阅读说明：本技术 基于机器学习的心电数据分类方法及装置 (Machine learning-based electrocardiogram data classification method and device ) 是由 徐啸 李晓宇 孙瑜尧 于 2021-08-31 设计创作，主要内容包括：本发明涉及人工智能技术领域,揭露一种基于机器学习的心电数据分类的方法,包括：获取对心电数据处理得到的多个等长的心跳数据；响应于心电分类任务的触发指令,将多个等长的心跳数据输入至预先构建的心电分类模型中进行预测,心电分类模型包括多个表征模块和心跳聚合模块,使用表征模块提取多个等长的心跳数据的向量表征,使用心跳聚合模块融合多个等长的心跳数据的向量表征之间的信息；将融合后心跳数据的向量表征处理到心跳数据在不同类别上的映射关系,输出心电数据的分类结果。本发明通过将心跳周期性融入数据形式与模型结构中,能够降低训练过程中模型解耦所需的计算复杂度,提高模型应用效果,使得心电数据分类结果具有更高的准确度。(The invention relates to the technical field of artificial intelligence, and discloses a machine learning-based electrocardiogram data classification method, which comprises the following steps: acquiring a plurality of heartbeat data with equal length obtained by processing the electrocardiographic data; responding to a triggering instruction of an electrocardio classification task, inputting a plurality of isometric heartbeat data into a pre-constructed electrocardio classification model for prediction, wherein the electrocardio classification model comprises a plurality of representation modules and a heartbeat aggregation module, the representation modules are used for extracting vector representations of the isometric heartbeat data, and the heartbeat aggregation module is used for fusing information among the vector representations of the isometric heartbeat data; and processing the vector representation of the fused heartbeat data to the mapping relation of the heartbeat data in different categories, and outputting the classification result of the electrocardio data. According to the method, the heartbeat is periodically fused into the data form and the model structure, so that the calculation complexity required by model decoupling in the training process can be reduced, the application effect of the model is improved, and the electrocardio data classification result has higher accuracy.)

1. A machine learning-based electrocardiogram data classification method is characterized by comprising the following steps:

acquiring a plurality of heartbeat data with equal length obtained by processing the electrocardiographic data, wherein the plurality of heartbeat data with equal length have the same time length;

responding to a triggering instruction of an electrocardio classification task, inputting the isometric heartbeat data into a pre-constructed electrocardio classification model for prediction, wherein the electrocardio classification model comprises a plurality of characterization modules and a heartbeat aggregation module, the characterization modules are used for extracting vector characterization of the isometric heartbeat data, and the heartbeat aggregation module is used for fusing information among the vector characterization of the isometric heartbeat data;

and processing the vector representation of the fused heartbeat data to the mapping relation of the heartbeat data in different categories, and outputting the classification result of the electrocardio data.

2. The method according to claim 1, wherein the acquiring a plurality of heartbeat data of equal length obtained by processing the electrocardiographic data specifically includes:

segmenting the electrocardiogram data to obtain electrocardiogram segments with different time lengths, wherein the electrocardiogram segments are a heartbeat cycle;

and performing resampling processing on the electrocardio segments with different time lengths to obtain a plurality of heartbeat data with equal length.

3. The method of claim 1, wherein before said triggering instruction in response to an electrocardiographic classification task inputs said plurality of heartbeat data of equal length into a pre-constructed electrocardiographic classification model for prediction, said method further comprises:

based on a preset electrocardiogram classification task label, inputting a plurality of heartbeat samples with equal length into a network model for training, and constructing an electrocardiogram classification model, wherein the network model comprises a plurality of characterization modules and a heartbeat aggregation module, the characterization modules are used for learning vector characterization of a single heartbeat sample, and the heartbeat aggregation module is used for fusing information among the vector characterization of the plurality of heartbeat samples;

the method comprises the following steps of inputting a plurality of heartbeat samples with equal length into a network model for training based on a preset electrocardio classification task label, and constructing an electrocardio classification model, wherein the method specifically comprises the following steps:

determining a category label and a corresponding resampling factor corresponding to each heartbeat sample in the plurality of heartbeat samples with equal length based on a preset electrocardio classification task label, wherein the resampling factor is a numerical value formed by resampling frequency proportion quantization;

splicing the heartbeat samples carrying the category labels with corresponding resampling factors, and inputting the heartbeat samples to different characterization modules for training to obtain vector characterization of each heartbeat sample;

and summarizing vector characteristics of the heartbeat samples output by the characteristic modules, inputting the vector characteristics to the aggregation module for training, obtaining the mapping relation of the heartbeat samples on different classes, and constructing the electrocardio classification model.

4. The method according to claim 3, wherein the characterization modules include a normalization layer and a self-attention layer, and the obtaining of the vector characterization of each heartbeat sample includes:

after being subjected to linear mapping, heartbeat samples carrying label categories are spliced with corresponding resampling factors and then are respectively input to a single heartbeat characterization module for training;

performing feature coding on the spliced heartbeat samples by using the self-attention layer of the characterization module to obtain the vector characterization of a single heartbeat sample;

mapping the vector representations of the single heartbeat sample to a stably distributed vector space using a normalization layer of the representation module such that the encoded vector representations have the same dimensions.

5. The method according to claim 4, wherein a windowing mechanism is added to the self-attention layer, and the feature coding is performed on the spliced heartbeat samples by using the self-attention layer of the characterization module to obtain the vector characterization of a single heartbeat sample, specifically including:

dividing the spliced heartbeat samples by using a preset window length by using the window mechanism to obtain heartbeat sample data of a plurality of windows;

and enabling the self-attention layer of the characterization module to act on the heartbeat sample data of each window to obtain the vector characterization of the single heartbeat sample.

6. The method according to claim 3, wherein the heartbeat aggregation module includes a convolution layer, a pooling layer and a normalization layer, and the vector characterization of the heartbeat samples output by each characterization module is summarized and then input to the aggregation module for training to obtain the mapping relationships of the heartbeat samples in different categories, so as to construct the electrocardiogram classification model, specifically including:

summarizing vector characteristics of the heartbeat samples output by the characterization modules, inputting the summarized vector characteristics into a convolution layer of the aggregation module, and extracting the characteristics of the vector characteristics of the summarized heartbeat samples;

mapping the vector representation after feature extraction to a stably distributed vector space by utilizing a normalization layer of the aggregation module, so that the vector representation after feature extraction has the same dimensionality;

and performing dimensionality reduction processing on the vector representations with the same dimensionality after feature extraction by using the pooling layer of the aggregation module to obtain the mapping relation of the heartbeat samples on different classes, and constructing the electrocardiogram classification model.

7. The method according to any one of claims 1-6, wherein after the processing the vector representation of the fused heartbeat data to the mapping relationship of the heartbeat data on different classes and outputting the classification result of the electrocardiographic data, the method further comprises:

recording the distribution characteristics of the electrocardiogram data according to the classification result of the electrocardiogram data;

and mapping the distribution characteristics of the electrocardiogram data to each dimension index set by the electrocardiogram classification task so as to perform abnormal evaluation on each dimension index.

8. An apparatus for machine learning-based classification of electrocardiographic data, the apparatus comprising:

the acquisition unit is used for acquiring a plurality of heartbeat data with equal length obtained by processing the electrocardiogram data, and the plurality of heartbeat data with equal length have the same time length;

the prediction unit is used for responding to a trigger instruction of an electrocardio classification task, inputting the isometric heartbeat data into a pre-constructed electrocardio classification model for prediction, wherein the electrocardio classification model comprises a plurality of characterization modules and a heartbeat aggregation module, the characterization modules are used for extracting vector characterization of the isometric heartbeat data, and the heartbeat aggregation module is used for fusing information among the vector characterization of the isometric heartbeat data;

and the output unit is used for processing the vector representation of the fused heartbeat data to the mapping relation of the heartbeat data in different categories and outputting the classification result of the electrocardio data.

9. A computer device comprising a memory and a processor, the memory storing a computer program, wherein the processor implements the steps of the method of any one of claims 1 to 7 when executing the computer program.

10. A computer storage medium on which a computer program is stored, characterized in that the computer program, when being executed by a processor, carries out the steps of the method of any one of claims 1 to 7.

Technical Field

The invention relates to the technical field of artificial intelligence, in particular to a method and a device for classifying electrocardiogram data based on machine learning, computer equipment and a computer storage medium.

Background

With the development of fire and heat in machine learning, more and more fields combine the machine learning method and the own field, and the medical field is no exception, for example: through the classification, prediction and the like of the electrocardiogram data and the analysis of a large amount of electrocardiogram data, doctors can be helped to find a plurality of heart diseases, such as atrial fibrillation, myocardial infarction, acute hypotension and the like, and further judge the heart condition of patients.

In the prior art, machine learning can well realize the description of a complex nonlinear function mapping relation by means of a multilayer neural network model, and can autonomously mine the hidden characteristics behind the electrocardio data, thereby completing the classification task of the electrocardio data. However, due to the rapid development of the novel sensing technology, the electrocardiographic data to be analyzed is not only huge in quantity but also complex in structure, the characteristics of the electrocardiographic data are not considered when the machine learning method is used for classifying the electrocardiographic data, and the calculation complexity required by the model structure in the training process is high, so that the application effect of the model is not ideal, and the accuracy of the electrocardiographic data classification result is affected.

Disclosure of Invention

In view of the above, the present invention provides a method, an apparatus, a computer device and a computer storage medium for classifying electrocardiographic data based on machine learning, and mainly aims to solve the problems that in the prior art, the method using machine learning requires high computational complexity for a model structure in a training process, so that the application effect of the model is not ideal, and the accuracy of the electrocardiographic data classification result is affected.

According to an aspect of the present invention, there is provided a method for classifying electrocardiographic data based on machine learning, the method comprising:

acquiring a plurality of heartbeat data with equal length obtained by processing the electrocardiographic data, wherein the plurality of heartbeat data with equal length have the same time length;

responding to a triggering instruction of an electrocardio classification task, inputting the isometric heartbeat data into a pre-constructed electrocardio classification model for prediction, wherein the electrocardio classification model comprises a plurality of characterization modules and a heartbeat aggregation module, the characterization modules are used for extracting vector characterization of the isometric heartbeat data, and the heartbeat aggregation module is used for fusing information among the vector characterization of the isometric heartbeat data;

and processing the vector representation of the fused heartbeat data to the mapping relation of the heartbeat data in different categories, and outputting the classification result of the electrocardio data.

In another embodiment of the present invention, the acquiring a plurality of heartbeat data with equal length obtained by processing electrocardiographic data specifically includes:

segmenting the electrocardiogram data to obtain electrocardiogram segments with different time lengths, wherein the electrocardiogram segments are a heartbeat cycle;

and performing resampling processing on the electrocardio segments with different time lengths to obtain a plurality of heartbeat data with equal length.

In another embodiment of the present invention, before the triggering instruction responding to the electrocardiographic classification task inputs the plurality of heartbeat data with equal length into a pre-constructed electrocardiographic classification model for prediction, the method further includes:

based on a preset electrocardiogram classification task label, inputting a plurality of heartbeat samples with equal length into a network model for training, and constructing an electrocardiogram classification model, wherein the network model comprises a plurality of characterization modules and a heartbeat aggregation module, the characterization modules are used for learning vector characterization of a single heartbeat sample, and the heartbeat aggregation module is used for fusing information among the vector characterization of the plurality of heartbeat samples;

the method comprises the following steps of inputting a plurality of heartbeat samples with equal length into a network model for training based on a preset electrocardio classification task label, and constructing an electrocardio classification model, wherein the method specifically comprises the following steps:

determining a category label and a corresponding resampling factor corresponding to each heartbeat sample in the plurality of heartbeat samples with equal length based on a preset electrocardio classification task label, wherein the resampling factor is a numerical value formed by resampling frequency proportion quantization;

splicing the heartbeat samples carrying the category labels with corresponding resampling factors, and inputting the heartbeat samples to different characterization modules for training to obtain vector characterization of each heartbeat sample;

and summarizing vector characteristics of the heartbeat samples output by the characteristic modules, inputting the vector characteristics to the aggregation module for training, obtaining the mapping relation of the heartbeat samples on different classes, and constructing the electrocardio classification model.

In another embodiment of the present invention, the characterization module includes a normalization layer and a self-attention layer, the heartbeat samples with the category labels are spliced with the corresponding resampling factors and then input to different characterization modules for training, so as to obtain a vector characterization of each heartbeat sample, which specifically includes:

after being subjected to linear mapping, heartbeat samples carrying label categories are spliced with corresponding resampling factors and then are respectively input to a single heartbeat characterization module for training;

performing feature coding on the spliced heartbeat samples by using the self-attention layer of the characterization module to obtain the vector characterization of a single heartbeat sample;

mapping the vector representations of the single heartbeat sample to a stably distributed vector space using a normalization layer of the representation module such that the encoded vector representations have the same dimensions.

In another embodiment of the present invention, a windowing mechanism is added to the self-attention layer, and the feature coding is performed on the spliced heartbeat samples by using the self-attention layer of the characterization module to obtain the vector characterization of a single heartbeat sample, which specifically includes:

dividing the spliced heartbeat samples by using a preset window length by using the window mechanism to obtain heartbeat sample data of a plurality of windows;

and enabling the self-attention layer of the characterization module to act on the heartbeat sample data of each window to obtain the vector characterization of the single heartbeat sample.

In another embodiment of the present invention, the heartbeat aggregation module includes a convolution layer, a pooling layer, and a normalization layer, and the vector characterizations of the heartbeat samples output by the respective characterization modules are summarized and then input to the aggregation module for training, so as to obtain mapping relationships of the heartbeat samples in different categories and construct the electrocardiogram classification model, which specifically includes:

summarizing vector characteristics of the heartbeat samples output by the characterization modules, inputting the summarized vector characteristics into a convolution layer of the aggregation module, and extracting the characteristics of the vector characteristics of the summarized heartbeat samples;

mapping the vector representation after feature extraction to a stably distributed vector space by utilizing a normalization layer of the aggregation module, so that the vector representation after feature extraction has the same dimensionality;

and performing dimensionality reduction processing on the vector representations with the same dimensionality after feature extraction by using the pooling layer of the aggregation module to obtain the mapping relation of the heartbeat samples on different classes, and constructing the electrocardiogram classification model.

In another embodiment of the present invention, after the processing the vector representation of the fused heartbeat data to the mapping relationship of the heartbeat data in different categories and outputting the classification result of the electrocardiographic data, the method further includes:

recording the distribution characteristics of the electrocardiogram data according to the classification result of the electrocardiogram data;

and mapping the distribution characteristics of the electrocardiogram data to each dimension index set by the electrocardiogram classification task so as to perform abnormal evaluation on each dimension index.

According to another aspect of the present invention, there is provided an apparatus for classifying electrocardiographic data based on machine learning, the apparatus comprising:

the acquisition unit is used for acquiring a plurality of heartbeat data with equal length obtained by processing the electrocardiogram data, and the plurality of heartbeat data with equal length have the same time length;

the prediction unit is used for responding to a trigger instruction of an electrocardio classification task, inputting the isometric heartbeat data into a pre-constructed electrocardio classification model for prediction, wherein the electrocardio classification model comprises a plurality of characterization modules and a heartbeat aggregation module, the characterization modules are used for extracting vector characterization of the isometric heartbeat data, and the heartbeat aggregation module is used for fusing information among the vector characterization of the isometric heartbeat data;

and the output unit is used for processing the vector representation of the fused heartbeat data to the mapping relation of the heartbeat data in different categories and outputting the classification result of the electrocardio data.

In another embodiment of the present invention, the obtaining unit includes:

the segmentation module is used for segmenting the electrocardiogram data to obtain electrocardiogram fragments with different time lengths, wherein the electrocardiogram fragments are a heartbeat cycle;

and the processing module is used for performing resampling processing on the electrocardio segments with different time lengths to obtain a plurality of isometric heartbeat data.

In another embodiment of the present invention, the apparatus further comprises:

the constructing unit is used for inputting a plurality of heartbeat samples with equal length into a network model for training and constructing the electrocardio classification model based on a preset electrocardio classification task label before inputting the plurality of heartbeat data with equal length into a pre-constructed electrocardio classification model for prediction in response to a triggering instruction of an electrocardio classification task, wherein the network model comprises a plurality of characterization modules and a heartbeat aggregation module, the characterization modules are used for learning vector characterization of a single heartbeat sample, and the heartbeat aggregation module is used for fusing information among the vector characterization of the plurality of heartbeat samples;

the construction unit includes:

the determination module is used for determining a category label and a corresponding resampling factor corresponding to each heartbeat sample in the plurality of heartbeat samples with equal lengths based on a preset electrocardio classification task label, wherein the resampling factor is a numerical value formed by the frequency scale quantization of resampling;

the training module is used for splicing the heartbeat samples carrying the category labels with the corresponding resampling factors and then respectively inputting the heartbeat samples to different characterization modules to obtain vector characterization of each heartbeat sample;

and the construction module is used for summarizing vector characteristics of the heartbeat samples output by the characterization modules and inputting the vector characteristics to the aggregation module to obtain the mapping relation of the heartbeat samples on different categories and construct the electrocardio classification model.

In another embodiment of the present invention, the characterization module includes a normalization layer and a self-attention layer, and the training module includes:

the splicing submodule is used for splicing the heartbeat samples carrying the label categories with corresponding resampling factors after linear mapping, and inputting the heartbeat samples to the single heartbeat characterization module for training;

the coding submodule is used for carrying out feature coding on the spliced heartbeat samples by utilizing the self-attention layer of the characterization module to obtain the vector characterization of a single heartbeat sample;

a first mapping sub-module, configured to map, by using a normalization layer of the characterization module, the vector characterization of the single heartbeat sample to a stably distributed vector space, so that the encoded vector characterizations have the same dimension.

In another embodiment of the present invention, the coding sub-module is specifically configured to divide the spliced heartbeat samples by using a preset window length by using the window mechanism, so as to obtain heartbeat sample data of multiple windows;

the coding submodule is specifically further configured to apply the self-attention layer of the characterization module to the heartbeat sample data of each window to obtain a vector characterization of a single heartbeat sample.

In another embodiment of the present invention, the heartbeat aggregation module includes a convolutional layer, a pooling layer, and a normalization layer, and the building module includes:

the extraction submodule is used for summarizing the vector characteristics of the heartbeat samples output by the characterization modules, inputting the summarized vector characteristics into the convolution layer of the aggregation module, and extracting the characteristics of the vector characteristics of the summarized heartbeat samples;

the second mapping submodule is used for mapping the vector representation after the feature extraction to a stably distributed vector space by utilizing the normalization layer of the aggregation module so as to enable the vector representation after the feature extraction to have the same dimensionality;

and the dimensionality reduction sub-module is used for performing dimensionality reduction processing on the vector representations with the same dimensionality after the features are extracted by utilizing the pooling layer of the aggregation module to obtain the mapping relation of the heartbeat samples on different classes and construct the electrocardiogram classification model.

In another embodiment of the present invention, the apparatus further comprises:

the recording unit is used for recording the distribution characteristics of the electrocardiogram data according to the classification result of the electrocardiogram data after the vector representation of the fused heartbeat data is processed to the mapping relation of the heartbeat data in different classes and the classification result of the electrocardiogram data is output;

and the evaluation unit is used for mapping the distribution characteristics of the electrocardiogram data to each dimension index set by the electrocardiogram classification task so as to carry out abnormal evaluation on each dimension index.

According to a further aspect of the invention, a computer device is provided, comprising a memory storing a computer program and a processor implementing the steps of the method for machine learning based classification of electrocardiographic data when executing the computer program.

According to a further aspect of the invention, a computer storage medium is provided, on which a computer program is stored which, when being executed by a processor, carries out the steps of the method for machine learning based classification of electrocardiographic data.

By means of the technical scheme, the invention provides a method and a device for classifying electrocardiographic data based on machine learning, wherein a plurality of heart beat data with equal length obtained by processing the electrocardiographic data are obtained, the heart beat data with equal length have the same time length, the heart beat data with equal length are input into a pre-constructed electrocardiographic classification model for prediction in response to a trigger instruction of an electrocardiographic classification task, the electrocardiographic classification model comprises a plurality of characterization modules and a heart beat aggregation module, vector representations of the heart beat data with equal length are extracted by the characterization modules, the heart beat aggregation module is used for fusing information among the vector representations of the heart beat data with equal length, the vector representations of the fused heart beat data are further processed to be mapping relations of the heart beat data in different classes, and classification results of the electrocardiographic data are output. Compared with the prior art that the classification process of the electrocardiogram data is carried out by means of a multilayer neural network structure, the periodic characteristics of the electrocardiogram data are integrated into a data form and a model structure, and the network model can provide higher calculation efficiency and better interpretability, so that the network model trained aiming at the heartbeat data with equal length has a better classification effect, the calculation complexity required by model decoupling in the training process can be reduced, and the classification result of the electrocardiogram data has higher accuracy.

Drawings

Various other advantages and benefits will become apparent to those of ordinary skill in the art upon reading the following detailed description of the preferred embodiments. The drawings are only for purposes of illustrating the preferred embodiments and are not to be construed as limiting the invention. Also, like reference numerals are used to refer to like parts throughout the drawings. In the drawings:

fig. 1 is a schematic flow chart illustrating a method for classifying electrocardiographic data based on machine learning according to an embodiment of the present invention;

FIG. 2 is a flow chart illustrating another method for machine learning-based classification of electrocardiographic data according to an embodiment of the present invention;

fig. 3 is a schematic structural diagram of an apparatus for classifying electrocardiographic data based on machine learning according to an embodiment of the present invention;

fig. 4 is a schematic structural diagram of another apparatus for classifying electrocardiographic data based on machine learning according to an embodiment of the present invention.

Detailed Description

Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be embodied in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

The embodiment of the application can acquire and process related data based on an artificial intelligence technology. Among them, Artificial Intelligence (AI) is a theory, method, technique and application system that simulates, extends and expands human Intelligence using a digital computer or a machine controlled by a digital computer, senses the environment, acquires knowledge and uses the knowledge to obtain the best result.

The artificial intelligence infrastructure generally includes technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technologies, operation/interaction systems, mechatronics, and the like. The artificial intelligence software technology mainly comprises a computer vision technology, a robot technology, a biological recognition technology, a voice processing technology, a natural language processing technology, machine learning/deep learning and the like.

The embodiment of the invention provides a machine learning-based electrocardiogram data classification method, which can reduce the computational complexity required by model decoupling in the training process and improve the application effect of a model by periodically integrating heartbeats into a data form and a model structure, so that an electrocardiogram data classification result has higher accuracy, as shown in figure 1, the method comprises the following steps:

101. and acquiring a plurality of heartbeat data with equal length obtained by processing the electrocardiogram data.

The plurality of heart beat data with the same length have the same time length, and the electrocardiogram data is equivalent to medical record information collected around the electrocardiogram, that is, the electrocardiogram data includes electrocardiogram signals and more comprehensive patient data, such as: specifically, the processing process of the electrocardiogram data mainly aims at the electrocardiogram signal, and the electrocardiogram signal has the periodic characteristic, so that the electrocardiogram signal can be divided by using the heartbeat period of the electrocardiogram signal to obtain electrocardiogram segments with different lengths, and then the electrocardiogram segments are resampled to obtain the heartbeat data with the same time length.

In the application, medical treatment can be connected to different medical equipment through cloud, such as electroencephalogram equipment, electromyogram equipment, dynamic blood pressure equipment and the like, and can be interconnected and intercommunicated with various systems in a hospital, such as electronic medical records, integrated platforms and the like, so that medical data in various expression forms can be acquired, and electrocardio data can be extracted from the medical data. The medical cloud is a medical health service cloud platform established by using cloud computing on the basis of new technologies such as cloud computing, mobile technology, multimedia, 4G communication, big data, Internet of things and the like and combining medical technology, so that sharing of medical resources and expansion of medical scope are realized. Due to the combination of the cloud computing technology, the medical cloud improves the efficiency of medical institutions and brings convenience to residents to see medical advice. Like the appointment register, the electronic medical record, the medical insurance and the like of the existing hospital are all products combining cloud computing and the medical field, and the medical cloud also has the advantages of data security, information sharing, dynamic expansion and overall layout.

In the embodiment of the invention, the execution main body can be a device for classifying the electrocardiographic data based on machine learning, and particularly can be applied to a medical platform server end suitable for disease prediction, such as intelligent medical treatment or medical treatment cloud, and the electrocardiographic data is processed by the medical platform server end to obtain a plurality of isometric heartbeat data.

The server may be an independent server, or may be a cloud server that provides basic cloud computing services such as a cloud service, a cloud database, cloud computing, a cloud function, cloud storage, a Network service, cloud communication, a middleware service, a domain name service, a security service, a Content Delivery Network (CDN), a big data and artificial intelligence platform, and the like.

102. And responding to a triggering instruction of the electrocardio classification task, and inputting the plurality of heartbeat data with equal length into a pre-constructed electrocardio classification model for prediction.

The electrocardio classification model comprises a plurality of representation modules and a heartbeat aggregation module, the representation modules are used for extracting vector representations of a plurality of pieces of heartbeat data with equal lengths, the heartbeat aggregation module is used for fusing information among the vector representations of the plurality of pieces of heartbeat data with equal lengths, it can be understood that the preset electrocardio classification model is obtained by training heartbeat samples through a network model, the representation modules can be trained through a coder network structure or other network structures with coding effects, and the heart-beat aggregation module is trained through a deep convolution network structure or other network structures with information fusion functions. The electrocardiogram classification tasks can be set according to actual application scenarios, and different electrocardiogram classification tasks need different predicted electrocardiogram classification models, for example, the application scenario of the electrocardiogram classification task is gender, an electrocardiogram classification model for the gender application scenario needs to be used, the application scenario of the electrocardiogram classification task is heart rate category judgment, an electrocardiogram classification model for the heart rate judgment application scenario needs to be used, and no limitation is made here.

Specifically, in the prediction process, the representation of heartbeat data in the electrocardio classification model is influenced to a certain extent by using different resampling frequencies for the electrocardio data, the resampling frequencies can form sampling factors which are spliced with one heartbeat data and then are transmitted to a representation module for a single heartbeat, similarly, other heartbeat data can also be spliced with the sampling factors formed by the corresponding resampling frequencies and then are input to the representation module for the single heartbeat, the heartbeat data are coded in each representation module to obtain the vector representation of the single heartbeat data, after passing through a plurality of representation modules, the vector representations of all the heartbeat data are aggregated and then are input to a heartbeat aggregation module for information fusion, and the vector representation of the fused heartbeat data is obtained.

103. And processing the vector representation of the fused heartbeat data to the mapping relation of the heartbeat data in different categories, and outputting the classification result of the electrocardio data.

In the application, the vector representation of the fused heartbeat data can be processed into the mapping relation of the heartbeat sample data on different classes, the mapping relation can be expressed as the probability value of the heartbeat data on different class labels, the higher the probability value is, the higher the probability of the electrocardio data on the corresponding class is, and the classification result of the electrocardio data is output according to the mapping relation.

The method can be understood that the classification effect of the electrocardio classification model depends on the correctness of feature extraction, the periodic features of the electrocardio data are considered, the electrocardio data are not directly transmitted into a deep learning network as one-dimensional multi-channel picture data, but the periodic features of the electrocardio data are considered, the electrocardio data are processed into the electrocardio classification model formed by training a plurality of isometric heartbeat samples, more accurate electrocardio features can be extracted, and the classification effect of the electrocardio data is improved.

In practical application, the classification result of the electrocardiographic data can use the probability value of the electrocardiographic data on each category label as an auxiliary basis for judgment according to the application scene of the electrocardiographic classification task, and the type with the highest probability value is determined as the reference of the electrocardiographic data category.

The embodiment of the invention provides a machine learning-based electrocardiogram data classification method, which comprises the steps of obtaining a plurality of pieces of isometric heartbeat data obtained by processing electrocardiogram data, enabling the plurality of pieces of isometric heartbeat data to have the same time length, responding to a trigger instruction of an electrocardiogram classification task, inputting the plurality of pieces of isometric heartbeat data into a pre-constructed electrocardiogram classification model for prediction, enabling the electrocardiogram classification model to comprise a plurality of characterization modules and a heartbeat aggregation module, extracting vector representations of the plurality of pieces of isometric heartbeat data by using the characterization modules, fusing information among the vector representations of the plurality of pieces of isometric heartbeat data by using the heartbeat aggregation module, further processing the vector representations of the fused heartbeat data into mapping relations of the heartbeat data on different classes, and outputting classification results of the electrocardiogram data. Compared with the prior art that the classification process of the electrocardiogram data is carried out by means of a multilayer neural network structure, the periodic characteristics of the electrocardiogram data are integrated into a data form and a model structure, and the network model can provide higher calculation efficiency and better interpretability, so that the network model trained aiming at the heartbeat data with equal length has a better classification effect, the calculation complexity required by model decoupling in the training process can be reduced, and the classification result of the electrocardiogram data has higher accuracy.

The embodiment of the invention provides another method for classifying electrocardiographic data based on machine learning, which can reduce the computational complexity required by model decoupling in the training process and improve the application effect of a model by periodically integrating the heartbeat into a data form and a model structure, so that the electrocardiographic data classification result has higher accuracy, as shown in fig. 2, and the method comprises the following steps:

201. and segmenting the electrocardiogram data to obtain electrocardiogram fragments with different time lengths.

The heartbeat segment is a heartbeat cycle, the electrocardio data can record the electrocardio segments of a plurality of heartbeat cycles within a set time, the waveform curve of the electrocardio data appearing in each heartbeat cycle has a certain rule, the characteristics in each heartbeat segment can be used as the judgment basis for electrocardio classification, the electrocardio data are divided according to the heartbeat cycle to obtain the electrocardio segments with different time lengths, characteristic analysis is carried out on the electrocardio data in each electrocardio segment, and the accuracy of the electrocardio classification task prediction result can be improved.

In one possible implementation, the electrocardiographic data may be represented as an electrocardiogram, which is a graph in which the heart is excited by the pacing point, the atrium, and the ventricle in succession in each heart cycle, and various forms of potential changes are drawn from the body surface by the electrocardiograph along with the changes of the electrocardiographic bioelectricity. For example, the electrocardiogram type of atrial fibrillation has irregular waveform characteristics in the heartbeat cycle, the duration of each heartbeat cycle of the electrocardiogram type of sinus bradycardia is greater than a set time value, and the duration of each heartbeat cycle of the electrocardiogram type of sinus tachycardia is less than the set time value.

202. And performing resampling processing on the electrocardio segments with different time lengths to obtain a plurality of heartbeat data with equal length.

Considering the duration interval of the heartbeat cycle, when resampling processing is performed on the electrocardiographic segments with different time lengths, the average value in the duration interval can be used, and each electrocardiographic segment is resampled into heartbeat data with equal length, for example, the time average of the heartbeat cycle is 0.6 second to 1 second, 0.8 second can be selected as the resampling time length, and a plurality of heartbeat data with the time length of 0.8 second are obtained.

203. And determining a category label and a corresponding resampling factor corresponding to each heartbeat sample in the plurality of heartbeat samples with equal lengths based on a preset electrocardio classification task label.

The preset electrocardiographic classification task tag corresponds to a category tag of electrocardiographic data, and may be specifically set according to an actual application scenario of the electrocardiographic classification task, for example, the application scenario of the electrocardiographic classification task is gender, the category tag of the electrocardiographic data may be male/female, the application scenario of the electrocardiographic classification task is a heart rate category judgment, the category tag of the electrocardiographic data may be atrial fibrillation/sinus tachycardia, and the like, and the category tag of the electrocardiographic data may be set in combination with the application scenario of the multi-electrocardiographic classification task, which is not limited herein.

It can be understood that before the heartbeat sample is input into the network model, the heartbeat sample needs to be labeled, and the classification effect of the network model is optimized by using the labeled class labels in the training process, because different electrocardio classification labels are influenced by application scenes, an electrocardio classification task label can be set according to the application scenes of the electrocardio classification task, the electrocardio classification task label is equivalent to the class label of electrocardio data, specifically, the electrocardio classification task label is set according to waveform change, characteristics of different wave bands in the heartbeat sample are considered, the heartbeat sample is labeled by using wave band characteristics in the heartbeat sample, the disease class representation of the heartbeat sample is considered by setting the electrocardio classification task label according to disease types, and the heartbeat sample is labeled by using the disease class representation. For example, the application scenario of the electrocardiographic classification task is gender, the category tag of the electrocardiographic data may be set as male/female, the application scenario of the electrocardiographic classification task is heart rate type judgment, the category tag of the electrocardiographic data may be atrial fibrillation/sinus tachycardia, and the like, and the category tag of the electrocardiographic data may be set in combination with the application scenario of the multi-electrocardiographic classification task, which is not limited herein.

Furthermore, because each heartbeat sample is heartbeat data with the same time length obtained by resampling the electrocardiographic segment, in order to ensure that the heartbeat samples have the same time length, different heartbeat samples correspond to different resampling frequency ratios, for example, an electrocardiographic segment with a long time needs to use a lower resampling frequency ratio, and an electrocardiographic segment with a short time needs to use a higher resampling frequency ratio, where the resampling frequency ratio can reflect morphological differences of electrocardiographic waves in the heartbeat segment, so as to reduce waveform feature loss in the heartbeat segment to a certain extent, it is necessary to determine a corresponding resampling frequency ratio for each heartbeat segment, and store the frequency ratio as a resampling factor, so as to use the resampling factor as a basis for waveform feature extraction in a subsequent model training process.

204. After the heartbeat samples carrying the category labels are spliced with the corresponding resampling factors, the heartbeat samples are respectively input to different characterization modules for training, and the vector characterization of each heartbeat sample is obtained.

The characterization module comprises a normalization layer and a self-attention layer, wherein each heartbeat sample is spliced with a corresponding resampling factor to form an input sample, the characterization module is arranged for each input sample, the plurality of input samples are namely arranged into the plurality of characterization modules, each input sample is respectively input into the characterization modules to be trained, each characterization module has the same network structure and is similar to the network structures of two encoders, namely, one normalization layer is superposed with one self-attention layer to form one sub block of the characterization module, one normalization layer is superposed with the other self-attention layer to form the other block of the characterization module, and the two sub blocks form one characterization module.

In consideration of the difference of vector spaces where different heartbeat samples are located, specifically, the heartbeat samples carrying the label categories are subjected to linear mapping and then spliced with corresponding resampling factors, then the heartbeat samples are respectively input to a characterization module of a single heartbeat for training, then feature coding is performed on the spliced heartbeat samples by using a self-attention layer of the characterization module to obtain vector characterization of the single heartbeat sample, and the vector characterization of the single heartbeat sample is mapped to a stably-distributed vector space by using a normalization layer of the characterization module, so that the coded vector characterization has the same dimensionality.

The method includes the steps that a window mechanism is added to the self-attention layer, specifically, the self-attention layer of the characterization module is utilized to perform feature coding on spliced heartbeat samples to obtain vector characterization of a single heartbeat sample, the spliced heartbeat samples can be divided by using preset window length by utilizing the window mechanism to obtain heartbeat sample data of a plurality of windows, then the window length is slid, the self-attention layer of the characterization module acts on the heartbeat sample data of each window, and the vector characterization of the single heartbeat sample is obtained. It should be noted that, for the network structure of the characterization module, there are two self-attention layers added with a window mechanism, after the first self-attention layer is added into the window mechanism, the first self-attention layer does not act on the whole input heartbeat sample, but divides the whole heartbeat sample by a fixed length, and applies the self-attention to the heartbeat sample of each window, and after the second self-attention layer is added into the window mechanism, the window needs to be slid before the self-attention acts on the heartbeat sample of each window, and then the heartbeat sample of each window after sliding is self-attentive, and the sliding length can be set to be half the window size.

205. And summarizing vector characteristics of the heartbeat samples output by the characteristic modules, inputting the vector characteristics to the aggregation module for training, obtaining the mapping relation of the heartbeat samples on different classes, and constructing the electrocardio classification model.

The heartbeat aggregation module comprises a convolution layer, a pooling layer and a normalization layer, and the vector characterization of the heartbeat samples output by the modules is reflected as the characteristic in a single heartbeat period and cannot reflect the integral category characteristic of the electrocardiogram.

Here, each heartbeat sample is spliced with a corresponding resampling factor to form an input sample, a characterization module is arranged for each input sample, the plurality of input samples are provided with the plurality of characterization modules, each input sample is input into the characterization modules respectively for training, each characterization module has the same network structure and is similar to the network structures of two encoders, namely, a normalization layer is superposed with a self-attention layer to form a sub-block of the characterization module, a normalization layer is superposed with another self-attention layer to form another sub-block of the characterization module, and the two sub-blocks form one characterization module.

The vector representations of the heartbeat samples output by the characterization modules are collected and input into a convolution layer of a polymerization module, feature extraction is carried out on the vector representations of the collected heartbeat samples, the vector representations after feature extraction are mapped to a stably-distributed vector space by utilizing a normalization layer of the polymerization module, so that the vector representations after feature extraction have the same dimension, dimension reduction processing is carried out on the vector representations with the same dimension after feature extraction by utilizing a pooling layer of the polymerization module, mapping relations of the heartbeat samples on different classes are obtained, and the electrocardio classification model is constructed.

It can be understood that, in order to ensure the training effect of the electrocardiograph classification model, in the process of training the network model by using the classification task tag corresponding to the heartbeat sample as the known class tag value, a loss function is required to calculate a deviation value formed by the predicted value of the electrocardiograph classification task and the known class tag value, and the deviation value is reversely transmitted in the network model to continuously adjust the model parameters of the network model.

206. And responding to a triggering instruction of the electrocardio classification task, and inputting the plurality of heartbeat data with equal length into a pre-constructed electrocardio classification model for prediction.

207. And processing the vector representation of the fused heartbeat data to the mapping relation of the heartbeat data in different categories, and outputting the classification result of the electrocardio data.

According to the method and the device, the heartbeat periodicity is integrated into the electrocardiogram data form and the model structure, so that the self-attention coding network constructed aiming at the electrocardiogram data can achieve a better classification effect, and the self-attention coding network can provide higher calculation efficiency and better interpretability. The classification model constructed by pre-training the self-attention coding network by using large-scale electrocardiogram data has a better recognition effect on an electrocardiogram classification task.

208. And recording the distribution characteristics of the electrocardiogram data according to the classification result of the electrocardiogram data.

It can be understood that the classification result of the electrocardiographic data can reflect various types of electrocardiographic representations to a certain extent based on different application scenarios, where the electrocardiographic representations may include, but are not limited to, potential and time changes of various parts of the heart, and the distribution characteristics of the target electrocardiographic data can be recorded by analyzing the potential and time changes of various parts of the heart from the classification result of the target electrocardiographic data.

209. And mapping the distribution characteristics of the electrocardiogram data to each dimension index set by the electrocardiogram classification task so as to perform abnormal evaluation on each dimension index.

Here, each dimension index set by the electrocardiographic classification task may be set for determining whether there is an abnormality in the electrocardiographic data, such as a heart rate index, a respiration rate index, a waveform index, and the like, and since the distribution characteristics of the electrocardiographic data are different in the expression values of different dimension indexes, the expression value of the electrocardiographic data on each dimension index can be obtained by mapping the distribution characteristics of the electrocardiographic data to each dimension index, it should be noted that some dimension indexes need to be calculated, for example, the heart rate index can be calculated by dividing the time interval of each heartbeat cycle, and some indexes can be directly extracted from the distribution characteristics, for example, the waveform index can be extracted from a waveform diagram.

It can be understood that each dimension index has clinical significance in an actual application scene, a normal numerical range is set, and abnormal evaluation on each dimension index can be realized by judging whether the expression numerical value of the dimension index is in the normal numerical range or not.

Further, as a specific implementation of the method shown in fig. 1, an embodiment of the present invention provides an apparatus for classifying electrocardiographic data based on machine learning, as shown in fig. 3, where the apparatus includes: an acquisition unit 31, a construction unit 32, an output unit 33.

The acquiring unit 31 may be configured to acquire a plurality of equal-length heartbeat data obtained by processing electrocardiographic data, where the plurality of equal-length heartbeat data have the same time length;

the prediction unit 32 may be configured to, in response to a trigger instruction of an electrocardiographic classification task, input the multiple pieces of heartbeat data with equal length into a pre-constructed electrocardiographic classification model for prediction, where the electrocardiographic classification model includes multiple characterization modules and a heartbeat aggregation module, and the characterization modules are used to extract vector characterizations of the multiple pieces of heartbeat data with equal length, and the heartbeat aggregation module is used to fuse information between the vector characterizations of the multiple pieces of heartbeat data with equal length;

the output unit 33 may be configured to process the vector representation of the fused heartbeat data to the mapping relationships of the heartbeat data in different categories, and output the classification result of the electrocardiographic data.

The device for classifying the electrocardiographic data based on machine learning, provided by the embodiment of the invention, acquires a plurality of isometric heartbeat data obtained by processing the electrocardiographic data, wherein the isometric heartbeat data have the same time length, and responds to a trigger instruction of an electrocardiographic classification task, the isometric heartbeat data are input into a pre-constructed electrocardiographic classification model for prediction, the electrocardiographic classification model comprises a plurality of characterization modules and a heartbeat aggregation module, the characterization modules are used for extracting vector representations of the isometric heartbeat data, the heartbeat aggregation module is used for fusing information among the vector representations of the isometric heartbeat data, the fused vector representations of the heartbeat data are further processed to the mapping relation of the heartbeat data on different classes, and the classification result of the electrocardiographic data is output. Compared with the prior art that the classification process of the electrocardiogram data is carried out by means of a multilayer neural network structure, the periodic characteristics of the electrocardiogram data are integrated into a data form and a model structure, and the network model can provide higher calculation efficiency and better interpretability, so that the network model trained aiming at the heartbeat data with equal length has a better classification effect, the calculation complexity required by model decoupling in the training process can be reduced, and the classification result of the electrocardiogram data has higher accuracy.

As a further description of the apparatus for classifying electrocardiographic data based on machine learning shown in fig. 3, fig. 4 is a schematic structural diagram of another apparatus for classifying electrocardiographic data based on machine learning according to an embodiment of the present invention, and as shown in fig. 4, the obtaining unit 31 includes:

the segmenting module 311 may be configured to segment the electrocardiographic data to obtain electrocardiographic segments with different time lengths, where the electrocardiographic segment is a heartbeat cycle;

the processing module 312 may be configured to perform resampling processing on the electrocardiographic segments with different time lengths to obtain a plurality of isometric heartbeat data.

In a specific application scenario, as shown in fig. 4, the apparatus further includes:

the constructing unit 34 may be configured to, before the triggering instruction responding to the electrocardiograph classification task inputs the multiple isometric heartbeat data into a pre-constructed electrocardiograph classification model for prediction, input multiple isometric heartbeat samples into a network model for training based on a preset electrocardiograph classification task tag, and construct the electrocardiograph classification model, where the network model includes multiple characterization modules and a heartbeat aggregation module, the characterization modules are configured to learn vector characterizations of a single heartbeat sample, and the heartbeat aggregation module is configured to fuse information between the vector characterizations of the multiple heartbeat samples;

the building unit 34 includes:

the determining module 341 is configured to determine, based on a preset electrocardiograph classification task tag, a category tag and a corresponding resampling factor corresponding to each heartbeat sample in the multiple heartbeat samples with equal length, where the resampling factor is a numerical value formed by frequency scale quantization of resampling;

the training module 342 may be configured to splice the heartbeat samples with the category labels with the corresponding resampling factors, and then input the heartbeat samples to different characterization modules, so as to obtain a vector characterization of each heartbeat sample;

the constructing module 343 may be configured to summarize the vector characterizations of the heartbeat samples output by the respective characterizing modules, and then input the summarized vectors to the aggregating module, so as to obtain mapping relationships of the heartbeat samples in different categories, and construct the electrocardiogram classification model.

In a specific application scenario, as shown in fig. 4, the characterization module includes a normalization layer and a self-attention layer, and the training module 342 includes:

the splicing submodule 3421 may be configured to splice the heartbeat samples carrying the label categories with corresponding resampling factors after linear mapping, and input the heartbeat samples to a single heartbeat characterization module for training;

the encoding submodule 3422 may be configured to perform feature encoding on the spliced heartbeat samples by using the self-attention layer of the characterization module, so as to obtain a vector characterization of a single heartbeat sample;

a first mapping sub-module 3423, which may be configured to map the vector characterization of the single heartbeat sample to a stably distributed vector space using a normalization layer of the characterization module, so that the encoded vector characterization has the same dimensions.

In a specific application scenario, the encoding sub-module 3422 may be specifically configured to divide the spliced heartbeat samples by using a preset window length by using the window mechanism to obtain heartbeat sample data of multiple windows;

the coding sub-module 3422 may be further configured to apply the self-attention layer of the characterization module to the heartbeat sample data of each window to obtain a vector characterization of a single heartbeat sample.

In a specific application scenario, as shown in fig. 4, the heartbeat aggregation module includes a convolutional layer, a pooling layer, and a normalization layer, and the building module 343 includes:

the extracting sub-module 3431 may be configured to summarize vector characterizations of the heartbeat samples output by the respective characterization modules, input the summarized vector characterizations to the convolution layer of the aggregation module, and perform feature extraction on the summarized vector characterizations of the heartbeat samples;

a second mapping sub-module 3432, which may be configured to map the feature-extracted vector representations to a stably-distributed vector space by using a normalization layer of the aggregation module, so that the feature-extracted vector representations have the same dimension;

the dimensionality reduction sub-module 3433 may be configured to perform dimensionality reduction processing on the vector representations with the same dimensionality after feature extraction by using the pooling layer of the aggregation module, obtain mapping relationships of the heartbeat samples in different classes, and construct an electrocardiogram classification model.

In a specific application scenario, as shown in fig. 4, the apparatus further includes:

the recording unit 35 may be configured to record the distribution characteristics of the electrocardiographic data according to the classification result of the electrocardiographic data after the vector characterization of the fused heartbeat data is processed to the mapping relationships of the heartbeat data in different categories and the classification result of the electrocardiographic data is output;

the evaluation unit 36 may be configured to map the distribution characteristics of the electrocardiographic data to each dimension index set by the electrocardiographic classification task, so as to perform anomaly evaluation on each dimension index.

It should be noted that other corresponding descriptions of the functional units related to the device for classifying electrocardiographic data based on machine learning provided in this embodiment may refer to the corresponding descriptions in fig. 1 and fig. 2, and are not described herein again.

Based on the methods shown in fig. 1 and fig. 2, correspondingly, the present embodiment further provides a storage medium, on which a computer program is stored, and the computer program, when executed by a processor, implements the method for classifying electrocardiographic data based on machine learning shown in fig. 1 and fig. 2.

Based on such understanding, the technical solution of the present application may be embodied in the form of a software product, which may be stored in a non-volatile storage medium (which may be a CD-ROM, a usb disk, a removable hard disk, etc.), and includes several instructions for enabling a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the method according to the implementation scenarios of the present application.

Based on the method shown in fig. 1 and fig. 2 and the virtual device embodiment shown in fig. 3 and fig. 4, in order to achieve the above object, an embodiment of the present application further provides a computer device, which may specifically be a personal computer, a server, a network device, and the like, where the entity device includes a storage medium and a processor; a storage medium for storing a computer program; processor for executing computer program to implement the above-mentioned method for classifying electrocardiographic data based on machine learning as shown in fig. 1 and 2

Optionally, the computer device may also include a user interface, a network interface, a camera, Radio Frequency (RF) circuitry, sensors, audio circuitry, a WI-FI module, and so forth. The user interface may include a Display screen (Display), an input unit such as a keypad (Keyboard), etc., and the optional user interface may also include a USB interface, a card reader interface, etc. The network interface may optionally include a standard wired interface, a wireless interface (e.g., a bluetooth interface, WI-FI interface), etc.

Those skilled in the art will appreciate that the physical device structure of the apparatus for classifying electrocardiographic data based on machine learning provided in the present embodiment does not constitute a limitation to the physical device, and may include more or less components, or combine some components, or arrange different components.

The storage medium may further include an operating system and a network communication module. The operating system is a program that manages the hardware and software resources of the computer device described above, supporting the operation of information handling programs and other software and/or programs. The network communication module is used for realizing communication among components in the storage medium and other hardware and software in the entity device.

Through the above description of the embodiments, those skilled in the art will clearly understand that the present application can be implemented by software plus a necessary general hardware platform, and can also be implemented by hardware. By applying the technical scheme of the application, compared with the prior art, the method has the advantages that the periodic characteristics of the electrocardiogram data are integrated into the data form and the model structure, and the network model can provide higher calculation efficiency and better interpretability, so that the network model trained aiming at the heartbeat data with equal length has better classification effect, the calculation complexity required by model decoupling in the training process can be reduced, and the electrocardiogram data classification result has higher accuracy.

Those skilled in the art will appreciate that the figures are merely schematic representations of one preferred implementation scenario and that the blocks or flow diagrams in the figures are not necessarily required to practice the present application. Those skilled in the art will appreciate that the modules in the devices in the implementation scenario may be distributed in the devices in the implementation scenario according to the description of the implementation scenario, or may be located in one or more devices different from the present implementation scenario with corresponding changes. The modules of the implementation scenario may be combined into one module, or may be further split into a plurality of sub-modules.

The above application serial numbers are for description purposes only and do not represent the superiority or inferiority of the implementation scenarios. The above disclosure is only a few specific implementation scenarios of the present application, but the present application is not limited thereto, and any variations that can be made by those skilled in the art are intended to fall within the scope of the present application.