terminal. The literature (Xing Xiaomin, Xu Hairui, Liao Mengke, Cao Xinhui. A comprehensive evaluation method for the health status of distribution terminals based on cloud model and DS evidence theory [J]. Power System Protection and Control, 2021, 49(13):72-81.) uses a cloud model that integrates the randomness and fuzziness information of objective things to solve the membership degree of the index. However, since the hardware status of the system itself in the working environment elements and the information of working environment factors contained in the evaluation index are generally not collected in the intelligent distribution system, this evaluation model cannot achieve the ideal goal in practical application. Therefore, there is an urgent need for a method of deep fusion of device waveform comparison to reflect the operating health status of deep fusion devices in real time. Summary of the Invention

[0003] The purpose of this invention is to overcome the shortcomings of the prior art and provide a waveform comparison method, system, device and storage medium for deep fusion device signals. By setting a similarity evaluation index and optimizing the waveform output of the deep fusion device after hierarchical clustering according to the analytic hierarchy process, the matching accuracy of waveform comparison similarity results is improved, and the operating health status of the deep fusion device is reflected in real time and accurately.

[0004] To achieve the above objectives, the present invention is implemented using the following technical solution:

[0005] In a first aspect, the present invention provides a waveform comparison method for deep fusion device signals, comprising:

[0006] Multiple sets of test samples are acquired, and a waveform similarity evaluation index is established; the test samples include multiple test waveforms output by the deep fusion device and standard waveforms output by the waveform recorder that correspond one-to-one with the test waveforms; the waveform similarity evaluation index is the interference factor when the deep fusion device outputs the test waveforms.

[0007] Based on the DTW algorithm, the minimum similarity distance of each group of test samples under the waveform similarity evaluation index is calculated;

[0008] Hierarchical clustering is performed on the minimum similarity distance of each group of test samples to obtain a similarity score;

[0009] The relative weights of the waveform similarity evaluation indicators are obtained by performing hierarchical analysis on the waveform similarity evaluation indicators.

[0010] The similarity score is optimized based on the relative weights to obtain the optimized score, and waveform comparison is performed based on the optimized score.

[0011] Furthermore, the waveform similarity evaluation index refers to the interference factors when the deep fusion device outputs the waveform to be tested, including vertical component factors and horizontal component factors. The vertical component factors include sampling rate, bandwidth and jitter, while the horizontal component factors include gain and noise.

[0012] Furthermore, the step of calculating the minimum similarity distance of each group of test samples under the waveform similarity evaluation index according to the DTW algorithm specifically includes:

[0013] Establish the second-order distance matrix D for each group of test samples;

[0014] Starting from the lower left corner of the second-order distance matrix D and ending at the upper right corner, a monotonic and continuous regular path W is established. n = (i, j), where the starting point w1 = (1, 1) and the ending point w k = (|X|, |Y|), i∈[1,|X|], j∈[1,|Y|], n∈[1,k], where i, j, n, k, |X|, and |Y| are all positive integers; where n is the number of steps and k is the normalized path W. n The length of X is the waveform to be measured, and Y is the standard waveform corresponding to the waveform to be measured. |X| and |Y| are the lengths of the waveform to be measured and the corresponding standard waveform, respectively. i and j are the components of the waveform to be measured X and the standard waveform in the normalized path after each step, respectively.

[0015] Based on the DTW algorithm and the normalized path W of the second-order distance matrix D. n The optimal path is obtained by calculating the second-order distance matrix D, using the following formula:

[0016]

[0017] In the formula, DTW represents the optimal path; K is the compensation coefficient;

[0018] Based on the optimal path, from the starting point to the ending point, the distance of each step under the optimal path is calculated using the first constraint. These distances are then accumulated to obtain the minimum similarity distance dist(X,Y) for each group of test samples. The formula for calculating the first constraint is:

[0019]

[0020] In the formula, g(i,j) is the similarity distance of the optimal path after each step; d(i,j) is the distance of each step.

[0021] Furthermore, the step of performing hierarchical clustering on the minimum similarity distance of each group of test samples to obtain a similarity score specifically includes:

[0022] The minimum similarity distance of each group of test samples is normalized to obtain the normalized distance of each group of test samples, and each normalized distance is set as each initial class;

[0023] Based on the normalized distance of each group of test samples, the inter-class distance between each pair of test samples is calculated using the sum of squared deviations method;

[0024] Merge two classes with small inter-class distance differences, and use the smaller inter-class distance between the two merged classes as the inter-class distance after merging. Construct a new class, and repeat the calculation of the inter-class distance between the new class and the initial class and the merging operation until the number of classes equals the preset number of classes, and obtain the cluster center.

[0025] The similarity score ω of the cluster centers is calculated based on the minimum similarity distance corresponding to the cluster centers. The formula for calculating the similarity score is as follows:

[0026]

[0027] In the formula, ω is the similarity score, and ctr is the normalized minimum similarity distance.

[0028] Furthermore, the number of cluster centers is two, and the smaller similarity score is ω1, and the larger similarity score is ω2.

[0029] Furthermore, the hierarchical analysis of the waveform similarity evaluation indicators yields their relative weights; specifically, this includes:

[0030] Based on waveform similarity evaluation indicators, a judgment matrix A is established to determine the degree of importance between each pair of waveform similarity evaluation indicators, and the consistency of judgment matrix A is verified. The degree of importance includes five levels from 1 to 5. In the formula, a αα Let α represent the importance of the αth similarity evaluation index relative to the αth similarity evaluation index, where α is the αth similarity evaluation index.

[0031] According to the geometric mean method, the relative weights of each row in the completed judgment matrix A are calculated using the following formula:

[0032]

[0033] In the formula, μ α α represents the relative weight of the α-th similarity evaluation index; p is the row number in judgment matrix A; q is the column number in judgment matrix A; t is the level of the similarity evaluation index.

[0034] Furthermore, the step of optimizing the similarity score based on the relative weights of the similarity evaluation indicators to obtain an optimized score, and then performing waveform comparison based on the optimized score, specifically includes:

[0035] The relative weight of each similarity evaluation index is multiplied by the smaller similarity score ω1 and summed to obtain the smaller optimization score ω1'. The larger optimization score ω2' is obtained by the same method.

[0036] Calculate the similarity score for all samples to be tested according to the aforementioned similarity calculation formula;

[0037] Based on the smaller optimization score ω1' and the larger optimization score ω2', the waveform similarity is divided into three categories. When the calculated similarity score of the waveform to be tested is between ω1' and ω2', the waveform is judged as suspected similar; when the similarity is between ω2' and 1, the waveform is judged as similar; when the similarity is between 0 and ω1', the waveform is judged as dissimilar.

[0038] Secondly, the present invention also provides a waveform comparison system for deep fusion device signals, comprising:

[0039] Data acquisition module: used to acquire multiple sets of test samples and establish waveform similarity evaluation index; the test samples include multiple test waveforms output by the deep fusion device and standard waveforms output by the waveform recorder that correspond one-to-one with the test waveforms; the waveform similarity evaluation index is the interference factor when the deep fusion device outputs the test waveforms;

[0040] Distance calculation module: used to calculate the minimum similarity distance of each group of test samples under the waveform similarity evaluation index according to the DTW algorithm;

[0041] Hierarchical clustering module: used to perform hierarchical clustering on the minimum similarity distance of each group of test samples to obtain a similarity score;

[0042] Hierarchical Analysis Module: Used to perform hierarchical analysis on waveform similarity evaluation indicators to obtain the relative weights of the waveform similarity evaluation indicators;

[0043] Waveform comparison module: used to optimize the similarity score according to the relative weight, obtain the optimized score, and perform waveform comparison based on the optimized score.

[0044] Thirdly, a waveform comparison device for deep fusion device signals includes a processor and a storage medium;

[0045] The storage medium is used to store instructions;

[0046] The processor is configured to operate according to the instructions to perform the steps of the method according to any of the foregoing.

[0047] Fourthly, a computer-readable storage medium having a computer program stored thereon that, when executed by a processor, implements the steps of any of the methods described above.

[0048] Compared with the prior art, the beneficial effects achieved by the present invention are as follows:

[0049] This invention compares the similarity scores of waveforms recorded by a deep fusion device and a waveform recorder, and performs waveform similarity analysis based on the similarity scores. Based on the advantages and existing problems of similarity analysis of waveform data, a dynamic time bending distance algorithm is introduced to solve the problems of inconsistent time intervals and durations in waveform sampling data. According to the health status of the power distribution terminal reflected by the waveform under different states, the relative weights of the influence of each similarity evaluation index on the waveform are calculated using hierarchical clustering and analytic hierarchy process (AHP). The similarity scores under the influence of multiple factors are optimized based on the relative weights, and the performance of the deep fusion device under test is evaluated, thereby improving the matching accuracy of waveform comparison similarity results. Attached Figure Description

[0050] Figure 1 This is a flowchart of the waveform comparison method provided in Embodiment 1 of the present invention;

[0051] Figure 2 This is a waveform similarity evaluation index classification diagram provided in Embodiment 1 of the present invention;

[0052] Figure 3 This is a flowchart of the hierarchical clustering method in Embodiment 1 of the present invention. Detailed Implementation

[0053] The present invention will be further described below with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and should not be used to limit the scope of protection of the present invention.

[0054] This invention provides a waveform comparison method for signals from deep fusion devices, such as... Figure 1 As shown, it specifically includes:

[0055] S1: Acquire multiple sets of test samples and establish waveform similarity evaluation index; the test samples include multiple test waveforms output by the deep fusion device and standard waveforms output by the waveform recorder that correspond one-to-one with the test waveforms; the waveform similarity evaluation index is the interference factor when the deep fusion device outputs the test waveform.

[0056] S2: Calculate the minimum similarity distance of each group of test samples under the waveform similarity evaluation index according to the DTW algorithm.

[0057] S3: Perform hierarchical clustering on the minimum similarity distance of each group of test samples to obtain a similarity score.

[0058] S4: Perform hierarchical analysis on the waveform similarity evaluation index to obtain the relative weights of the waveform similarity evaluation index.

[0059] S5: Optimize the similarity score based on the relative weight to obtain the optimized score, and then perform waveform comparison based on the optimized score.

[0060] This invention compares the similarity scores of waveforms recorded by a deep fusion device and a waveform recorder, and performs waveform similarity analysis based on the similarity scores. Based on the advantages and existing problems of similarity analysis of waveform data, it introduces the Dynamic Time Warping Distance (DTW) algorithm to solve the problems of inconsistent time intervals and durations in waveform sampling data. According to the health status of the power distribution terminal reflected by the waveform under different states, it calculates and sets the relative weights of each similarity evaluation index on the waveform using hierarchical clustering and analytic hierarchy process (AHP). Based on the relative weights, it optimizes the similarity scores under the influence of multiple factors, and performs performance evaluation on the deep fusion device under test, thereby improving the matching accuracy of waveform comparison similarity results.

[0061] Example 1:

[0062] This embodiment provides a specific example of the above waveform comparison method, specifically including three aspects: the establishment of waveform similarity evaluation index, the calculation of the impact of waveform similarity evaluation index on waveform comparison, and the judgment of waveform comparison.

[0063] Waveform acquisition is performed on the output signal of the deep fusion device to obtain multiple sets of waveforms to be tested. The waveform acquired by the waveform recorder at the same time is used as the standard waveform, and the waveforms to be tested and the standard waveforms acquired at the same time are used as a set of test samples. The acquisition of the waveforms to be tested and the acquisition of the standard waveforms are existing technologies and will not be described in detail here.

[0064] I. Establishment of Waveform Similarity Evaluation Indicators

[0065] like Figure 2As shown in this embodiment, the waveform similarity evaluation index mainly includes two levels. The first-level index includes the influence of horizontal components and the influence of vertical components. Under the first-level index, in order to more accurately calculate the influence of various factors in the waveform comparison process, it is further subdivided into second-level indexes. Among them, the influence of horizontal components includes gain and noise, and the influence of vertical components includes sampling rate, bandwidth, and jitter. By analyzing the similarity scores of waveforms under the influence of each second-level index, the accuracy of waveform comparison can be effectively improved.

[0066] II. Calculation of the Impact of Waveform Similarity Evaluation Indicators on Waveform Comparison

[0067] In this stage, the DTW algorithm is first used to calculate the dynamic time warping distance between the standard waveform and the waveform under test (i.e., the waveform of the interfered terminal) under different secondary indicators, forming a large amount of sample data. The dynamic time warping distance is the minimum similarity distance of the optimal path between the waveform under test and the standard waveform. For example, under the influence of noise, the minimum similarity distance of a set of samples under test is calculated. Specific methods include:

[0068] S21: Establish the second-order distance matrix D of the samples to be tested. In a set of samples to be tested, the waveform to be tested is set as X, and its corresponding standard waveform is set as Y. Their lengths are |X| and |Y|, respectively, both of which are positive integers.

[0069] S22: Starting from the lower left corner of the second-order distance matrix D and ending at the upper right corner, establish a monotonic and continuous regular path W. n = (i, j), where the starting point w1 = (1, 1) and the ending point w k = (|X|, |Y|), i∈[1,|X|], j∈[1,|Y|], n∈[1,k], where i, j, n, and k are all positive integers; where n is the number of steps and k is the normalized path W. n The length of X is the waveform to be measured, and Y is the standard waveform corresponding to the waveform to be measured. |X| and |Y| are the lengths of the waveform to be measured and the corresponding standard waveform, respectively. i and j are the components of the waveform to be measured X and the standard waveform in the normalized path after each step, respectively.

[0070] S23: The normalized path W based on the DTW algorithm and the second-order distance matrix D n The optimal path is obtained from the second-order distance matrix D, from the starting point w1 = (1, 1) to the ending point w. k The formula for calculating the optimal path for (|X|, |Y|) is:

[0071]

[0072] In the formula, DTW represents the optimal path; K is the compensation coefficient.

[0073] S24: Based on the optimal path, starting from the starting point w1 = (1, 1) and ending point w k The process ends with =(|X|,|Y|). The distance per step in the optimal path is calculated using the first constraint. These steps are then accumulated to obtain the minimum similarity distance dist(X,Y) for the group of samples to be tested. The formula for calculating the first constraint is:

[0074]

[0075] In the formula, g(i,j) is the similarity distance of the optimal path after each step; d(i,j) is the distance of each step. Under the first constraint, from the starting point w1 = (1,1) to the ending point w k The process ends at (|X|, |Y|), and the calculated g(i,j) at the endpoint is the minimum similarity distance dist(X,Y) of the test samples. By continuously processing each group of test samples using S21-S24, the minimum similarity distance of each group of test samples under noise influence can be obtained. Similarly, the minimum similarity distance of each group of test samples under gain, sampling rate, bandwidth, and jitter influence can be obtained, thus forming a large amount of sample data.

[0076] After obtaining a large amount of sample data, hierarchical clustering is performed on the minimum similarity distance of each group of test samples under the influence of noise to obtain a similarity score, such as... Figure 3 As shown, it specifically includes:

[0077] S31: Normalize the minimum similarity distance of each group of test samples to obtain the normalized distance of each group of test samples, and set each normalized distance as each initial class.

[0078] S32: Based on the normalized distance of each group of test samples, calculate the inter-class distance between each pair of test samples using the sum of squared deviations method.

[0079] S33: Merge two classes with small inter-class distance differences, and use the smaller inter-class distance between the two merged classes as the merged inter-class distance to construct a new class. Repeat the calculation of inter-class distances between the new class and the initial class, as well as the merging operation, until the number of classes equals the preset number of classes, thus obtaining the cluster centers. In this embodiment, the waveform comparison results need to be divided into three cases: similar, suspected similar, and dissimilar. Therefore, the preset number of classes is 3, containing two cluster centers.

[0080] S34: Calculate the similarity score ω of the cluster centers based on the minimum similarity distances corresponding to the cluster centers. The general formula for calculating the similarity score is:

[0081]

[0082] In the formula, ω represents the similarity score, and ctr is the normalized minimum similarity distance. Among the similarity scores of cluster centers, ω1 represents the smaller score and ω2 represents the larger score. This yields the similarity score of cluster centers under the influence of noise. Similarly, the similarity scores of cluster centers under the influence of gain, sampling rate, bandwidth, and jitter can be obtained respectively. The results are shown in Table 1. This similarity score can initially reflect the waveform comparison results under the influence of various waveform similarity evaluation indicators. However, in a complex environment where various influencing factors interact, this result cannot comprehensively evaluate the accurate comparison results between the waveform under test output by the deep fusion device and the standard waveform, and cannot be used as a standard for judging the global terminal health status. Therefore, further processing of the comprehensive influence between factors is needed.

[0083] Table 1. Similarity scores of cluster centers after hierarchical clustering based on the sum of squared deviations algorithm

[0084]

[0085] In this embodiment, hierarchical analysis is required to process all waveform similarity evaluation indicators to obtain the relative weight of each waveform similarity evaluation indicator. The specific method includes:

[0086] S41: Based on the waveform similarity evaluation index, establish a judgment matrix A for the degree of importance between each pair of waveform similarity evaluation indexes, and perform consistency verification on the judgment matrix A. The degree of importance includes five levels from 1 to 5. In the formula, a αα Let α represent the importance of the α-th similarity evaluation index relative to the α-th similarity evaluation index. The importance level is as follows: 1 indicates that the first and second similarity evaluation indices are equally important; 2 indicates that the first is slightly more important than the second; 3 indicates that the first is significantly more important than the second; 4 indicates that the first is strongly more important than the second; and 5 indicates that the first is extremely more important than the second. Note that the table contains reciprocals. For example, if the importance of the first similarity evaluation index relative to the second is 5, then the importance of the second similarity evaluation index relative to the first is 1 / 5. For instance, in the second-order judgment matrix A, α... 12 and a 21 They are reciprocals of each other.

[0087] S42: Calculate the relative weights of each row in the completed second-order judgment matrix A using the geometric mean method. The calculation formula is as follows:

[0088]

[0089] In the formula, μ α Let be the relative weight of the α-th similarity evaluation index; p be the row number in the second-order judgment matrix A; q be the column number in the second-order judgment matrix A; and t be the level of the similarity evaluation index. Thus, the relative weights of noise in the five similarity evaluation indices of this embodiment can be obtained. The relative weights of gain, sampling rate, bandwidth, and jitter can be obtained using the same method, and the results are shown in Table 2.

[0090] Table 2. Relative weights of each similarity evaluation index after analytic hierarchy process.

[0091]

[0092] III. Waveform Comparison Judgment

[0093] In this embodiment, after obtaining each similarity evaluation index and its corresponding two similarity scores, the similarity scores are optimized according to the relative weights of the similarity evaluation indexes to obtain optimized scores, and waveform comparison is performed based on the optimized scores; the specific method includes:

[0094] S51: Multiply the relative weight of each similarity evaluation index with the smaller similarity score ω1 and sum them up to obtain the smaller optimization score ω1'. The same method is used to obtain the larger optimization score ω2'.

[0095] S52: Calculate the similarity score of all samples to be tested according to the similarity calculation formula;

[0096] S53: Based on the smaller optimization score ω1' and the larger optimization score ω2', the waveform similarity is divided into three categories. When the calculated similarity score of the waveform to be tested is between ω1' and ω2', the waveform is judged as suspected similar; when the similarity is between ω2' and 1, the waveform is judged as similar; when the similarity is between 0 and ω1', the waveform is judged as dissimilar.

[0097] After optimizing the similarity scores of cluster centers after hierarchical clustering in Table 1 and the relative weights of various similarity evaluation indicators after hierarchical analysis in Table 2 using the above method, a similarity evaluation index suitable for global comparison of sampled waveform signals of deep fusion devices is obtained. The calculated ω1' is 67.85%, and the calculated ω2' is 84.81%. In practical applications, the similarity between the source signal waveform and the waveform acquired by the terminal is calculated. When the waveform similarity is between 67.85% and 84.81%, the waveform is judged as potentially similar; when the similarity is greater than 84.81%, the waveform is judged as similar; and when the similarity is less than 67.85%, the waveform is judged as dissimilar. In this way, the health status of deep fusion devices is assessed through waveform comparison.

[0098] Example 2:

[0099] This embodiment provides a waveform comparison system for deep fusion device signals, including:

[0100] Data acquisition module: used to acquire multiple sets of test samples and establish waveform similarity evaluation index; the test samples include multiple test waveforms output by the deep fusion device and standard waveforms output by the waveform recorder that correspond one-to-one with the test waveforms; the waveform similarity evaluation index is the interference factor when the deep fusion device outputs the test waveforms;

[0101] Distance calculation module: used to calculate the minimum similarity distance of each group of test samples under the waveform similarity evaluation index according to the DTW algorithm;

[0102] Hierarchical clustering module: used to perform hierarchical clustering on the minimum similarity distance of each group of test samples to obtain a similarity score;

[0103] Hierarchical Analysis Module: Used to perform hierarchical analysis on waveform similarity evaluation indicators to obtain the relative weights of the waveform similarity evaluation indicators;

[0104] Waveform comparison module: used to optimize the similarity score according to the relative weight, obtain the optimized score, and perform waveform comparison based on the optimized score.

[0105] The specific operating principles and methods of each module have been clearly explained in Example 1, and will not be elaborated further here.

[0106] Example 3:

[0107] This invention also provides a waveform comparison device for deep fusion device signals, including a processor and a storage medium;

[0108] The storage medium is used to store instructions;

[0109] The processor is configured to operate according to the instructions to perform the steps of the following method:

[0110] Multiple sets of test samples are acquired, and a waveform similarity evaluation index is established; the test samples include multiple test waveforms output by the deep fusion device and standard waveforms output by the waveform recorder that correspond one-to-one with the test waveforms; the waveform similarity evaluation index is the interference factor when the deep fusion device outputs the test waveforms.

[0111] Based on the DTW algorithm, the minimum similarity distance of each group of test samples under the waveform similarity evaluation index is calculated;

[0112] Hierarchical clustering is performed on the minimum similarity distance of each group of test samples to obtain a similarity score;

[0113] The relative weights of the waveform similarity evaluation indicators are obtained by performing hierarchical analysis on the waveform similarity evaluation indicators.

[0114] The similarity score is optimized based on the relative weights to obtain the optimized score, and waveform comparison is performed based on the optimized score.

[0115] The specific operating principles and methods have been clearly explained in Example 1, and will not be elaborated further here.

[0116] Example 4:

[0117] This invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the following method:

[0118] Multiple sets of test samples are acquired, and a waveform similarity evaluation index is established; the test samples include multiple test waveforms output by the deep fusion device and standard waveforms output by the waveform recorder that correspond one-to-one with the test waveforms; the waveform similarity evaluation index is the interference factor when the deep fusion device outputs the test waveforms.

[0119] Based on the DTW algorithm, the minimum similarity distance of each group of test samples under the waveform similarity evaluation index is calculated;

[0120] Hierarchical clustering is performed on the minimum similarity distance of each group of test samples to obtain a similarity score;

[0121] The relative weights of the waveform similarity evaluation indicators are obtained by performing hierarchical analysis on the waveform similarity evaluation indicators.

[0122] The similarity score is optimized based on the relative weights to obtain the optimized score, and waveform comparison is performed based on the optimized score.

[0123] The specific operating principles and methods have been clearly explained in Example 1, and will not be elaborated further here.

[0124] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0125] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0126] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0127] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0128] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A waveform comparison method for signals from a deep fusion device, characterized in that, include: Multiple sets of test samples are acquired, and a waveform similarity evaluation index is established; the test samples include multiple test waveforms output by the deep fusion device and standard waveforms output by the waveform recorder that correspond one-to-one with the test waveforms; the waveform similarity evaluation index is the interference factor when the deep fusion device outputs the test waveforms. Based on the DTW algorithm, the minimum similarity distance of each group of test samples under the waveform similarity evaluation index is calculated; Hierarchical clustering is performed on the minimum similarity distance of each group of test samples to obtain a similarity score; The relative weights of the waveform similarity evaluation indicators are obtained by performing hierarchical analysis on the waveform similarity evaluation indicators. The similarity score is optimized based on the relative weights to obtain the optimized score, and waveform comparison is performed based on the optimized score. Specifically, the step of performing hierarchical clustering on the minimum similarity distance of each group of test samples to obtain a similarity score includes: The minimum similarity distance of each group of test samples is normalized to obtain the normalized distance of each group of test samples, and each normalized distance is set as each initial class; Based on the normalized distance of each group of test samples, the inter-class distance between each pair of test samples is calculated using the sum of squared deviations method; Merge two classes with small inter-class distance differences, and use the smaller inter-class distance between the two merged classes as the inter-class distance after merging. Construct a new class, and repeat the calculation of the inter-class distance between the new class and the initial class and the merging operation until the number of classes equals the preset number of classes, and obtain the cluster center. Calculate the similarity score of the cluster centers based on the minimum similarity distance corresponding to the cluster centers. ω The formula for calculating the similarity score is: 0< ω <1， In the formula, ω To score the similarity, ctr This is the minimum similarity distance after normalization.

2. The waveform comparison method for deep fusion device signals according to claim 1, characterized in that, The waveform similarity evaluation index refers to the interference factors when the deep fusion device outputs the waveform to be tested, including vertical component factors and horizontal component factors. The vertical component factors include sampling rate, bandwidth and jitter, while the horizontal component factors include gain and noise.

3. The waveform comparison method for deep fusion device signals according to claim 1, characterized in that, The step involves calculating the minimum similarity distance for each group of test samples under the waveform similarity evaluation index using the DTW algorithm; specifically, this includes: Establish the second-order distance matrix D for each group of test samples; Starting from the bottom left corner of the second-order distance matrix D and ending at the top right corner, a monotonic and continuous regular path is established. W n =( i, j ), where the starting point w 1 = (1, 1), Endpoint w k = ( |X|,|Y| ), i ∈[1 , |X| ], j ∈[1 , |Y| ],n∈[1 , k ], i, j, n, k, |X| and |Y| All are positive integers; in the formula, n is the number of steps, and k is the normalized path. W n The length of X is given by X, where X is the waveform to be measured and Y is the standard waveform corresponding to the waveform to be measured. |X| and |Y| These are the lengths of the waveform to be measured and the corresponding standard waveform, respectively. i and j These are the components of the waveform X to be measured and the components of the standard waveform in the regularized path after each step. Based on the DTW algorithm and the normalized path of the second-order distance matrix D W n The optimal path is obtained by calculating the second-order distance matrix D, using the following formula: ; In the formula, DTW represents the optimal path; K is the compensation coefficient; Let n be the step distance for n steps, where n∈[1, ..., n] k ], k is the normalized path W n Length; Based on the optimal path, from the starting point to the ending point, the distance of each step under the optimal path is calculated using the first constraint. These distances are then accumulated to obtain the minimum similarity distance for each group of test samples. The formula for calculating the first constraint is: ， In the formula, g ( i , j () represents the similarity distance of the optimal path after each step; This represents the distance traveled in each step.

4. The waveform comparison method for deep fusion device signals according to claim 1, characterized in that, The number of cluster centers is two, and the smaller of the resulting similarity scores is... ω 1 The larger one is ω 2 .

5. The waveform comparison method for deep fusion device signals according to claim 4, characterized in that, The process of performing hierarchical analysis on the waveform similarity evaluation indicators to obtain the relative weights of the waveform similarity evaluation indicators specifically includes: Based on waveform similarity evaluation indicators, a judgment matrix A is established to determine the degree of importance between each pair of waveform similarity evaluation indicators, and the consistency of judgment matrix A is verified. The degree of importance includes five levels from 1 to 5. In the formula, α This represents the sequence number of the similarity evaluation index. a αα For the first α The first similarity evaluation metric is relative to the second. α The degree of importance of each similarity evaluation indicator; According to the geometric mean method, the relative weights of each row in the completed judgment matrix A are calculated using the following formula: ， In the formula, μ α For the first α The relative weights of the similarity evaluation indicators; p is the number of rows in judgment matrix A; q is the number of columns in judgment matrix A; t is the level of the similarity evaluation indicator.

6. The waveform comparison method for deep fusion device signals according to claim 5, characterized in that, The process involves optimizing the similarity score based on the relative weights of the similarity evaluation metrics to obtain an optimized score, and then performing waveform comparison based on the optimized score; specifically, this includes: The relative weight of each similarity evaluation metric is combined with the smaller similarity score. ω 1 Multiply and sum them to get a smaller optimization score. ω 1 ’ The same method yields a larger optimization score. ω 2 ’ ; Calculate the similarity score for all samples to be tested according to the aforementioned similarity calculation formula; Based on smaller optimization score ω 1 ’ and larger optimization score ω 2 ’ The waveform similarity is divided into three categories. When the calculated similarity score of the waveform to be tested is within a certain range... ω 1 ’ and ω 2 ’ When the similarity is between 0 and 1, the waveform is judged as potentially similar; when the similarity is between 0 and 1, the waveform is judged as potentially similar. ω 2 ’ When the similarity is between 0 and 1, the waveforms are considered similar; when the similarity is between 0 and 1, the waveforms are considered similar. ω 1 ’ When the waveforms are in a certain range, they are judged to be dissimilar.

7. A waveform comparison system for deep fusion device signals, characterized in that, include: Data acquisition module: used to acquire multiple sets of test samples and establish waveform similarity evaluation index; the test samples include multiple test waveforms output by the deep fusion device and standard waveforms output by the waveform recorder that correspond one-to-one with the test waveforms; the waveform similarity evaluation index is the interference factor when the deep fusion device outputs the test waveforms; Distance calculation module: used to calculate the minimum similarity distance of each group of test samples under the waveform similarity evaluation index according to the DTW algorithm; Hierarchical clustering module: used to perform hierarchical clustering on the minimum similarity distance of each group of test samples to obtain a similarity score; Hierarchical Analysis Module: Used to perform hierarchical analysis on waveform similarity evaluation indicators to obtain the relative weights of the waveform similarity evaluation indicators; Waveform comparison module: used to optimize the similarity score according to the relative weight, obtain the optimized score, and perform waveform comparison based on the optimized score; Specifically, the step of performing hierarchical clustering on the minimum similarity distance of each group of test samples to obtain a similarity score includes: The minimum similarity distance of each group of test samples is normalized to obtain the normalized distance of each group of test samples, and each normalized distance is set as each initial class; Based on the normalized distance of each group of test samples, the inter-class distance between each pair of test samples is calculated using the sum of squared deviations method; Merge two classes with small inter-class distance differences, and use the smaller inter-class distance between the two merged classes as the inter-class distance after merging. Construct a new class, and repeat the calculation of the inter-class distance between the new class and the initial class and the merging operation until the number of classes equals the preset number of classes, and obtain the cluster center. Calculate the similarity score of the cluster centers based on the minimum similarity distance corresponding to the cluster centers. ω The formula for calculating the similarity score is: , 0< ω <1， In the formula, ω To score the similarity, ctr This is the minimum similarity distance after normalization.

8. A waveform comparison device for signals from a deep fusion device, characterized in that, Including processor and storage media; The storage medium is used to store instructions; The processor is configured to operate according to the instructions to perform the steps of the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the program implements the steps of the method according to any one of claims 1 to 6.

Images