Multi-user sleep monitoring method based on millimeter wave radar beam linkage analysis

Abstract

The application relates to the field of sleep quality evaluation, and discloses a multi-user sleep monitoring method based on millimeter wave radar beam linkage analysis, which comprises the following steps: firstly, presetting the emission angle of a millimeter wave beam, constructing a three-dimensional model of an intelligent bed, combining the beam with the three-dimensional model for joint analysis, positioning the position of a user on the intelligent bed in real time and collecting corresponding physiological data, analyzing the physiological data, and obtaining a sleep quality report of the user. The application can capture the tiny fluctuations of the chest cavity caused by breathing and heartbeat, which are in millimeter or even sub-millimeter levels, so that the breathing rate and the heart rate can be accurately calculated, and the purpose of sleep quality monitoring is achieved.

Description

Technical Field

[0001] This invention relates to the field of sleep quality assessment, and in particular to a multi-user sleep monitoring method based on millimeter-wave radar beam linkage analysis. Background Technology

[0002] Millimeter-wave radar beams are extremely high-frequency electromagnetic waves. Millimeter waves refer to electromagnetic waves with wavelengths between 1 and 10 millimeters. Their extremely short wavelengths allow radar systems to be made very small while maintaining extremely high resolution and accuracy. Radar antennas use beamforming technology to concentrate emitted energy in a specific, very narrow direction, rather than scattering it in all directions. This forms a beam. The beam is directional and controllable. Based on millimeter-wave radar beam linkage for user sleep monitoring, contactless and imperceptible monitoring can be achieved, along with spatial allocation capabilities. This means that by forming multiple independent beams, the spatial positions of two users can be simultaneously locked and tracked. Even if they move slightly, the system can continue to track them through electronic beam deflection. Millimeter waves are extremely sensitive to micro-movements, capturing minute fluctuations in the chest cavity at the millimeter or even sub-millimeter level caused by breathing and heartbeat, thus accurately calculating respiratory and heart rates. This is crucial because sleep quality is closely related to cardiovascular health, the immune system, cognitive function, and emotional stability. Long-term sleep deprivation or abnormal sleep structure is an independent risk factor for many major diseases such as hypertension, diabetes, depression, and Alzheimer's disease. Furthermore, given the advantages of millimeter-wave radar beams, it is necessary to conduct multi-user sleep quality analysis based on millimeter-wave radar beam linkage. Summary of the Invention

[0003] This invention overcomes the shortcomings of the prior art and provides a multi-user sleep monitoring method based on millimeter-wave radar beam linkage 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 multi-user sleep monitoring method based on millimeter-wave radar beam linkage analysis, comprising the following steps:

[0006] The system transmits a linked millimeter-wave beam through a millimeter-wave radar system to perform full-domain modeling of the smart bed, and presets the angle of the millimeter-wave beam by combining historical data.

[0007] The millimeter-wave beam of the target smart bed is collected and analyzed in real time, and the data preprocessing of the millimeter-wave beam is performed simultaneously.

[0008] By using a convolutional neural network model, the target analysis beam signal and the updated real-time target 3D model are jointly processed and analyzed to achieve real-time positioning of the user target location of the smart bed.

[0009] The number of users on the target smart bed is determined, and the user's chest tracking beam is adaptively filtered according to the number of users;

[0010] The sleep quality of all users of the target smart bed is analyzed in real time according to the user available spectrum characteristics, and the sleep quality report of different users is specified according to the analysis result.

[0011] Further, in a preferred embodiment of the present application, the millimeter wave radar system transmits a millimeter wave beam to perform global modeling of the smart bed, and the angle of the millimeter wave beam is preset according to historical data, specifically:

[0012] The smart bed used for multi-user sleep monitoring is calibrated as a target smart bed, wherein the target smart bed is built-in millimeter wave radar sensor, associated sensor and main controller;

[0013] The millimeter wave radar sensor and the associated sensor are powered on, and communication with the main controller is established to ensure that the main controller can control the millimeter wave radar sensor to emit millimeter wave beams, and to ensure that the main controller can control the associated sensor to operate;

[0014] The millimeter wave radar sensor emits millimeter wave beams to scan the bed body of the target smart bed through the main controller, wherein when scanning the bed body of the target smart bed, a spatial clustering algorithm based on density analysis is introduced to construct point cloud data of the bed body of the target smart bed in combination with the millimeter wave beam;

[0015] The spatial clustering algorithm based on density analysis automatically identifies the spatial density of different point cloud data, outputs different point cloud data clusters, presets the standard point cloud data cluster density range of the smart bed, selects the point cloud data cluster whose density is maintained within the standard point cloud data cluster density range of the smart bed, and calibrates it as the target point cloud data cluster;

[0016] The point cloud connection is performed on the target point cloud data cluster to generate a three-dimensional model of the bed body of the target smart bed in real time, which is calibrated as a real-time target three-dimensional model, and the specification data of the real-time target three-dimensional model is calculated;

[0017] The user end software connected with the target smart bed is acquired, and the common sleep posture and the historical preferred sleep position of the user of the target smart bed are determined in the user end software, and the historical preferred sleep position is located in the real-time target three-dimensional model;

[0018] The angle of the millimeter wave beam is preliminarily set on the target smart bed in combination with the historical preferred sleep position and the common sleep posture of the user, to ensure that the millimeter wave beam points to the chest position of the user on the historical preferred sleep position on the target smart bed, and the millimeter wave beam is a steerable beam, which is calibrated as a user chest tracking beam;

[0019] Meanwhile, a millimeter wave wide beam for identifying the real-time sleep position of the user is emitted, and is calibrated as a user position positioning beam.

[0020] Further, in a preferred embodiment of the present application, the millimeter wave beams of the target intelligent bed are synchronously collected and analyzed in real time, and the data preprocessing of the millimeter wave beams is performed.

[0021] The millimeter wave radar sensor is used as a master clock source to emit a synchronization pulse signal to all associated sensors, and the associated sensors are operated simultaneously after receiving the synchronization pulse signal, thereby forming clock synchronization of the millimeter wave radar sensor and all associated sensors.

[0022] When the associated sensors receive the synchronization pulse signal, the disconnection of all associated sensors and the master controller is immediately performed, and all associated sensors are restarted through the master controller immediately after the disconnection, thereby realizing clock synchronization of the millimeter wave radar sensor and all associated sensors.

[0023] For the millimeter wave beams emitted by the millimeter wave radar sensor, different independent beam channels are obtained, wherein an independent beam channel has an independent millimeter wave beam, including a user chest cavity tracking beam and a user position positioning beam.

[0024] The echo signals of the millimeter wave beams in different independent beam channels are received by the master controller, and the echo signals are subjected to signal preprocessing, wherein the signal preprocessing is noise reduction processing and signal filtering processing of the echo signals by filtering, and the echo signals are subjected to frequency spectrum analysis to analyze whether the reflection frequency range of the echo signals is maintained within a predetermined range.

[0025] If not, the echo signals are subjected to continuous noise reduction processing and signal filtering processing until the reflection frequency range of the echo signals is maintained within the predetermined range.

[0026] The real-time target three-dimensional model is updated in combination with the sensor data output by the associated sensors, the output sensor data on the real-time target three-dimensional model is ensured, and the echo signals of the preprocessed millimeter wave beams are calibrated as target analysis beam signals.

[0027] Further, in a preferred embodiment of the present application, the target analysis beam signals and the updated real-time target three-dimensional model are jointly processed and analyzed by the convolutional neural network model, thereby realizing real-time positioning of the user target position of the target intelligent bed.

[0028] The convolutional neural network model is introduced into the master controller of the target intelligent bed, and is used for convolutional prediction analysis of the target analysis beam signals and the updated target three-dimensional model, and is calibrated as a convolutional neural network model.

[0029] The convolutional neural network model internally exists a sensing data branch and a beam signal branch, a four-dimensional data cube is constructed for the target analysis beam signal in the beam signal branch, and the four-dimensional data cube is composed of an independent beam channel corresponding to the target analysis beam signal, an independent beam distance, an independent beam position and an echo time;

[0030] The four-dimensional data cube is convolved and slid in space-time dimensions by a convolution kernel in the beam signal branch, the space-time features of the target analysis beam signal are captured, and global average pooling processing is performed on the target analysis beam signal, so that global statistical information of the target analysis beam signal is obtained;

[0031] A big data network is introduced, the attention weight of different spectral features of the beam signal is searched, and the attention weight distribution of the global statistical information of the target analysis beam signal is performed, so that the spectral feature related to the breathing and heartbeat of the target intelligent bed user is generated, and the user target spectral feature is calibrated;

[0032] Periodic convolution analysis is performed on the user target spectral feature, and the periodic characteristics of the user target spectral feature are outputted;

[0033] The sensing data branch is combined, the updated target three-dimensional model is analyzed, and the user target position orientation of the target intelligent bed is realized in real time by combining the user target spectral feature and the corresponding periodic characteristics.

[0034] Further, in a preferred embodiment of the present application, the sensing data branch is combined, the updated target three-dimensional model is analyzed, and the user target position orientation of the target intelligent bed is realized in real time by combining the user target spectral feature and the corresponding periodic characteristics, specifically:

[0035] The sensing data of the updated target three-dimensional model is acquired, wherein the sensing data includes pressure data and temperature data;

[0036] The sensing data of the target three-dimensional model is introduced into the sensing data branch, wherein the pressure data is analyzed to determine the pressure distribution on the target intelligent bed, and a two-dimensional pressure grayscale image is constructed;

[0037] The two-dimensional pressure grayscale image is analyzed in real time by a convolution kernel in the sensing data branch to identify the real-time activity state of the user on the target intelligent bed, and the temperature data is analyzed to output a heat distribution map of the user on the target intelligent bed;

[0038] The heat distribution map describes the correlation state of the human body temperature and the environment temperature, and the real-time orientation and position of the user on the target intelligent bed are determined by combining the real-time activity state of the user on the target intelligent bed;

[0039] In combination with the real-time orientation and position of the user on the target smart bed, the user chest tracking beam and the user position locating beam are adaptively adjusted in angle, and the periodic characteristics of the user target spectral feature are adaptively updated after the adaptive adjustment, to generate an updated user spectral feature.

[0040] Further, in a preferred embodiment of the present application, the user chest tracking beam and the user position locating beam are adaptively adjusted in angle, and the periodic characteristics of the user target spectral feature are adaptively updated after the adaptive adjustment, to generate an updated user spectral feature, specifically:

[0041] According to the real-time orientation and position of the user on the smart bed, a user position-orientation joint analysis model is constructed, named as a user sleep state real-time model;

[0042] The user sleep state real-time model is updated through the user position locating beam, that is, the chest position of the user is searched through the user position locating beam, to realize real-time determination of the spatial coordinates of the chest position of the user, which is calibrated as chest real-time spatial coordinates;

[0043] Wherein, when the user sleep state real-time model changes, that is, the real-time orientation and position of the user on the smart bed changes, the user position locating beam is adaptively adjusted in angle, to realize secondary determination of the real-time orientation and position of the user;

[0044] For the user chest tracking beam, the user chest tracking beam is controlled to be adjusted in real-time positioning according to the chest real-time spatial coordinates;

[0045] Wherein, a redundant positioning range of the chest real-time spatial coordinates is predetermined, and a ratio table of the redundant positioning range of the chest real-time spatial coordinates and the signal strength of the user chest tracking beam is preset, which is calibrated as a positioning-signal strength ratio table, wherein the redundant positioning range of the chest real-time spatial coordinates is negatively correlated with the signal strength of the user chest tracking beam;

[0046] Based on the positioning-signal strength ratio table, the beam width of the user chest tracking beam is timely adjusted;

[0047] During the adjustment of the user chest tracking beam and the user position locating beam, the periodic characteristics of the user target spectral feature are adaptively updated, wherein the adaptive update is first to maintain the continuity of beam switching, and the user target spectral feature in the time period without user body movement is collected;

[0048] The system performs periodic heartbeat frequency drift analysis on the collected user target spectral features and determines the heartbeat frequency drift window based on the drift range, so as to maintain the periodic heartbeat frequency drift of users within a window at a predetermined value.

[0049] By combining the heartbeat frequency drift window, environmental compensation is performed on the collected user target spectrum features to obtain the adaptively updated user target spectrum features, which are then labeled as user updated spectrum features.

[0050] Furthermore, in a preferred embodiment of the present invention, the step of determining the number of users on the target smart bed and adaptively filtering the user chest cavity tracking beam based on the number of users specifically involves:

[0051] Analyze the number of users on the target smart bed. If the number of users is 1, directly output the user update spectrum characteristics.

[0052] If the number of users is greater than 1, the chest cavity distance for different users is calculated based on the real-time spatial coordinates of the chest cavity, and the minimum range of chest cavity distance is preset.

[0053] If the interthoracic distance between users is not less than the minimum interthoracic distance range, then directly output the updated spectral features of the users;

[0054] If the distance between chest cavities between users is less than the minimum range of chest cavities, the user update spectrum features are analyzed, the respiratory rate is extracted, and an adaptive filtering algorithm is used to filter the interference components of different user chest cavity tracking beams to obtain purified different user chest cavity tracking beams.

[0055] By combining the user location positioning beam with the purified different user chest cavity tracking beams, the user update spectrum features are updated a second time, and together with the directly output user update spectrum features, they are collectively referred to as the user available spectrum features.

[0056] Furthermore, in a preferred embodiment of the present invention, the analysis of available spectral characteristics of users, the real-time analysis of sleep quality for all users of the target smart bed, and the generation of sleep quality reports for different users based on the analysis results, specifically involves:

[0057] The available spectrum characteristics of users are analyzed to extract physiological data of all users on the target smart bed. The physiological data includes the user's respiratory rate and heart rate.

[0058] The main controller creates identity IDs for different users of the target smart bed and matches the physiological data of the users corresponding to the identity IDs.

[0059] The physiological data comparison analysis knowledge graph records different physiological data corresponding to sleep stages, and sleep quality corresponding to different sleep stage duration lengths;

[0060] The physiological data of all users is imported into the physiological data comparison analysis knowledge graph, the current sleep stage of different users is output, and the identity ID is matched and processed, and the duration of the user in different sleep stages is analyzed to obtain the sleep quality of different users;

[0061] The sleep quality of different users is matched with the corresponding identity ID, and the sleep quality report of different users is output in the main controller.

[0062] The technical defects in the background art are solved, and the present application has the following advantages: first, the transmission angle of the millimeter wave beam is preset, and a three-dimensional model of the smart bed is constructed, the beam and the three-dimensional model are combined for joint analysis, the position of the user on the smart bed is located in real time, and the corresponding physiological data is collected, the physiological data is analyzed, and the sleep quality report of the user is obtained. The present application can capture the chest millimeter level or even sub-millimeter level micro fluctuations caused by breathing and heartbeat, so as to accurately calculate the respiratory rate and heart rate, and achieve the purpose of sleep quality monitoring. BRIEF DESCRIPTION OF DRAWINGS

[0063] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments or the prior art description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and those skilled in the art can obtain other drawings of embodiments according to these drawings without creative labor.

[0064] Figure 1 A flowchart of a multi-user sleep monitoring method based on millimeter wave radar beam linkage analysis is shown;

[0065] Figure 2 A flowchart of a method for real-time positioning of the target position of the user of the target smart bed is shown. DETAILED DESCRIPTION

[0066] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments or the prior art description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and those skilled in the art can obtain other drawings of embodiments according to these drawings without creative labor.

[0067] In the following description, a lot of specific details are set forth in order to provide a thorough understanding of the present application, however, the present application can be practiced in other manners different from those described herein, therefore, the scope of protection of the present application is not limited by the specific embodiments disclosed below.

[0068] Figure 1 A flowchart of a multi-user sleep monitoring method based on millimeter wave radar beam linkage analysis is shown, including the following steps:

[0069] S102: Emit linkage millimeter wave beams through a millimeter wave radar system, perform global modeling processing of the smart bed, and preset the angles of the millimeter wave beams in combination with historical data;

[0070] S104: Real-time synchronous acquisition and analysis of the millimeter wave beams of the target smart bed, while performing data preprocessing of the millimeter wave beams;

[0071] S106: Joint processing and analysis of the target analysis beam signal and the updated real-time target three-dimensional model through a convolutional neural network model, to realize real-time positioning of the user target position of the target smart bed;

[0072] S108: Determine the number of users on the target smart bed, and perform adaptive filtering on the user chest tracking beam according to the number of users;

[0073] S110: Analyze the user available spectrum features, perform real-time analysis of the sleep quality of all users of the target smart bed, and specify sleep quality reports for different users according to the analysis results.

[0074] Further, in a preferred embodiment of the present application, the S102 specifically comprises:

[0075] The smart bed used for multi-user sleep monitoring is calibrated as a target smart bed, wherein the target smart bed is built-in with a millimeter wave radar sensor, an associated sensor, and a main controller;

[0076] Power on the millimeter wave radar sensor and the associated sensor, and establish communication with the main controller, to ensure that the main controller can control the millimeter wave beam emission of the millimeter wave radar sensor, and to ensure that the main controller can control the operation of the associated sensor;

[0077] Control the millimeter wave radar sensor to emit millimeter wave beams through the main controller, to scan the bed body of the target smart bed, wherein when scanning the bed body of the target smart bed, introduce a spatial clustering algorithm based on density analysis in combination with the millimeter wave beams, to construct point cloud data of the bed body of the target smart bed;

[0078] The spatial clustering algorithm based on density analysis automatically identifies the spatial density of different point cloud data, outputs different point cloud data clusters, predefines a standard point cloud data cluster density range of the smart bed, selects a point cloud data cluster whose density is maintained within the standard point cloud data cluster density range of the smart bed, and labels the point cloud data cluster as a target point cloud data cluster.

[0079] The point cloud connection is performed on the target point cloud data cluster, a three-dimensional model of a bed body of the target smart bed is generated in real time, is labeled as a real-time target three-dimensional model, and specification data of the real-time target three-dimensional model is calculated.

[0080] The user end software connected with the target smart bed is acquired, and a common sleep posture of a user of the target smart bed and a historical preferred sleep position are determined in the user end software, and the historical preferred sleep position is located in the real-time target three-dimensional model.

[0081] The angle of the millimeter wave beam is preliminarily set on the target smart bed in combination with the historical preferred sleep position and the common sleep posture of the user, to ensure that the millimeter wave beam points to the chest position of the user on the historical preferred sleep position on the target smart bed, and the millimeter wave beam is a steerable beam, which is labeled as a user chest tracking beam.

[0082] Meanwhile, a millimeter wave wide beam for identifying the real-time sleep position of the user is emitted, which is labeled as a user position positioning beam.

[0083] It should be noted that the three-dimensional model of the target smart bed needs to be constructed first, which is used to preliminarily set the initial positions of different millimeter wave beams. The main controller is used to control the millimeter wave beam emission of all millimeter wave radar sensors, which can improve the emission efficiency of the millimeter wave. The three-dimensional model is constructed by point cloud data through the beam, that is, three-dimensional point cloud data, and the three-dimensional model can be generated by collecting all point cloud data. The densities of the point cloud data clusters are different, and the point cloud data cluster whose density is maintained within the standard point cloud data cluster range of the smart bed is selected, because the densities of the data clusters corresponding to different positions and different planes on the bed are different, and the classification is performed through the density, which can improve the accuracy of the model construction and identify all candidate planes of the smart bed. The user has a preferred sleep position when sleeping, and the historical preferred sleep position is located in combination with the three-dimensional model, and the millimeter wave beam is aimed at the historical preferred sleep position, which reduces the step of adjusting the angle of the millimeter wave beam and improves the sleep monitoring efficiency.

[0084] Further, in a preferred embodiment of the present application, the S104 specifically comprises:

[0085] The millimeter wave radar sensor is used as a main clock source, and a synchronization pulse signal is emitted to all associated sensors, the associated sensors are operated at the same time after receiving the synchronization pulse signal, and the clock synchronization of the millimeter wave radar sensor and all associated sensors is formed.

[0086] wherein, when the associated sensor receives the synchronization pulse signal, the disconnection of all associated sensors with the main controller is immediately performed, and all associated sensors are restarted through the main controller immediately after disconnection, so as to realize the clock synchronization of the millimeter wave radar sensor and all associated sensors;

[0087] For the millimeter wave beam emitted by the millimeter wave radar sensor, different independent beam channels are obtained, wherein there is an independent millimeter wave beam in an independent beam channel, including a user chest cavity tracking beam and a user position positioning beam.

[0088] The echo signal of the millimeter wave beam in the different independent beam channels is received by the main controller, and the echo signal is preprocessed, wherein the preprocessing is noise reduction processing and signal filtering processing of the echo signal by filtering, and the echo signal is subjected to frequency spectrum analysis to analyze whether the reflection frequency range of the echo signal is maintained within a predetermined range.

[0089] If not, the echo signal is subjected to continuous noise reduction processing and signal filtering processing until the reflection frequency range of the echo signal is maintained within a predetermined range.

[0090] The real-time target three-dimensional model is updated in combination with the sensor data output by the associated sensor, the output sensor data on the real-time target three-dimensional model is ensured, and the echo signal of the preprocessed millimeter wave beam is calibrated as a target analysis beam signal.

[0091] It should be noted that the data analysis of the occurring beam is realized by receiving the echo signal, and the echo signal reflects the sleep state of the user. The signal beam needs to be synchronously transmitted to ensure the real-time and accuracy of monitoring the sleep state of the user. Through the clock synchronization method, that is, taking the synchronization pulse signal as the reference signal, the signal beam is transmitted after receiving the synchronization pulse signal, so as to realize the purpose of transmitting all signals at the same time. Different independent beam channels are obtained, wherein there is an independent millimeter wave beam in an independent beam channel, and the user chest cavity tracking beam and the user position positioning beam are both beams for monitoring the sleep of the user. The signal after echo needs to be preprocessed, because there may be noise, so noise reduction and signal filtering processing are needed to obtain the target analysis beam signal.

[0092] Further, in a preferred embodiment of the present application, the S108 specifically comprises:

[0093] The number of users on the target intelligent bed is analyzed, and if the number of users is 1, the user update spectrum feature is directly output;

[0094] If the number of users is greater than 1, the chest cavity distance between different users is calculated according to the real-time spatial coordinates of the chest cavity, and a minimum range of the chest cavity distance is preset.

[0095] If the chest cavity distance between the users is not less than the minimum range of the chest cavity distance, the user updated frequency spectrum feature is directly outputted;

[0096] If the chest cavity distance between the users is less than the minimum range of the chest cavity distance, the user updated frequency spectrum feature is analyzed, the breathing frequency is extracted, and the adaptive filtering algorithm is combined to filter the interference components of the different user chest cavity tracking beams, so that the purified different user chest cavity tracking beams are obtained;

[0097] The user updated frequency spectrum feature is updated again by combining the user position positioning beam and the purified different user chest cavity tracking beam, and is collectively referred to as the user available frequency spectrum feature.

[0098] It should be noted that the number of bed users affects the accuracy of the current user updated frequency spectrum feature. If there is only one person, the signal will not be affected by others, that is, the signal will not be disturbed, and the user updated frequency spectrum feature can be directly used. If there are multiple people, but the distance between two people is large, the signal will not cross and will not be disturbed, and the user updated frequency spectrum feature can be directly used. When there are multiple people and the distance between people is short, the signal may cross and the environment may also affect it, so the breathing frequency is used as a reference because the breathing frequencies of different people are different. By using the adaptive filtering method, the breathing frequency is combined to distinguish different user chest cavity tracking beams, and the environment is filtered by interference components to realize the purification of the user chest cavity tracking beam, and the user updated frequency spectrum feature is updated again after the purification.

[0099] Further, in a preferred embodiment of the present application, the S110 specifically comprises:

[0100] The user available frequency spectrum feature is analyzed to extract physiological data of all users of the target smart bed, wherein the physiological data includes the breathing frequency and the heartbeat frequency of the user;

[0101] The identity ID of different users of the target smart bed is created by the main controller, and the physiological data of the user corresponding to the identity ID is matched and processed;

[0102] The physiological data comparison and analysis knowledge graph is introduced, wherein the physiological data comparison and analysis knowledge graph records different physiological data corresponding to sleep stages, and sleep quality corresponding to different sleep stage durations;

[0103] The physiological data of all users is imported into the physiological data comparison and analysis knowledge graph, the current sleep stage of different users is outputted, and the identity ID is matched and processed, and the duration of the user in different sleep stages is analyzed to obtain the sleep quality of different users;

[0104] The sleep quality of different users is matched with corresponding identity IDs, and a sleep quality report of different users is output in the main controller.

[0105] It should be noted that the sleep quality of the user can be determined by analyzing physiological data, and the breathing frequency and heartbeat frequency at different sleep times represent different sleep stages, and the sleep quality of light sleep is different from that of deep sleep, so the physiological data are compared and analyzed with the knowledge graph to determine the sleep quality of different users, and the corresponding identity IDs are combined to output the sleep quality report of different users, thereby directly determining the sleep quality of the user.

[0106] Figure 2 A method flowchart for real-time positioning of a target user position of a target smart bed is shown, including the following steps:

[0107] S202: The target analysis beam signal and the updated real-time target three-dimensional model are jointly processed and analyzed by a convolutional neural network model to realize real-time positioning of the target user position of the target smart bed;

[0108] S204: The updated target three-dimensional model is analyzed in combination with the sensing data branch, and the real-time positioning of the target user position of the target smart bed is realized in combination with the user target spectrum feature and the corresponding periodic characteristic;

[0109] S206: The angle of the user chest tracking beam and the user position positioning beam is adaptively adjusted, and the periodic characteristic of the user target spectrum feature is adaptively updated after the adaptive adjustment to generate the updated user spectrum feature.

[0110] Further, in a preferred embodiment of the present application, the S202 is specifically:

[0111] A convolutional neural network model is introduced in the main controller of the target smart bed, which is used for convolutional prediction analysis of the target analysis beam signal and the updated target three-dimensional model, and is calibrated as a convolutional neural network model;

[0112] The convolutional neural network model includes a sensing data branch and a beam signal branch, and the target analysis beam signal is constructed into a four-dimensional data cube in the beam signal branch, and the four-dimensional data cube is composed of an independent beam channel, an independent beam distance, an independent beam position and an echo time corresponding to the target analysis beam signal.

[0113] The four-dimensional data cube is convolved in space-time dimensions by a convolution kernel in the beam signal branch to capture the space-time features of the target analysis beam signal, and the target analysis beam signal is globally averaged and pooled to obtain global statistical information of the target analysis beam signal;

[0114] The big data network is introduced, the attention weight of different spectrum characteristics of the beam signal is retrieved, and the attention weight distribution of the global statistical information of the target analysis beam signal is generated, and the spectrum characteristics related to the breathing and heartbeat of the target intelligent bed user are generated, and the user target spectrum characteristics are calibrated;

[0115] Periodic convolution analysis is performed on the user target spectrum characteristics, and the periodic characteristics of the user target spectrum characteristics are output.

[0116] It should be noted that the convolutional neural network model is used for spectrum characteristic prediction analysis of the user, and the spectrum characteristics include information related to breathing and heartbeat, which is used for analysis and determination of the sleep quality of the user. The four-dimensional data cube is constructed for the target analysis beam signal, that is, the joint analysis of different beam channels, beam echo distances, echo times and echo positions is performed. These data represent the state of the beam echo, and the four-dimensional data cube is convolved and slid in the time and space dimensions through the convolution kernel in the beam signal branch, so that the periodicity of the vital signs is captured, and the spectrum characteristics related to the breathing and heartbeat are captured. Global average pooling processing can obtain global statistical information of each feature channel, and the attention weight of different spectrum characteristics, that is, the weight allocated to different features, is combined to generate spectrum characteristics related to the breathing and heartbeat of the target intelligent bed user.

[0117] Further, in a preferred embodiment of the present application, the S204, specifically:

[0118] The sensing data of the updated target three-dimensional model is acquired, wherein the sensing data includes pressure data and temperature data;

[0119] The sensing data of the target three-dimensional model is introduced into the sensing data branch, wherein the pressure data is analyzed to determine the pressure distribution on the target intelligent bed, and a two-dimensional pressure grayscale image is constructed;

[0120] The two-dimensional pressure grayscale image is analyzed in real time through the convolution kernel in the sensing data branch to identify the real-time activity state of the user on the target intelligent bed, and the temperature data is analyzed to output a heat distribution map of the user on the target intelligent bed;

[0121] The heat distribution map describes the correlation state of the human body temperature and the environment temperature, and the real-time activity state of the user on the target intelligent bed is combined to determine the real-time orientation and position of the user on the target intelligent bed.

[0122] It's important to note that combining pressure and temperature data allows for initial positioning of the user's sleeping position in bed. Different pressure and temperature distributions reflect a person's location and activity level because ambient and body temperatures differ. The thermal distribution map illustrates the relationship between body temperature and ambient temperature. Combined with the user's real-time activity on the target smart bed, the user's real-time orientation and position on the target smart bed can be determined.

[0123] Furthermore, in a preferred embodiment of the present invention, S206 specifically includes:

[0124] Based on the user's real-time orientation and location on the smart bed, a joint analysis model of user location and orientation is constructed and named the real-time model of user sleep state.

[0125] The user's sleep state model is updated in real time by using the user's location positioning beam. Specifically, the user's chest cavity location is searched by the user's location positioning beam to determine the spatial coordinates of the user's chest cavity location in real time and calibrate them as the real-time spatial coordinates of the chest cavity.

[0126] Among them, the user's chest cavity location is searched by the user location positioning beam. When the user's sleep state model changes in real time, that is, when the user's real-time orientation and position on the smart bed change, the user location positioning beam adaptively adjusts the angle to achieve secondary determination of the user's real-time orientation and position.

[0127] For the user's chest cavity tracking beam, the user's chest cavity tracking beam is controlled to perform real-time positioning and adjustment based on the real-time spatial coordinates of the chest cavity;

[0128] Among them, the redundant positioning range of the real-time spatial coordinates of the chest cavity is predetermined, and a ratio table of the redundant positioning range of the real-time spatial coordinates of the chest cavity and the signal strength of the user's chest cavity tracking beam is preset and calibrated as a positioning-signal strength ratio table. Among them, the redundant positioning range of the real-time spatial coordinates of the chest cavity is negatively correlated with the signal strength of the user's chest cavity tracking beam.

[0129] Based on the aforementioned positioning-signal strength ratio table, the beamwidth of the user's chest cavity tracking beam is adjusted in a timely manner.

[0130] During the adjustment of the user chest cavity tracking beam and the user location positioning beam, the periodic characteristics of the user target spectrum feature are adaptively updated. The adaptive update firstly maintains the continuity of beam switching and collects the user target spectrum feature during the time period when there is no user body movement.

[0131] The system performs periodic heartbeat frequency drift analysis on the collected user target spectral features and determines the heartbeat frequency drift window based on the drift range, so as to maintain the periodic heartbeat frequency drift of users within a window at a predetermined value.

[0132] In combination with the heartbeat frequency drift window, the collected user target spectrum feature is compensated for the environment, that is, the adaptive updated user target spectrum feature is obtained, which is calibrated as the user updated spectrum feature.

[0133] It should be noted that the user position positioning beam function is to determine the approximate position of the chest cavity in combination with the real-time orientation and position of the user in the smart bed during the initial positioning. After the real-time orientation and position of the user in the smart bed are initially determined, because the deviation values of temperature and pressure are large, the position and orientation of the user need to be determined again to improve the angle adjustment efficiency of the user chest cavity tracking beam. The user sleep state real-time model is a model for real-time positioning of the real-time orientation and position of the user, which realizes the secondary determination of the real-time orientation and position of the user in combination with the user position positioning beam. The user chest cavity tracking beam is used to analyze the frequency of the chest cavity heartbeat, and to predefine the redundant positioning range of the chest cavity real-time space coordinates. When the signal strength is weak, the positioning range can be large, and at this time the positioning accuracy will be reduced. Because the signal strength is weak and cannot support the positioning accuracy, it is easy to consume energy furniture and waste electricity. On the contrary, when the signal strength is strong, it can be directly positioned accurately. During the adjustment process of the user chest cavity tracking beam and the user position positioning beam, the periodic characteristics of the user target spectrum feature are adaptively updated. Because the data collected after the angle adjustment will have deviation, adaptive updating is needed to maintain the continuity of beam switching and improve the collection accuracy. The collected user target spectrum feature is analyzed for periodic heartbeat frequency drift, and the heartbeat frequency drift window is determined according to the drift range. The purpose is to maintain the stability of the heartbeat frequency. Because the heartbeat will not always remain unchanged, there may be a momentary heartbeat acceleration, etc. Therefore, the heartbeat frequency drift window is determined to ensure that the heartbeat frequency in the window remains stable, which is convenient for analyzing the current sleep quality. The collected user target spectrum feature is compensated for the environment, and finally the user updated spectrum feature is obtained.

[0134] The above is only a specific embodiment of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of changes or replacements within the technical range disclosed by the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A multi-user sleep monitoring method based on millimeter-wave radar beam linkage analysis, characterized in that, Includes the following steps: A millimeter-wave radar system emits a millimeter-wave beam to perform full-domain modeling of the target intelligent bed, and the angle of the millimeter-wave beam is preset based on historical data. Specifically: The smart bed used for multi-user sleep monitoring is designated as the target smart bed, wherein the target smart bed has a built-in millimeter-wave radar sensor, a correlation sensor and a main controller; Power on the millimeter-wave radar sensor and associated sensors, and establish communication with the main controller to ensure that the main controller controls the millimeter-wave radar sensor's millimeter-wave beam transmission and controls the operation of the associated sensors. The main controller controls the millimeter-wave radar sensor to emit a millimeter-wave beam to scan the target smart bed. When scanning the target smart bed, a spatial clustering algorithm based on density analysis is introduced in combination with the millimeter-wave beam to construct point cloud data of the target smart bed. Among them, the spatial clustering algorithm based on density analysis automatically identifies the spatial density of different point cloud data, outputs different point cloud data clusters, presets the density range of standard point cloud data clusters for smart beds, selects point cloud data clusters whose density is within the density range of standard point cloud data clusters for smart beds, and marks them as target point cloud data clusters. The target point cloud data cluster is connected to generate a 3D model of the target smart bed in real time, which is then calibrated as a real-time target 3D model, and the specification data of the real-time target 3D model is calculated. The user-end software connected to the target smart bed is acquired, and the user's common sleeping posture and historical preferred sleeping position are determined in the user-end software. The historical preferred sleeping position is then located in the real-time target 3D model. Based on historical preferred sleeping positions and the user's common sleeping postures, the millimeter wave beam is initially set at an angle on the target smart bed to ensure that the millimeter wave beam points to the user's chest cavity position in the historical preferred sleeping position on the target smart bed, and that the millimeter wave beam is a steerable beam, calibrated as the user's chest cavity tracking beam. Simultaneously, a millimeter-wave wide beam is emitted to identify the user's real-time sleep location, calibrated as the user location positioning beam; The millimeter-wave beam of the target smart bed is acquired and analyzed in real time, and the data preprocessing of the millimeter-wave beam is performed simultaneously, specifically as follows: Using a millimeter-wave radar sensor as the master clock source, a synchronization pulse signal is transmitted to all associated sensors. All associated sensors operate simultaneously after receiving the synchronization pulse signal, thus achieving clock synchronization between the millimeter-wave radar sensor and all associated sensors. When all associated sensors receive the synchronization pulse signal, all associated sensors are immediately disconnected from the main controller, and all associated sensors are immediately restarted through the main controller after disconnection, so as to realize the clock synchronization between the millimeter-wave radar sensor and all associated sensors. For the millimeter-wave beam emitted by the millimeter-wave radar sensor, different independent beam channels are acquired. Among them, there is an independent millimeter-wave beam within an independent beam channel. The independent millimeter-wave beam includes the user chest cavity tracking beam and the user position positioning beam. The main controller receives echo signals from millimeter-wave beams in different independent beam channels and performs signal preprocessing on the echo signals. The signal preprocessing involves filtering the echo signals and performing spectrum analysis on the echo signals to determine whether the reflection frequency of the echo signals is within a predetermined range. If not, the echo signal will continue to be filtered until the reflection frequency of the echo signal is within the predetermined range. By combining the sensor data output from all associated sensors, the real-time target 3D model is updated to ensure that sensor data is output on the real-time target 3D model, and the echo signal of the preprocessed millimeter wave beam is calibrated as the target analysis beam signal. By jointly processing and analyzing the target analysis beam signal and the updated real-time target 3D model using a convolutional neural network model, real-time positioning of the user target orientation of the smart bed is achieved. Specifically: A convolutional neural network model is introduced into the main controller of the target intelligent bed to perform convolutional prediction analysis on the target analysis beam signal and the updated real-time target 3D model. The convolutional neural network model contains a sensing data branch and a beam signal branch. Within the beam signal branch, a four-dimensional data cube is constructed for the target analysis beam signal. The four-dimensional data cube consists of the independent beam channel, independent beam distance, independent beam position, and echo time corresponding to the target analysis beam signal. By performing spatiotemporal dimension convolution sliding on the four-dimensional data cube using convolution kernels within the beam signal branch, the spatiotemporal characteristics of the target analysis beam signal are captured. Global average pooling is then performed on the target analysis beam signal to obtain global statistical information of the target analysis beam signal. By introducing a big data network, attention weights of different spectral features of the beam signal are retrieved, and attention weights are allocated to the global statistical information of the target analysis beam signal to generate spectral features related to the breathing and heartbeat of the target smart bed user, which are then labeled as the user target spectral features. Perform periodic convolution analysis on the user target spectral features and output the periodic characteristics of the user target spectral features; By combining the sensor data branch, the updated real-time target 3D model is analyzed using sensor data. Combined with the user target spectral characteristics and corresponding periodic characteristics, the real-time positioning of the user target orientation of the smart bed is achieved. Based on the user's real-time orientation and position on the target smart bed, the angle of the user's chest cavity tracking beam and the user's position positioning beam is adaptively adjusted, and the periodic characteristics of the user's target spectral features are adaptively updated after the adaptive adjustment to generate updated user spectral features. The number of users is determined on the target smart bed, and adaptive filtering is applied to the user chest cavity tracking beam based on the number of users. Specifically: Analyze the number of users on the target smart bed. If the number of users is 1, directly output the user update spectrum characteristics. If the number of users is greater than 1, the distance between different users' chest cavities is calculated based on the real-time spatial coordinates of the chest cavity, and a minimum value for the distance between chest cavities is preset. If the interthoracic distance between different users is not less than the minimum interthoracic distance, then directly output the updated spectral features of the users; If the interthoracic distance between different users is less than the minimum interthoracic distance, the user update spectrum features are analyzed, the respiratory rate is extracted, and an adaptive filtering algorithm is used to filter the interference components of the chest tracking beams of different users to obtain purified chest tracking beams of different users. By combining the user location positioning beam with the purified different user chest cavity tracking beams, the user update spectrum features are updated a second time, and together with the directly output user update spectrum features, they are collectively referred to as the user available spectrum features. The system analyzes the available spectrum characteristics of users, performs real-time sleep quality analysis on all users of the target smart bed, and generates sleep quality reports for different users based on the analysis results.

2. The multi-user sleep monitoring method based on millimeter-wave radar beam linkage analysis as described in claim 1, characterized in that, The method involves combining sensor data branches to perform sensor data analysis on the updated real-time target 3D model, and combining the user target spectral characteristics and corresponding periodic characteristics to achieve real-time positioning of the user target orientation of the smart bed. Specifically: Acquire the updated real-time target 3D model sensor data, which includes pressure data and temperature data; Import the sensor data of the real-time target 3D model into the sensor data branch. Among them, perform distribution analysis on the pressure data to determine the pressure distribution on the target smart bed and construct a two-dimensional pressure grayscale image. By performing real-time convolutional analysis of human contour features on a two-dimensional pressure grayscale image using convolutional kernels within the sensor data branch, the system identifies the user's real-time activity status on the target smart bed and analyzes temperature data to output a thermal distribution map of the user on the target smart bed. The thermal distribution map describes the correlation between human body temperature and ambient temperature. Combined with the user's real-time activity status on the target smart bed, it determines the user's real-time orientation and position on the target smart bed.

3. The multi-user sleep monitoring method based on millimeter-wave radar beam linkage analysis as described in claim 2, characterized in that, The process involves adaptively adjusting the angles of the user's chest cavity tracking beam and the user's location positioning beam, and then adaptively updating the periodic characteristics of the user's target spectral features to generate updated user spectral features. Specifically: Based on the user's real-time orientation and location on the target smart bed, a user location-orientation joint analysis model is constructed and named the real-time user sleep state model. The user's sleep state model is updated in real time by using the user's location positioning beam. Specifically, the user's chest cavity location is searched by the user's location positioning beam, and the spatial coordinates of the user's chest cavity location are determined in real time and calibrated as the real-time spatial coordinates of the chest cavity. Among them, the user's chest cavity location is searched by the user location positioning beam: when the user's sleep state model changes in real time, that is, when the user's real-time orientation and position on the target smart bed change, the user location positioning beam adaptively adjusts the angle to achieve secondary determination of the user's real-time orientation and position on the target smart bed. For the user's chest cavity tracking beam, the user's chest cavity tracking beam is controlled to perform real-time positioning and adjustment based on the real-time spatial coordinates of the chest cavity; Among them, the redundant positioning range of the real-time spatial coordinates of the chest cavity is predetermined, and a ratio table of the redundant positioning range of the real-time spatial coordinates of the chest cavity and the signal strength of the user's chest cavity tracking beam is preset and calibrated as a positioning-signal strength ratio table. Among them, the redundant positioning range of the real-time spatial coordinates of the chest cavity is negatively correlated with the signal strength of the user's chest cavity tracking beam. Based on the aforementioned positioning-signal strength ratio table, the beamwidth of the user's chest cavity tracking beam is adjusted in a timely manner. During the adjustment of the user chest cavity tracking beam and the user location positioning beam, the periodic characteristics of the user target spectrum feature are adaptively updated. The adaptive update firstly maintains the continuity of beam switching and collects the user target spectrum feature during the time period when there is no user body movement. The system performs periodic heartbeat frequency drift analysis on the collected user target spectral features and determines the heartbeat frequency drift window based on the drift range, so as to maintain the periodic heartbeat frequency drift of users within a window at a predetermined value. By combining the heartbeat frequency drift window, environmental compensation is performed on the collected user target spectrum features to obtain the adaptively updated user target spectrum features, which are then labeled as user updated spectrum features.

4. The multi-user sleep monitoring method based on millimeter-wave radar beam linkage analysis as described in claim 1, characterized in that, The analysis utilizes available spectral characteristics to perform real-time sleep quality analysis on all users of the target smart bed, and generates sleep quality reports for different users based on the analysis results. Specifically: The available spectrum characteristics of users are analyzed to extract physiological data of all users on the target smart bed. The physiological data includes the user's respiratory rate and heart rate. The main controller creates identity IDs for different users of the target smart bed and matches the physiological data of the users corresponding to the identity IDs. A physiological data comparison and analysis knowledge graph is introduced, which records the sleep stages corresponding to different physiological data, as well as the sleep quality corresponding to the duration of different sleep stages. Import all users' physiological data into the physiological data comparison and analysis knowledge graph, output the current sleep stage of different users, match it with their identity ID, and analyze the duration of different sleep stages to obtain the sleep quality of different users; The system matches the sleep quality of different users with their corresponding identity IDs and outputs sleep quality reports for each user in the main controller.

Images