Sleep monitoring and early warning method and system based on respiratory data analysis and medium

Abstract

The present application relates to sleep monitoring and early warning method, system and medium based on respiratory data analysis, belong to technical field, the present application is through sleep state and sleep apnea hypopnea index prediction model to the respiratory data of user is predicted, the sleep state and sleep apnea hypopnea index of user are estimated, finally, according to the sleep state and sleep apnea hypopnea index of user, early warning is carried out, simultaneously, according to early warning information, relevant treatment scheme is generated.The application is through pre-training a model named deep learning model, the model is in unified framework simultaneously carries out sleep staging and sleep apnea hypopnea index estimation, effectively decouples the internal correlation of sleep macrostructure and respiratory event.Then, through the domain adversarial training mechanism, the sleep respiratory feature knowledge learned from the contact respiratory signal is migrated to the millimeter wave radar signal, and the problem of insufficient model generalization ability caused by the scarcity of radar data labeling is solved.

Description

Technical Field

[0001] This invention relates to the field of sleep monitoring technology, and in particular to sleep monitoring and early warning methods, systems and media based on respiratory data analysis. Background Technology

[0002] Sleep monitoring is a key tool for assessing and diagnosing various sleep disorders. It primarily determines sleep quality by recording physiological signals during sleep (such as electroencephalography, respiration, and heart rate). Sleep monitoring is an objective medical examination method for evaluating an individual's nighttime sleep status and is widely used to identify problems such as sleep apnea, nocturnal epileptic seizures, insomnia, and behavioral disorders. The sleep apnea-hypopnea index changes with the user's physical and mental state. The sleep apnea-hypopnea index refers to the number of apneas and hypopneas per hour of sleep. Apnea is defined as the complete cessation of airflow through the mouth and nose for more than 10 seconds during sleep; hypopnea is defined as a reduction in respiratory airflow intensity (amplitude) of more than 50% compared to baseline levels during sleep, accompanied by a decrease in blood oxygen saturation of ≥4% or microarousals. Traditional contact-based sleep monitoring devices suffer from poor comfort and lack consideration for user-specific factors, resulting in low monitoring accuracy. Summary of the Invention

[0003] This invention overcomes the shortcomings of the prior art and provides a method, system and medium for sleep monitoring and early warning based on respiratory data analysis.

[0004] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0005] The first aspect of this invention provides a sleep monitoring and early warning method based on respiratory data analysis, comprising the following steps:

[0006] Millimeter-wave radar is used to capture body surface movements caused by breathing by emitting electromagnetic waves and receiving echoes reflected by the slight movements of the human chest and abdomen, thus forming real-time respiratory data.

[0007] By preprocessing real-time respiratory data, respiratory harmonic components are separated, and thoracic breathing and abdominal breathing are distinguished. A prediction model for sleep state and sleep apnea-hypopnea index is constructed based on a deep learning network.

[0008] Real-time collection of user breathing data; prediction of user breathing data using sleep state and sleep apnea-hypopnea index prediction model; estimation of user sleep state and sleep apnea-hypopnea index.

[0009] The system issues warnings based on the user's sleep status and sleep apnea-hypopnea index, and generates relevant treatment plans based on the warning information.

[0010] Furthermore, in sleep monitoring and early warning methods based on respiratory data analysis, millimeter-wave radar is used to capture body surface movements caused by breathing by emitting electromagnetic waves and receiving echoes reflected from the subtle movements of the human chest and abdomen, thus forming real-time respiratory data. Specifically:

[0011] By installing millimeter-wave radar in a predetermined area, the millimeter-wave radar operates in the 60-64 GHz frequency band, uses frequency-modulated continuous wave to transmit electromagnetic waves and receives the echoes reflected by the slight movements of the human chest and abdomen to capture the surface movements caused by breathing.

[0012] Configure the number of transmitting and receiving antennas in the millimeter-wave radar, initialize the number of transmitting and receiving antennas, and use a multi-input multi-output array based on the number of transmitting and receiving antennas to collect and test respiratory data that captures body surface movements caused by breathing. Through the test, determine the spatial resolution of the current data.

[0013] Set a spatial resolution threshold, determine whether the spatial resolution of the current data is greater than the spatial resolution threshold, and if the spatial resolution of the current data is greater than the spatial resolution threshold, arrange the current number of transmitting and receiving antennas and collect real-time respiratory data.

[0014] If the spatial resolution of the current data is not greater than the spatial resolution threshold, adjust the number of current transmitting and receiving antennas until the spatial resolution of the current data is greater than the spatial resolution threshold.

[0015] Furthermore, in sleep monitoring and early warning methods based on respiratory data analysis, real-time respiratory data is preprocessed to separate respiratory harmonic components and distinguish between thoracic and abdominal breathing, specifically:

[0016] By filtering and denoising real-time respiratory data to remove noise, preprocessed respiratory data is obtained. Then, a phase extraction algorithm is used to process the preprocessed respiratory data to obtain phase change data caused by respiration.

[0017] Multipath cancellation technology is used to process the phase change data caused by respiration to reduce environmental reflection interference, and variational mode decomposition algorithm is introduced to decompose the phase change data caused by respiration.

[0018] By decomposing, respiratory harmonic components are separated, and thoracic breathing and abdominal breathing are distinguished based on these components.

[0019] Furthermore, in the sleep monitoring and early warning method based on respiratory data analysis, a predictive model for sleep state and sleep apnea-hypopnea index is constructed based on a deep learning network, specifically as follows:

[0020] A prediction model for sleep state and sleep apnea-hypopnea index is constructed based on a deep learning network. Sleep state data, thoracic breathing data and abdominal breathing data are collected to form a large-scale contact breathing signal database. Sleep stage and sleep apnea-hypopnea index are estimated simultaneously within a unified framework.

[0021] By estimating, the intrinsic relationship between the macroscopic structure of sleep and respiratory events is decoupled, and the knowledge of sleep breathing characteristics learned from contact breathing signals is transferred to millimeter-wave radar signals through a domain adversarial training mechanism.

[0022] The thoracic and abdominal breathing data are used as model inputs, and the sleep stages and sleep apnea-hypopnea index are used as model outputs. The learning rate and number of training sessions are set.

[0023] The sleep state and sleep apnea-hypopnea index prediction model is trained based on the learning rate and the number of training sessions. When the prediction accuracy of the sleep state and sleep apnea-hypopnea index prediction model reaches the preset prediction accuracy data, the sleep state and sleep apnea-hypopnea index prediction model is output.

[0024] Furthermore, in the sleep monitoring and early warning method based on respiratory data analysis, the user's respiratory data is collected in real time, and the user's respiratory data is predicted using a sleep state and sleep apnea-hypopnea index prediction model to estimate the user's sleep state and sleep apnea-hypopnea index, specifically as follows:

[0025] Real-time collection of user breathing data, and input of the user breathing data into the sleep state and sleep apnea-hypopnea index prediction model for prediction;

[0026] The system estimates and obtains the user's sleep state and sleep apnea-hypopnea index, and outputs the user's sleep state and sleep apnea-hypopnea index.

[0027] Furthermore, in the sleep monitoring and early warning method based on respiratory data analysis, an early warning is issued based on the user's sleep state and sleep apnea-hypopnea index, specifically as follows:

[0028] Set the user's sleep state and the range of the sleep apnea-hypopnea index evaluation indicators, and determine whether the user's sleep state and sleep apnea-hypopnea index are within the range of the user's sleep state and sleep apnea-hypopnea index evaluation indicators.

[0029] If the user's sleep state and sleep apnea-hypopnea index are within the range of the user's sleep state and sleep apnea-hypopnea index evaluation indicators, then normal information is generated and displayed in a preset manner.

[0030] If the user's sleep state and sleep apnea-hypopnea index are not within the range of the user's sleep state and sleep apnea-hypopnea index evaluation indicators, an early warning message will be generated and displayed in a preset manner.

[0031] Furthermore, in sleep monitoring and early warning methods based on respiratory data analysis, relevant treatment plans are generated based on the early warning information, specifically as follows:

[0032] Based on the warning information, the changes in the user's sleep state characteristics and sleep apnea-hypopnea index data within a preset time period are obtained, and a real-time sleep structure map is constructed based on the changes in the user's sleep state characteristics within the preset time period.

[0033] Users upload various sleep structure diagrams and historical treatment plans under sleep apnea-hypopnea index data to build a database, and input the various sleep structure diagrams and historical treatment plans under sleep apnea-hypopnea index data into the database for storage;

[0034] Real-time sleep structure graphs and sleep apnea-hypopnea index data are input into the database for data matching. Through data matching, the optimal treatment plan under the current real-time sleep structure graph and sleep apnea-hypopnea index data is obtained.

[0035] Based on the current real-time sleep structure map and sleep apnea-hypopnea index data, the optimal treatment plan is generated.

[0036] Furthermore, in sleep monitoring and early warning methods based on respiratory data analysis, noise is removed by filtering and denoising the real-time respiratory data. Specifically:

[0037] The real-time respiratory data is processed by Hampel filtering to remove outliers from the original CSI sequence, and Gaussian filtering and linear interpolation are used to smooth the data and handle possible data loss.

[0038] The respiratory frequency band signal was extracted using a Butterworth low-pass filter, and signal anomalies caused by large-scale body movement were identified and suppressed by Z-Score anomaly detection.

[0039] The filtered signal is peaked by local peak detection, the breathing rate is calculated, and noise is removed.

[0040] A second aspect of the present invention provides a sleep monitoring and early warning system based on respiratory data analysis, including a memory and a processor. The memory includes a sleep monitoring and early warning method program based on respiratory data analysis. When the processor executes the sleep monitoring and early warning method program based on respiratory data analysis, it implements the steps of any of the sleep monitoring and early warning methods based on respiratory data analysis described in the present invention.

[0041] A third aspect of the present invention provides a computer-readable storage medium including a sleep monitoring and early warning method program based on respiratory data analysis, wherein when the sleep monitoring and early warning method program based on respiratory data analysis is executed by a processor, it implements the steps of the sleep monitoring and early warning method based on respiratory data analysis as described in any one of the present invention.

[0042] This invention addresses the shortcomings of the prior art and has the following beneficial effects:

[0043] This invention utilizes millimeter-wave radar to capture body surface movements caused by respiration by emitting electromagnetic waves and receiving echoes reflected from subtle movements of the human chest and abdomen, forming real-time respiratory data. This data is then preprocessed to separate respiratory harmonic components and distinguish between thoracic and abdominal breathing. A sleep state and sleep apnea-hypopnea index (PAH-index) prediction model is constructed based on a deep learning network. This allows for real-time acquisition of the user's respiratory data, prediction of the user's sleep state and PAH-index using the model, and finally, early warning based on the user's sleep state and PAH-index. Simultaneously, relevant treatment plans are generated based on the warning information. This invention pre-trains a deep learning model that simultaneously performs sleep staging and PAH estimation within a unified framework, effectively decoupling the intrinsic relationship between the macroscopic structure of sleep and respiratory events. Then, through a domain adversarial training mechanism, sleep respiratory feature knowledge learned from contact respiratory signals is transferred to millimeter-wave radar signals, addressing the problem of insufficient model generalization ability caused by the scarcity of radar data annotations. Attached Figure Description

[0044] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other embodiments can be obtained from these drawings without creative effort.

[0045] Figure 1 A flowchart illustrating the overall process of sleep monitoring and early warning methods based on respiratory data analysis is provided.

[0046] Figure 2 A system block diagram of a sleep monitoring and early warning system based on respiratory data analysis is shown. Detailed Implementation

[0047] To better understand the above-mentioned objectives, features, and advantages of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.

[0048] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and therefore the scope of protection of the invention is not limited to the specific embodiments disclosed below.

[0049] like Figure 1 As shown, the first aspect of the present invention provides a sleep monitoring and early warning method based on respiratory data analysis, comprising the following steps:

[0050] Millimeter-wave radar is used to capture body surface movements caused by breathing by emitting electromagnetic waves and receiving echoes reflected by the slight movements of the human chest and abdomen, thus forming real-time respiratory data.

[0051] By preprocessing real-time respiratory data, respiratory harmonic components are separated, and thoracic breathing and abdominal breathing are distinguished. A prediction model for sleep state and sleep apnea-hypopnea index is constructed based on a deep learning network.

[0052] Real-time collection of user breathing data; prediction of user breathing data using sleep state and sleep apnea-hypopnea index prediction model; estimation of user sleep state and sleep apnea-hypopnea index.

[0053] The system provides early warnings based on the user's sleep status and sleep apnea-hypopnea index, and generates relevant treatment plans based on the warning information.

[0054] It should be noted that this invention pre-trains a deep learning model on a large-scale contact breathing signal database (such as a multi-center database containing 15,785 nights). This model simultaneously estimates sleep stages and the apnea-hypopnea index (AHI) within a unified framework, effectively decoupling the intrinsic relationship between the macroscopic structure of sleep and respiratory events. Then, through a domain adversarial training mechanism, the sleep breathing feature knowledge learned from contact breathing signals is transferred to millimeter-wave radar signals, solving the problem of insufficient model generalization ability caused by the scarcity of radar data annotations.

[0055] Furthermore, in sleep monitoring and early warning methods based on respiratory data analysis, millimeter-wave radar is used to capture body surface movements caused by breathing by emitting electromagnetic waves and receiving echoes reflected from the subtle movements of the human chest and abdomen, thus forming real-time respiratory data. Specifically:

[0056] By installing millimeter-wave radar in a predetermined area, the millimeter-wave radar operates in the 60-64 GHz frequency band, uses frequency-modulated continuous wave to transmit electromagnetic waves and receives the echoes reflected by the slight movements of the human chest and abdomen to capture the surface movements caused by breathing.

[0057] Configure the number of transmitting and receiving antennas in the millimeter-wave radar, initialize the number of transmitting and receiving antennas, and use a multi-input multi-output array based on the number of transmitting and receiving antennas to collect and test respiratory data that captures body surface movements caused by breathing. Through the test, determine the spatial resolution of the current data.

[0058] Set a spatial resolution threshold to determine if the spatial resolution of the current data is greater than the spatial resolution threshold. If the spatial resolution of the current data is greater than the spatial resolution threshold, arrange the antennas according to the current number of transmitting and receiving antennas and collect real-time respiratory data.

[0059] If the spatial resolution of the current data is not greater than the spatial resolution threshold, adjust the number of current transmit and receive antennas until the spatial resolution of the current data is greater than the spatial resolution threshold.

[0060] It should be noted that millimeter-wave radar operates in the 60-64 GHz frequency band, employing frequency-modulated continuous wave mode. It captures body surface movements caused by respiration by emitting electromagnetic waves and receiving echoes reflected from subtle movements of the human chest and abdomen. By integrating a predetermined number of transmitting and receiving antennas at the radar front end, optimizing the number of transmitting and receiving antennas, and improving the spatial resolution of the multiple-input multiple-output (MIMO) array, the rationality of millimeter-wave radar data acquisition can be enhanced.

[0061] Furthermore, in sleep monitoring and early warning methods based on respiratory data analysis, real-time respiratory data is preprocessed to separate respiratory harmonic components and distinguish between thoracic and abdominal breathing, specifically:

[0062] By filtering and denoising real-time respiratory data to remove noise, preprocessed respiratory data is obtained. Then, a phase extraction algorithm is used to process the preprocessed respiratory data to obtain phase change data caused by respiration.

[0063] Multipath cancellation technology is used to process the phase change data caused by respiration to reduce environmental reflection interference, and variational mode decomposition algorithm is introduced to decompose the phase change data caused by respiration.

[0064] By decomposing the components, respiratory harmonic components are separated, and thoracic breathing and abdominal breathing are distinguished based on these components.

[0065] It should be noted that by filtering and denoising the real-time breathing data, noise is removed, specifically as follows:

[0066] Outlier removal was performed on the raw CSI sequence by using Hampel filtering on real-time respiratory data, and Gaussian filtering and linear interpolation were used to smooth the data and handle possible data loss.

[0067] The respiratory frequency band signal was extracted using a Butterworth low-pass filter, and signal anomalies caused by large-scale body movement were identified and suppressed by Z-Score anomaly detection.

[0068] The filtered signal is peaked by local peak detection, the breathing rate is calculated, and noise is removed.

[0069] It should be noted that after preprocessing, the phase changes caused by breathing are obtained through a phase extraction algorithm from the raw radar signal. Subsequently, multipath cancellation technology is used to reduce environmental reflection interference, and then variational mode decomposition (VMD) is used to separate the breathing harmonic components, effectively distinguishing between thoracic breathing and abdominal breathing.

[0070] Furthermore, in the sleep monitoring and early warning method based on respiratory data analysis, a predictive model for sleep state and sleep apnea-hypopnea index is constructed based on a deep learning network, specifically as follows:

[0071] A prediction model for sleep state and sleep apnea-hypopnea index is constructed based on a deep learning network. Sleep state data, thoracic breathing data and abdominal breathing data are collected to form a large-scale contact breathing signal database. Sleep stage and sleep apnea-hypopnea index are estimated simultaneously within a unified framework.

[0072] By estimating, the intrinsic relationship between the macroscopic structure of sleep and respiratory events is decoupled, and the knowledge of sleep breathing characteristics learned from contact breathing signals is transferred to millimeter-wave radar signals through a domain adversarial training mechanism.

[0073] Use thoracic breathing and abdominal breathing data as model input, and sleep stages and sleep apnea-hypopnea index as model output, and set the learning rate and number of training sessions.

[0074] The sleep state and sleep apnea-hypopnea index prediction model is trained based on the learning rate and the number of training sessions. When the prediction accuracy of the sleep state and sleep apnea-hypopnea index prediction model reaches the preset prediction accuracy data, the sleep state and sleep apnea-hypopnea index prediction model is output.

[0075] It should be noted that a deep learning model is pre-trained on a large-scale contact respiratory signal database (such as a multi-center database containing 15,785 nights). This model simultaneously performs sleep staging and apnea-hypopnea index (AHI) estimation within a unified framework, effectively decoupling the intrinsic relationship between the macroscopic structure of sleep and respiratory events. Then, through a domain adversarial training mechanism, the sleep respiratory feature knowledge learned from contact respiratory signals is transferred to millimeter-wave radar signals, solving the problem of insufficient model generalization ability caused by the scarcity of radar data annotations. The domain adversarial training mechanism is the core technology for improving the model's cross-domain generalization ability by eliminating the distribution difference between the source and target domains through adversarial learning. Its key lies in the dynamic game between the feature extractor and the domain classifier. It mainly consists of three parts: a feature extractor, a threshold classifier, and a gradient inversion layer. In the feature extractor, the input is mapped to the feature space, with the goal of generating "domain-independent features" that the domain classifier cannot distinguish. In the threshold classifier, it determines whether the feature comes from the source or target domain, with the goal of maximizing the classification accuracy. In the gradient inversion layer, the gradient sign is reversed during backpropagation, creating an adversarial relationship between the feature extractor and the domain classifier—the former attempts to "confuse" the latter, while the latter tries to "decipher" the domain origin.

[0076] Furthermore, in the sleep monitoring and early warning method based on respiratory data analysis, the user's respiratory data is collected in real time. The user's respiratory data is then predicted using a sleep state and sleep apnea-hypopnea index prediction model to estimate the user's sleep state and sleep apnea-hypopnea index. Specifically:

[0077] Real-time collection of users' breathing data; inputting users' breathing data into the sleep state and sleep apnea-hypopnea index prediction model for prediction.

[0078] The system estimates and obtains the user's sleep state and sleep apnea-hypopnea index, and outputs the user's sleep state and sleep apnea-hypopnea index.

[0079] Furthermore, in sleep monitoring and early warning methods based on respiratory data analysis, early warnings are issued based on the user's sleep state and sleep apnea-hypopnea index, specifically as follows:

[0080] Set the user's sleep status and the range of evaluation indicators for sleep apnea-hypopnea index, and determine whether the user's sleep status and sleep apnea-hypopnea index are within the range of evaluation indicators.

[0081] If the user's sleep status and sleep apnea-hypopnea index are within the range of the user's sleep status and sleep apnea-hypopnea index evaluation indicators, then normal information is generated and displayed according to the preset method;

[0082] If the user's sleep state and sleep apnea-hypopnea index are outside the evaluation range of the user's sleep state and sleep apnea-hypopnea index, an early warning message will be generated and displayed according to the preset method.

[0083] It should be noted that sleep states include light sleep, deep sleep, and wakefulness. The Sleep Apnea-Hypopnea Index (AHI) refers to the number of apneas and hypopneas per hour of sleep. Apnea is defined as the complete cessation of airflow through the mouth and nose for more than 10 seconds during sleep; hypopnea is defined as a reduction in airflow intensity (amplitude) of more than 50% compared to baseline levels during sleep, accompanied by a decrease in blood oxygen saturation of ≥4% or micro-awakening. The system can be configured to generate an alert and display it according to a preset method when the patient is in deep sleep and the AHI is ≥5 times per hour. Otherwise, a normal alert will be generated.

[0084] Furthermore, in sleep monitoring and early warning methods based on respiratory data analysis, relevant treatment plans are generated based on the early warning information, specifically as follows:

[0085] Based on the early warning information, the changes in the user's sleep state characteristics and sleep apnea-hypopnea index data within a preset time period are obtained, and a real-time sleep structure map is constructed based on the changes in the user's sleep state characteristics within a preset time period.

[0086] Users upload various sleep structure diagrams and historical treatment plans under sleep apnea-hypopnea index data to build a database, and input the various sleep structure diagrams and historical treatment plans under sleep apnea-hypopnea index data into the database for storage;

[0087] Real-time sleep structure graphs and sleep apnea-hypopnea index data are input into the database for data matching. Through data matching, the optimal treatment plan is obtained under the current real-time sleep structure graph and sleep apnea-hypopnea index data.

[0088] Based on the current real-time sleep structure map and sleep apnea-hypopnea index data, the optimal treatment plan is generated.

[0089] It should be noted that this method can recommend relevant treatment plans, such as recommending continuous positive airway pressure (CPAP) therapy for patients with severe sleep apnea.

[0090] In addition, this method also includes:

[0091] An array of multiple miniature microphones is deployed in the bedroom to collect snoring and breathing sounds. Beamforming technology is used to amplify the sound signal from the direction of the bed, and a deep neural network is introduced. A random forest algorithm is used to classify audio segments, identify different types of snoring sounds, and associate these different types of snoring sounds with sleep apnea events to generate correlation data. This correlation data is then input into the neural network for training. The neural network is used to identify regular changes in breathing patterns during different sleep stages, and this regularity is used to construct the user's original sleep data. With the user's authorization, the user's original sleep data is processed using a Markov model to generate the user's original sleep data state sequence change characteristics, and the system determines whether there are any abnormal states in these characteristics. When abnormal states are found in the user's original sleep data state sequence change characteristics, an early warning is issued.

[0092] It should be noted that the random forest algorithm is used to classify audio segments, identify different types of snoring sounds (such as simple snoring, partial airway obstruction sounds, etc.), and associate them with sleep apnea events. Since autonomic nervous system regulation in different sleep stages causes regular changes in breathing patterns, these changes can also be reflected in sound characteristics. The system further optimizes the accuracy of sleep monitoring by analyzing the rhythm and intensity of breathing sounds.

[0093] In addition, this method also includes:

[0094] The sleep monitor incorporates millimeter-wave radar to monitor the patient's sleep state in real time, identifying wakefulness, difficulty falling asleep, or shallow sleep based on the monitored sleep state. The user uploads their personal pathological information, and feature extraction is performed based on this information to determine if central sleep apnea or sleep rhythm disorders are present. When central sleep apnea or sleep rhythm disorders are present, the intensity, frequency, and duration of magnetic stimulation from the sleep monitor are automatically optimized based on the identified wakefulness, difficulty falling asleep, or shallow sleep. During stimulation intervals, the sleep monitor guides the user through specific breathing pattern training, and their compliance and physiological responses are monitored in real time, forming a closed-loop intervention.

[0095] It should be noted that this method monitors the user's breathing signals in real time. When central sleep apnea (characterized by the loss of breathing effort) or sleep rhythm disorder is detected, the pulsed magnetic stimulation system is automatically activated. By applying a magnetic field to the relevant areas of the brain's respiratory center (such as the brainstem), the excitability of neurons is regulated, helping to restore a normal breathing rhythm.

[0096] A second aspect of the present invention provides a sleep monitoring and early warning system based on respiratory data analysis, including a memory and a processor. The memory includes a sleep monitoring and early warning method program based on respiratory data analysis. When the sleep monitoring and early warning method program based on respiratory data analysis is executed by the processor, it implements the steps of any one of the sleep monitoring and early warning methods based on respiratory data analysis.

[0097] A third aspect of the present invention provides a computer-readable storage medium including a sleep monitoring and early warning method program based on respiratory data analysis, wherein when the sleep monitoring and early warning method program based on respiratory data analysis is executed by a processor, it implements the steps of any one of the sleep monitoring and early warning methods based on respiratory data analysis.

[0098] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods, such as: multiple units or components can be combined, or integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the various components shown or discussed can be through some interfaces, and the indirect coupling or communication connection between devices or units can be electrical, mechanical, or other forms.

[0099] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units. They may be located in one place or distributed across multiple network units. Some or all of the units may be selected to achieve the purpose of this embodiment according to actual needs.

[0100] In addition, in the various embodiments of the present invention, each functional unit can be integrated into one processing unit, or each unit can be a separate unit, or two or more units can be integrated into one unit; the integrated unit can be implemented in hardware or in the form of hardware plus software functional units.

[0101] Those skilled in the art will understand that all or part of the steps of the above method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it performs the steps of the above method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0102] Alternatively, if the integrated units of this invention are implemented as software functional modules and sold or used as independent products, they can also be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the embodiments of this invention, or the parts that contribute to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, ROM, RAM, magnetic disks, or optical disks.

[0103] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A sleep monitoring and early warning method based on respiratory data analysis, characterized in that, Includes the following steps: Millimeter-wave radar is used to capture body surface movements caused by breathing by emitting electromagnetic waves and receiving echoes reflected by the slight movements of the human chest and abdomen, thus forming real-time respiratory data. By preprocessing real-time respiratory data, respiratory harmonic components are separated, and thoracic breathing and abdominal breathing are distinguished. A prediction model for sleep state and sleep apnea-hypopnea index is constructed based on a deep learning network. Real-time collection of user breathing data; prediction of user breathing data using sleep state and sleep apnea-hypopnea index prediction model; estimation of user sleep state and sleep apnea-hypopnea index. The system provides early warnings based on the user's sleep status and sleep apnea-hypopnea index, and generates relevant treatment plans based on the warning information. By preprocessing real-time respiratory data, respiratory harmonic components are separated, and thoracic breathing and abdominal breathing are distinguished, specifically: By filtering and denoising real-time respiratory data to remove noise, preprocessed respiratory data is obtained. The preprocessed respiratory data is then processed using a phase extraction algorithm to obtain phase change data caused by respiration. Multipath cancellation technology is used to process the phase change data caused by respiration to reduce environmental reflection interference, and variational mode decomposition algorithm is introduced to decompose the phase change data caused by respiration. By decomposing, respiratory harmonic components are separated, and thoracic breathing and abdominal breathing are distinguished based on these respiratory harmonic components. A prediction model for sleep state and sleep apnea-hypopnea index based on deep learning network is constructed as follows: A prediction model for sleep state and sleep apnea-hypopnea index is constructed based on a deep learning network. Sleep state data, thoracic breathing data and abdominal breathing data are collected to form a large-scale contact breathing signal database. Sleep stage and sleep apnea-hypopnea index are estimated simultaneously within a unified framework. By estimating, the intrinsic relationship between the macroscopic structure of sleep and respiratory events is decoupled, and the knowledge of sleep breathing characteristics learned from contact breathing signals is transferred to millimeter-wave radar signals through a domain adversarial training mechanism. The thoracic and abdominal breathing data are used as model inputs, and the sleep stages and sleep apnea-hypopnea index are used as model outputs. The learning rate and number of training sessions are set. The sleep state and sleep apnea-hypopnea index prediction model is trained based on the learning rate and the number of training sessions. When the prediction accuracy of the sleep state and sleep apnea-hypopnea index prediction model reaches the preset prediction accuracy data, the sleep state and sleep apnea-hypopnea index prediction model is output.

2. The sleep monitoring and early warning method based on respiratory data analysis according to claim 1, characterized in that, Millimeter-wave radar is used to capture body surface movements caused by respiration by emitting electromagnetic waves and receiving echoes reflected from subtle movements of the human chest and abdomen, thus generating real-time respiratory data. Specifically: By installing millimeter-wave radar in a predetermined area, the millimeter-wave radar operates in the 60-64 GHz frequency band, uses frequency-modulated continuous wave to transmit electromagnetic waves and receives the echoes reflected by the slight movements of the human chest and abdomen to capture the surface movements caused by breathing. Configure the number of transmitting and receiving antennas in the millimeter-wave radar, initialize the number of transmitting and receiving antennas, and use a multi-input multi-output array based on the number of transmitting and receiving antennas to collect and test respiratory data that captures body surface movements caused by breathing. Through the test, determine the spatial resolution of the current data. Set a spatial resolution threshold, determine whether the spatial resolution of the current data is greater than the spatial resolution threshold, and if the spatial resolution of the current data is greater than the spatial resolution threshold, arrange the current number of transmitting and receiving antennas and collect real-time respiratory data. If the spatial resolution of the current data is not greater than the spatial resolution threshold, adjust the number of current transmitting and receiving antennas until the spatial resolution of the current data is greater than the spatial resolution threshold.

3. The sleep monitoring and early warning method based on respiratory data analysis according to claim 1, characterized in that, Real-time collection of the user's respiratory data, and prediction of the user's sleep state and sleep apnea-hypopnea index using a sleep state and sleep apnea-hypopnea index prediction model, specifically: Real-time collection of user breathing data, and input of the user breathing data into the sleep state and sleep apnea-hypopnea index prediction model for prediction; The system estimates and obtains the user's sleep state and sleep apnea-hypopnea index, and outputs the user's sleep state and sleep apnea-hypopnea index.

4. The sleep monitoring and early warning method based on respiratory data analysis according to claim 1, characterized in that, The warning is issued based on the user's sleep status and sleep apnea-hypopnea index, specifically as follows: Set the user's sleep state and the range of the sleep apnea-hypopnea index evaluation indicators, and determine whether the user's sleep state and sleep apnea-hypopnea index are within the range of the user's sleep state and sleep apnea-hypopnea index evaluation indicators. If the user's sleep state and sleep apnea-hypopnea index are within the range of the user's sleep state and sleep apnea-hypopnea index evaluation indicators, then normal information is generated and displayed in a preset manner. If the user's sleep state and sleep apnea-hypopnea index are not within the range of the user's sleep state and sleep apnea-hypopnea index evaluation indicators, an early warning message will be generated and displayed in a preset manner.

5. The sleep monitoring and early warning method based on respiratory data analysis according to claim 1, characterized in that, Based on the early warning information, a relevant treatment plan is generated, specifically as follows: Based on the warning information, the changes in the user's sleep state characteristics and sleep apnea-hypopnea index data within a preset time period are obtained, and a real-time sleep structure map is constructed based on the changes in the user's sleep state characteristics within the preset time period. Users upload various sleep structure diagrams and historical treatment plans under sleep apnea-hypopnea index data to build a database, and input the various sleep structure diagrams and historical treatment plans under sleep apnea-hypopnea index data into the database for storage; Real-time sleep structure graphs and sleep apnea-hypopnea index data are input into the database for data matching. Through data matching, the optimal treatment plan under the current real-time sleep structure graph and sleep apnea-hypopnea index data is obtained. Based on the current real-time sleep structure map and the sleep apnea-hypopnea index data, the optimal treatment plan is generated.

6. The sleep monitoring and early warning method based on respiratory data analysis according to claim 1, characterized in that, By filtering and denoising the real-time respiratory data, noise is removed, specifically as follows: Hampel filtering was used to remove outliers from the original CSI sequence of the real-time respiratory data, and Gaussian filtering and linear interpolation were used to smooth the data and handle possible data loss. The respiratory frequency band signal was extracted using a Butterworth low-pass filter, and signal anomalies caused by large-scale body movement were identified and suppressed by Z-Score anomaly detection. The filtered signal is peaked by local peak detection, the breathing rate is calculated, and noise is removed.

7. A sleep monitoring and early warning system based on respiratory data analysis, characterized in that, The device includes a memory and a processor. The memory includes a sleep monitoring and early warning method program based on respiratory data analysis. When the processor executes the sleep monitoring and early warning method program based on respiratory data analysis, it implements the steps of the sleep monitoring and early warning method based on respiratory data analysis as described in any one of claims 1-6.

8. A computer-readable storage medium, characterized in that, The method includes a sleep monitoring and early warning method program based on respiratory data analysis, which, when executed by a processor, implements the steps of the sleep monitoring and early warning method based on respiratory data analysis as described in any one of claims 1-6.

Images