Electrocardiosignal abnormity identification and analysis system based on multilevel clustering algorithm

Abstract

An electrocardiosignal anomaly recognition and analysis system based on a multilevel clustering algorithm is characterized by comprising an electrocardiosignal acquisition module, a data transmission module, a data storage management module, a data processing module, an electrocardiosignal denoising module, an electrocardiosignal analysis module and an electrocardiosignal judgment module. According to the electrocardiosignal anomaly recognition and analysis system based on the multi-level clustering algorithm, electrocardiosignal data are collected through the external lead electrodes, noise reduction is conducted on the collected electrocardiosignal data through the reconstruction noise reduction algorithm, then the electrocardiosignal data are clustered through the multi-level clustering algorithm, and the electrocardiosignal anomaly recognition and analysis system based on the multi-level clustering algorithm is obtained. And performing abnormal electrocardiosignal data output based on the standard abnormal electrocardiosignal data.

Description

Technical Field

[0001] The present invention relates to the field of intelligent medicine, and particularly to an electrocardiogram signal abnormality recognition and analysis system based on a multi-level clustering algorithm. Background Art

[0002] With the rapid development of biomedical sensing technology, Internet of Things devices, and artificial intelligence algorithms, electrocardiogram signal monitoring and analysis systems are gradually extending from hospital professional scenarios to the directions of household, portable, and intelligent. As one of the chronic diseases with the highest fatality rate globally, cardiovascular diseases have an increasingly urgent need for early screening and real-time monitoring. According to the statistics of "Report on Cardiovascular Health and Diseases in China", more than 80% of cases have identifiable electrocardiogram abnormality characteristics before the onset of the disease. This severe situation drives the continuous innovation of electrocardiogram monitoring devices in sampling accuracy, analysis efficiency, and diagnostic accuracy. However, in the process of achieving high-precision dynamic monitoring and intelligent diagnosis, the existing technology system still faces multiple technical bottlenecks, restricting the clinical application effect of electrocardiogram abnormality recognition.

[0003] Traditional electrocardiogram signal processing methods mostly adopt threshold-based algorithms and simple feature extraction techniques, which often cannot meet complex clinical requirements. When dealing with complex multi-dimensional signals, these methods may lead to signal loss and reduced processing accuracy. Therefore, the research and development of new data analysis technologies have become a popular direction in current electrocardiogram signal research. The application of technologies based on machine learning and deep learning in various signal processing is gradually heating up. These technologies can achieve a more in-depth analysis and accurate judgment of electrocardiogram signals by learning and mining deep-level features in the data. The current electrocardiogram abnormality recognition technology has insufficient signal fidelity in a noisy environment, resulting in distorted feature extraction. The abnormality recognition and analysis of electrocardiogram signals are not only a technical requirement in the field of medical diagnosis but also a comprehensive challenge to information processing capabilities, data storage management capabilities, and application algorithms. With the help of modern technical means, especially the processing capabilities based on computational intelligence, we have reason to believe that the system proposed in the present invention will open up a new situation for the automation and intelligence of electrocardiogram signal analysis and promote the progress of the entire cardiovascular disease detection technology. Summary of the Invention

[0004] The purpose of the present invention is to provide an electrocardiogram signal abnormality recognition and analysis system based on a multi-level clustering algorithm, aiming to solve the problems mentioned in the above background.

[0005] In order to achieve the above-mentioned purpose, the present invention proposes an ECG signal abnormality identification and analysis system based on a multi-level clustering algorithm, which is characterized in that it includes an ECG signal acquisition module, a data transmission module, a data storage management module, a data processing module, an ECG signal denoising module, an ECG signal analysis module, and an ECG signal judgment module; the ECG signal acquisition module collects the user's ECG signal in real time through hardware equipment; the data transmission module transmits the collected ECG signal data through a network protocol; the data storage management module encrypts and stores the collected ECG signal data in a computing server; the data processing module decrypts and pre-processes the stored ECG signal data; the ECG signal denoising module proposes a reconstruction denoising algorithm to denoise the ECG signal data; the ECG signal analysis module proposes a multi-level clustering algorithm to cluster the ECG signal data and perform abnormal analysis on the ECG signal data; the ECG signal judgment module makes an abnormal judgment on the collected ECG signal data based on the results of the ECG signal analysis module.

[0006] Furthermore, the ECG signal acquisition module converts the current changes generated by the contraction and relaxation of the user's heart into digital signals through external lead electrodes, and collects the user's ECG signals in real time.

[0007] Furthermore, the data transmission module establishes a data transmission channel between the external lead electrodes and the computing server through a network protocol to transmit the collected ECG signal data.

[0008] Furthermore, the data storage management module encrypts and stores the received ECG signal data in a computing server with computing and storage functions.

[0009] Furthermore, the data processing module decrypts the encrypted ECG signal data through the computing server, restores it to plain text, and then removes abnormal data.

[0010] Furthermore, a noise reduction algorithm is reconstructed to construct a decomposition matrix based on the collected ECG signal data, calculate the decomposition coefficients using the objective function, and reconstruct the data based on the decomposition coefficients to complete the noise reduction of the ECG signal data.

[0011] Furthermore, the noise reduction of ECG signal data is as follows:

[0012] ECG signal is a time series. The collected ECG signal data is constructed into a time series, which can be expressed as:

[0013] Xori=[x1,x2,…,xi,…,xn]

[0014] Among them, Xori represents the original electrocardiogram signal data set. x1, x2, xi, and xn represent the data at the 1st moment, the data at the 2nd moment, the data at the ith moment, and the data at the nth moment in the original electrocardiogram signal data set respectively. Based on the original electrocardiogram signal data set, a decomposition matrix Xtra is constructed, and the elements in the decomposition matrix are represented as follows:

[0015] Xtra,i = [xi, xi+1, …, xi+l-1]T

[0016] Among them, Xtra,i represents the ith element in the decomposition matrix Xtra. xi+1 and xi+l-1 represent the data at the (i + 1)th moment and the data at the (i + l - 1)th moment in the original electrocardiogram signal data set respectively. Where l < n, a dimensionality reduction matrix D is defined. Based on the decomposition matrix Xtra, the dimensionality reduction matrix is constructed and represented as follows:

[0017]

[0018] Among them, represents the transpose matrix of the decomposition matrix A target function is constructed to calculate the decomposition coefficients of the dimensionality reduction matrix. The target function is as follows:

[0019]

[0020] Among them, F represents the target function, β represents the decomposition coefficients of the dimensionality reduction matrix, y represents the decomposition vector. Find the set of values of the decomposition coefficients when the target function is zero. Based on the decomposition coefficients, the reconstruction matrix B is calculated. The formula is as follows:

[0021]

[0022] Among them, βi represents the ith decomposition coefficient of the dimensionality reduction matrix, yi represents the ith decomposition vector of the dimensionality reduction matrix, and Bi represents the ith column vector of the reconstruction matrix. Based on the reconstruction matrix B, the noise reduction vector H of the collected electrocardiogram signal data is obtained. The calculation formula is as follows:

[0023]

[0024] Among them, hk represents the k-th value of the noise reduction vector H, and bi,k-i+1 represents the value in the i-th row and the (k-i+1)-th column of the reconstruction matrix B, obtaining the denoised electrocardiogram signal data. The reconstruction denoising algorithm proposed by the present invention constructs a time series based on the electrocardiogram signal data, uses the time series to obtain a decomposition matrix, constructs an objective function based on the decomposition matrix, calculates the decomposition coefficients, and then performs data reconstruction to obtain a reconstruction matrix. Based on the reconstruction matrix, the denoising operation of the electrocardiogram signal is completed. The reconstruction denoising algorithm proposed by the present invention decomposes and reconstructs the data, uses the reconstructed matrix for denoising, and improves the denoising effect.

[0025] Furthermore, the multi-level clustering algorithm uses a neural network to extract the characteristics of electrocardiogram signal data, obtains a feature representation, and based on the feature representation, selects an abnormal analysis clustering center for abnormal clustering analysis.

[0026] Furthermore, the abnormal clustering analysis of electrocardiogram signal data is as follows in detail:

[0027] For the denoised electrocardiogram signal data, first, feature extraction is performed. Using a neural network, the electrocardiogram signal data is input to obtain an output, and the output is represented as follows:

[0028]

[0029] Among them, zmid,j represents the value of the j-th neuron in the middle layer of the neural network, zi represents the value of the i-th neuron in the input layer of the neural network, ωj,i represents the weight between the j-th neuron in the middle layer of the neural network and the i-th neuron in the input layer of the neural network. A multi-layer middle layer is constructed to perform multiple input-output operations on the input data to obtain the value of the output layer, and then through recursion, a feature representation is obtained. The calculation formula is as follows:

[0030]

[0031] Among them, zout represents the vector composed of the values of the output layer of the neural network, and vout represents the feature representation. respectively represent the vector composed of the values of the (t-1)-th middle layer and the vector composed of the values of the t-th middle layer. Based on the feature representations of all collected electrocardiogram signal data, abnormal analysis is performed. The feature representations of all electrocardiogram signal data are

[0032]

[0033] Among them, respectively represent the feature representation of the first segment of electrocardiogram signal data, the feature representation of the second segment of electrocardiogram signal data, and the feature representation of the u-th segment of electrocardiogram signal data. Nu feature representations of electrocardiogram signal data are selected as abnormal analysis clustering centers in the electrocardiogram signal data for abnormal clustering analysis. The formula is represented as follows:

[0034]

[0035] Among them, vcer,r represents the r-th abnormal analysis clustering center, and αu,r represents the clustering coefficient between the feature representation of the u-th electrocardiogram signal data and the r-th abnormal analysis clustering center vcer,r. Based on the clustering coefficient, clustering of the electrocardiogram signal data feature representation is performed. The clustering center related to the lowest clustering coefficient of the electrocardiogram signal feature representation is selected as the clustering cluster for clustering division. Then, the means in different clustering clusters are selected as the new abnormal analysis clustering centers, and clustering is performed again until the clustering clusters no longer change, and the electrocardiogram signal data is completed with clustering. The multi-level clustering algorithm proposed by the present invention constructs a multi-layer neural network, uses the multi-layer neural network to extract features from the electrocardiogram signal data to obtain a feature representation, selects the data in the feature representation as the abnormal analysis clustering center for clustering, calculates the clustering coefficient, and based on the clustering coefficient, clusters the feature representation. The multi-level clustering algorithm proposed by the present invention extracts the features of the electrocardiogram signal data through a multi-level neural network, improving the accuracy of the clustering result.

[0036] Furthermore, the electrocardiogram signal judgment module compares the feature representation of the electrocardiogram signal data after clustering with the feature representation of the standard abnormal electrocardiogram signal, and automatically outputs the corresponding type of electrocardiogram signal abnormality.

[0037] Beneficial effects

[0038] 1. The reconstruction and noise reduction algorithm proposed by the present invention constructs a time series based on the electrocardiogram signal data, uses the time series to obtain a decomposition matrix, constructs an objective function based on the decomposition matrix, calculates the decomposition coefficient, and then performs data reconstruction to obtain a reconstruction matrix. Based on the reconstruction matrix, the noise reduction operation of the electrocardiogram signal is completed. The reconstruction and noise reduction algorithm proposed by the present invention decomposes and reconstructs the data, and uses the reconstructed matrix to perform noise reduction, improving the noise reduction effect.

[0039] 2. The multi-level clustering algorithm proposed by the present invention constructs a multi-layer neural network, uses the multi-layer neural network to extract features from the electrocardiogram signal data to obtain a feature representation, selects the data in the feature representation as the abnormal analysis clustering center for clustering, calculates the clustering coefficient, and based on the clustering coefficient, clusters the feature representation. The multi-level clustering algorithm proposed by the present invention extracts the features of the electrocardiogram signal data through a multi-level neural network, and compared with the traditional clustering algorithm, improves the accuracy of the clustering result. Description of the drawings

[0040] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0041] Figure 1 is a schematic diagram of the system of the present invention; Specific embodiments

[0042] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0043] To achieve the above object, the present invention proposes an electrocardiogram signal abnormality recognition and analysis system based on a multi-level clustering algorithm, which is characterized by including an electrocardiogram signal acquisition module, a data transmission module, a data storage and management module, a data processing module, an electrocardiogram signal denoising module, an electrocardiogram signal analysis module, and an electrocardiogram signal judgment module; the electrocardiogram signal acquisition module uses hardware devices to collect the electrocardiogram signals of users in real time; the data transmission module transmits the collected electrocardiogram signal data through network protocols; the data storage and management module encrypts and stores the collected electrocardiogram signal data in a computing server; the data processing module decrypts and preprocesses the stored electrocardiogram signal data; the electrocardiogram signal denoising module proposes a reconstruction denoising algorithm to denoise the electrocardiogram signal data; the electrocardiogram signal analysis module proposes a multi-level clustering algorithm to cluster the electrocardiogram signal data and perform abnormal analysis of the electrocardiogram signal data; the electrocardiogram signal judgment module makes an abnormal judgment on the collected electrocardiogram signal data based on the results of the electrocardiogram signal analysis module.

[0044] Specifically, the electrocardiogram signal acquisition module converts the current changes generated during the contraction and relaxation of the user's heart into digital signals through external lead electrodes to collect the user's electrocardiogram signals in real time.

[0045] Specifically, the data transmission module establishes a data transmission channel between the external lead electrodes and the computing server through network protocols to transmit the collected electrocardiogram signal data.

[0046] Specifically, the data storage and management module encrypts and stores the received electrocardiogram signal data in a computing server with computing and storage functions.

[0047] Specifically, the data processing module decrypts the encrypted electrocardiogram signal data through a computing server, restores it to plaintext, and then eliminates abnormal data.

[0048] Specifically, the reconstruction and noise reduction algorithm constructs a decomposition matrix based on the collected electrocardiogram signal data, calculates the decomposition coefficients using the objective function, and reconstructs the data based on the decomposition coefficients to complete the noise reduction of the electrocardiogram signal data. The detailed process is as follows:

[0049] The electrocardiogram signal is a time series. The collected electrocardiogram signal data is constructed into a time series, expressed as:

[0050] Xori = [x1, x2, …, xi, …, xn]

[0051] where Xori represents the set of original electrocardiogram signal data, x1, x2, xi, and xn respectively represent the data at the 1st moment, the data at the 2nd moment, the data at the i-th moment, and the data at the n-th moment in the set of original electrocardiogram signal data. Based on the set of original electrocardiogram signal data, a decomposition matrix Xtra is constructed, and the elements in the decomposition matrix are expressed as follows:

[0052] Xtra,i = [xi, xi+1, …, xi+l-1]T

[0053] where Xtra,i represents the i-th element in the decomposition matrix Xtra, xi+1 and xi+l-1 respectively represent the data at the (i + 1)-th moment and the data at the (i + l - 1)-th moment in the set of original electrocardiogram signal data, where l < n. A dimensionality reduction matrix D is defined, and based on the decomposition matrix Xtra, a dimensionality reduction matrix is constructed, expressed as follows:

[0054]

[0055] where, represents the transpose matrix of the decomposition matrix An objective function is constructed to calculate the decomposition coefficients of the dimensionality reduction matrix. The objective function is as follows:

[0056]

[0057] where F represents the objective function, β represents the decomposition coefficients of the dimensionality reduction matrix, y represents the decomposition vector, and the set of values of the decomposition coefficients when the objective function is zero is obtained. Based on the decomposition coefficients, the reconstruction matrix B is calculated. The formula is as follows:

[0058]

[0059] Among them, βi represents the i-th decomposition coefficient of the dimensionality reduction matrix, yi represents the i-th decomposition vector of the dimensionality reduction matrix, and Bi represents the i-th column vector of the reconstruction matrix. Based on the reconstruction matrix B, the noise reduction vector H of the collected electrocardiogram signal data is obtained, and the calculation formula is as follows:

[0060]

[0061] Among them, hk represents the k-th value of the noise reduction vector H, and bi,k-i+1 represents the value in the (k-i+1)-th column of the i-th row of the reconstruction matrix B, and the denoised electrocardiogram signal data is obtained.

[0062] Specifically, for the multi-level clustering algorithm, a neural network is used to extract the features of the electrocardiogram signal data to obtain a feature representation. Based on the feature representation, an abnormal analysis clustering center is selected for abnormal clustering analysis. The detailed process is as follows:

[0063] For the denoised electrocardiogram signal data, feature extraction is first performed. Using a neural network, the electrocardiogram signal data is input to obtain an output, and the output is represented as follows:

[0064]

[0065] Among them, zmid,j represents the value of the j-th neuron in the middle layer of the neural network, zi represents the value of the i-th neuron in the input layer of the neural network, ωj,i represents the weight between the j-th neuron in the middle layer of the neural network and the i-th neuron in the input layer of the neural network. A multi-layer middle layer is constructed to perform multiple input-output operations on the input data to obtain the value of the output layer, and then through recursion, the feature representation is obtained. The calculation formula is as follows:

[0066]

[0067] Among them, zout represents the vector composed of the values of the output layer of the neural network, and vout represents the feature representation. respectively represent the vector composed of the values of the (t-1)-th middle layer and the vector composed of the values of the t-th middle layer. Based on the feature representations of all the collected electrocardiogram signal data, abnormal analysis is performed. The feature representations of all the electrocardiogram signal data are

[0068]

[0069] Among them, respectively represent the feature representation of the first segment of electrocardiogram signal data, the feature representation of the second segment of electrocardiogram signal data, and the feature representation of the u-th segment of electrocardiogram signal data. Nu feature representations of electrocardiogram signal data are selected as the abnormal analysis clustering centers in the electrocardiogram signal data for abnormal clustering analysis. The formula is expressed as follows:

[0070]

[0071] Among them, vcer,r represents the r-th abnormal analysis clustering center, and αu,r represents the clustering coefficient between the data feature representation of the u-th electrocardiogram signal segment and the r-th abnormal analysis clustering center vcer,r. Based on the clustering coefficient, clustering is performed on the data feature representation of the electrocardiogram signal. The clustering center related to the lowest clustering coefficient of the electrocardiogram signal feature representation is selected as the clustering cluster for clustering division. Then, the means in different clustering clusters are selected as the new abnormal analysis clustering centers, and clustering is performed again until the clustering clusters no longer change, and the clustering of the electrocardiogram signal data is completed.

[0072] Specifically, the electrocardiogram signal judgment module compares the data feature representation of the electrocardiogram signal after clustering with the standard abnormal electrocardiogram signal feature representation, and automatically outputs the corresponding electrocardiogram signal abnormal type.

[0073] In a specific embodiment, the user first uses an external lead electrode to convert the current change generated during the contraction and relaxation of the user's heart into a digital signal, and real-time collects the user's electrocardiogram signal data; then, through a network protocol, a data transmission channel is established between the external lead electrode and the computing server for data transmission, and the electrocardiogram signal data is encrypted and stored in the computing server; then, for the restored electrocardiogram signal data, a reconstruction noise reduction algorithm is used to perform noise reduction on the electrocardiogram signal data based on the decomposition and reconstruction of the data; then, a multi-level clustering algorithm is used to extract the data features of the electrocardiogram signal to obtain a feature representation, and based on the feature representation, an abnormal analysis clustering center is selected for abnormal clustering analysis; finally, the clustered electrocardiogram signal data is compared and analyzed with the standard abnormal electrocardiogram signal data, and the corresponding electrocardiogram signal abnormal type is automatically output.

[0074] It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not expressly listed, or also includes elements inherent to such process, method, article or device. Without further limitation, an element defined by the statement "including a..." does not exclude the existence of additional identical elements in the process, method, article or device including the element.

[0075] Each embodiment in this specification is described in a related manner. For the same or similar parts among the embodiments, reference can be made to each other. Each embodiment focuses on the differences from other embodiments. In particular, for system embodiments, since they are basically similar to method embodiments, the description is relatively simple. For related parts, reference can be made to the corresponding description in the method embodiments.

[0076] The above are only the preferred embodiments of the present invention, and are not intended to limit the protection scope of the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention are all included in the protection scope of the present invention.

Claims

1. An electrocardiogram signal abnormality recognition and analysis system based on a multi-level clustering algorithm, characterized in that, It includes an electrocardiogram (ECG) signal acquisition module, a data transmission module, a data storage and management module, a data processing module, an ECG signal denoising module, an ECG signal analysis module, and an ECG signal judgment module; the ECG signal acquisition module uses hardware devices to collect the user's ECG signals in real time; the data transmission module transmits the collected ECG signal data through network protocols; the data storage and management module encrypts and stores the collected ECG signal data in a computing server; the data processing module decrypts and preprocesses the stored ECG signal data; the ECG signal denoising module proposes a reconstruction denoising algorithm to denoise the ECG signal data; the ECG signal analysis module proposes a multi-level clustering algorithm to cluster the ECG signal data and perform anomaly analysis on the ECG signal data; the ECG signal judgment module makes an anomaly judgment on the collected ECG signal data based on the results of the ECG signal analysis module.

2. The electrocardiogram signal abnormality recognition and analysis system based on a multi-level clustering algorithm according to claim 1, wherein The ECG signal acquisition module uses external lead electrodes to convert the current changes generated during the user's heart contraction and relaxation into digital signals for real-time acquisition of the user's ECG signals.

3. The electrocardiogram signal abnormality recognition and analysis system based on the multi-level clustering algorithm according to claim 1, wherein, The data transmission module uses network protocols to establish a data transmission channel between the external lead electrodes and the computing server to transmit the collected ECG signal data.

4. The electrocardiogram signal abnormal recognition and analysis system based on a multi-level clustering algorithm according to claim 1, characterized in that, The data storage and management module encrypts and stores the received ECG signal data in a computing server with computing and storage functions.

5. The electrocardiogram signal abnormality recognition and analysis system based on a multi-level clustering algorithm according to claim 4, wherein The data processing module decrypts the encrypted ECG signal data through the computing server to restore it to plaintext, and then eliminates the abnormal data.

6. The electrocardiogram signal abnormal recognition and analysis system based on a multi-level clustering algorithm according to claim 1, characterized in that, The reconstruction denoising algorithm constructs a decomposition matrix based on the collected ECG signal data, calculates the decomposition coefficients using the objective function, and performs data reconstruction based on the decomposition coefficients to complete the denoising of the ECG signal data.

7. An electrocardiogram signal abnormality recognition and analysis system based on a multi-level clustering algorithm according to claim 6, characterized in that The detailed process of the denoising of the ECG signal data is as follows: The ECG signal is a time series. The collected ECG signal data is constructed into a time series, expressed as: Xori = [x1, x2, …, xi, …, xn] where Xori represents the set of original ECG signal data, x1, x2, xi, and xn represent the data at the 1st moment, the data at the 2nd moment, the data at the i-th moment, and the data at the n-th moment in the set of original ECG signal data respectively. Based on the set of original ECG signal data, a decomposition matrix Xtra is constructed, and the elements in the decomposition matrix are expressed as follows: Xtra,i = [xi, xi+1, …, xi+l-1]T where Xtra,i represents the i-th element in the decomposition matrix Xtra, xi+1 and xi+l-1 represent the data at the (i + 1)-th moment and the data at the (i + l - 1)-th moment in the set of original ECG signal data respectively, where l < n. A dimensionality reduction matrix D is defined, and based on the decomposition matrix Xtra, the dimensionality reduction matrix is constructed, expressed as: Among them, represents the transposed matrix of the decomposition matrix Construct the objective function, calculate the decomposition coefficients of the dimensionality reduction matrix, and the objective function is as follows: Among them, F represents the objective function, β represents the decomposition coefficient of the dimensionality reduction matrix, y represents the decomposition vector, and the set of values of the decomposition coefficient when the objective function is zero is obtained. Based on the decomposition coefficient, the reconstruction matrix B is calculated as follows: Among them, βi represents the i-th decomposition coefficient of the dimensionality reduction matrix, yi represents the i-th decomposition vector of the dimensionality reduction matrix, and Bi represents the i-th column vector of the reconstruction matrix. Based on the reconstruction matrix B, the noise reduction vector H of the collected electrocardiogram signal data is obtained, and the calculation formula is as follows: Among them, hk represents the k-th value of the noise reduction vector H, and bi,k-i+1 represents the value in the i-th row and the (k-i+1)-th column of the reconstruction matrix B, and the electrocardiogram signal data after noise reduction is obtained.

8. A system for abnormal recognition and analysis of electrocardiogram signals based on a multi-level clustering algorithm according to claim 1, characterized in that, The multi-level clustering algorithm uses a neural network to extract the characteristics of electrocardiogram signal data to obtain a feature representation. Based on the feature representation, an abnormal analysis clustering center is selected for abnormal clustering analysis.

9. An electrocardiogram signal abnormality recognition and analysis system based on a multi-level clustering algorithm according to claim 8, characterized in that, The detailed process of the abnormal clustering analysis of the electrocardiogram signal data is as follows: For the noise-reduced electrocardiogram signal data, feature extraction is first performed. Using a neural network, the electrocardiogram signal data is input to obtain an output, and the output is represented as follows: Among them, zmid,j represents the value of the j-th neuron in the middle layer of the neural network, zi represents the value of the i-th neuron in the input layer of the neural network, ωj,i represents the weight between the j-th neuron in the middle layer of the neural network and the i-th neuron in the input layer of the neural network. A multi-layer middle layer is constructed, and multiple input-output operations are performed on the input data to obtain the value of the output layer. Then, through recursion, the feature representation is obtained, and the calculation formula is as follows: Among them, zout represents the vector composed of the values of the output layer of the neural network, and vout represents the feature representation. They respectively represent the vector composed of the values of the (t - 1)-th intermediate layer and the vector composed of the values of the t-th intermediate layer. Anomaly analysis is performed based on the feature representations of all collected electrocardiogram signal data, and all electrocardiogram signal data feature representations are Among them, respectively represent the feature representation of the first segment of electrocardiogram signal data, the feature representation of the second segment of electrocardiogram signal data, and the feature representation of the u-th segment of electrocardiogram signal data. Nu feature representations of electrocardiogram signal data are selected as the abnormal analysis clustering centers in the electrocardiogram signal data for abnormal clustering analysis. The formula is as follows: Among them, vcer,r represents the r-th abnormal analysis clustering center, and αu,r represents the clustering coefficient between the feature representation of the u-th segment of electrocardiogram signal data and the r-th abnormal analysis clustering center vcer,r. Based on the clustering coefficient, clustering of the electrocardiogram signal data feature representation is performed, and the clustering center related to the lowest clustering coefficient of the electrocardiogram signal feature representation is selected as the clustering cluster. Then, the mean values in different clustering clusters are selected as the new abnormal analysis clustering centers, and clustering is performed again until the clustering cluster no longer changes, and the electrocardiogram signal data is completed for clustering.

10. A system for abnormal recognition and analysis of electrocardiogram signals based on a multi-level clustering algorithm according to claim 1, characterized in that, The electrocardiogram signal judgment module compares the feature representation of the electrocardiogram signal data after clustering with the feature representation of the standard abnormal electrocardiogram signal, and automatically outputs the corresponding type of electrocardiogram signal abnormality.

Images