Disclosure of Invention
The invention aims to solve the problems of complex data management, low practicality, large labor cost investment, poor accuracy, poor functionality and poor user experience in the prior art, and provides a heart rate monitoring method and system based on a neural network.
The technical scheme adopted by the invention is as follows:
a neural network-based heart rate monitoring method, comprising the steps of:
Based on a cloud computing center, constructing a user portrait generation model, a knowledge graph, a heart rate abnormality recognition model and a heart rate monitoring optimization model by using a neural network algorithm;
based on the user terminal, receiving input real-time postoperative clinical data and real-time heart rate acquisition data sent by the heart rate acquisition equipment, and uploading the data to a cloud computing center;
Based on the cloud computing center, generating a user portrait by using a user portrait generation model according to real-time postoperative clinical data to obtain a real-time user portrait;
Based on the cloud computing center, carrying out heart rate abnormality recognition according to real-time postoperative clinical data, real-time heart rate acquisition data and real-time user portraits by using a heart rate abnormality recognition model to obtain a real-time heart rate abnormality recognition result;
Based on the cloud computing center, according to the real-time heart rate abnormality recognition result and the real-time user image, performing heart rate monitoring optimization by using a heart rate monitoring optimization model to obtain a real-time heart rate monitoring optimization strategy;
based on a cloud computing center, a knowledge graph is used for adjusting a real-time heart rate monitoring optimization strategy to obtain an adjusted real-time heart rate monitoring optimization strategy;
based on the cloud computing center, generating postoperative rehabilitation advice by using a knowledge graph according to the real-time user portrait and the real-time heart rate abnormality recognition result to obtain real-time postoperative rehabilitation advice;
based on the cloud computing center, sending the real-time heart rate abnormality identification result, the adjusted real-time heart rate monitoring optimization strategy and the real-time postoperative rehabilitation suggestion to a corresponding user terminal;
Based on the user terminal, generating a real-time alarm signal according to the real-time heart rate abnormality recognition result, and visualizing the real-time alarm signal, the real-time heart rate abnormality recognition result and the real-time postoperative rehabilitation suggestion;
Based on the user terminal, generating a real-time heart rate monitoring instruction according to the adjusted real-time heart rate monitoring optimization strategy, and sending the real-time heart rate monitoring instruction to the heart rate acquisition equipment.
Further, based on the cloud computing center, a neural network algorithm is used for constructing a user portrait generation model, a knowledge graph, a heart rate abnormality identification model and a heart rate monitoring optimization model, and the method comprises the following steps:
based on a cloud computing center, collecting a plurality of heart rate monitoring and postoperative rehabilitation knowledge, a plurality of historical postoperative clinical data and a plurality of historical heart rate collecting data;
Preprocessing to obtain heart rate monitoring and postoperative rehabilitation knowledge after intervention processing, a plurality of preprocessed historical postoperative clinical data and a plurality of preprocessed historical heart rate acquisition data;
Performing data dimension reduction on the plurality of preprocessed historical postoperative clinical data to obtain a plurality of dimension-reduced historical postoperative clinical data and a key clinical index set;
according to the post-operation clinical data of the post-operation histories after the dimension reduction and the post-processing historic heart rate acquisition data, constructing a user portrait generation model by using a neural network algorithm, and generating a plurality of historic user portraits;
According to the heart rate monitoring after pretreatment and the postoperative rehabilitation knowledge, a natural language processing algorithm is used for constructing a knowledge graph;
according to the post-operation clinical data of the post-operation history, the post-pretreatment history heart rate acquisition data and the history user portraits, constructing a heart rate abnormality recognition model by using a neural network algorithm, and generating history heart rate abnormality recognition results;
and constructing a heart rate monitoring optimization model by using a reinforcement learning algorithm according to the historical heart rate abnormality recognition results and the historical user portraits.
Further, according to the heart rate monitoring after pretreatment and the rehabilitation knowledge after operation, a natural language processing algorithm is used for constructing a knowledge graph, and the method comprises the following steps:
Extracting a plurality of heart rate monitoring named entities and a plurality of postoperative rehabilitation named entities of the preprocessed heart rate monitoring and postoperative rehabilitation knowledge by using a pre-trained named entity extraction model;
Extracting a heart rate monitoring entity relationship among a plurality of heart rate monitoring named entities and a postoperative rehabilitation entity relationship among a plurality of postoperative rehabilitation named entities by using a pre-trained entity relationship extraction model;
And constructing a knowledge graph according to the heart rate monitoring named entities, the heart rate monitoring entity relationships, the postoperative rehabilitation named entities and the postoperative rehabilitation entity relationships.
Further, a user portrait generation model is constructed based on an N-GAN-attribute-LSTM algorithm;
the heart rate abnormality recognition model is constructed based on an RNN-GNN-MLP algorithm;
The heart rate monitoring optimization model is constructed based on the DQN algorithm.
Further, based on the cloud computing center, according to real-time postoperative clinical data, user portrait generation is performed by using a user portrait generation model, and a real-time user portrait is obtained, which comprises the following steps:
based on a cloud computing center, performing data dimension reduction on real-time postoperative clinical data according to the key clinical index set to obtain dimension-reduced real-time postoperative clinical data;
inputting the real-time post-operation clinical data after dimension reduction into a user portrait generation model, and extracting N real-time key dimension features;
According to the preset attention weight, carrying out weighted splicing on N real-time key dimension features to obtain real-time weighted spliced features;
and according to the real-time weighting splicing characteristics, predicting the user labels to obtain a plurality of real-time user labels, and according to the plurality of real-time user labels, generating the user portrait to obtain the real-time user portrait.
Further, based on the cloud computing center, according to real-time postoperative clinical data, real-time heart rate acquisition data and real-time user portraits, heart rate abnormality recognition is performed by using a heart rate abnormality recognition model, and a real-time heart rate abnormality recognition result is obtained, and the method comprises the following steps:
Based on a cloud computing center, inputting real-time post-operation clinical data, real-time heart rate acquisition data and real-time user portraits after dimension reduction into a heart rate abnormality recognition model;
extracting real-time postoperative clinical data characteristics of the real-time postoperative clinical data after dimension reduction, real-time heart rate acquisition data characteristics of the real-time heart rate acquisition data and real-time image characteristics of the real-time user image;
Performing feature fusion on the real-time postoperative clinical data feature, the real-time heart rate acquisition data feature and the real-time map feature to obtain a real-time fusion feature;
And carrying out heart rate abnormality recognition according to the real-time fusion characteristics to obtain a real-time heart rate abnormality prediction tag, namely a real-time heart rate abnormality recognition result.
Further, based on the cloud computing center, according to the real-time heart rate abnormality recognition result and the real-time user image, using a heart rate monitoring optimization model to perform heart rate monitoring optimization to obtain a real-time heart rate monitoring optimization strategy, the method comprises the following steps:
based on a cloud computing center, extracting a plurality of history heart rate monitoring optimization experiences matched in an experience playback pool of a heart rate monitoring optimization model according to real-time user images;
Updating model parameters of the heart rate monitoring optimization model according to a plurality of historical heart rate monitoring optimization experiences and real-time heart rate abnormality recognition results to obtain updated model parameters;
And carrying out heart rate monitoring optimization by using a heart rate monitoring optimization model according to the updated model parameters to obtain a plurality of real-time execution actions, and obtaining a real-time heart rate monitoring optimization strategy according to the plurality of real-time execution actions.
Further, based on the cloud computing center, the real-time heart rate monitoring optimization strategy is adjusted by using the knowledge graph, and the adjusted real-time heart rate monitoring optimization strategy is obtained, and the method comprises the following steps of;
Based on the cloud computing center, according to action naming entities corresponding to the real-time execution actions in the real-time heart rate monitoring optimization strategy, searching and matching are carried out in heart rate monitoring naming entities of the knowledge graph, and a plurality of matching heart rate monitoring naming entities are obtained;
And according to the plurality of the heart rate monitoring named entities and the heart rate monitoring entity relation of the heart rate monitoring named entities and other heart rate monitoring named entities, adjusting the real-time heart rate monitoring optimization strategy to obtain the adjusted real-time heart rate monitoring optimization strategy.
Further, based on a cloud computing center, according to the real-time user portrait and the real-time heart rate abnormality recognition result, using a knowledge graph to generate postoperative rehabilitation advice, and obtaining real-time postoperative rehabilitation advice, wherein the method comprises the following steps;
Based on a cloud computing center, searching and matching are carried out in postoperative rehabilitation named entities of a knowledge graph according to a user named entity corresponding to a real-time user tag in a real-time user image and an abnormal named entity corresponding to a real-time heart rate abnormal prediction tag in a real-time heart rate abnormal recognition result, so as to obtain a plurality of matched postoperative rehabilitation named entities;
And according to a plurality of postoperative rehabilitation named entities and the postoperative rehabilitation entity relation between the postoperative rehabilitation named entities and other postoperative rehabilitation named entities, generating postoperative rehabilitation suggestions, and obtaining real-time postoperative rehabilitation suggestions.
The heart rate monitoring system based on the neural network is used for realizing a heart rate monitoring method, and comprises a cloud computing center, a plurality of user terminals and a plurality of heart rate acquisition devices, wherein the cloud computing center is respectively in communication connection with the plurality of user terminals, and each user terminal is in communication connection with one heart rate acquisition device in a one-to-one correspondence manner;
The cloud computing center comprises a model building unit, a user portrait generating unit, a heart rate abnormality identifying unit, a heart rate monitoring optimizing unit, an optimizing strategy adjusting unit, a postoperative rehabilitation suggestion generating unit and a data transmitting unit which are sequentially connected.
The beneficial effects of the invention are as follows:
The invention discloses a heart rate monitoring method and system based on a neural network, which combine real-time postoperative clinical data and real-time heart rate acquisition data sent by heart rate acquisition equipment, more comprehensively reflect health condition information of a user, improve data comprehensiveness, perform unified management and analysis on the data based on a cloud computing center, improve practicability, build a user portrait generation model and a heart rate monitoring optimization model, generate personalized heart rate monitoring schemes according to specific health conditions and heart rate data of the user, realize customization functions of different rehabilitation situations and different users, improve user experience, build a heart rate abnormality identification model, combine the real-time postoperative clinical data, the real-time heart rate acquisition data and the real-time user portrait, perform heart rate abnormality identification, reduce labor cost investment, improve heart rate abnormality identification accuracy, build a knowledge graph, generate real-time postoperative rehabilitation suggestions according to the real-time user portrait and the real-time heart rate abnormality identification result, and improve functionality.
Other advantageous effects of the present invention will be further described in the detailed description.
Detailed Description
The invention is further illustrated by the following description of specific embodiments in conjunction with the accompanying drawings.
Example 1:
As shown in fig. 1 and fig. 2 together, the present embodiment provides a heart rate monitoring method based on a neural network, including the following steps:
S1, constructing a user portrait generation model, a knowledge graph, a heart rate abnormality identification model and a heart rate monitoring optimization model by using a neural network algorithm based on a cloud computing center, wherein the method comprises the following steps of:
S1-1, acquiring a plurality of heart rate monitoring and postoperative rehabilitation knowledge, a plurality of historical postoperative clinical data and a plurality of historical heart rate acquisition data based on a cloud computing center;
heart rate monitoring and postoperative rehabilitation knowledge comprises heart rate-related medical knowledge and postoperative rehabilitation-related medical knowledge, wherein the heart rate-related medical knowledge comprises heart rate definition, heart rate abnormal symptoms, heart rate monitoring rules and the like, and the postoperative rehabilitation-related medical knowledge comprises heart rate abnormal range, postoperative rehabilitation rules and the like;
the postoperative clinical data comprise basic information, health conditions, living habits, exercise preferences, physiological data, clinical consultation, operation projects and the like of the user;
heart rate acquisition data includes electrocardiogram data, photoplethysmogram data, motion sensor data, pulse waveform data, phonocardiogram data, etc.;
S1-2, preprocessing to obtain heart rate monitoring and postoperative rehabilitation knowledge after intervention processing, a plurality of preprocessed historical postoperative clinical data and a plurality of preprocessed historical heart rate acquisition data;
The preprocessing comprises the steps of sequentially carrying out noise removal, data screening, data cleaning and normalization on the acquired data, eliminating noise interference, removing repeated and error data, eliminating data magnitude difference and improving data quality;
S1-3, performing data dimension reduction on a plurality of preprocessed historical postoperative clinical data by using a principal component analysis (PRINCIPAL COMPONENT ANALYSIS, PCA) method to obtain a plurality of dimension-reduced historical postoperative clinical data and a key clinical index set, wherein the method comprises the following steps of:
S1-3-1, performing matrix conversion on a plurality of preprocessed historical postoperative clinical data to obtain a corresponding historical postoperative clinical data matrix X= [ X1,x2,...xp,...,xn]T ] wherein Xp is the p-th preprocessed historical postoperative clinical data row vector, p is a row vector indication quantity, n is the total number of the preprocessed historical postoperative clinical data, an initial row vector of the historical postoperative clinical data matrix is the preprocessed historical postoperative clinical data, an initial column vector of the historical postoperative clinical data matrix is the historical postoperative clinical index data, the historical postoperative clinical index data comprises postoperative clinical indexes including basic information indexes, health condition indexes, life habit indexes, movement preference indexes, physiological data indexes, clinical inquiry indexes, operation item indexes and the like, and each index data corresponds to a data value of each column in the historical postoperative clinical data;
s1-3-2, carrying out standardization processing on the historical postoperative clinical data matrix to obtain a corresponding standardized processed historical postoperative clinical data matrix;
The formula is:
Wherein X' is a standardized post-treatment historical post-operation clinical data matrix, mu is the mean value of the historical post-operation clinical data matrix, sigma is the variance of the historical post-operation clinical data matrix;
S1-3-3, acquiring a covariance matrix of a standardized post-treatment historical post-operation clinical data matrix, and acquiring a corresponding alternative principal component matrix according to the standardized post-treatment historical post-operation clinical data matrix and the covariance matrix, wherein alternative row vectors of the alternative principal component matrix are pre-treatment historical post-operation clinical data, and alternative column vectors of the alternative principal component matrix are historical post-operation clinical alternative index data;
The formula is:
wherein D is covariance matrix of the history post-operation clinical data matrix after standardized treatment, Y is alternative principal component matrix, P is conversion matrix, E is unit eigenvector matrix, n is total number of the history post-operation clinical data after pretreatment;
Y=PX'
wherein Y is an alternative principal component matrix, P is a conversion matrix, and X' is a standardized post-treatment historical post-operation clinical data matrix;
S1-3-4, taking a plurality of candidate column vectors Y 'l with the 90% of the cumulative contribution rate of the differences in the candidate principal component matrix Y= [ Y1,y2,...yl,...,yL ] as a plurality of corresponding principal component column vectors to obtain a post-dimension-reduction historical post-operation clinical data matrix Y ' = [ Y '1,y'2,...y'l,...,yK ]. A key row vector of the post-dimension-reduction historical post-operation clinical data matrix is post-dimension-reduction historical post-operation clinical data, and a data index corresponding to the key column vector of the post-dimension-reduction historical post-operation clinical data matrix is a key clinical index to obtain a key clinical index set;
The formula is:
Wherein G is the variance accumulation contribution rate, lambdal is the variance of the first alternative principal component yl, L is the indication quantity of the alternative principal component, L is the total number of the alternative principal components, K is the total number of the principal components;
S1-3-5, performing inverse matrix conversion on the post-dimension reduction historical post-operation clinical data matrix to obtain a plurality of post-dimension reduction historical post-operation clinical data;
S1-4, constructing a user portrait generation model by using an N-GAN-Attention-LSTM algorithm according to a plurality of post-dimension reduction historical postoperative clinical data and a plurality of post-preprocessing historical heart rate acquisition data, and generating a plurality of historical user portraits;
The user portrait generation model comprises N key dimension feature extraction modules constructed based on a generation countermeasure network (GENERATIVE ADVERSARIAL Networks, GAN) algorithm, an Attention weight generation module constructed based on an Attention mechanism and a user portrait generation module constructed based on a Long Short-Term Memory network (LSTM), wherein N is the total number of key dimensions, and the key dimensions comprise an operation grade dimension, a rehabilitation difficulty dimension, a heart rate risk dimension and a postoperative state dimension;
In the key dimension feature extraction, GAN can be used for generating data matched with the real data features to realize the feature extraction of key dimensions, the attention weight generation module is used for splicing key dimension features and integrating the scattered features through preset attention weights, the LSTM network is a special cyclic neural network, the special cell state design makes the LSTM network excellent in processing and predicting time sequence data, long-term dependency relation can be captured, and the data matched with the real data features are suitable for predicting the spliced data features;
s1-5, constructing a knowledge graph by using a natural language processing algorithm according to the heart rate monitoring after pretreatment and the rehabilitation knowledge after operation, wherein the method comprises the following steps of:
S1-5-1, extracting a plurality of heart rate monitoring named entities and a plurality of postoperative rehabilitation named entities of the preprocessed heart rate monitoring and postoperative rehabilitation knowledge by using a pre-trained named entity extraction model;
The named entity extraction model is constructed based on a BERT-BiLSTM-CRF algorithm, a pre-training language sub-model (BERT, bidirectional Encoder Representation from Transformers) expressed by Bi-directional coding from Transformers is used for carrying out vector characterization on heart rate monitoring after pretreatment and postoperative rehabilitation knowledge to obtain a knowledge data vector, a Bi-directional long-short Term Memory network (BiLSTM, bi-directional Long Short-Term Memory) is used for extracting semantic features of the knowledge data vector, a linear chain member random field module (CRF, conditional Random Field) is used for marking named entities according to the semantic features to obtain a plurality of corresponding heart rate monitoring named entities and a plurality of postoperative rehabilitation named entities;
s1-5-2, extracting heart rate monitoring entity relations among a plurality of heart rate monitoring named entities and postoperative rehabilitation entity relations of a plurality of postoperative rehabilitation named entities by using a pre-trained entity relation extraction model;
The entity relation extraction model is constructed based on a BERT-BiGRU-Attention algorithm, a BERT model is used for carrying out vector characterization on the preprocessed heart rate monitoring and postoperative rehabilitation knowledge to obtain a knowledge data vector, a bidirectional circulating neural network (BiGRU, bidirectional Recurrent Neural Network) is used for extracting semantic features of the knowledge data vector, an Attention mechanism is used for distributing Attention weight to each channel of the bidirectional circulating neural network, influence caused by knowledge of relation labeling errors is reduced, and based on the BiGRU network, the heart rate monitoring entity relation among a plurality of heart rate monitoring named entities and a plurality of postoperative rehabilitation named entities, the knowledge data vector and the Attention weight are extracted according to the heart rate monitoring entity relation among a plurality of heart rate monitoring named entities and the postoperative rehabilitation entity relation among a plurality of postoperative rehabilitation named entities;
S1-5-3, constructing a knowledge graph according to a plurality of heart rate monitoring named entities, a plurality of heart rate monitoring entity relations, a plurality of postoperative rehabilitation named entities and a plurality of postoperative rehabilitation entity relations;
S1-6, constructing a heart rate abnormality recognition model by using an RNN-GNN-MLP algorithm according to a plurality of post-dimension reduction historical postoperative clinical data, a plurality of post-preprocessing historical heart rate acquisition data and a plurality of historical user portraits, and generating a plurality of historical heart rate abnormality recognition results;
The heart rate abnormality recognition model comprises a data feature extraction module constructed based on a recurrent neural network (Recurrent Neural Network, RNN) algorithm, a graph feature extraction module constructed based on a graph neural network (Graph Neural Network, GNN) algorithm and a fusion prediction module constructed based on a Multi-Layer Perceptron (MLP) algorithm;
The RNN network is a special neural network architecture, allows information to flow in time sequence data, can process and forecast the sequence data, is used for extracting postoperative clinical data characteristics, and is a neural network specially used for processing image data, wherein the image consists of nodes and edges, is used for extracting image characteristics of user images, comprises node information of user labels and edge information between the user labels, and is characterized in that neurons between layers are all connected, are used for carrying out characteristic fusion on different characteristics, obtaining corresponding fusion characteristics, improving data characterization capability, carrying out characteristic fusion, and carrying out heart rate abnormality recognition according to the fusion characteristics;
S1-7, constructing a heart rate monitoring optimization model by using a Deep Q Network (DQN) algorithm according to a plurality of historical heart rate abnormality recognition results and a plurality of historical user portraits, wherein the heart rate monitoring optimization model comprises the following steps of:
S1-7-1, generating a simulation environment serving as an DQN algorithm by using a heart rate monitoring optimization scheme, and constructing an intelligent body and an experience playback pool;
s1-7-2, defining a state space of the DQN algorithm according to each heart rate state type corresponding to the historical heart rate abnormality recognition result, wherein parameters of the state space correspond to each heart rate state, such as over-high heart rate, normal heart rate, over-slow heart rate and the like;
S1-7-3, defining an action space of the DQN algorithm according to actions required to be output by a heart rate monitoring optimization strategy, for example, reducing monitoring frequency, increasing monitoring duration, processing abnormal conditions and the like;
s1-7-4, defining a reward function of the DQN algorithm according to the possible influence condition of each action in the action space, and evaluating the advantages and disadvantages or the influence of the actions;
S1-7-5, constructing an input layer, a plurality of hidden layers and an output layer of a depth Q network, connecting the input layer to a state space, and connecting the output layer to an action space;
s1-7-6, optimizing training is carried out on the depth Q network and the intelligent agent based on a state space, an action space and a reward function according to a plurality of historical heart rate anomaly recognition results, a heart rate monitoring optimization model is constructed, and a plurality of generated historical heart rate monitoring optimization experiences are constructed;
S1-7-7, taking a historical user portrait as a retrieval tag of a historical heart rate monitoring optimization experience, and storing a plurality of historical heart rate monitoring optimization experiences provided with the retrieval tag into an experience playback pool;
s2, based on a user terminal, receiving input real-time postoperative clinical data and real-time heart rate acquisition data sent by heart rate acquisition equipment, and uploading the data to a cloud computing center;
S3, based on a cloud computing center, generating a user portrait by using a user portrait generation model according to real-time postoperative clinical data to obtain a real-time user portrait, wherein the method comprises the following steps of:
S3-1, performing data dimension reduction on real-time postoperative clinical data based on a cloud computing center according to a key clinical index set to obtain dimension-reduced real-time postoperative clinical data;
s3-2, inputting the real-time post-operation clinical data after the dimension reduction into a user portrait generation model, and extracting real-time key dimension characteristics of the real-time post-operation clinical data after the dimension reduction in N key dimensions by using a key dimension characteristic extraction module;
s3-3, according to the preset attention weight, using an attention weight generation module to carry out weighted splicing on the N real-time key dimension features to obtain real-time weighted spliced features;
S3-4, according to the real-time weighting splicing characteristics, using a user portrait generation module to conduct user label prediction to obtain a plurality of real-time user labels, and according to the plurality of real-time user labels, conducting user portrait generation to obtain a real-time user portrait;
S4, based on a cloud computing center, carrying out heart rate abnormality recognition by using a heart rate abnormality recognition model according to real-time postoperative clinical data, real-time heart rate acquisition data and real-time user portraits to obtain a real-time heart rate abnormality recognition result, wherein the method comprises the following steps of:
s4-1, inputting real-time post-operation clinical data, real-time heart rate acquisition data and real-time user portraits after dimension reduction into a heart rate abnormality recognition model based on a cloud computing center;
S4-2, extracting real-time postoperative clinical data characteristics of real-time postoperative clinical data after dimension reduction by using a data characteristic extraction module, extracting real-time heart rate acquisition data characteristics of real-time heart rate acquisition data by using a data characteristic extraction module, and extracting real-time graph characteristics of real-time user images by using a graph characteristic extraction module;
S4-3, performing feature fusion on the real-time postoperative clinical data feature, the real-time heart rate acquisition data feature and the real-time map feature by using a fusion prediction module to obtain a real-time fusion feature;
s4-4, carrying out heart rate abnormality recognition according to the real-time fusion characteristics to obtain a real-time heart rate abnormality prediction tag, namely a real-time heart rate abnormality recognition result;
S5, based on a cloud computing center, according to a real-time heart rate abnormality recognition result and a real-time user image, performing heart rate monitoring optimization by using a heart rate monitoring optimization model to obtain a real-time heart rate monitoring optimization strategy, wherein the method comprises the following steps of:
S1-3-1, based on a cloud computing center, matching in a plurality of retrieval labels according to real-time user images, and extracting a plurality of history heart rate monitoring optimization experiences matched in an experience playback pool of a heart rate monitoring optimization model;
S1-3-2, updating model parameters of a heart rate monitoring optimization model according to a plurality of historical heart rate monitoring optimization experiences and real-time heart rate abnormality recognition results to obtain updated model parameters, wherein the method comprises the following steps of:
S1-3-2-1, updating the action space of a heart rate monitoring optimization model according to a plurality of historical heart rate monitoring optimization experiences to obtain an updated action space A ' = [ a '1,...,a'j",...,a'I ], wherein a 'j" is an updated j ' action value, j ' is an action indication quantity, and I is the total number of dimensions of the action space;
s1-3-2-2, updating a state space of a heart rate monitoring optimization model according to a real-time heart rate abnormality recognition result to obtain an updated state space S '= [ S'1,...,s'i",...,s'I' ], wherein S 'i" is an updated I' state value, I 'is a state indication quantity, and I' is the total number of state space dimensions;
and S1-3-3, performing heart rate monitoring optimization according to an updated action space A '= [ a'1,...,a'j",...,a'I ] and an updated state space S '= [ S'1,...,s'i",...,s'I' ] by using a heart rate monitoring optimization model to obtain a plurality of real-time execution actions, and obtaining a real-time heart rate monitoring optimization strategy according to the plurality of real-time execution actions, wherein the method comprises the following steps:
S1-3-3-1, taking the updated state space S '= [ S'1,...,s'i',...,s'I ] as input of a heart rate monitoring optimization model, and generating a Q value of each possible action in the updated action space A '= [ a'1,...,a'j',...,a'I ] by using a depth Q network;
S1-3-3-2, acquiring a reward value of each possible action in the updated action space by using a reward function, and updating the Q value of the possible action according to the reward value to obtain an updated Q value of the possible action;
The formula is:
Q(s'p',a'p')＝(1-α")·Q(sp',ap')+α"·(R(sp',ap',s'p')+γ·Qmax(sp',ap'))
Wherein Q (s 'p',a'p') is an updated Q value corresponding to the updated state value s'p' and the updated action value a 'p', Q (sp',ap') is a predicted Q value corresponding to the state value sp' and the action value ap', alpha' is a learning rate, and Qmax(sp',ap') is the highest predicted Q value;
S1-3-3-3, repeating the steps until reaching the iteration frequency threshold, using a greedy strategy, taking the possible action corresponding to the highest updated Q value as an execution action, and outputting the execution action as a real-time heart rate monitoring optimization strategy, wherein the steps comprise reducing the monitoring frequency and increasing the monitoring duration;
s6, based on a cloud computing center, using a knowledge graph to adjust a real-time heart rate monitoring optimization strategy to obtain an adjusted real-time heart rate monitoring optimization strategy, wherein the method comprises the following steps of;
S6-1, based on a cloud computing center, performing search matching in heart rate monitoring named entities of a knowledge graph according to action named entities corresponding to real-time execution actions in a real-time heart rate monitoring optimization strategy, including 'reducing', 'monitoring', 'frequency', 'increasing', 'duration' and the like, so as to obtain a plurality of matched heart rate monitoring named entities, including 'reducing', 'monitoring equipment', 'frequency collecting', 'monitoring', 'frequency', 'duration of sampling', and the like;
S6-2, according to a plurality of heart rate monitoring named entities and the heart rate monitoring entity relation of the heart rate monitoring named entities and other heart rate monitoring named entities, adjusting a real-time heart rate monitoring optimization strategy to obtain an adjusted real-time heart rate monitoring optimization strategy, wherein the adjusting real-time heart rate monitoring optimization strategy comprises the steps of reducing the triggering frequency of central electrogram data acquisition of heart rate acquisition equipment, increasing the sampling time length of a motion sensor in the heart rate acquisition equipment, enhancing the motion acquisition, reducing the electrocardiograph acquisition and avoiding the motion from influencing the accuracy of electrocardiograph data;
S7, based on a cloud computing center, generating postoperative rehabilitation advice by using a knowledge graph according to the real-time user portrait and the real-time heart rate abnormality recognition result to obtain real-time postoperative rehabilitation advice, wherein the method comprises the following steps of;
S7-1, based on a cloud computing center, searching and matching in postoperative rehabilitation named entities of a knowledge graph according to a user named entity corresponding to a real-time user tag in a real-time user image and an abnormal named entity corresponding to a real-time heart rate abnormal prediction tag in a real-time heart rate abnormal recognition result to obtain a plurality of matched postoperative rehabilitation named entities;
S7-2, performing postoperative rehabilitation suggestion generation according to a plurality of matched postoperative rehabilitation named entities and postoperative rehabilitation entity relations of the matched postoperative rehabilitation named entities and other postoperative rehabilitation named entities to obtain real-time postoperative rehabilitation suggestions;
s8, based on a cloud computing center, sending real-time heart rate abnormality identification results, an adjusted real-time heart rate monitoring optimization strategy and real-time postoperative rehabilitation suggestions to corresponding user terminals;
S9, based on the user terminal, generating a real-time alarm signal according to the real-time heart rate abnormality recognition result, and visualizing the real-time alarm signal, the real-time heart rate abnormality recognition result and the real-time postoperative rehabilitation suggestion;
If the real-time abnormal heart rate identification result is that the heart rate is too low, generating a real-time alarm signal, performing red marking and highlighting treatment on the real-time alarm signal, and distributing the real-time alarm signal to a mobile terminal of medical staff and the like to realize automatic reminding;
and S10, based on the user terminal, generating a real-time heart rate monitoring instruction according to the adjusted real-time heart rate monitoring optimization strategy, and sending the real-time heart rate monitoring instruction to heart rate acquisition equipment.
Example 2:
As shown in fig. 3, the present embodiment provides a heart rate monitoring system based on a neural network, for implementing a heart rate monitoring method, where the system includes a cloud computing center, a plurality of user terminals, and a plurality of heart rate acquisition devices, where the cloud computing center is respectively connected with the plurality of user terminals in a communication manner, and each user terminal is correspondingly connected with one heart rate acquisition device in a communication manner;
The system comprises a user terminal, a cloud computing center, a real-time alarm signal generation device, a real-time heart rate abnormality recognition device, a real-time postoperative rehabilitation suggestion generation device, a real-time heart rate monitoring device and a heart rate monitoring device, wherein the user terminal is used for receiving input real-time postoperative clinical data and real-time heart rate acquisition data sent by the heart rate acquisition device and uploading the real-time heart rate acquisition data to the cloud computing center;
The heart rate acquisition equipment acquires real-time heart rate acquisition data of a user and sends the data to the user terminal;
The cloud computing center comprises a model building unit, a user portrait generating unit, a heart rate abnormality identifying unit, a heart rate monitoring optimizing unit, an optimizing strategy adjusting unit, a postoperative rehabilitation suggestion generating unit and a data transmitting unit which are connected in sequence;
The model building unit is used for building a user portrait generation model, a knowledge graph, a heart rate abnormality recognition model and a heart rate monitoring optimization model by using a neural network algorithm;
A user portrait generation unit for generating a user portrait by using a user portrait generation model according to real-time postoperative clinical data to obtain a real-time user portrait;
The heart rate abnormality identification unit is used for carrying out heart rate abnormality identification by using a heart rate abnormality identification model according to real-time postoperative clinical data, real-time heart rate acquisition data and real-time user portraits to obtain a real-time heart rate abnormality identification result;
The heart rate monitoring optimization unit is used for performing heart rate monitoring optimization by using a heart rate monitoring optimization model according to the real-time heart rate abnormality recognition result and the real-time user image to obtain a real-time heart rate monitoring optimization strategy;
The optimization strategy adjustment unit is used for adjusting the real-time heart rate monitoring optimization strategy by using the knowledge graph to obtain an adjusted real-time heart rate monitoring optimization strategy;
the postoperative rehabilitation suggestion generation unit is used for generating postoperative rehabilitation suggestions by using a knowledge graph according to the real-time user portrait and the real-time heart rate abnormality recognition result to obtain real-time postoperative rehabilitation suggestions;
And the data sending unit is used for sending the real-time heart rate abnormality identification result, the adjusted real-time heart rate monitoring optimization strategy and the real-time postoperative rehabilitation advice to the corresponding user terminal.
The invention discloses a heart rate monitoring method and system based on a neural network, which combine real-time postoperative clinical data and real-time heart rate acquisition data sent by heart rate acquisition equipment, more comprehensively reflect health condition information of a user, improve data comprehensiveness, perform unified management and analysis on the data based on a cloud computing center, improve practicability, build a user portrait generation model and a heart rate monitoring optimization model, generate personalized heart rate monitoring schemes according to specific health conditions and heart rate data of the user, realize customization functions of different rehabilitation situations and different users, improve user experience, build a heart rate abnormality identification model, combine the real-time postoperative clinical data, the real-time heart rate acquisition data and the real-time user portrait, perform heart rate abnormality identification, reduce labor cost investment, improve heart rate abnormality identification accuracy, build a knowledge graph, generate real-time postoperative rehabilitation suggestions according to the real-time user portrait and the real-time heart rate abnormality identification result, and improve functionality.
The invention is not limited to the alternative embodiments described above, but any person may derive other various forms of products in the light of the present invention. The above detailed description should not be construed as limiting the scope of the invention, which is defined in the claims and the description may be used to interpret the claims.