A non-contact microwave monitoring physiological parameter whole-process management method and system

Abstract

The application provides a whole-process management method and system for non-contact microwave monitoring of physiological parameters, which comprises the following steps: collecting microwave signals of physiological parameters of a user, pre-processing and feature extraction processing the microwave signals to obtain human vital sign parameters; transmitting the human vital sign parameters to a terminal of the user or a nurse station in a wireless manner to evaluate the human vital sign parameters; comparing the evaluation results with a health rule library to generate corresponding health suggestions and push the health suggestions to the terminal of the user; when the user is in a night sleep state, analyzing the human vital sign parameters in the terminal by using a pre-trained model, judging whether the user is in an abnormal breathing state according to the analysis results; if yes, sending a control instruction to a corresponding sickbed of the user to control the sickbed to perform an angle adjustment operation; and specifically, the whole-process management method of collecting signs-state judgment-linkage intervention is used to realize real-time response and intervention on the health state of the patient.

Description

Technical Field

[0001] This invention belongs to the field of medical monitoring full-process management technology, and specifically relates to a non-contact microwave monitoring physiological parameter full-process management method and system. Background Technology

[0002] Traditional health monitoring methods mostly rely on contact-based devices, such as electrocardiogram monitors, blood pressure monitors, and wearable devices. While these devices can provide relatively accurate physiological parameters, long-term use may lead to problems such as inconvenience in wearing them, strong dependence on them, and low patient comfort.

[0003] In recent years, microwave radar technology has emerged as a new means of monitoring vital signs due to its advantages such as non-contact operation, high precision, and continuous monitoring. By analyzing the subtle movements of the human body reflected by radar signals, it is possible to perceive and extract vital signs such as respiration, heart rate, and even blood pressure. Compared to traditional methods, microwave radar does not rely on physical contact and has greater adaptability and applicability to various scenarios.

[0004] However, existing technologies still have certain limitations in non-contact vital sign monitoring. For example, most systems only reach the data collection and display stage, lacking the ability to generate intelligent health suggestions and intervene in behaviors, and cannot make in-depth judgments on the user's health status based on real-time vital sign data. At the same time, current systems generally lack linkage mechanisms with surrounding medical devices, and cannot automatically respond when abnormal conditions are detected, thus missing the best intervention opportunity.

[0005] Especially when users are asleep, certain physiological abnormalities (such as snoring) are often overlooked, and if such problems are not addressed in time, they may pose potential risks to the user's health. Summary of the Invention

[0006] Based on this, the present invention provides a method and system for full-process management of non-contact microwave monitoring of physiological parameters. It aims to utilize high-precision non-contact monitoring technology of microwave radar, combined with wireless data transmission, pre-trained artificial intelligence models and medical device linkage control strategies, to achieve real-time response and intervention on the patient's health status. In particular, during the user's sleep, it can effectively alleviate breathing disorders such as snoring, improve the user's quality of life and the level of intelligent medical care.

[0007] The first aspect of this invention provides a method for the full-process management of non-contact microwave monitoring of physiological parameters, applicable to multi-medical institution scenarios with microwave radar devices, the method comprising:

[0008] Microwave signals of the user's physiological parameters are collected by a microwave radar device installed at a preset position on the hospital bed. The microwave signals are preprocessed and feature extracted to obtain human vital signs parameters. The microwave signals are used to determine whether the user has left the hospital bed. Based on the determination result, energy-saving operation is performed.

[0009] The human vital signs parameters are transmitted wirelessly to the user's or nurse's station terminal for evaluation.

[0010] The assessment results are compared with the health rule base to generate corresponding health recommendations, which are then pushed to the user's terminal.

[0011] When the user is asleep at night, the terminal analyzes the human vital signs parameters through a pre-trained model. Based on the analysis results, it is determined whether the user is in an abnormal breathing state. The pre-trained model is an LSTM model based on federated learning attention mechanism.

[0012] If so, a control command is sent to the corresponding user's bed to control the bed to perform an angle adjustment operation;

[0013] The training process of the pre-trained model includes:

[0014] Step 1: The control center node sends the same initial model to all edge nodes. The center node is deployed by a third-party platform and is used to distribute the initial model, aggregate the model parameters of each institution, and update the global model. The edge nodes include each medical service institution.

[0015] Step 2: Control each edge node to train the initial model using local data;

[0016] Step 3: The control center node receives the local parameters from each edge node and obtains the global model parameters by dynamically weighting and aggregating them according to a dual-dimensional method of data quality and model contribution.

[0017] Step four: The control center node distributes the global model parameters to each edge node. Each edge node uses the distributed global model parameters to continue local training and repeats steps two and three until the model accuracy converges.

[0018] Furthermore, the control center node receives local parameters from each edge node and obtains global model parameters using a dynamic weighted aggregation method based on both data quality and model contribution. The steps include:

[0019] The control center node distributes a global verification set to each edge node and defines three quality dimensions for the local data of each edge node, including signal integrity, label consistency, and scene diversity. The data quality score of the edge node is calculated by weighting the signal integrity, label consistency, and scene diversity.

[0020] Calculate the accuracy improvement value of the local model on the global validation set, and calculate the contribution of the edge nodes based on the accuracy improvement value;

[0021] Based on the data quality score and the contribution degree, the final weight of each edge node is calculated, and the parameters are aggregated to obtain the global model. The data quality score and contribution degree are updated after each round of training.

[0022] Furthermore, the input of the pre-trained model includes at least respiratory status data, user basic information, and medical history. The respiratory status data includes at least respiratory rate and blood oxygen saturation. The output is the adjustment angle of the hospital bed. Specifically, in the input layer, the respiratory status data is encoded to obtain respiratory status temporal features. The user basic information is passed through an embedding layer to obtain a basic information embedding vector. The medical history is passed through a convolutional layer to obtain medical history features. Then, the respiratory status temporal features, the basic information embedding vector, and the medical history features are concatenated into a composite input. In the attention-enhanced LSTM layer, a bidirectional LSTM is used to extract the temporal features of the composite input, and a self-attention mechanism is used to capture key temporal sequences to obtain an attention context vector. Then, the attention context vector is concatenated with the LSTM hidden state to obtain an enhanced feature expression. In the output layer, the enhanced feature expression is aggregated into a global vector, and the global vector is mapped to a continuous adjustment angle.

[0023] Furthermore, the human vital signs parameters include at least respiratory rate, heart rate, and blood pressure.

[0024] Furthermore, in the step of assessing the human vital signs parameters, the respiratory rhythm abnormalities are assessed based on the respiratory rate; heart rate changes are assessed based on the heart rate; and blood pressure is assessed to determine whether the blood pressure exceeds a threshold.

[0025] Furthermore, in the step of comparing the evaluation results with the health rule base, generating corresponding health suggestions, and pushing them to the user terminal, the combination of abnormal respiratory rhythm, heart rate changes, and whether blood pressure exceeds the threshold is matched with the health rule base to generate corresponding health suggestions.

[0026] Furthermore, when the abnormal breathing rhythm reaches a preset level, the following step is executed: when the user is in a sleep state at night, the terminal analyzes the human vital signs parameters through a pre-trained model, and determines whether the user is in an abnormal breathing state based on the analysis results.

[0027] A second aspect of this invention provides a full-process management system for non-contact microwave monitoring of physiological parameters, used to implement the full-process management method for non-contact microwave monitoring of physiological parameters described in the first aspect, the system comprising:

[0028] The acquisition module is used to acquire microwave signals of the user's physiological parameters through a microwave radar device installed at a preset position on the hospital bed, and to preprocess and extract features from the microwave signals to obtain human vital signs parameters. The microwave signals are used to determine whether the user has left the hospital bed, and energy-saving operations are performed based on the determination result.

[0029] The assessment module is used to wirelessly transmit the human vital signs parameters to the user's or nurse's station's terminal for assessment.

[0030] The comparison module is used to compare the assessment results with the health rule base, generate corresponding health suggestions, and push them to the user's terminal.

[0031] The judgment module is used to analyze the human vital signs parameters in the terminal when the user is in a sleep state at night through a pre-trained model, and to determine whether the user is in an abnormal breathing state based on the analysis results. The pre-trained model is an LSTM model based on a federated learning attention mechanism.

[0032] The control module is used to send control commands to the corresponding user's bed when it is determined that the user is in an abnormal breathing state, and control the bed to perform angle adjustment operations.

[0033] The training process of the pre-trained model includes:

[0034] Step 1: The control center node sends the same initial model to all edge nodes. The center node is deployed by a third-party platform and is used to distribute the initial model, aggregate the model parameters of each institution, and update the global model. The edge nodes include each medical service institution.

[0035] Step 2: Control each edge node to train the initial model using local data;

[0036] Step 3: The control center node receives the local parameters from each edge node and obtains the global model parameters by dynamically weighting and aggregating them according to a dual-dimensional method of data quality and model contribution.

[0037] Step four: The control center node distributes the global model parameters to each edge node. Each edge node uses the distributed global model parameters to continue local training and repeats steps two and three until the model accuracy converges.

[0038] A third aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the full-process management method for non-contact microwave monitoring of physiological parameters provided in the first aspect.

[0039] A fourth aspect of the present invention provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the full-process management method for non-contact microwave monitoring of physiological parameters provided in the first aspect.

[0040] This invention provides a non-contact microwave monitoring method and system for managing physiological parameters throughout the entire process. The method involves using a microwave radar device installed at a pre-defined location on the patient's bed to collect microwave signals of the user's physiological parameters. These signals are then pre-processed and feature extracted to obtain vital sign parameters. These parameters are wirelessly transmitted to a terminal at the user's or nurse's station for evaluation. The evaluation results are compared with a health rule base to generate corresponding health suggestions, which are then pushed to the user's terminal. When the user is asleep at night, the terminal analyzes the vital sign parameters using a pre-trained model. Based on the analysis, it determines whether the user is experiencing abnormal breathing. If so, a control command is sent to the corresponding patient's bed to adjust its angle. Specifically, this complete process management method, from vital sign collection to status assessment and intervention, enables real-time response and intervention to the patient's health status. Attached Figure Description

[0041] Figure 1 The flowchart illustrates the implementation of a non-contact microwave monitoring method for managing physiological parameters, as provided in Embodiment 1 of the present invention.

[0042] Figure 2 This is a structural block diagram of a full-process management system for non-contact microwave monitoring of physiological parameters provided in Embodiment 2 of the present invention;

[0043] Figure 3 This is a structural block diagram of an electronic device provided in Embodiment 3 of the present invention. Detailed Implementation

[0044] To facilitate understanding of the present invention, a more complete description will be given below with reference to the accompanying drawings. Several embodiments of the invention are illustrated in the drawings. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete.

[0045] It should be noted that when a component is said to be "fixed to" another component, it can be directly on the other component or there may be an intervening component. When a component is said to be "connected to" another component, it can be directly connected to the other component or there may be an intervening component. The terms "vertical," "horizontal," "left," "right," and similar expressions used in this document are for illustrative purposes only.

[0046] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0047] Example 1

[0048] According to an embodiment of the present invention, a method for full-process management of non-contact microwave monitoring of physiological parameters is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0049] This first embodiment provides a method for the entire process management of non-contact microwave monitoring of physiological parameters, which can be used in electronic devices, such as computers. Please refer to... Figure 1 , Figure 1 The flowchart of the implementation of a non-contact microwave monitoring physiological parameter full-process management method provided in Embodiment 1 of the present invention is shown, specifically including steps S01 to S05.

[0050] Step S01: Microwave signals of the user's physiological parameters are collected by a microwave radar device installed at a preset position on the hospital bed, and the microwave signals are preprocessed and feature extracted to obtain human vital signs parameters.

[0051] This system integrates microwave radar for non-contact monitoring of the human body, extracting vital signs including respiratory rate, heart rate, and blood pressure. The collected raw microwave signals are denoised and enhanced by a signal processing module, transforming them into standardized physiological signal data. This data is then transmitted in real-time to user terminals (such as nurse station screens, medical staff mobile terminals, and user smartphones) using a low-power wireless communication module (BLE or Wi-Fi), ensuring data continuity and remote visualization.

[0052] Specifically, a 77GHz non-contact millimeter-wave microwave radar is used as the core sensor, deployed near the user's bed or wheelchair. It utilizes the Doppler effect and a phased array antenna to achieve highly sensitive detection of minute surface displacements of the human body (such as chest and abdominal movements, heart vibrations, etc.). The radar-collected data undergoes signal processing (including filtering, noise reduction, and FFT transformation) to extract respiratory rate, heart rate, and blood pressure characteristic parameters. The data acquisition terminal uploads the standardized vital signs data to the nurse station screen, a mobile app, or a cloud server via BLE (Bluetooth Low Energy) or Wi-Fi modules, enabling 24-hour remote monitoring.

[0053] In addition, the microwave signal is used to determine whether the user has left the bed, and energy-saving operations are performed based on the determination result.

[0054] Step S02: The human vital signs parameters are transmitted wirelessly to the user's or nurse's station terminal for evaluation.

[0055] It should be noted that the assessment of human vital signs parameters is completed by evaluating respiratory rate to determine abnormalities in respiratory rhythm, heart rate to assess heart rate changes, and blood pressure to determine whether blood pressure exceeds a threshold, thereby obtaining the assessment results.

[0056] Step S03: Compare the assessment results with the health rule base, generate corresponding health recommendations, and push them to the user's terminal.

[0057] Specifically, the system matches a combination of abnormal respiratory rhythm, heart rate changes, and blood pressure exceeding a threshold with a health rule base to generate corresponding health recommendations. More specifically, it assesses vital sign data by setting multi-level health management rules (such as threshold judgment and trend analysis) and pushes personalized recommendations to users and caregivers. For example, when a significantly elevated heart rate, abnormal respiratory rhythm, or blood pressure exceeding a warning value is detected, the system can push rest reminders, hydration prompts, and medical advice.

[0058] More specifically, the user terminal has a built-in health advice push module that analyzes and evaluates received vital sign data in real time based on a preset rule base. The advice pushes are divided into two categories:

[0059] Health status notifications: If the heart rate is ≥110 bpm or the respiratory rate is abnormally high (≥24 breaths / minute), the system will send reminders such as "Please remain rested and avoid strenuous activity" or "It is recommended to drink water or rest." If the health indicators remain abnormal, the system will send a message suggesting "Go to the hospital for examination" or activate the family doctor remote contact mechanism.

[0060] Sleep status notifications: When frequent snoring, sleep apnea, or a downward trend in blood oxygen saturation is detected at night, notifications such as "Abnormal snoring detected, we suggest you adjust your posture" or "We have adjusted the angle of your bed, please pay attention to your comfort" will be sent to ensure the user's sleep safety.

[0061] The push rules can be based on an initial template developed by experts, or dynamically adjusted and optimized as data accumulates to improve the level of intelligence.

[0062] Step S04: When the user is in a sleep state at night, the terminal analyzes the human vital signs parameters through a pre-trained model. Based on the analysis results, it is determined whether the user is in an abnormal breathing state. If so, step S05 is executed.

[0063] In this embodiment of the invention, the terminal integrates a pre-trained model, which is an LSTM model based on federated learning attention mechanism. This model accurately identifies abnormal states such as snoring and sleep apnea by continuously analyzing the respiratory signal waveforms during the user's sleep process. During model training, real-time data such as respiration, heart rate, and blood pressure are incorporated, along with the collection of basic user information (such as gender, age, weight, and height) and medical history (such as whether the user has respiratory diseases, cardiovascular problems, and snoring frequency). By constructing a dataset containing multi-dimensional features such as respiratory status, user tags, and bed angle, a deep learning model is trained to achieve personalized mapping between respiratory status and bed angle.

[0064] It should be noted that traditional models (such as the attention-based LSTM model) rely on data from a single scenario (e.g., a hospital) for training, resulting in poor generalization (e.g., they cannot be adapted to elderly patients in community nursing homes). Furthermore, data privacy restrictions between hospitals prevent the sharing of training data. This invention introduces federated learning to achieve the goal of multiple institutions training the model together without sharing data, thereby improving the model's adaptability to different scenarios (hospitals, nursing homes, home-based care) while protecting data privacy.

[0065] In this embodiment of the invention, the training process of the pre-trained model includes:

[0066] Step 1: The control center node sends the same initial model to all edge nodes. The control center node is deployed by a third-party platform (such as a regional health information center) and is used to distribute the initial model, aggregate the model parameters of each institution, and update the global model. The edge nodes include various medical service institutions, such as hospitals, nursing homes, and home-based care service providers.

[0067] Step 2: Control each edge node to train the initial model using local data (such as postoperative patient data from hospital nodes and disabled elderly data from nursing home nodes). Only the updated values ​​of the model parameters are uploaded, not the original data, to avoid data leakage.

[0068] Step 3: The control center node receives the local parameters from each edge node and obtains the global model parameters by dynamically weighting and aggregating them according to a dual-dimensional method of data quality and model contribution.

[0069] Step four: The control center node distributes the global model parameters to each edge node. Each edge node uses the distributed global model parameters to continue local training and repeats steps two and three until the model accuracy converges (e.g., test set accuracy ≥ 92%).

[0070] It should be noted that in step three above, the traditional method uses a weighted average based on the percentage of data volume at each node (e.g., a node with 30% of the data volume has a parameter weight of 0.3) to obtain the global model parameters. This method ignores the differences in data quality. Understandably, the traditional method only uses data volume as the weighting criterion, but if the local data of a certain edge node is noisy (e.g., microwave radar signals are frequently interfered with), its parameters may reduce the accuracy of the global model. To address this, the solution in step three is proposed, which quantifies the relationship between data quality and model contribution, assigning dynamic weights to each node to achieve the goal of high-quality data contributing more and low-quality data having less impact.

[0071] The control center node receives local parameters from each edge node and obtains global model parameters using a dynamic weighted aggregation method based on both data quality and model contribution. The steps include:

[0072] The control center node distributes a global validation set to each edge node and defines three quality dimensions for the local data of each edge node, including signal integrity, label consistency, and scene diversity. Based on the signal integrity, label consistency, and scene diversity, the data quality score of the edge node is calculated by weighting. Here, signal integrity refers to the proportion of effective physiological signals in microwave radar data, label consistency refers to the matching degree between manually labeled abnormal respiratory states and model prediction results, and scene diversity refers to the richness of user types (age, underlying diseases) and environmental scenes (temperature and humidity range) covered by local data. In this embodiment of the invention, each dimension is scored from 0 to 10, and the data quality score Q is calculated by weighting.

[0073] Calculate the accuracy improvement of the local model on the global validation set (public test data held by the central node that does not contain private data of each edge node), and calculate the contribution of the edge nodes based on the accuracy improvement. The calculation formula is as follows:

[0074] ;

[0075] Where C represents the contribution level. The value represents the improvement in accuracy, and N represents the amount of local data.

[0076] Based on the data quality score and contribution degree, the final weight of each edge node is calculated, and the parameters are aggregated to obtain the global model. The data quality score and contribution degree are updated after each training round. The formula for calculating the final weight is:

[0077] ;

[0078] Where W is the final weight, α and β are coefficients, and .

[0079] Specifically, the input to the pre-trained model includes at least respiratory status data, user basic information, and medical history. The respiratory status data includes at least respiratory rate and blood oxygen saturation. The output is the adjustment angle of the hospital bed.

[0080] More specifically, respiratory rate and blood oxygen saturation are used as time series inputs, and a time series is constructed using a sliding window. The time series is then standardized to construct a two-dimensional time series feature vector. Where T is the time step, For the real number field,

[0081] ;

[0082] Norm(•) is a normalization function, such as Z-score or Min-Max normalization; RF is blood oxygen saturation; and SpO2 is respiratory rate. Let be the respiratory state feature vector at time step t. Let be the respiratory rate at time step t. Let be the blood oxygen saturation at time step t;

[0083] Furthermore, after normalizing the user's basic information (such as age, gender, height, weight, etc.), it is mapped into a low-dimensional dense vector through the embedding layer. ,

[0084] ;

[0085] For discrete features of the input, For embedding layer parameters, Let be the dimension of the low-dimensional dense vector, and Embedding()• be the embedding layer operation;

[0086] Furthermore, medical history data is treated as a sequence of events to construct time-series features. , The time steps of the medical history data. For each time step's feature dimension, local features are extracted using convolutional layers:

[0087] ;

[0088] W c b c Here are the convolution kernel parameters, and Conv1D(•) is a 1D convolution operation;

[0089] Finally, the outputs of the three branches are concatenated. Specifically, the temporal features of respiratory status, the basic information embedding vector, and the medical history features are concatenated into a composite input:

[0090] ;

[0091] D in The feature dimensions after fusion;

[0092] In this embodiment of the invention, a bidirectional LSTM (Bi-LSTM) is used to capture past and future context information and output a hidden state sequence:

[0093] ;

[0094] H represents the dimension of the unidirectional LSTM hidden layer. This represents the hidden state of the forward LSTM in step 1. This represents the hidden state of the forward LSTM at step T. This represents the hidden state of the inverse LSTM in step 1. This represents the hidden state of the inverse LSTM at step T. For a forward LSTM (moving from step 1 to step T), the hidden state at step t is... For a reverse LSTM (going backwards from step T to step 1), the hidden state at step t is represented by Concat, which is the concatenation operation. For the final hidden state at step t, attention weights are calculated on the hidden state sequence output by the Bi-LSTM to highlight key time steps related to bedside adjustment.

[0095] ;

[0096] , as well as These represent the Query, Key, and Value vectors, respectively. Let t be the query vector at step t. Let be the key vector at step t. Let t be the value vector at step t. Let D be the attention parameter matrix. a For the attention dimension;

[0097] Attention scores and weights are represented as follows:

[0098] ;

[0099] Let be the similarity between the Query at step t and the Key at step s. It is the transpose symbol. Let be the transpose of the key vector at step s. Let be the attention weight from step s to step t;

[0100] The context vector is represented as:

[0101] ;

[0102] c t For the context-aware features at step t, Let s be the value vector at step s;

[0103] It should be noted that concatenating the attention context vector with the LSTM hidden state enhances feature representation:

[0104] ;

[0105] To enhance feature representation, Concat() is used for concatenation, followed by layer normalization:

[0106] ;

[0107] For the layer normalization result, LayerNorm()• is the layer normalization operation. In the output layer of the model, the temporal features are aggregated into a global vector through the pooling layer. :

[0108] ;

[0109] The global features are then mapped to a continuous adjustment angle θ (in degrees) through a fully connected layer, which is the original unconstrained prediction value:

[0110] ;

[0111] ;

[0112] This is the weight matrix. The bias term is represented by Linear(•), which is a linear transformation operation, σ is the Sigmoid function, and θ is the bias term. max To maximize the adjustable angle, This is the final constrained angle prediction value.

[0113] During model training, mean squared error (MSE) is used to measure the difference between the predicted angle and the true angle θ. * Differences:

[0114] ;

[0115] Where N is the loss value and N is the number of samples. Let be the predicted angle of the model for the i-th sample. Let be the true angle of the i-th sample. In addition, dropout is used to prevent overfitting by adding a dropout layer between the LSTM layer and the fully connected layer.

[0116] In step S05, a control command is sent to the corresponding user's bed to control the bed to perform an angle adjustment operation.

[0117] Understandably, the detected abnormal breathing states are used as trigger signals. Combined with the "user breathing state - bed posture angle" mapping established in the big data model, control commands are output to the intelligent wheelchair bed to automatically adjust the bed angle (e.g., raising the back by 15°~45°). This strategy aims to effectively alleviate upper airway obstruction by scientifically changing the user's sleeping posture, thereby alleviating or preventing snoring and sleep apnea. For example, if mild snoring is detected: a control signal is output to raise the bed back angle to 15°; if sleep apnea persists: the model outputs a suggested angle increase to 30°~45°, and the strategy is gradually optimized based on user feedback. After responding to the control signal, the bed control system executes the angle adjustment through a stepper motor drive structure and feeds back the current state to update training data and improve model accuracy.

[0118] In other embodiments of the present invention, when the abnormal respiratory rhythm reaches a preset level, the step of analyzing the human vital signs parameters in the terminal through a pre-trained model when the user is in a nighttime sleep state, and determining whether the user is in an abnormal respiratory state based on the analysis results, is executed. That is, in order to achieve the long-term stable operation and energy-saving requirements of the system, the following strategy is adopted:

[0119] Layered power consumption management: The system divides the microwave radar, data transmission, model analysis, and bed-linked control modules into three layers: "real-time operation layer," "standby monitoring layer," and "deep sleep layer." During non-critical periods, it automatically enters low-power or sleep mode, retaining only basic heart rate monitoring and abnormal status wake-up mechanisms.

[0120] Event triggering mechanism: The system only activates the terminal push and bed control modules when certain abnormal indicator trends are met (such as a decrease in respiratory amplitude exceeding 30%, or a continuous increase in snoring intensity), avoiding frequent and ineffective operation. It can also determine whether the user has left the bed via microwave signals; if the system detects that the user has left the bed, it stops monitoring.

[0121] In addition, edge computing optimization can be adopted, that is, the initial calculation of the model can be completed locally on the terminal or bedside device, reducing the frequency of cloud communication, reducing network and computing power consumption, and improving real-time response speed.

[0122] In summary, the non-contact microwave monitoring physiological parameter management method described in the above embodiments of the present invention collects microwave signals of the user's physiological parameters through a microwave radar device installed at a preset position on the hospital bed, and performs preprocessing and feature extraction on the microwave signals to obtain human vital sign parameters; the human vital sign parameters are wirelessly transmitted to the user's or nurse's station terminal for evaluation; the evaluation results are compared with a health rule base to generate corresponding health suggestions, which are then pushed to the user's terminal; when the user is asleep at night, the terminal analyzes the human vital sign parameters through a pre-trained model, and determines whether the user is in an abnormal breathing state based on the analysis results; if so, a control command is sent to the corresponding user's hospital bed to control the bed to perform angle adjustment operations. Specifically, through a complete process management method from vital sign collection to state judgment to linkage intervention, real-time response and intervention to the patient's health status are achieved.

[0123] Example 2

[0124] Please see Figure 2 , Figure 2 This is a structural block diagram of a non-contact microwave monitoring physiological parameter end-to-end management system 200 provided in Embodiment 2 of the present invention. This system is used to implement the above embodiments and preferred embodiments; details already described will not be repeated. As used below, the term "module" can refer to a combination of software and / or hardware that performs a predetermined function. Although the apparatus described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0125] Specifically, the non-contact microwave monitoring physiological parameter end-to-end management system 200 includes: a data acquisition module 21, an evaluation module 22, a comparison module 23, a judgment module 24, and a control module 25, wherein:

[0126] The acquisition module 21 is used to acquire microwave signals of the user's physiological parameters through a microwave radar device installed at a preset position on the hospital bed, and to preprocess and extract features from the microwave signals to obtain human vital signs parameters. The microwave signals are used to determine whether the user has left the hospital bed, and energy-saving operation is performed based on the determination result. The human vital signs parameters include at least respiratory rate, heart rate and blood pressure.

[0127] The assessment module 22 is used to wirelessly transmit the human vital signs parameters to the user's or nurse's station terminal, and to assess the human vital signs parameters. Based on the respiratory rate, it assesses abnormal respiratory rhythm; based on the heart rate, it assesses heart rate changes; and based on the blood pressure, it assesses whether the blood pressure exceeds the threshold.

[0128] The comparison module 23 is used to compare the assessment results with the health rule base, generate corresponding health suggestions, and push them to the user terminal. It matches the combination of abnormal respiratory rhythm, heart rate changes, and whether blood pressure exceeds the threshold with the health rule base to generate corresponding health suggestions.

[0129] The judgment module 24 is used to analyze the human vital signs parameters in the terminal through a pre-trained model when the user is in a sleep state at night, and to determine whether the user is in an abnormal breathing state based on the analysis results. The pre-trained model is an LSTM model based on an attention mechanism network. In addition, the input of the pre-trained model includes at least respiratory status data, user basic information and medical history. The respiratory status data includes at least respiratory rate and blood oxygen. The output is the adjustment angle of the hospital bed.

[0130] Furthermore, when the abnormal breathing rhythm reaches a preset level, the following step is executed: when the user is in a sleep state at night, the human vital signs parameters are analyzed in the terminal through a pre-trained model, and the user is judged to be in an abnormal breathing state based on the analysis results.

[0131] The control module 25 is used to send control commands to the corresponding user's bed when it is determined that the user is in an abnormal breathing state, and control the bed to perform angle adjustment operations.

[0132] Example 3

[0133] In another aspect, the present invention also proposes an electronic device, please refer to [link to relevant documentation]. Figure 3 The image shows an electronic device according to Embodiment 3 of the present invention, including a memory 20, a processor 10, and a computer program 30 stored in the memory and executable on the processor. When the processor 10 executes the computer program 30, it implements the above-described method for full-process management of non-contact microwave monitoring of physiological parameters.

[0134] In some embodiments, the processor 10 may be a central processing unit (CPU), controller, microcontroller, microprocessor or other data processing chip, used to run program code stored in memory 20 or process data, such as executing access restriction programs.

[0135] The memory 20 includes at least one type of readable storage medium, such as flash memory, hard disk, multimedia card, card-type memory (e.g., SD or DX memory), magnetic memory, magnetic disk, optical disk, etc. In some embodiments, the memory 20 can be an internal storage unit of an electronic device, such as the hard disk of the electronic device. In other embodiments, the memory 20 can also be an external storage device of the electronic device, such as a plug-in hard disk, SmartMediaCard (SMC), SecureDigital (SD) card, FlashCard, etc., equipped on the electronic device. Furthermore, the memory 20 can include both internal and external storage units of the electronic device. The memory 20 can be used not only to store application software and various types of data of the electronic device, but also to temporarily store data that has been output or will be output.

[0136] It should be pointed out that, Figure 3 The structure shown does not constitute a limitation on the electronic device. In other embodiments, the electronic device may include fewer or more components than shown, or combine certain components, or have different component arrangements.

[0137] This invention also proposes a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method for the full-process management of non-contact microwave monitoring of physiological parameters.

[0138] Those skilled in the art will understand that the logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as a ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can mean any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0139] More specific examples of computer-readable media (a non-exhaustive list) include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0140] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0141] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0142] The above embodiments merely illustrate several implementation methods of the present invention, and their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this patent should be determined by the appended claims.

Claims

1. A full-process management system for non-contact microwave monitoring of physiological parameters, characterized in that, The system, applicable to multi-medical institution scenarios equipped with microwave radar devices, includes: The acquisition module is used to acquire microwave signals of the user's physiological parameters through a microwave radar device installed at a preset position on the hospital bed, and to preprocess and extract features from the microwave signals to obtain human vital signs parameters. The microwave signals are used to determine whether the user has left the hospital bed, and energy-saving operations are performed based on the determination result. The assessment module is used to wirelessly transmit the human vital signs parameters to the user's or nurse's station's terminal for assessment. The comparison module is used to compare the assessment results with the health rule base, generate corresponding health suggestions, and push them to the user's terminal. The judgment module is used to analyze the human vital signs parameters in the terminal when the user is in a sleep state at night through a pre-trained model, and to determine whether the user is in an abnormal breathing state based on the analysis results. The pre-trained model is a Bi-LSTM model based on federated learning. The control module is used to send control commands to the corresponding user's bed when it is determined that the user is in an abnormal breathing state, and control the bed to perform angle adjustment operations. The training process of the pre-trained model includes: Step 1: The control center node sends the same initial model to all edge nodes. The center node is deployed by a third-party platform and is used to distribute the initial model, aggregate the model parameters of each institution, and update the global model. The edge nodes include each medical service institution. Step 2: Control each edge node to train the initial model using local data; Step 3: The control center node receives the local parameters from each edge node and obtains the global model parameters by dynamically weighting and aggregating them according to a dual-dimensional method of data quality and model contribution. Step 4: The control center node distributes the global model parameters to each edge node. Each edge node uses the distributed global model parameters to continue local training and repeats steps 2 and 3 until the model accuracy converges. The control center node receives local parameters from each edge node and obtains global model parameters using a dynamic weighted aggregation method based on both data quality and model contribution. The steps include: The control center node distributes a global verification set to each edge node and defines three quality dimensions for the local data of each edge node, including signal integrity, label consistency, and scene diversity. The data quality score of the edge node is calculated by weighting the signal integrity, label consistency, and scene diversity. Calculate the accuracy improvement value of the local model on the global validation set, and calculate the contribution of the edge nodes based on the accuracy improvement value; Based on the data quality score and the contribution degree, the final weight of each edge node is calculated, and the parameters are aggregated to obtain the global model. The data quality score and contribution degree are updated after each round of training.

2. The full-process management system for non-contact microwave monitoring of physiological parameters according to claim 1, characterized in that, The input to the pre-trained model includes at least respiratory status data, user basic information, and medical history. The respiratory status data includes at least respiratory rate and blood oxygen saturation. The output is the adjustment angle of the hospital bed. Specifically, in the input layer, the respiratory status data is encoded to obtain respiratory status temporal features. The user basic information is embedded through an embedding layer to obtain a basic information embedding vector. The medical history is passed through a convolutional layer to obtain medical history features. Then, the respiratory status temporal features, the basic information embedding vector, and the medical history features are concatenated into a composite input. Bi-LSTM is used to extract the temporal features of the composite input, and a self-attention mechanism is used to capture key temporal sequences to obtain an attention context vector. Then, the attention context vector is concatenated with the Bi-LSTM hidden state to obtain an enhanced feature expression. In the output layer, the enhanced feature expression is aggregated into a global vector, and the global vector is mapped to a continuous adjustment angle.

3. The full-process management system for non-contact microwave monitoring of physiological parameters according to claim 2, characterized in that, The human vital signs parameters include at least respiratory rate, heart rate, and blood pressure.

4. The full-process management system for non-contact microwave monitoring of physiological parameters according to claim 3, characterized in that, In the step of assessing the human vital signs parameters, respiratory rhythm abnormalities are assessed based on respiratory rate; heart rate changes are assessed based on heart rate; and blood pressure is assessed to determine whether blood pressure exceeds a threshold.

5. The full-process management system for non-contact microwave monitoring of physiological parameters according to claim 4, characterized in that, In the step of comparing the assessment results with the health rule base, generating corresponding health suggestions, and pushing them to the user terminal, the combination of abnormal respiratory rhythm, heart rate changes, and whether blood pressure exceeds the threshold is matched with the health rule base to generate corresponding health suggestions.

6. The full-process management system for non-contact microwave monitoring of physiological parameters according to claim 5, characterized in that, When the abnormal breathing rhythm reaches a preset level, the following step is executed: when the user is in a sleep state at night, the terminal analyzes the human vital signs parameters through a pre-trained model, and determines whether the user is in an abnormal breathing state based on the analysis results.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by the processor, the program implements the execution steps in the full-process management system for non-contact microwave monitoring of physiological parameters as described in any one of claims 1-6.

8. An electronic device, characterized in that, It includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the execution steps in the full-process management system for non-contact microwave monitoring of physiological parameters as described in any one of claims 1-6.

Images