CAS-based chronic disease health monitoring method and system

Abstract

The invention discloses a CAS-based chronic disease health monitoring method and system, and relates to the technical field of intelligent monitoring. The method comprises the following steps: acquiring heartbeat data and respiration data of a target user through a CAS system, and performing respiration event detection on the target user according to the respiration data to obtain an activity state and respiration signal data; performing adaptive heartbeat signal extraction on the heartbeat data according to the activity state and the respiration signal data to obtain heartbeat signal data; and performing activity state early warning on the target user according to the heartbeat signal data. Non-contact breathing and heartbeat monitoring, heartbeat signal extraction, breathing event recognition and breathing interference suppression in heartbeat signals are achieved through a CAS system, and accurate early warning and dynamic parameter adjustment are achieved through multi-dimensional physiological feature analysis so as to enhance signal separation robustness and reduce the false alarm rate.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent monitoring, and particularly relates to a chronic disease health monitoring method and system based on CAS. Background Art

[0002] With the global population aging and the increasing number of chronic disease patients, the traditional medical monitoring mode is facing challenges. Chronic disease patients need long-term monitoring, but the traditional method is time-consuming and resource-consuming, and it is difficult to track in real time. In recent years, the development of Internet of Things, big data and artificial intelligence technologies has promoted the chronic disease monitoring based on the Continuous Health Monitoring System (CAS). CAS collects vital sign data in real time through wearable devices and sensors, uses algorithms for analysis and early warning, provides a non-invasive and convenient monitoring solution, improves the quality of life of patients, and optimizes the allocation of medical resources, becoming an important development direction for chronic disease management.

[0003] Publication number CN119523442A discloses an intelligent elderly health monitoring system. The heart rate module collects heart rate signals, identifies local peaks in the signals to detect heartbeats, measures the time interval between consecutive peaks to determine the inter-beat interval, and calculates the difference between adjacent intervals to evaluate variability. In the present invention, through the synchronous analysis of multi-dimensional data of comprehensive heart rate, skin conductance and facial expressions, the comprehensive understanding and evaluation ability of the health status of the elderly is enhanced. It does not rely solely on any physiological index, but accurately captures and analyzes the complex relationship between emotional changes and physiological responses by integrating multi-source data. Through this method, health warning signals can be identified in subtle changes and early warnings can be given in a timely manner. In addition, it can also dynamically adjust environmental factors such as lighting, music and fragrance according to the real-time monitored emotions and health status to provide a more personalized and optimized environmental response.

[0004] In the prior art, the frequency bands of respiratory movement and heartbeat signals overlap, and it is difficult for traditional filtering methods to dynamically separate them. Contact monitoring brings inconvenience and limitations, and existing non-contact monitoring technologies have large signal interference and low monitoring accuracy in complex environments, resulting in false early warnings. Summary of the Invention

[0005] The object of the present invention is to solve the problems that the frequency bands of respiratory movement and heartbeat signals overlap, it is difficult for traditional filtering methods to dynamically separate them, contact monitoring brings inconvenience and limitations, and existing non-contact monitoring technologies have large signal interference and low monitoring accuracy in complex environments, resulting in false early warnings, and to propose a chronic disease health monitoring method and system based on CAS.

[0006] In the first aspect of the implementation of the present invention, a chronic disease health monitoring method based on CAS is first proposed. The method includes:

[0007] Obtain the heartbeat data and breathing data of the target user through the CAS system, and perform breathing event detection on the target user according to the breathing data to obtain the activity status and breathing signal data; the CAS system is a data acquisition system;

[0008] Extract the adaptive heartbeat signal from the heartbeat data according to the activity status and the breathing signal data to obtain the heartbeat signal data;

[0009] Perform activity status warning on the target user according to the heartbeat signal data.

[0010] Optionally, performing breathing event detection on the target user according to the breathing data to obtain the activity status and breathing signal data includes:

[0011] Step 1, extract amplitude data and phase data according to the breathing data, and normalize the amplitude data and the phase data to obtain target amplitude data and target phase data;

[0012] Step 2, concatenate the target amplitude data and the target phase data to obtain concatenated data, and use the concatenated data as the input of the encoder to obtain an embedding vector;

[0013] Step 3, generate an embedding matrix according to the embedding vector, and input the embedding matrix into the multi-head attention module to obtain a first feature vector;

[0014] Step 4, add noise to the breathing data to obtain contaminated breathing data, perform Steps 1-3 on the contaminated breathing data to obtain a second feature vector, and calculate the similarity between the first feature vector and the second feature vector through a contrast loss function to obtain the similarity;

[0015] Step 5, determine the breathing event detection weight according to the similarity, update the weight of the breathing event detection according to the breathing event detection weight, perform breathing event detection on the breathing data to obtain the activity status, and determine the breathing signal data corresponding to the activity status; the activity status includes: exercise, sleep, and daily.

[0016] Optionally, extracting the adaptive heartbeat signal from the heartbeat data according to the activity status and the breathing signal data includes:

[0017] Initialize the parameter combination of the optimization algorithm and the VMD algorithm, and use the parameter combination of the VMD algorithm as an individual of the population; the parameters of the optimization algorithm are the population size and the maximum number of iterations, and the parameters of the VMD algorithm are the number of modes K and the penalty parameter α;

[0018] Calculate the fitness of each individual, determine the optimal individual according to the fitness, use the optimal individual as the update target to update the population, and determine the number of iterations;

[0019] If the number of iterations meets the preset detection condition, update the modal spectrum, and perform central frequency update and Lagrange multiplier update;

[0020] If the maximum number of iterations is reached or when the update values of all modes converge to a preset value, stop the update and output the first signal data and the second signal data; otherwise, repeat the update.

[0021] Optionally, after outputting the first signal data and the second signal data, it further includes:

[0022] Calculate the similarity between the first signal data and the second signal data and the respiratory signal data respectively to obtain the similarity, and determine the heartbeat signal data according to the similarity;

[0023] Modal spectrum iterative separation formula:

[0024]

[0025] Among them, u t+1 (ω) represents the updated spectrum, f(ω) represents the spectrum of the original signal, λ(ω) represents the Lagrange multiplier, u other represents the spectra of other modes, α represents the penalty factor, ω k represents the central frequency of the k-th mode, t represents the number of iterations, and ω0 represents the standard central frequency.

[0026] Optionally, perform activity status warning on the target user according to the heartbeat signal data, including:

[0027] Determine the heart rate of the target user according to the heartbeat signal data. If the heart rate is within the preset interval, determine that the target user is in a normal state;

[0028] If the heart rate is outside the preset interval, determine that the target user is in an abnormal state.

[0029] In the second aspect of the implementation of the present invention, a chronic disease health monitoring system based on CAS is proposed, including: a data acquisition module, a data extraction module, and an activity warning module:

[0030] The data acquisition module is used to obtain the heartbeat data and respiratory data of the target user through the CAS system, and perform respiratory event detection on the target user according to the respiratory data to obtain the activity status and respiratory signal data; the CAS system is a data acquisition system;

[0031] The data extraction module is used to adaptively extract a heartbeat signal from the heartbeat data according to the activity status and the respiratory signal data to obtain heartbeat signal data;

[0032] The activity warning module is used to give an activity status warning to the target user according to the heartbeat signal data.

[0033] Optionally, the data acquisition module includes: a first execution module, a second execution module, a third execution module, a fourth execution module, and a fifth execution module

[0034] The first execution module is used to extract amplitude data and phase data from the respiratory data, and normalize the amplitude data and the phase data to obtain target amplitude data and target phase data;

[0035] The second execution module is used to concatenate the target amplitude data and the target phase data to obtain concatenated data, and use the concatenated data as the input of an encoder to obtain an embedding vector;

[0036] The third execution module is used to generate an embedding matrix according to the embedding vector, and input the embedding matrix into a multi-head attention module to obtain a first feature vector;

[0037] The fourth execution module is used to add noise to the respiratory data to obtain contaminated respiratory data, execute the first execution module, the second execution module, and the third execution module on the contaminated respiratory data to obtain a second feature vector, and calculate the similarity between the first feature vector and the second feature vector through a contrast loss function to obtain a similarity;

[0038] The fifth execution module is used to determine a respiratory event detection weight according to the similarity, update the weight of respiratory event detection according to the respiratory event detection weight, perform respiratory event detection on the respiratory data to obtain an activity status, and determine the respiratory signal data corresponding to the activity status; The activity status includes: exercise, sleep, and daily life.

[0039] Optionally, the data extraction module includes: an initialization module, a population target determination module, a model update module, and an update judgment module:

[0040] The initialization module is used to initialize the parameter combinations of the optimization algorithm and the VMD algorithm, and use the parameter combinations of the VMD algorithm as the individuals of the population; the parameters of the optimization algorithm are the population size and the maximum number of iterations, and the parameters of the VMD algorithm are the number of modes K and the penalty parameter α;

[0041] The population target determination module is used to calculate the fitness of each individual, determine the optimal individual according to the fitness, use the optimal individual as the update target for population update, and determine the number of iterations;

[0042] The model update module is used to update the modal spectrum, and perform central frequency update and Lagrange multiplier update if the number of iterations meets the preset detection conditions;

[0043] The update judgment module is used to stop the update and output the first signal data and the second signal data if the maximum number of iterations is reached or when the update values of all modes converge to a preset value; otherwise, the update is repeated.

[0044] Optionally, the update judgment module is further used to calculate the similarity between the first signal data and the second signal data and the respiration signal data respectively to obtain the similarity, and determine the heartbeat signal data according to the similarity;

[0045] Modal spectrum iterative separation formula:

[0046]

[0047] Among them, u t+1 (ω) represents the updated spectrum, f(ω) represents the spectrum of the original signal, λ(ω) represents the Lagrange multiplier, u other represents the spectra of other modes, α represents the penalty factor, ω k represents the central frequency of the k-th mode, t represents the number of iterations, and ω0 represents the standard central frequency.

[0048] Optionally, the activity warning module includes: a first warning module and a second warning module:

[0049] The first warning module is used to determine the heart rate of the target user according to the heartbeat signal data, and if the heart rate is within the preset interval, it is determined that the target user is in a normal state;

[0050] The second warning module is used to determine that the target user is in an abnormal state if the heart rate is outside the preset interval.

[0051] Advantages of the present invention:

[0052] The present invention proposes a chronic disease health monitoring method based on CAS. Heartbeat data and breathing data of a target user are obtained through the CAS system. Respiratory event detection is performed on the target user according to the breathing data to obtain the activity status and breathing signal data. Adaptive heartbeat signal extraction is performed on the heartbeat data according to the activity status and breathing signal data to obtain heartbeat signal data. Activity status warning is performed on the target user according to the heartbeat signal data. Non-contact breathing and heartbeat monitoring are realized through the CAS system, the heartbeat signal is extracted, respiratory events are identified, the respiratory interference in the heartbeat signal is suppressed, and accurate warning is realized through multi-dimensional physiological feature analysis. The parameters are dynamically adjusted to enhance the robustness of signal separation and reduce the false alarm rate. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] The present invention will be further described below with reference to the accompanying drawings.

[0054] Figure 1 It is a flowchart of the chronic disease health monitoring method based on CAS provided by an embodiment of the present invention;

[0055] Figure 2 It is a framework diagram of the chronic disease health monitoring system based on CAS provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0056] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. The term "and / or" in this article is only a description of the associated relationship of the associated objects, indicating that there can be three relationships. For example, A and B can represent: A exists alone, A and B exist simultaneously, and B exists alone. In addition, the descriptions such as "first" and "second" in the present invention are only for the purpose of description, and cannot be understood as indicating or implying their relative importance or implicitly indicating the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include at least one of the features. In addition, the technical solutions between the various embodiments can be combined with each other, but it must be based on the fact that those skilled in the art can implement them. When the combination of technical solutions appears to be contradictory or unable to be implemented, it should be considered that such a combination of technical solutions does not exist and is not within the protection scope required by the present invention.

[0057] All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0058] An embodiment of the present invention provides a chronic disease health monitoring method based on CAS. Refer to Figure 1 , Figure 1Flowchart of the chronic disease health monitoring method based on CAS provided by the embodiments of the present invention. The method includes the following steps:

[0059] S101, obtain the heartbeat data and respiration data of the target user through the CAS system, and perform respiration event detection on the target user according to the respiration data to obtain the activity state and respiration signal data;

[0060] S102, adaptively extract the heartbeat signal from the heartbeat data according to the activity state and respiration signal data to obtain the heartbeat signal data;

[0061] S103, perform activity state early warning on the target user according to the heartbeat signal data.

[0062] The CAS system is a data acquisition system;

[0063] Based on the chronic disease health monitoring method based on CAS provided by the embodiments of the present invention, non-contact respiration and heartbeat monitoring are realized through the CAS system, the heartbeat signal is extracted, respiration events are identified, the respiration interference in the heartbeat signal is suppressed, and through multi-dimensional physiological feature analysis, accurate early warning is realized, and the parameters are dynamically adjusted to enhance the robustness of signal separation and reduce the false alarm rate.

[0064] In one implementation, the CAS system collects the physiological data of patients (such as blood glucose, blood pressure, heart rate) in real time through technologies such as sensors, wearable devices, and mobile applications, and analyzes the abnormal trends through algorithms to provide early warning alerts; obtain the heartbeat data and respiration data of the target user through the CAS system, where the heartbeat data is obtained through a millimeter-wave radar and the respiration data: the millimeter-wave radar (operating frequency band 77 - 81 GHz) emits a high-frequency frequency-modulated continuous wave (FMCW) signal, receives the tiny displacement signal reflected by the human chest cavity, and uses moving average filtering and static interference elimination in the preprocessing stage to locate the heart activity area, and then separates the heartbeat signal through an adaptive variational mode decomposition algorithm (suppressing the interference of respiration and body movement), and finally estimates the heart rate based on the spectral peak search.

[0065] In one implementation, the ultra-wideband (UWB) technology utilizes the non-contact reflection characteristics of wideband short pulse signals (center frequency 7.3 GHz) to detect the thoracic respiration displacement by analyzing the signal phase and amplitude changes. The system removes noise through background elimination and Savitzky-Golay filtering, extracts the respiration period by combining the autocorrelation function, and uses a deep learning model (fusing CNN and multi-head self-attention mechanism) to identify respiration events. The advantage of UWB is its high time resolution, which can penetrate obstacles to capture subtle movements and is suitable for respiration signal monitoring in complex environments.

[0066] In one implementation, respiratory event classification (such as sleep apnea) and abnormal heartbeat detection (such as sudden increase in heart rate) achieve accurate activity status warning through multi-dimensional physiological characteristics (respiration-heartbeat coupling relationship); for example: when apnea is accompanied by heart rate fluctuations, a high-risk warning is triggered; while pure exercise interference (without respiratory abnormalities) reduces the false alarm rate.

[0067] In one implementation, respiratory harmonics and body movement interference are dynamically identified through respiratory event detection, and signal aliasing caused by respiration is specifically removed through the time-frequency feature matching of respiratory rate and heartbeat signals.

[0068] In one implementation, the parameters of the VMD algorithm (such as the number of modes K and the penalty factor α) are dynamically adjusted through respiratory signal data (such as apnea and deep breathing states) to enhance the robustness of heartbeat signal separation. For example: when the respiratory rate changes rapidly, the spectral peak search range of the heartbeat signal is corrected through respiratory phase information to avoid misjudging respiratory harmonics as the heart rate fundamental frequency.

[0069] In one embodiment, respiratory event detection is performed on a target user based on respiratory data to obtain an activity status and respiratory signal data, including:

[0070] Step 1, data extraction is performed on the respiratory data to obtain amplitude data and phase data, and the amplitude data and phase data are normalized to obtain target amplitude data and target phase data;

[0071] Step 2, the target amplitude data and target phase data are concatenated to obtain concatenated data, and the concatenated data is used as the input of the encoder to obtain an embedding vector;

[0072] Step 3, an embedding matrix is generated based on the embedding vector, and the embedding matrix is input into the multi-head attention module to obtain a first feature vector;

[0073] Step 4, noise is added to the respiratory data to obtain contaminated respiratory data, steps 1-3 are performed on the contaminated respiratory data to obtain a second feature vector, and the similarity between the first feature vector and the second feature vector is calculated through a contrast loss function to obtain a similarity;

[0074] Step 5, the respiratory event detection weight is determined based on the similarity, the weight of respiratory event detection is updated according to the respiratory event detection weight, respiratory event detection is performed on the respiratory data to obtain an activity status, and the respiratory signal data corresponding to the activity status is determined; the activity status includes: exercise, sleep, and daily.

[0075] In one implementation, by decomposing the respiration data into amplitude data and phase data and normalizing them, the core features of the respiration signal can be effectively extracted, while the dimensional difference and numerical range difference between different data sources are eliminated. The normalized target amplitude data and target phase data have a unified numerical range, which not only improves the consistency and comparability of the data, but also prepares high-quality data for the subsequent model input, enabling the model to more accurately learn and identify the patterns in the respiration signal, and contributing to the subsequent respiration event detection and classification tasks.

[0076] In one implementation, the normalized target amplitude data and target phase data are concatenated to form concatenated data. This operation integrates two key features of the respiration signal, enabling the model to consider both amplitude and phase information simultaneously. The embedding vector obtained by inputting the concatenated data into the encoder not only retains the details of the original data, but also converts the complex respiration signal into a lower-dimensional embedding space representation that is easier to process and analyze through the feature extraction ability of the encoder. The embedding vector can more effectively capture the internal structure and features of the respiration signal, providing more powerful support for the subsequent processing of deep learning models, and contributing to improving the detection accuracy and generalization ability of the model for respiration events.

[0077] In one implementation, an embedding matrix is generated based on the embedding vector and input into the multi-head attention module, which can make full use of the advantages of the multi-head attention mechanism to learn the complex patterns and long-term and short-term dependencies in the respiration signal sequence from multiple subspaces simultaneously. The multi-head attention module can capture features at different frequencies and time scales through parallel processing of multiple attention heads, thereby more comprehensively understanding the dynamic changes of the respiration signal. The obtained first feature vector not only contains rich semantic information, but also can highlight the key features of the respiration event, providing a more accurate feature representation for the subsequent respiration event detection and classification, and further enhancing the model's recognition ability and discrimination ability for different respiration states.

[0078] In one implementation, by actively introducing interference data into the respiration data (adding noise to the respiration data to obtain contaminated respiration data), the effect verification of the respiration detection model can be accelerated, that is, quickly determine the attention weights of the model (whether the model's attention is biased towards the noise or the model is not biased towards the noise). If the model is biased towards the noise, the model needs to be further trained. If the model is not biased towards the noise, it means that the model has been trained and the training can be terminated in advance (reducing the model training amount); a high similarity indicates that the model fails to distinguish different category signals (for example: motion, sleep, and daily), and there may be missed detections or false detections due to feature space confusion; a low similarity indicates that the model effectively separates different categories, has strong feature discriminability, and can achieve accurate classification.

[0079] In one implementation, after the first feature vector and the second feature vector pass through the multi-head attention module, global average pooling (AvgPooling) is performed to obtain a feature vector (used to represent the global features of the respiration signal, such as the respiration cycle and amplitude change); the second feature vector is the feature vector corresponding to the contaminated version of the first feature vector (such as adding noise and time shift) (used to force the model to learn the essential features of the respiration signal (such as the invariant waveform pattern) and ignore the noise interference); the similarity between the first feature vector and the second feature vector is calculated through a contrastive loss function to obtain a similarity; sim(,) cosine similarity is used to quantify the similarity degree of two feature vectors, ranging from [-1, 1], and the larger the value, the more similar; exp() represents the exponential function used to amplify the similarity difference; N represents the number of samples included in each batch (that is, the size of the training batch); τ represents the temperature hyperparameter used to control the sharpness of the similarity distribution, τ→0: amplify the similarity difference, and the model pays more attention to difficult samples (samples with similar similarities), τ→∞: smooth the similarity distribution and alleviate overfitting; Z li Used to widen the feature distance between different samples through contrastive learning and enhance the model's ability to distinguish heterogeneous signals;

[0080] Contrastive loss function:

[0081] Where L ctlf () represents the contrastive loss function, Z l1 represents the first feature vector, Z l2 represents the second feature vector, exp() represents the exponential function, τ represents the temperature hyperparameter, sim(,) represents the cosine similarity, N represents the training batch, Z li represents the feature vector of other samples in the same batch.

[0082] In one embodiment, adaptive heartbeat signal extraction is performed on the heartbeat data according to the activity state and respiration signal data to obtain heartbeat signal data, including:

[0083] Initialize the parameter combinations of the optimization algorithm and the VMD algorithm, and use the parameter combinations of the VMD algorithm as the individuals of the population; the parameters of the optimization algorithm are the population size and the maximum number of iterations, and the parameters of the VMD algorithm are the number of modes K and the penalty parameter α;

[0084] Calculate the fitness of each individual, determine the optimal individual according to the fitness, use the optimal individual as the update target for population update and determine the number of iterations;

[0085] If the number of iterations meets the preset detection conditions, update the modal spectrum, and perform central frequency update and Lagrange multiplier update;

[0086] If the maximum number of iterations is reached or when the updated values of all modes converge to a preset value, stop the update and output the first signal data and the second signal data; otherwise, repeat the update.

[0087] In one implementation, optimization algorithms such as: Whale Optimization Algorithm, Genetic Algorithm, Dung Beetle Algorithm, etc.; taking the parameter combinations of the VMD algorithm as individuals of the population can effectively utilize the global search ability of the WOA to quickly find the optimal VMD parameter combination. This parameter optimization method significantly improves the efficiency of parameter adjustment, reduces the time and workload of manual parameter adjustment; calculates the fitness of each individual, determines the optimal individual according to the fitness, and uses the optimal individual as the update target to update the population. This process ensures that the VMD algorithm can dynamically adapt to signal changes and extract more accurate modal components. Through iterative update, the VMD algorithm can better suppress noise and interference and improve the accuracy of signal decomposition.

[0088] In one implementation, if the number of iterations meets the preset detection conditions (for example: the maximum number of iterations is 100 times, and the preset detection condition is to update the central frequency and the Lagrange multiplier every ten iterations), then update the modal spectrum and perform the central frequency update and the Lagrange multiplier update. This adaptive update mechanism enables the VMD algorithm to dynamically adjust the central frequency and the Lagrange multiplier of the mode, further improving the accuracy and robustness of modal decomposition; the central frequency represents the main frequency component of each mode, and by updating the central frequency, the frequency characteristics of each mode can be more accurately described; the Lagrange multiplier is used to handle the constraint conditions to ensure that the modal decomposition satisfies certain physical or mathematical constraints;

[0089]

[0090] Among them, λ t+1 (ω) represents the updated value of the Lagrange multiplier at frequency ω in the (b + 1)-th iteration, λ t (ω) represents the value of the Lagrange multiplier at frequency ω in the b-th iteration, ∈ represents the noise tolerance parameter as the natural constant, f(ω) represents the spectrum of the original signal at frequency ω, ∑ k u t+1 (ω) represents the sum of the spectra of all modes at frequency ω in the (b + 1)-th iteration.

[0091] In one implementation, if the maximum number of iterations is reached or when the updated values of all modes converge to a preset value, stop the update; otherwise, repeat the update. This termination condition ensures that the algorithm stops when it reaches the optimal solution or converges to a stable state, avoids unnecessary calculations, and improves the convergence and stability of the algorithm.

[0092] In one embodiment, after outputting the first signal data and the second signal data, it further includes:

[0093] Calculating the similarity between the first signal data and the second signal data and the respiratory signal data respectively to obtain the similarity, and determining the heartbeat signal data according to the similarity;

[0094] Modal spectrum iterative separation formula:

[0095]

[0096] Wherein, u t+1 (ω) represents the updated spectrum, f(ω) represents the spectrum of the original signal, λ(ω) represents the Lagrange multiplier, u other represents the spectra of other modes, α represents the penalty factor, ω k represents the central frequency of the k-th mode, t represents the number of iterations, and ω0 represents the standard central frequency.

[0097] In one implementation, u t+1 (ω) is the updated value of the modal spectrum at frequency ω in the (t + 1)-th iteration, representing the signal component of each mode at a specific frequency in the current iteration; f(ω) is the spectrum of the original signal at frequency ω. This is the input of the algorithm, representing the energy distribution of the signal at different frequencies; ∑ j≠k u other is the sum of the spectra of other modes except the k-th mode at frequency ω. This step ensures that the update of each mode is not affected by other modes.

[0098] In one implementation, λ(ω) is the Lagrange multiplier, which is used to constrain the optimization problem to ensure that the modal decomposition satisfies certain physical or mathematical constraints. The role of the Lagrange multiplier is to balance the accuracy and smoothness of the signal decomposition; α: the penalty factor, which is used to control the smoothness of the modal decomposition. A larger α will make the mode smoother and reduce the influence of noise; ω k is the central frequency of the k-th mode, representing the main frequency component of this mode. The update of the central frequency helps to more accurately describe the frequency characteristics of each mode; t is the number of iterations, representing the current iteration step. The increase in the number of iterations helps the algorithm to gradually converge to the optimal solution. By iteratively updating the spectra of each mode, the algorithm can gradually separate the heartbeat signal and the respiratory signal, reduce the interference between signals, and improve the accuracy and robustness of signal decomposition.

[0099] In one embodiment, performing activity state warning on the target user according to the heartbeat signal data includes:

[0100] Determining the heart rate of the target user according to the heartbeat signal data. If the heart rate is within the preset interval, it is determined that the target user is in a normal state;

[0101] If the heart rate is outside the preset range, it is determined that the target user is in an abnormal state.

[0102] In one implementation, a user in a normal state can continue with the current activity without alarm intervention; for a user in an abnormal state, alarms are issued, including voice alarms and cloud alarms; by analyzing the heart rate signal data in real time, the heart rate of the target user can be quickly and accurately determined. This real-time monitoring function enables the system to react immediately when the user's heart rate is abnormal, improving the response speed to sudden health problems. For example: in a sports monitoring scenario, the user can adjust the exercise intensity in a timely manner when the heart rate is too high, avoiding health risks caused by overexertion.

[0103] In one implementation, according to the user's state, the system provides different types of alarms, such as voice alarms and cloud alarms. This diverse alarm mechanism can meet the needs of different users, ensuring that the alarm information can be conveyed to the user or relevant medical personnel in a timely manner. For example, for the elderly living alone at home, a voice alarm can directly remind them to pay attention to their physical condition; while a cloud alarm can notify medical staff or family members so that timely measures can be taken.

[0104] In one implementation, the collected heart rate signal is preprocessed to remove noise and interference, and then through frequency domain analysis methods such as Fourier transform, the signal is transformed from the time domain to the frequency domain, thereby obtaining the spectrum of the signal. In the spectrum, the heart rate signal usually appears as an obvious main frequency peak, and its frequency value corresponds to the heart rate. By detecting the position of this main frequency peak, the heart rate can be determined, and then it is converted to the heart rate, that is, the number of heartbeats per minute.

[0105] Based on the same inventive concept, the embodiments of the present invention also provide a chronic disease health monitoring system based on CAS. See Figure 2 , Figure 2 FIG. is a schematic structural diagram of the chronic disease health monitoring system based on CAS provided by the embodiments of the present invention, including: a data acquisition module, a data extraction module, and an activity warning module:

[0106] The data acquisition module is used to obtain the heart rate data and respiratory data of the target user through the CAS system, and perform respiratory event detection on the target user according to the respiratory data to obtain the activity state and respiratory signal data; the CAS system is a data acquisition system;

[0107] The data extraction module is used to adaptively extract the heart rate signal from the heart rate data according to the activity state and respiratory signal data to obtain the heart rate signal data;

[0108] The activity warning module is used to perform activity state warning on the target user according to the heart rate signal data.

[0109] Based on the chronic disease health monitoring system based on CAS provided by the embodiments of the present invention, non-contact respiration and heartbeat monitoring are realized through the CAS system, heartbeat signals are extracted, respiration events are identified, respiration interference in the heartbeat signals is suppressed, and through multi-dimensional physiological feature analysis, accurate early warning is realized, and parameters are dynamically adjusted to enhance the robustness of signal separation and reduce the false alarm rate.

[0110] In one embodiment, the data acquisition module includes: a first execution module, a second execution module, a third execution module, a fourth execution module, and a fifth execution module

[0111] The first execution module is used to extract amplitude data and phase data according to respiration data, and normalize the amplitude data and phase data to obtain target amplitude data and target phase data;

[0112] The second execution module is used to concatenate the target amplitude data and the target phase data to obtain concatenated data, and use the concatenated data as the input of the encoder to obtain an embedding vector;

[0113] The third execution module is used to generate an embedding matrix according to the embedding vector, and input the embedding matrix into the multi-head attention module to obtain a first feature vector;

[0114] The fourth execution module is used to add noise to the respiration data to obtain contaminated respiration data, execute the first execution module, the second execution module, and the third execution module for the contaminated respiration data to obtain a second feature vector, and calculate the similarity between the first feature vector and the second feature vector through a contrast loss function to obtain a similarity;

[0115] The fifth execution module is used to determine the respiration event detection weight according to the similarity, update the weight of the respiration event detection according to the respiration event detection weight, perform respiration event detection on the respiration data to obtain an activity state, and determine the respiration signal data corresponding to the activity state; The activity states include: exercise, sleep, and daily life.

[0116] In one embodiment, the data extraction module includes: an initialization module, a population target determination module, a model update module, and an update judgment module:

[0117] The initialization module is used to initialize the parameter combinations of the optimization algorithm and the VMD algorithm, and use the parameter combinations of the VMD algorithm as the individuals of the population; the parameters of the optimization algorithm are the population size and the maximum number of iterations, and the parameters of the VMD algorithm are the number of modes K and the penalty parameter α;

[0118] The population target determination module is used to calculate the fitness of each individual, determine the optimal individual according to the fitness, use the optimal individual as the update target to update the population, and determine the number of iterations;

[0119] A model update module, configured to update the modal spectrum if the number of iterations meets a preset detection condition, and perform central frequency update and Lagrange multiplier update;

[0120] An update judgment module, configured to stop the update and output first signal data and second signal data if the maximum number of iterations is reached or when the update values of all modes converge to a preset value; otherwise, repeat the update.

[0121] In one embodiment, the update judgment module is further configured to calculate the similarity between the first signal data and the second signal data and the respiration signal data respectively to obtain a similarity, and determine the heartbeat signal data according to the similarity;

[0122] Modal spectrum iteration separation formula:

[0123]

[0124] where u t+1 (ω) represents the updated spectrum, f(ω) represents the spectrum of the original signal, λ(ω) represents the Lagrange multiplier, u other represents the spectra of other modes, α represents the penalty factor, ω k represents the central frequency of the k-th mode, t represents the number of iterations, and ω0 represents the standard central frequency.

[0125] In one embodiment, the activity warning module includes: a first warning module and a second warning module:

[0126] The first warning module is configured to determine the heart rate of the target user according to the heartbeat signal data, and determine that the target user is in a normal state if the heart rate is within a preset interval;

[0127] The second warning module is configured to determine that the target user is in an abnormal state if the heart rate is outside the preset interval.

[0128] The above has described an embodiment of the present invention in detail, but the content is only a preferred embodiment of the present invention, and cannot be considered as limiting the scope of implementation of the present invention. All equivalent changes and improvements made within the scope of the application of the present invention shall still fall within the scope covered by the patent of the present invention.

Claims

1. A chronic disease health monitoring method based on CAS, characterized in that, The method includes: Obtaining the heartbeat data and respiration data of a target user through a CAS system, and performing respiration event detection on the target user according to the respiration data to obtain an activity state and respiration signal data; the CAS system is a data acquisition system; Performing adaptive heartbeat signal extraction on the heartbeat data according to the activity state and the respiration signal data to obtain heartbeat signal data; Performing activity state early warning on the target user according to the heartbeat signal data.

2. The CAS-based chronic disease health monitoring method according to claim 1, wherein Performing respiration event detection on the target user according to the respiration data to obtain an activity state and respiration signal data, including: Step 1, performing data extraction on the respiration data to obtain amplitude data and phase data, and normalizing the amplitude data and the phase data to obtain target amplitude data and target phase data; Step 2, concatenating the target amplitude data and the target phase data to obtain concatenated data, and using the concatenated data as the input of an encoder to obtain an embedding vector; Step 3, generating an embedding matrix according to the embedding vector, and inputting the embedding matrix into a multi-head attention module to obtain a first feature vector; Step 4, adding noise to the respiration data to obtain contaminated respiration data, performing Steps 1-3 on the contaminated respiration data to obtain a second feature vector, and calculating the similarity between the first feature vector and the second feature vector through a contrast loss function to obtain a similarity; Step 5, determining a respiration event detection weight according to the similarity, updating the weight of respiration event detection according to the respiration event detection weight, performing respiration event detection on the respiration data to obtain an activity state, and determining the respiration signal data corresponding to the activity state; the activity state includes: exercise, sleep, and daily life.

3. The method for chronic disease health monitoring based on CAS according to claim 1, wherein Performing adaptive heartbeat signal extraction on the heartbeat data according to the activity state and the respiration signal data to obtain heartbeat signal data, including: Initializing the parameter combination of an optimization algorithm and a VMD algorithm, and using the parameter combination of the VMD algorithm as an individual of a population; the parameters of the optimization algorithm are the population size and the maximum number of iterations, and the parameters of the VMD algorithm are the number of modes K and the penalty parameter α; Calculating the fitness of each individual, determining the optimal individual according to the fitness, updating the population with the optimal individual as the update target and determining the number of iterations; If the number of iterations meets a preset detection condition, then update the modal spectrum, and perform central frequency update and Lagrange multiplier update; If the maximum number of iterations is reached or when the update values of all modes converge to a preset value, stop the update and output first signal data and second signal data; otherwise, repeat the update.

4. The method for chronic disease health monitoring based on CAS according to claim 3, wherein After outputting the first signal data and the second signal data, it further includes: Calculating the similarity between the first signal data and the second signal data and the respiration signal data respectively to obtain a similarity, and determining the heartbeat signal data according to the similarity; Modal spectrum iterative separation formula: Among them, u t+1 (ω) represents the updated spectrum, f(ω) represents the spectrum of the original signal, λ(ω) represents the Lagrange multiplier, u other represents the spectra of other modes, α represents the penalty factor, ω k represents the central frequency of the k-th mode, t represents the number of iterations, and ω0 represents the standard central frequency.

5. The method for chronic disease health monitoring based on CAS according to claim 1, wherein Performing activity state early warning on the target user according to the heartbeat signal data, including: Determine the heart rate of the target user based on the heartbeat signal data. If the heart rate is within the preset range, determine that the target user is in a normal state; If the heart rate is outside the preset range, determine that the target user is in an abnormal state.

6. The chronic disease health monitoring system based on CAS is characterized in that The system includes: a data acquisition module, a data extraction module, and an activity warning module: The data acquisition module is used to obtain the heartbeat data and respiratory data of the target user through the CAS system, and perform respiratory event detection on the target user according to the respiratory data to obtain the activity state and respiratory signal data; the CAS system is a data acquisition system; The data extraction module is used to adaptively extract the heartbeat signal from the heartbeat data according to the activity state and the respiratory signal data to obtain the heartbeat signal data; The activity warning module is used to perform activity state warning on the target user according to the heartbeat signal data.

7. The CAS-based chronic disease health monitoring system according to claim 6, characterized in that, The data acquisition module includes: a first execution module, a second execution module, a third execution module, a fourth execution module, and a fifth execution module The first execution module is used to extract data from the respiratory data to obtain amplitude data and phase data, and normalize the amplitude data and the phase data to obtain target amplitude data and target phase data; The second execution module is used to concatenate the target amplitude data and the target phase data to obtain concatenated data, and use the concatenated data as the input of the encoder to obtain an embedding vector; The third execution module is used to generate an embedding matrix according to the embedding vector, and input the embedding matrix into the multi-head attention module to obtain a first feature vector; The fourth execution module is used to add noise to the respiratory data to obtain contaminated respiratory data, execute the first execution module, the second execution module, and the third execution module on the contaminated respiratory data to obtain a second feature vector, and calculate the similarity between the first feature vector and the second feature vector through a contrast loss function to obtain the similarity; The fifth execution module is used to determine the respiratory event detection weight according to the similarity, update the weight of the respiratory event detection according to the respiratory event detection weight, perform respiratory event detection on the respiratory data to obtain the activity state, and determine the respiratory signal data corresponding to the activity state; the activity state includes: exercise, sleep, and daily life.

8. The CAS-based chronic disease health monitoring system according to claim 6, wherein, The data extraction module includes: an initialization module, a population target determination module, a model update module, and an update judgment module: The initialization module is used to initialize the parameter combination of the optimization algorithm and the VMD algorithm, and use the parameter combination of the VMD algorithm as an individual of the population; the parameters of the optimization algorithm are the population size and the maximum number of iterations, and the parameters of the VMD algorithm are the number of modes K and the penalty parameter α; The population target determination module is used to calculate the fitness of each individual, determine the optimal individual according to the fitness, update the population with the optimal individual as the update target, and determine the number of iterations; The model update module is used to update the modal spectrum, and perform central frequency update and Lagrange multiplier update if the number of iterations meets the preset detection conditions; The update judgment module is used to stop the update and output the first signal data and the second signal data if the maximum iteration number is reached or when the update values of all modalities converge to a preset value; otherwise, the update is repeatedly executed.

9. The CAS-based chronic disease health monitoring system according to claim 8, characterized in that, The update judgment module is further used to calculate the similarity between the first signal data and the second signal data and the respiratory signal data respectively to obtain the similarity, and determine the heartbeat signal data according to the similarity; Modal spectrum iterative separation formula: where, u t+1 (ω) represents the updated spectrum, f(ω) represents the spectrum of the original signal, λ(ω) represents the Lagrange multiplier, u other represents the spectra of other modes, α represents the penalty factor, ω k represents the central frequency of the k-th mode, t represents the number of iterations, and ω0 represents the standard central frequency.

10. The CAS-based chronic disease health monitoring system according to claim 6, characterized in that, The activity warning module includes: a first warning module and a second warning module: The first warning module is used to determine the heart rate of the target user according to the heartbeat signal data, and if the heart rate is within a preset interval, it is determined that the target user is in a normal state; The second warning module is used to determine that the target user is in an abnormal state if the heart rate is outside the preset interval.

Images