Electrocardiogram signal detection method and system based on dynamic learning and multi-scale analysis

Through dynamic learning and multi-scale analysis, the ECG signal detection methods are solved, and the problems of inaccurate STT interception and insufficient information extraction in ECG signal detection are improved, and the accuracy of myocardial infarction detection is achieved, and more abundant feature extraction and Improved detection performance.

CN115517685BInactive Publication Date: 2024-11-05SHANDONG UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202211121295.9
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-09-15
Publication Date
2024-11-05
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

The prior art has problems such as inaccurate STT interception and insufficient information extraction in electrocardiogram signal detection, which leads to the poor detection performance of myocardial infarction, especially when the ECG signals of early myocardial infarction are diverse and not obvious, making it difficult to accurately diagnose.

Method used

Using the ECG signal detection method based on dynamic learning and multi-scale analysis, the 12-lead ECG data is preprocessed and converted into a 3-lead ECG vector graph, and dynamic learning algorithm is used to perform dynamic modeling, and in multiple Five-level discrete wavelet transformation decomposition is performed under the resolution framework, statistical, frequency domain, chaotic invariant and nonlinear features are extracted, and feature selection is performed in combination with mutual information to form a subset of low-dimensional features to improve detection accuracy.

Benefits of technology

It improves the accuracy of electrocardiogram signal detection, includes changes in cardiac depolarization and repolarization, extracts richer information, avoids the problem of inaccurate STT interception, and significantly improves detection performance.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN115517685B_ABST
    Figure CN115517685B_ABST
Patent Text Reader

Abstract

The present invention discloses an electrocardiogram signal detection method and system based on dynamic learning and multi-scale analysis, and relates to the technical field of electrocardiogram signal detection. The electrocardiogram signal is modeled dynamically using a dynamic learning algorithm to obtain a nonlinear dynamic mode to generate an electrocardiogram; secondly, the electrocardiogram signal is analyzed multi-scale using discrete wavelets to extract the global and local features of the electrocardiogram; then, four effective features are selected from the above features by combining a variety of feature selection methods to form a low-dimensional feature subset; finally, the feature subset is used to form an analysis and detection for various electrocardiogram signals. The method can be well applied to various instruments for detecting and monitoring heart changes, such as electrocardiographs, dynamic electrocardiographs, and 24-hour electrocardiogram monitors, and has a good application scenario.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of electrocardiogram signal detection, and in particular to an electrocardiogram signal detection method and system based on dynamic learning and multi-scale analysis. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] Myocardial infarction is one of the leading causes of death and disability worldwide and is the most serious cardiovascular disease threatening human life and health. According to the Global Burden of Disease report, millions of people die from myocardial infarction every year. Timely and accurate detection of myocardial infarction is of great significance to reducing patient mortality and improving the quality of life. Currently, clinical diagnostic methods include electrocardiogram, coronary angiography and functional examination, among which electrocardiogram is the most convenient and non-invasive initial diagnosis method. The electrocardiogram of myocardial infarction mainly manifests as ST segment deviation and T wave inversion, but these changes are usually weak and diverse in the electrocardiogram, which poses a major challenge to the accurate diagnosis of myocardial infarction based on electrocardiogram.

[0004] With the development of artificial intelligence, research on myocardial infarction detection based on machine learning and deep learning algorithms has made some progress. However, due to the limitations of the methods, such as inaccurate STT interception during ECG signal analysis, insufficient information extraction, inability to mine deep and useful feature information, and the diverse and unobvious characteristics of early myocardial infarction ECG signals, the detection performance of myocardial infarction is not ideal, so it is rarely used in clinical practice. Therefore, how to achieve multi-scale dynamic detection of ECG signals has become an urgent problem to be solved in the medical field. Summary of the invention

[0005] In view of the deficiencies in the prior art, the purpose of the present invention is to provide an ECG signal detection method and system based on dynamic learning and multi-scale analysis, which improves the accuracy of diagnostic ECG signal detection.

[0006] In order to achieve the above object, the present invention is implemented through the following technical solutions:

[0007] The first aspect of the present disclosure provides an electrocardiogram signal detection method based on dynamic learning and multi-scale analysis, comprising the following steps:

[0008] Step (1): preprocessing the 12-lead electrocardiogram data and converting it into a 3-lead vectorcardiogram;

[0009] Step (2): sampling the electrocardiogram (ECV) signal of the three-lead ECG vector diagram, using a dynamic learning algorithm to perform accurate dynamic modeling, and visualizing the modeling results in three dimensions to obtain an electrocardiogram (ECD);

[0010] Step (3): Under the multi-resolution framework, five-level discrete wavelet transform decomposition is performed on each dimension of the electrocardiogram;

[0011] Step (4): further extracting statistical, frequency domain, chaotic invariant and nonlinear characteristics from the data of each dimension and wavelet coefficients at each level of the electrocardiogram;

[0012] Step (5): Normalize the extracted wavelet features, then use mutual information to select features, and detect and analyze the selected features.

[0013] Furthermore, the specific method of step (1) is: converting the 12-lead electrocardiogram into a 3-lead vectorcardiogram through a Kors matrix, using the DWT of the mother wavelet function to remove noise and baseline drift, and then modeling based on the entire electrocardiogram signal.

[0014] Furthermore, the specific method of step (2) is: sampling the 3D ECG vector signal, using a radial basis function neural network identifier to perform accurate neural network modeling, and obtaining the dynamic characteristics of the ECG vector signal.

[0015] Furthermore, a radial basis function neural network identifier is used to accurately model the system dynamics of the entire ECG signal including depolarization and repolarization.

[0016] Furthermore, the specific method of step (3) is: taking the electrocardiogram as an input signal, and using the db4 wavelet function to perform a five-level discrete wavelet transform decomposition on the data of each dimension.

[0017] Furthermore, in step (4), the following features are extracted for the electrocardiogram: mean, standard deviation, approximate entropy, sample entropy, frequency domain features and Lyapunov exponent.

[0018] Furthermore, the feature selection method is combined to screen four effective features from the features extracted from the electrocardiogram to form a low-dimensional feature subset.

[0019] A second aspect of the present disclosure provides an electrocardiogram signal detection system based on dynamic learning and multi-scale analysis, comprising:

[0020] A preprocessing module is configured to perform data preprocessing on the 12-lead electrocardiogram and convert it into a 3-lead vectorcardiogram;

[0021] The dynamic model building module is configured to sample the electrocardiogram (ECV) signal of the 3-lead ECG VCG, use the dynamic learning algorithm to perform accurate dynamic modeling, and visualize the modeling results in three dimensions to obtain an electrocardiogram (ECG);

[0022] The transform decomposition module is configured to perform five-level discrete wavelet transform decomposition on each dimension of the electrocardiodynamic graph in a multi-resolution framework;

[0023] A feature extraction module is configured to further extract statistical, frequency domain, chaotic invariant and nonlinear features from the data of each dimension and wavelet coefficients of each level of the electrocardiogram;

[0024] The diagnosis module is configured to normalize the extracted wavelet features, then perform feature selection using mutual information, and detect and analyze the selected features.

[0025] The third aspect of the present disclosure provides a medium on which a program is stored. When the program is executed by a processor, the steps in the electrocardiogram signal detection method based on dynamic learning and multi-scale analysis as described in the first aspect of the present disclosure are implemented.

[0026] The fourth aspect of the present disclosure provides a device, including a memory, a processor, and a program stored in the memory and executable on the processor. When the processor executes the program, the steps in the electrocardiogram signal detection method based on dynamic learning and multi-scale analysis as described in the first aspect of the present disclosure are implemented.

[0027] The beneficial effects of the above embodiments of the present invention are as follows:

[0028] The present invention collects a twelve-lead electrocardiogram, performs data preprocessing, converts the twelve-lead electrocardiogram into a three-dimensional vectorcardiogram using a Kors matrix, performs dynamic modeling on an electrocardiogram signal using a dynamic learning algorithm, obtains a nonlinear dynamic mode to generate an electrocardiogram, then uses a discrete wavelet to perform multi-scale analysis on the electrocardiogram signal, extracts statistical features, spectrum fitting index, sample entropy, approximate entropy and Lyapunov index of the electrocardiogram as global features, extracts statistical features, approximate entropy and sample entropy of each offspring as local features, then combines a plurality of feature selection methods to screen four effective features from the above features, and forms a low-dimensional feature subset, thereby improving the accuracy of electrocardiogram signal detection.

[0029] Compared with the traditional method, the method of the present invention utilizes the entire electrocardiogram signal for dynamic modeling, avoiding a series of problems caused by inaccurate STT interception, while including the changes in cardiac depolarization and repolarization, and contains richer information; secondly, it utilizes multi-level discrete wavelets to extract more useful features, which is conducive to improving detection performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0031] Figure 1 This is a flow chart of an electrocardiogram signal detection method based on dynamic learning and multi-scale analysis in Embodiment 1 of the present invention;

[0032] Figure 2 The electrocardiogram of a patient with myocardial infarction and a healthy subject in Example 1 of the present invention;

[0033] Figure 3 The dimensional signals of the electrocardiogram of a certain myocardial infarction patient and a healthy subject and the db4 discrete wavelet decomposition coefficient diagram in the first embodiment of the present invention;

[0034] Figure 4 The central electrodynamics of the embodiment of the present invention Figure 3 Distribution difference of wavelet features of dimensional data between myocardial infarction patients and healthy subjects;

[0035] Figure 5 This is a graph showing the relationship between feature selection using the mutual information algorithm, mutual information ranking, and the number of features and myocardial infarction detection performance in Example 1 of the present invention. DETAILED DESCRIPTION

[0036] It should be noted that the following detailed descriptions are illustrative and are intended to provide further explanation of the present application. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present application belongs.

[0037] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present application. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "include" and / or "include" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or their combinations;

[0038] Embodiment 1:

[0039] The first embodiment of the present disclosure provides an electrocardiogram (ECG) signal detection method based on dynamic learning and multi-scale analysis. Taking the early ECG signal detection of myocardial infarction as an example, Figure 1 As shown, the following steps are included:

[0040] Step (1): Preprocess the 12-lead electrocardiogram data and convert it into a 3-lead vectorcardiogram using the Kors matrix.

[0041] As a further technical solution, the 12-lead ECG is first converted into a 3-lead vectorcardiogram through the Kors matrix, the DWT of the mother wavelet function is used to remove noise and baseline drift, and then the model is built based on the entire ECG signal.

[0042] Step (2): Sampling the electrocardiogram (ECV) signal of the three-lead ECG vector diagram, using a dynamic learning algorithm to perform accurate dynamic modeling, and visualizing the modeling results in three dimensions to obtain an electrocardiogram (ECD).

[0043] As a further technical solution, the 3D ECG vector signal is first sampled, and the radial basis function neural network identifier is used to accurately model the system dynamics of the entire ECG signal, including depolarization and repolarization, to obtain the dynamic characteristics of the ECG vector signal. Figure 2 shown.

[0044] Preferably, the network identifier is:

[0045]

[0046] Where X(t) = [x1(t), x2(t), x3(t)] T ∈R 3 represents the three-dimensional ECG ST-T vector loop, F(X(t)) = [F1(X(t)), F2(X(t)), F3(X(t))] is the nonlinear system dynamics, representing the inherent dynamic characteristics of the ECG vector signal. Represents the dynamic changes of X(t).

[0047] Step (3): Under the multi-resolution framework, perform five-level discrete wavelet transform decomposition on each dimension of the electrocardiogram.

[0048] As a further technical solution, the electrocardiogram is used as the input signal X(k), and the db4 wavelet function is used to perform five-level discrete wavelet transform decomposition on its dimensional data. The dimensional signals of the electrocardiogram of myocardial infarction patients and healthy subjects and the db4 discrete wavelet decomposition coefficients are shown in Figure 2. Figure 3 Shown

[0049] Step (4): further extract statistical, frequency domain, chaotic invariant and nonlinear characteristics from the data of each dimension and wavelet coefficients at each level of the electrocardiogram.

[0050] As a further technical solution, the following features were extracted for the electrocardiodynamic graph: mean, standard deviation, approximate entropy, sample entropy, frequency domain features and Lyapunov exponent;

[0051] The mean value E of the electrocardiogram after wavelet decomposition is calculated by formula (2): h,l (X):

[0052]

[0053] Where h = 1, 2, 3 represents the dimension of the electrocardiogram, l = 1, 2, 3, 4, 5 represents the wavelet transform coefficient; n is the sampling point; x i is the value corresponding to the i-th sampling point.

[0054] The standard deviation S of the electrocardiogram after wavelet decomposition is calculated by formula (3): h,l :

[0055]

[0056] Among them, h = 1, 2, 3 represents the dimension of the electrocardiogram, and l = 1, 2, 3, 4, 5 represents the wavelet transform coefficient.

[0057] The approximate entropy ApEn of each dimension and each level of wavelet coefficients of the electrocardiogram is calculated by formula (4): h,l :

[0058]

[0059] Among them, A i is the number of sampling points that the data sequence matches with the i-th template vector of length m+1, B i is the number of sampling points that the data sequence matches with the i-th template vector of length m. Where h = 1, 2, 3 represents the dimension of the electrocardiogram, and l = 1, 2, 3, 4, 5 represents the wavelet transform coefficient.

[0060] The sample entropy SampEn of each dimension and each level of wavelet coefficients of the electrocardiogram is calculated by formula (5): h,l :

[0061]

[0062] in, is the matching probability of the data sequence with the template vector of length m+1 within the tolerance r, is the matching probability of the data sequence with the template vector of length m within the tolerance r. Wherein, h = 1, 2, 3 represents the dimension of the electrocardiogram, and l = 1, 2, 3, 4, 5 represents the wavelet transform coefficient.

[0063] The mean square value ESFI of the electrocardiogram fitting index is calculated by formula (6):

[0064]

[0065] in, represents a set of negative exponential functions fitting the spectral data, k represents the three dimensions of the electrocardiogram, and F max is the maximum value of the spectrum data, w = 1, 2, ..., N, and N is the number of spectrum data points.

[0066] The Lyapunov index LYE of the electrocardiogram is calculated by formula (7):

[0067]

[0068] Where K is the number of iterations; F is the system, electrocardiodynamics Figure 3 The distribution difference of wavelet features of dimensional data between myocardial infarction patients and healthy subjects is as follows Figure 4 shown.

[0069] Step (5): Normalize the extracted wavelet features, then use mutual information to select features, and detect and analyze the selected features.

[0070] Taking the detection of early myocardial infarction ECG signals as an example, the mutual information algorithm is used for feature selection, and four effective features are selected from the features extracted from the electrocardiogram to form a low-dimensional feature subset. The relationship between the mutual information ranking and the number of features and the performance of myocardial infarction detection is shown in Figure 2. Figure 5 shown.

[0071] After feature selection, SVM based on linear kernel and Gaussian radial basis function can be used to detect myocardial infarction. For the instance (x, y), the decision function is defined as

[0072]

[0073] Among them, α i is the Lagrangian factor, b is the offset term, m1 is the number of samples, and κ(·,·) is the kernel function, including:

[0074] (1) Linear kernel: κ(x i ,x j )

[0075] (2) Gaussian kernel: σ is the bandwidth of the Gaussian kernel.

[0076] Take the early myocardial infarction ECG signal detection as an example for experimental verification:

[0077] In this embodiment, a total of 549 records of 290 subjects (average age of 57.2 years, 209 males) are used. Among them, 52 healthy subjects and 148 myocardial infarction subjects are included. Each ECG record contains 12 conventional leads (I, II, III, AVR, AVL, AVF, V1, V2, V3, V4, V5, V6) and 3 Frank leads (VX, VY, VZ) ECG signals, with a sampling rate of 1000 Hz, 16-bit resolution, and a range of 16.384 mV. Four classifiers, logistic regression (LR), SVM-RBF, K nearest neighbor method (KNN), and SVM-Linea, were used for testing, and good results were achieved.

[0078] The above embodiments are preferred implementations of the present invention, but the implementations of the present invention are not limited to the above embodiments.

[0079] Embodiment 2:

[0080] Embodiment 2 of the present disclosure provides an electrocardiogram signal detection system based on dynamic learning and multi-scale analysis, including:

[0081] A preprocessing module is configured to perform data preprocessing on the 12-lead electrocardiogram and convert it into a 3-lead vectorcardiogram;

[0082] The dynamic model building module is configured to sample the electrocardiogram (ECV) signal of the 3-lead ECG VCG, use the dynamic learning algorithm to perform accurate dynamic modeling, and visualize the modeling results in three dimensions to obtain an electrocardiogram (ECG);

[0083] The transform decomposition module is configured to perform five-level discrete wavelet transform decomposition on each dimension of the electrocardiodynamic graph in a multi-resolution framework;

[0084] A feature extraction module is configured to further extract statistical, frequency domain, chaotic invariant and nonlinear features from the data of each dimension and wavelet coefficients of each level of the electrocardiogram;

[0085] The diagnosis module is configured to normalize the extracted wavelet features, then perform feature selection using mutual information, and detect and analyze the selected features.

[0086] Embodiment three:

[0087] Embodiment 3 of the present disclosure provides a medium on which a program is stored. When the program is executed by a processor, the steps in the electrocardiogram signal detection method based on dynamic learning and multi-scale analysis as described in Embodiment 1 of the present disclosure are implemented. The steps are:

[0088] Step (1): preprocessing the 12-lead electrocardiogram data and converting it into a 3-lead vectorcardiogram;

[0089] Step (2): sampling the electrocardiogram (ECV) signal of the three-lead ECG vector diagram, using a dynamic learning algorithm to perform accurate dynamic modeling, and visualizing the modeling results in three dimensions to obtain an electrocardiogram (ECD);

[0090] Step (3): Under the multi-resolution framework, five-level discrete wavelet transform decomposition is performed on each dimension of the electrocardiogram;

[0091] Step (4): further extracting statistical, frequency domain, chaotic invariant and nonlinear characteristics from the data of each dimension and wavelet coefficients at each level of the electrocardiogram;

[0092] Step (5): Normalize the extracted wavelet features, then use mutual information to select features, and detect and analyze the selected features.

[0093] Embodiment 4:

[0094] Embodiment 4 of the present disclosure provides a device, including a memory, a processor, and a program stored in the memory and executable on the processor. When the processor executes the program, the steps in the electrocardiogram signal detection method based on dynamic learning and multi-scale analysis as described in Embodiment 1 of the present disclosure are implemented. The steps are:

[0095] Step (1): preprocessing the 12-lead electrocardiogram data and converting it into a 3-lead vectorcardiogram;

[0096] Step (2): sampling the electrocardiogram (ECV) signal of the three-lead ECG vector diagram, using a dynamic learning algorithm to perform accurate dynamic modeling, and visualizing the modeling results in three dimensions to obtain an electrocardiogram (ECD);

[0097] Step (3): Under the multi-resolution framework, five-level discrete wavelet transform decomposition is performed on each dimension of the electrocardiogram;

[0098] Step (4): further extracting statistical, frequency domain, chaotic invariant and nonlinear characteristics from the data of each dimension and wavelet coefficients at each level of the electrocardiogram;

[0099] Step (5): Normalize the extracted wavelet features, then use mutual information to select features, and detect and analyze the selected features.

[0100] The steps involved in the apparatuses of the above embodiments 2, 3 and 4 correspond to the method embodiment 1, and the specific implementation methods can refer to the relevant description part of embodiment 1. The term "computer-readable storage medium" should be understood as a single medium or multiple media including one or more instruction sets; it should also be understood to include any medium that can store, encode or carry an instruction set for execution by a processor and enable the processor to execute any method in the present invention.

[0101] Those skilled in the art should understand that the modules or steps of the present invention described above can be implemented by a general-purpose computer device, or alternatively, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. The present invention is not limited to any specific combination of hardware and software.

[0102] Although the above describes the specific implementation mode of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art on the basis of the technical solution of the present invention without creative work are still within the scope of protection of the present invention.

Claims

1. A method for detecting electrocardiogram signals based on dynamic learning and multi-scale analysis, characterized in that: The following steps are involved: Step (1): preprocessing the 12-lead electrocardiogram data and converting it into a 3-lead vectorcardiogram; Step (2): sampling the electrocardiogram (ECV) signal of the three-lead ECG vector diagram, using a dynamic learning algorithm to perform accurate dynamic modeling, and visualizing the modeling results in three dimensions to obtain an electrocardiogram (ECD); The radial basis function neural network identifier is used to accurately model the system dynamics of the entire ECG signal, including depolarization and repolarization; Step (3): Under the multi-resolution framework, five-level discrete wavelet transform decomposition is performed on each dimension of the electrocardiogram; Step (4): further extracting statistical, frequency domain, chaotic invariant and nonlinear characteristics from the data of each dimension and wavelet coefficients at each level of the electrocardiogram; In step (4), the following features are extracted for the electrocardiogram: mean, standard deviation, approximate entropy, sample entropy, frequency domain features and Lyapunov exponent; The mean value E of the electrocardiogram after wavelet decomposition is calculated by formula (2): h,l (X): Where h = 1, 2, 3 represents the dimension of the electrocardiogram, l = 1, 2, 3, 4, 5 represents the wavelet transform coefficient; n is the sampling point; x i is the value corresponding to the i-th sampling point; The standard deviation S of the electrocardiogram after wavelet decomposition is calculated by formula (3): h,l : Among them, h = 1, 2, 3 represents the dimension of the electrocardiogram, and l = 1, 2, 3, 4, 5 represents the wavelet transform coefficient; The approximate entropy ApEn of each dimension and each level of wavelet coefficients of the electrocardiogram is calculated by formula (4): h,l : Among them, A i is the number of sampling points that the data sequence matches with the i-th template vector of length m+1, B i is the number of sampling points that the data sequence matches with the i-th template vector of length m; where h = 1, 2, 3 represents the dimension of the electrocardiogram, and l = 1, 2, 3, 4, 5 represents the wavelet transform coefficient; The sample entropy SampEn of each dimension and each level of wavelet coefficients of the electrocardiogram is calculated by formula (5): h,l : in, is the matching probability of the data sequence with the template vector of length m+1 within the tolerance r, is the matching probability of the data sequence with the template vector of length m within the tolerance r; where h = 1, 2, 3 represents the dimension of the electrocardiogram, and l = 1, 2, 3, 4, 5 represents the wavelet transform coefficient; The mean square value ESFI of the electrocardiogram fitting index is calculated by formula (6): in, represents a set of negative exponential functions fitting the spectral data, k represents the three dimensions of the electrocardiogram, and F max is the maximum value of the spectrum data, w = 1, 2, ..., N, and N is the number of data points of the spectrum; The Lyapunov index LYE of the electrocardiogram is calculated by formula (7): Where K is the number of iterations; F is the system; Step (5): Normalize the extracted wavelet features, then use mutual information to select features, and detect and analyze the selected features.

2. The electrocardiogram signal detection method based on dynamic learning and multi-scale analysis as claimed in claim 1, characterized in that: The specific method of step (1) is as follows: the 12-lead electrocardiogram is converted into a 3-lead vectorcardiogram through the Kors matrix, the DWT of the mother wavelet function is used to remove noise and baseline drift, and then a model is built based on the entire electrocardiogram signal.

3. The electrocardiogram signal detection method based on dynamic learning and multi-scale analysis according to claim 1, characterized in that: The specific method of step (2) is: sampling the 3D ECG vector signal, using a radial basis function neural network identifier to perform accurate neural network modeling, and obtaining the dynamic characteristics of the ECG vector signal.

4. The electrocardiogram signal detection method based on dynamic learning and multi-scale analysis according to claim 1, characterized in that: The specific method of step (3) is: taking the electrocardiogram as the input signal, and using the db4 wavelet function to perform five-level discrete wavelet transform decomposition on its various dimensional data.

5. The electrocardiogram signal detection method based on dynamic learning and multi-scale analysis as claimed in claim 1, characterized in that: The feature selection method is combined to screen four effective features from the features extracted from the electrocardiogram to form a low-dimensional feature subset.

6. An electrocardiogram signal detection system based on dynamic learning and multi-scale analysis, characterized in that: include: A preprocessing module is configured to perform data preprocessing on the 12-lead electrocardiogram and convert it into a 3-lead vectorcardiogram; The dynamic model building module is configured to sample the electrocardiogram (ECV) signal of the 3-lead ECG VCG, use the dynamic learning algorithm to perform accurate dynamic modeling, and visualize the modeling results in three dimensions to obtain an electrocardiogram (ECG); The radial basis function neural network identifier is used to accurately model the system dynamics of the entire ECG signal, including depolarization and repolarization; The transform decomposition module is configured to perform five-level discrete wavelet transform decomposition on each dimension of the electrocardiodynamic graph in a multi-resolution framework; A feature extraction module is configured to further extract statistical, frequency domain, chaotic invariant and nonlinear features from the data of each dimension and wavelet coefficients of each level of the electrocardiogram; The following features were extracted for the electrocardiogram: mean, standard deviation, approximate entropy, sample entropy, frequency domain features, and Lyapunov exponent; The mean value E of the electrocardiogram after wavelet decomposition is calculated by formula (2): h,l (X): Where h = 1, 2, 3 represents the dimensions of the electrocardiogram, l = 1, 2, 3, 4, 5 represents the wavelet transform coefficients; n is the sampling point; x i is the value corresponding to the i-th sampling point; The standard deviation S of the electrocardiogram after wavelet decomposition is calculated by formula (3): h,l : Among them, h = 1, 2, 3 represents the dimension of the electrocardiogram, and l = 1, 2, 3, 4, 5 represents the wavelet transform coefficient; The approximate entropy ApEn of each dimension and each level of wavelet coefficients of the electrocardiogram is calculated by formula (4): h,l : Among them, A i is the number of sampling points that the data sequence matches with the i-th template vector of length m+1, B i is the number of sampling points that the data sequence matches with the i-th template vector of length m; where h = 1, 2, 3 represents the dimension of the electrocardiodynamic diagram, and l = 1, 2, 3, 4, 5 represents the wavelet transform coefficient; The sample entropy SampEn of each dimension and each level of wavelet coefficients of the electrocardiogram is calculated by formula (5): h,l : in, is the matching probability of the data sequence with the template vector of length m+1 within the tolerance r, is the matching probability of the data sequence with the template vector of length m within the tolerance r; where h = 1, 2, 3 represents the dimension of the electrocardiogram, and l = 1, 2, 3, 4, 5 represents the wavelet transform coefficient; The mean square value ESFI of the electrocardiogram fitting index is calculated by formula (6): in, represents a set of negative exponential functions fitting the spectral data, k represents the three dimensions of the electrocardiogram, and F max is the maximum value of the spectrum data, w = 1, 2, ..., N, and N is the number of data points of the spectrum; The Lyapunov index LYE of the electrocardiogram is calculated by formula (7): Where K is the number of iterations; F is the system; The diagnosis module is configured to normalize the extracted wavelet features, then perform feature selection using mutual information, and detect and analyze the selected features.

7. A computer-readable storage medium, characterized in that: A plurality of instructions are stored therein, and the instructions are suitable for being loaded by a processor of a terminal device and executing the electrocardiogram signal detection method based on dynamic learning and multi-scale analysis as described in any one of claims 1-5.

8. A terminal device, characterized in that: It includes a processor and a computer-readable storage medium, the processor is used to implement each instruction; the computer-readable storage medium is used to store multiple instructions, and the instructions are suitable for being loaded by the processor and executing the electrocardiogram signal detection method based on dynamic learning and multi-scale analysis according to any one of claims 1-5.