Multi-view electrocardiogram hash coding method and system, medium and equipment

Abstract

The invention discloses a multi-view electrocardiogram hash coding method and system, a medium and equipment, and relates to the technical field of intelligent data learning. The method comprises the following steps: acquiring a multi-lead electrocardiogram original signal, and constructing a first view according to the original signal; constructing a second view by using an autocorrelation function of the original signal, and constructing a third view of the original signal by using fast Fourier transform; extracting signal feature representation in each view by using a convolutional neural network; and fusing the three view features by adopting a linear projection technology to obtain a hash code of the electrocardiogram. According to the method, the defects of insufficient electrocardiogram single view signal features and low fusion efficiency can be overcome.

Description

Technical Field

[0001] The present invention relates to the field of intelligent data learning technology, and in particular to a multi-view electrocardiogram hash coding method, system, medium and device. 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] Electrocardiogram (ECG) is a diagnostic tool used to reflect the periodicity and time-dependence of surface myocardial electrical activity. ECG has become an indispensable means for clinical diagnosis of cardiovascular diseases due to its cost-effectiveness and efficiency. In recent years, with the rapid development of intelligent medical technology, machine learning-based ECG detection methods have become an important auxiliary tool for medical professionals to diagnose ECG. Although significant progress has been made in this field, developing lightweight models to further improve the accuracy of ECG abnormality detection remains a major challenge.

[0004] The current research field of multi-view electrocardiogram (ECG) learning is mostly based on the differences in heart positions reflected by different leads, and multiple leads of the ECG are regarded as different views. For example, a 12-lead ECG is divided into 6 signal views, and then a feature extractor is designed for each view, and later integration is achieved through fusion technology. However, the problem of insufficient signal features of a single view is prone to occur, resulting in unsatisfactory final results.

[0005] In addition, the difference in fusion technology also seriously affects the final result of ECG learning. Therefore, the effective fusion of multi-view features is crucial to improving the performance of deep networks. Existing fusion strategies mainly focus on directly weighted averaging the output features of each sub-network or introducing adaptive parameters to achieve the fusion of different view features, aiming to effectively integrate heterogeneous information. However, these methods often fail to fully consider the diversity and differences between different view features, resulting in the fused features failing to capture the core discriminative information.

[0006] In summary, how to overcome the defects of insufficient signal features of a single ECG view and low fusion efficiency has become an urgent problem to be solved in the existing technology. Summary of the invention

[0007] In view of the shortcomings of the prior art, the purpose of the present invention is to provide a multi-view electrocardiogram hash coding method, system, medium and device, which constructs a second view of the ECG signal through autocorrelation technology, constructs a third view using fast Fourier transform, and then constructs a convolutional neural network (CNN) feature extraction framework based on these three different perspectives, and finally uses a machine learning algorithm to generate a unified ECG hash code to support subsequent classification and detection tasks.

[0008] In order to achieve the above object, the present invention is implemented through the following technical solutions: A first aspect of the present invention provides a multi-view electrocardiogram hash coding method, comprising the following steps: Acquire multi-lead electrocardiogram original signals, and construct a first view according to the original signals; A second view is constructed using an autocorrelation function of the original signal, and a third view of the original signal is constructed using a fast Fourier transform; A convolutional neural network is used to extract the signal feature representation in each view, and the first view feature, the second view feature and the third view feature are obtained respectively; Linear projection technology is used to fuse the features of the three views and obtain the hash code of the ECG.

[0009] Furthermore, the autocorrelation function of the original signal is used to capture the linear correlation of the signal at any different time, and to find the periodic features submerged by the noise and the discarded fundamental frequency.

[0010] Furthermore, the specific steps of constructing the third view of the original signal using fast Fourier transform are: Perform fast Fourier transform on the original signal to achieve conversion from time domain to frequency domain and obtain frequency domain signal; Use filtering technology to reduce noise on frequency domain signals; The frequency domain signal after noise reduction is converted back to the time domain by inverse fast Fourier transform, so as to obtain the denoised electrocardiogram signal as the third view.

[0011] Furthermore, the convolutional neural network structure includes four convolutional layers, three pooling layers, three fully connected layers, and two Dropout layers.

[0012] Furthermore, the specific steps of using convolutional neural network to extract signal feature representation in each view are as follows: The convolutional layer is used to extract the local features of each lead in each view; Reduce the feature dimension of local features through a downsampling pooling layer; Use convolutional layers to perform cross-channel feature fusion on the features after dimensionality reduction; The cross-lead fused features are input into the convolutional layer and the fully connected layer to complete the convolutional neural network feature extraction.

[0013] Furthermore, the specific steps of using linear projection technology to fuse the features of the three views are as follows: The three view features are mapped to the Hamming space respectively using linear projection technology; Construct the objective function, solve the objective function according to the optimization strategy, and obtain the hash code of the electrocardiogram.

[0014] Furthermore, the specific steps of constructing the objective function are: Use matrix decomposition function to capture the hash code of each view and get the preliminary objective function; Define a shared hash code and establish a mapping relationship between the hash code of each view and the shared hash code; The residual factor is introduced, and the label information is used for learning to obtain the final objective function.

[0015] A second aspect of the present invention provides a multi-view electrocardiogram hash coding system, comprising: A data acquisition module is configured to acquire original electrocardiogram signals of multiple leads and construct a first view according to the original signals; A heterogeneous view construction module is configured to construct a second view using an autocorrelation function of the original signal and to construct a third view of the original signal using a fast Fourier transform; A feature extraction module is configured to extract a signal feature representation in each view using a convolutional neural network to obtain a first view feature, a second view feature, and a third view feature respectively; The feature fusion and hash coding module is configured to fuse the three view features using linear projection technology to obtain the hash code of the ECG.

[0016] A third aspect of the present invention provides a medium having a program stored thereon, which, when executed by a processor, implements the steps in the multi-view electrocardiogram hash coding method as described in the first aspect of the present invention.

[0017] The fourth aspect of the present invention provides a device, including a memory, a processor, and a program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps in the multi-view electrocardiogram hash coding method as described in the first aspect of the present invention are implemented.

[0018] One or more of the above technical solutions have the following beneficial effects: The present invention discloses a multi-view electrocardiogram hash coding method, system, medium and device. The multi-view electrocardiogram (ECG) learning in the current research field is mostly based on the differences in heart positions reflected by different leads, and multiple leads of the electrocardiogram are regarded as different views. For example, a 12-lead electrocardiogram is divided into 6 signal views, and then a feature extractor is designed for each view, and later integration is achieved through fusion technology. Different from the above method, the present invention proposes a new strategy, which is to construct multiple views from the original ECG signal, and then capture key information in combination with the original signal. This method can effectively supplement the deficiencies of single view signal features.

[0019] In the field of ECG signal preprocessing, noise reduction technology is widely used to eliminate various artifacts such as baseline drift, power line interference, electrode movement, and muscle artifacts. From the perspective of spectrum analysis, autocorrelation processing can be used as an effective strategy to alleviate white noise interference. The present invention uses an autocorrelation function to process the original signal, and obtains a second view of the signal rich in periodic components through autocorrelation operations. In addition, the fast Fourier transform (FFT) can not only realize the conversion of the signal from the time domain to the frequency domain, but also fully reveal the characteristics and structure of the signal through spectrum analysis. At the same time, with the help of processing methods such as filtering, FFT can effectively remove noise and significantly improve the anti-interference performance of the signal.

[0020] The present invention extracts features from three views by constructing a convolutional neural network feature extraction framework, and finally uses a machine learning algorithm to generate a unified ECG hash code. By introducing a learnable residual factor to construct an objective function and using an optimization strategy to solve it, a more efficient feature fusion is achieved, which greatly improves the stability and accuracy of hash code generation.

[0021] Advantages of additional aspects of the present invention will be given in part in the following description, and in part will become obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] 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.

[0023] Figure 1 This is a comparison diagram of the classification effects of three different signals of CNN in the PhysioNet / CinC Challenge 2017 in Example 1 of the present invention; Figure 2 This is a multi-view ECG hash coding framework diagram in Embodiment 1 of the present invention; Figure 3 This is a graph showing the percentage of samples in each category of the data set in the first embodiment of the present invention; Figure 4 This is a graph showing the ACC comparison results of each method in each category in Example 1 of the present invention. DETAILED DESCRIPTION

[0024] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present invention. 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 invention belongs.

[0025] 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 invention. 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; Embodiment 1: Embodiment 1 of the present invention provides a multi-view ECG hash coding method. In the field of ECG signal preprocessing, noise reduction technology is widely used to eliminate various artifacts such as baseline drift, power line interference, electrode movement, and muscle artifacts. From the perspective of spectrum analysis, autocorrelation processing can be used as an effective strategy to alleviate white noise interference. This embodiment uses an autocorrelation function to process the original signal, and obtains a second view of the signal rich in periodic components through autocorrelation operations. In addition, the fast Fourier transform (FFT) can not only realize the conversion of the signal from the time domain to the frequency domain, but also fully reveal the characteristics and structure of the signal through spectrum analysis. At the same time, with the help of processing methods such as filtering, FFT can effectively remove noise and significantly improve the anti-interference performance of the signal. Therefore, this embodiment uses FFT to construct a third view of the signal. Experiments based on the data set PhysioNet / CinC Challenge 2017 have proved that, for example Figure 1 As shown in Figure 3, different views contribute significantly to the recognition rate of each category, so combining multiple views helps capture more discriminative features.

[0026] The multi-view ECG hash coding method of this embodiment includes: first, constructing two new views of the original signal through the autocorrelation function and fast Fourier transform (FFT) of the signal; second, using a convolutional neural network (CNN) to extract the signal feature representation in each view; third, using linear projection technology to fuse the three extracted features; finally, realizing denoised and periodic signal hash coding, and establishing a mapping relationship between the hash dictionary and label information.

[0027] To achieve the above goals, given an ECG dataset ,in Indicates that it contains The training set of examples, represents the set of real numbers, Indicates Examples: Represents the corresponding onehot tag set, Indicates The labels corresponding to the samples are represents the number of categories; and represents the feature dimension of the training examples, Indicates the number of channels or leads. is the number of training signals. The goal of this invention is to find a mapping The coding representation of ECG signals is realized, and then the signal recognition and detection can be realized quickly. The whole method consists of three parts: heterogeneous view construction module, CNN feature extraction module, feature fusion and hash coding module.

[0028] like Figure 2 As shown, the specific steps include: Step 1: Obtain multi-lead electrocardiogram original signals and construct a first view based on the original signals.

[0029] Specifically, the original signals of multiple leads are constructed into a view, namely the first view. The purpose of this is to maintain the correlation between multiple leads of the same signal. Before inputting into the convolutional neural network, the first view is normalized.

[0030] Step 2: Use the autocorrelation function of the original signal to construct the second view, and use the fast Fourier transform to construct the third view of the original signal.

[0031] Step 2.1: The autocorrelation function of the original signal is used to capture the linear correlation of the signal at any different time, looking for periodic features that may be submerged by noise and fundamental frequencies that may be discarded.

[0032] In this embodiment, the autocorrelation function is defined as: (1).

[0033] in, is the autocorrelation function, is the expectation operator, and Represents two different moments. The above function can be used to calculate the second view of the original signal ,from Figure 1 The results show that the signal under this view has a particularly obvious noise reduction effect.

[0034] (2).

[0035] Step 2.2: Construct a third view of the original signal using Fast Fourier Transform.

[0036] Step 2.2.1: Perform fast Fourier transform (FFT) on the original signal to achieve conversion from time domain to frequency domain and obtain frequency domain signal.

[0037] FFT technology can reveal features that are difficult to identify in the time domain. By converting the signal to the frequency domain, it can capture key spectral information, thereby significantly improving the distinguishability of features.

[0038] Step 2.2.2: Use filtering technology to reduce the noise of the frequency domain signal and effectively remove the noise component.

[0039] Step 2.2.3: Use the inverse fast Fourier transform (IFFT) to convert the denoised frequency domain signal back to the time domain to obtain the denoised ECG signal as the third view. .

[0040] (3).

[0041] in represents the fast Fourier transform operator, represents the filter, Represents the inverse fast Fourier transform operator.

[0042] This allows us to construct three views of the ECG signal, thereby constructing a three-view dataset , represents the view number, and , represents training examples, Indicates the corresponding multi-view sharing label.

[0043] Step 3: Use a convolutional neural network to extract the signal feature representation in each view, and obtain the first view feature, the second view feature, and the third view feature respectively.

[0044] In this embodiment, three feature extraction networks are included. The three CNN feature extraction networks have the same structure. For each view, the convolutional neural network structure includes four convolutional layers, three pooling layers, three fully connected layers, and two Dropout layers. The network integrates the features of electrocardiogram (ECG) signals of different leads through the following steps: Step 3.1: Use convolutional layers to extract local features of each lead in each view.

[0045] Specifically, multi-lead ECG heartbeat signals are input into two convolutional layers with a convolution kernel size of 1*3 to extract local features of each lead.

[0046] Step 3.2: Reduce the feature dimension of local features through a downsampling pooling layer to simplify the model and reduce the risk of overfitting.

[0047] Step 3.3: Use the convolutional layer to perform cross-channel feature fusion on the features after dimensionality reduction.

[0048] A convolution layer with a convolution kernel size of 2*1 is used for feature fusion, so that the features of any two leads can be merged.

[0049] Step 3.4: Input the cross-lead fused features into the convolutional layer and the fully connected layer to complete the convolutional neural network feature extraction.

[0050] The cross-lead fused features are input into the 1*1 convolutional layer and the subsequent three fully connected layers to complete the convolutional neural network feature extraction. The features extracted by CNN are .

[0051] Step 4: Linear projection technology is used to fuse the first view features, the second view features and the third view features to obtain the hash code of the electrocardiogram.

[0052] Step 4.1: Use linear projection technology to map the first view feature, the second view feature and the third view feature to the Hamming space respectively.

[0053] This embodiment aims to transform CNN features Represented as a shared hash code. Given the essential differences in the features of each view, directly mapping them to the shared hash code will cause energy loss. Therefore, the features of each view are mapped to the Hamming space and then fused.

[0054] Step 4.2: Construct the objective function, solve the objective function according to the optimization strategy, and obtain the hash code of the electrocardiogram.

[0055] Step 4.2.1: Use the matrix decomposition function to capture the hash code of each view and obtain the preliminary objective function.

[0056] The matrix decomposition function is used to capture the hash code of each view. The specific objective function to be minimized is for: (4).

[0057] in for The unit matrix, is the orthogonal projection factor, A hash representation of each view.

[0058] Step 4.2.2: Define a shared hash code and establish a mapping relationship between the hash code of each view and the shared hash code.

[0059] Assume the shared hash code is , establish a mapping relationship between the hash code of each view and the shared hash code as shown below: (5).

[0060] in represents the objective function to be minimized, is the correction factor between shared hash coding and single view hash coding.

[0061] Step 4.2.3: Introduce the residual factor and use the label information for learning to obtain the final objective function.

[0062] In order to further improve the discriminability of hash coding, label information is used for learning. Given the label matrix is a one-hot matrix with most of its elements being zero. The label matrix is ​​directly combined with the hash code Establishing a mapping relationship may lead to significant energy loss and is not conducive to improving the discriminative performance of hash coding. Therefore, a learnable residual factor is introduced , in order to reduce the difference between the two and optimize the discriminative power of hash coding. The specific method is as follows: (6).

[0063] in represents the objective function to be minimized, is an orthogonal basis used to learn semantic hash coding representations. is a learnable residual factor. In order to increase the difference between heterogeneous classes and reduce the numerical difference between the Hamming space and the label space, the initial To simplify the optimization, let represents the label with residual factors, then the target can be rewritten as: (7).

[0064] in Represents the regularization parameter to prevent over-learning.

[0065] In summary, the overall objective function It can be defined as: (8).

[0066] in, They represent different objective functions that need to be minimized.

[0067] Step 4.2.4: Solve the objective function according to the optimization strategy.

[0068] Next, give the goal The binary discrete constraint of the Hamming space causes the objective function to be non-differentiable, which leads to , are difficult to optimize, and and All of them have orthogonal constraints, and gradient descent cannot be used directly to optimize the target. Therefore, different optimization schemes are set for each variable to achieve the target. overall optimization.

[0069] renew :Eliminate the objective function with Irrelevant items, fixed and , you can target It can be simplified into the following optimization subproblem: (9).

[0070] use The gradient of According to the definition of the residual factor, the expected tag learned is expected to contain the real label element, so according to To revise The update rule is: (10).

[0071] renew :Eliminate the objective function with Irrelevant items, fixed , you can target It can be simplified into the following optimization subproblem: (11).

[0072] because ,therefore , so the target Simplified again to: (12).

[0073] question Corresponding to the classic orthogonal Procrustes problem, singular value decomposition (SVD) can be used to achieve the optimization goal. , so we can use calculate : (13).

[0074] renew :Eliminate the objective function with Irrelevant items, fixed ,and , so the target It can be simplified into the following optimization subproblem: (14).

[0075] because , so , so the target can be simplified to: (15).

[0076] From the inequality relationship, we can know that: (16).

[0077] renew :Eliminate the objective function with Irrelevant items, fixed ,Target It can be simplified into the following optimization subproblem: (17).

[0078] The discrete constraints lead to the goal Difficult to optimize, so The constraints are relaxed so that , then the above objective can be derived: (18).

[0079] Utilization We can get: (19).

[0080] renew :Eliminate the objective function with Irrelevant items, fixed , you can target It can be simplified into the following optimization subproblem: (20).

[0081] because ， , so , the above formula can be simplified to: (twenty one).

[0082] Through something similar to The SVD update strategy is Then there is the following update rule: (twenty two).

[0083] renew :Eliminate the objective function with Irrelevant items, fixed ,according to , you can target It can be simplified into the following optimization subproblem: (twenty three).

[0084] From the inequality relationship, we can know that: (twenty four).

[0085] In order to verify the effectiveness of the proposed multi-view ECG collaborative hash coding, the public 2017 PhysioNet / CinC Challenge ECG dataset is used for verification. The dataset consists of 8528 30-60 seconds ECG segmentations, including four categories: normal sinus rhythm , atrial fibrillation , other heart rhythms , Noise ECG .

[0086] In order to verify the discriminative ability of ECG co-hash coding, 6000 samples were selected for training and the rest for testing. SVM classifier and several popular deep models were selected for comparison, including BiLSTM, CNN, and CNNBiLSTM. Figure 3 As shown, normal sinus rhythm accounts for 59%, atrial fibrillation accounts for 9%, other rhythms account for 29%, and noisy ECG accounts for 3%. For this purpose, multiple metrics are selected for experiments. The metrics include: ACC, F1, G_means, Precise, Recall, and Specificity. By comparing the predicted category with the actual category, four key parameters are obtained: true positive (TP), false positive (FP), true negative (TN), and false negative (FN). Then the above metrics are defined as follows: .

[0087] .

[0088] .

[0089] Table 1 Comparison of various methods under multiple evaluation criteria

[0090] The experimental results are shown in Table 1, which shows that the multi-view ECG hash coding method proposed in this embodiment can effectively make up for the shortcomings of single view signal features by fusing multi-view features. Under multiple evaluation criteria, this method is significantly better than the current popular deep model. Multi-view ECG hash coding fully considers the diversity and differences between different view features, effectively integrates heterogeneous information, and thus can capture the most core discriminant information. Figure 4 The experimental results show that multi-view feature fusion is particularly effective in improving the recognition effect of low recognition rate categories, which is consistent with Figure 1 The experimental results are consistent.

[0091] Embodiment 2: Embodiment 2 of the present invention provides a multi-view electrocardiogram hash coding system, including: A data acquisition module is configured to acquire original electrocardiogram signals of multiple leads and construct a first view according to the original signals; A heterogeneous view construction module is configured to construct a second view using an autocorrelation function of the original signal and to construct a third view of the original signal using a fast Fourier transform; A feature extraction module is configured to extract a signal feature representation in each view using a convolutional neural network to obtain a first view feature, a second view feature, and a third view feature respectively; The feature fusion and hash coding module is configured to fuse the features of the three views using a linear projection technique to obtain a hash code of the electrocardiogram.

[0092] Embodiment three: Embodiment 3 of the present invention provides a medium on which a program is stored. When the program is executed by a processor, the steps in the multi-view electrocardiogram hash coding method as described in Embodiment 1 of the present invention are implemented.

[0093] Embodiment 4: Embodiment 4 of the present invention 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 multi-view electrocardiogram hash encoding method as described in Embodiment 1 of the present invention are implemented.

[0094] The steps involved in the above embodiments 2, 3 and 4 correspond to the method embodiment 1. For the specific implementation methods, please refer to the relevant description part of embodiment 1.

[0095] 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.

[0096] 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 multi-view electrocardiogram hash coding method, characterized in that: The following steps are involved: Acquire multi-lead electrocardiogram original signals, and construct a first view according to the original signals; A second view is constructed using an autocorrelation function of the original signal, and a third view of the original signal is constructed using a fast Fourier transform; A convolutional neural network is used to extract the signal feature representation in each view, and the first view feature, the second view feature and the third view feature are obtained respectively; Linear projection technology is used to fuse the features of the three views and obtain the hash code of the ECG.

2. The multi-view electrocardiogram hash coding method according to claim 1, characterized in that: The autocorrelation function of the original signal is used to capture the linear correlation of the signal at any different time, and to find the periodic features submerged by the noise and the discarded fundamental frequency.

3. The multi-view electrocardiogram hash coding method according to claim 1, characterized in that: The specific steps of constructing the third view of the original signal using fast Fourier transform are: Perform fast Fourier transform on the original signal to achieve conversion from time domain to frequency domain and obtain frequency domain signal; Use filtering technology to reduce noise on frequency domain signals; The frequency domain signal after noise reduction is converted back to the time domain by inverse fast Fourier transform, so as to obtain the denoised electrocardiogram signal as the third view.

4. The multi-view electrocardiogram hash coding method according to claim 1, characterized in that: The convolutional neural network structure consists of four convolutional layers, three pooling layers, three fully connected layers, and two Dropout layers.

5. The multi-view electrocardiogram hash coding method as claimed in claim 4, characterized in that: The specific steps of using convolutional neural network to extract signal feature representation in each view are: The convolutional layer is used to extract the local features of each lead in each view; Reduce the feature dimension of local features through a downsampling pooling layer; Use convolutional layers to perform cross-channel feature fusion on the features after dimensionality reduction; The cross-lead fused features are input into the convolutional layer and the fully connected layer to complete the convolutional neural network feature extraction.

6. The multi-view electrocardiogram hash coding method according to claim 1, characterized in that: The specific steps of fusing the three view features using linear projection technology are as follows: The three view features are mapped to the Hamming space respectively using linear projection technology; Construct the objective function, solve the objective function according to the optimization strategy, and obtain the hash code of the electrocardiogram.

7. The multi-view electrocardiogram hash coding method as claimed in claim 1, characterized in that: The specific steps to construct the objective function are: Use matrix decomposition function to capture the hash code of each view and get the preliminary objective function; Define a shared hash code and establish a mapping relationship between the hash code of each view and the shared hash code; The residual factor is introduced, and the label information is used for learning to obtain the final objective function.

8. A multi-view electrocardiogram hash coding system, characterized in that: include: A data acquisition module is configured to acquire original electrocardiogram signals of multiple leads and construct a first view according to the original signals; A heterogeneous view construction module is configured to construct a second view using an autocorrelation function of the original signal and to construct a third view of the original signal using a fast Fourier transform; A feature extraction module is configured to extract a signal feature representation in each view using a convolutional neural network to obtain a first view feature, a second view feature, and a third view feature respectively; The feature fusion and hash coding module is configured to fuse the three view features using linear projection technology to obtain the hash code of the ECG.

9. 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 multi-view electrocardiogram hash coding method described in any one of claims 1-7.

10. A terminal device, characterized in that: The invention comprises a processor and a computer-readable storage medium, wherein the processor is used to implement various instructions; and the computer-readable storage medium is used to store a plurality of instructions, wherein the instructions are suitable for being loaded by the processor and executing the multi-view electrocardiogram hash coding method according to any one of claims 1 to 7.

Images