Radar non-inductive respiration monitoring system and method based on passive folded air bag

Abstract

The invention provides a radar non-inductive respiration monitoring system and method based on a passive wrinkled airbag, and the system comprises a sleep monitoring pillow and a respiration recognition and abnormity monitoring subsystem, the respiration recognition and abnormity monitoring subsystem comprises a signal monitoring and coupling module which processes original monitoring data, and transmits the processed data to the sleep monitoring pillow; a human body respiration direct measurement signal and an air bag coupling amplification signal are obtained; the state recognition and inversion module is used for carrying out in-place state recognition, sleeping posture recognition and airbag state inversion on the coupled double-domain fusion respiration candidate data to obtain state constraint data; the respiration reconstruction and feature extraction module is used for performing screening and feature extraction on the double-domain fusion respiration candidate data to obtain respiration feature data; and the abnormity identification and confirmation module is used for identifying the apnea candidate event and the abnormal event, obtaining an abnormity judgment result and carrying out graded alarm pushing. According to the invention, through the non-inductive pillow type structure and radar-air bag double-domain fusion perception, breathing abnormity intelligent monitoring with high robustness and low false alarm rate is realized.

Description

Technical Field

[0001] This invention relates to the field of respiratory monitoring technology, and in particular to a radar-based non-intrusive respiratory monitoring system and method based on a passive pleated airbag. Background Technology

[0002] According to the 2025 China Elderly Health Monitoring Report, there are over 120 million elderly people aged 60 and above living alone in my country, 65% of whom suffer from chronic respiratory diseases. The incidence of weak breathing, shortness of breath, and sudden respiratory arrest during sleep at night reaches 31.2%. Family members and caregivers cannot be on duty 24 hours a day, and the golden rescue time (4-6 minutes) is easily missed. Existing solutions are either medical wearable devices, which are uncomfortable and have poor compliance, or ordinary pillow-based monitoring devices, which have insufficient accuracy in capturing weak respiratory signals and require frequent maintenance and replacement, which cannot meet the needs of long-term home use and large-scale use in institutions. Therefore, focusing on the core needs of elderly people living alone, people undergoing home rehabilitation for COPD and asthma, and elderly care institutions, existing respiratory monitoring technologies have the following problems: 1) Traditional respiratory monitoring devices, such as chest straps and nasal cannulas, affect sleep comfort; single sensors, such as radar or pressure sensors alone, have unstable signals when there is body movement or changes in sleeping posture; 2) They cannot identify whether the user is in bed or in a sleeping position, leading to misjudgments of respiratory abnormalities; 3) Respiratory abnormality identification relies on manual methods or simple thresholds, resulting in a high false alarm rate and a lack of quantitative analysis of respiratory waveform details. Summary of the Invention

[0003] To overcome the shortcomings of the prior art, the purpose of this invention is to provide a radar-based non-intrusive respiratory monitoring system and method based on a passive pleated airbag. This system can achieve intelligent monitoring of respiratory abnormalities with high robustness and low false alarm rate through non-intrusive pillow-type structure and radar-airbag dual-domain fusion perception.

[0004] To achieve the above objectives, the present invention provides the following solution: a radar-based non-intrusive respiratory monitoring system based on a passive pleated airbag, comprising a sleep monitoring pillow and a respiratory recognition and anomaly monitoring subsystem connected to the sleep monitoring pillow, wherein the respiratory recognition and anomaly monitoring subsystem includes: The signal monitoring and coupling module is used to acquire raw monitoring data through the sleep monitoring pillow, and based on the raw monitoring data, to perform static background elimination, micro-motion signal enhancement and regional signal layering analysis to obtain the human respiratory direct measurement signal and the airbag coupling amplification signal. The state recognition and inversion module is used to couple the body breathing direct measurement signal and the airbag coupled amplification signal to obtain dual-domain fused breathing candidate data. Based on the dual-domain fused breathing candidate data, in-situ state recognition, sleeping posture recognition and airbag state inversion are performed to obtain state constraint data. The respiratory reconstruction and feature extraction module is used to filter and extract features from the dual-domain fused respiratory candidate data based on the state constraint data to obtain respiratory feature data. The anomaly identification and confirmation module is used to combine the state constraint data and the respiratory feature data to identify candidate events and abnormal events of sleep apnea, obtain abnormal candidate data, perform secondary confirmation on the abnormal candidate data to obtain an anomaly judgment result, and then push graded alarms based on the anomaly judgment result. The signal monitoring and coupling module, the state recognition and inversion module, the respiratory reconstruction and feature extraction module, and the anomaly recognition and confirmation module are interconnected.

[0005] Optionally, the sleep monitoring pillow includes a pillow core, a first central groove disposed on the upper surface of the pillow core, a second central groove disposed in the upper part of the first central groove, a double-layer passive quilted transverse pleated airbag disposed on the first central groove, and a millimeter-wave radar monitoring module disposed on the second central groove.

[0006] Optionally, the double-layer passive simulated quilted transverse pleated airbag includes an inner airtight layer, an outer protective layer disposed on the outer surface of the inner airtight layer, and an air chamber disposed inside the inner airtight layer. The inner airtight layer is a 0.12mm high-barrier polyester TPU film, and the outer protective layer is a 0.08mm TPU composite 40D nylon knitted fabric. The inner airtight layer and the outer protective layer are welded together by high-frequency heat sealing. The air chamber is filled with 99.9% dry nitrogen.

[0007] Optionally, the double-layer passive quilted transverse pleated airbag divides the air chamber into 12 independent deformation chambers through 3 horizontal and 2 vertical continuous heat-sealing ribs. The width of the continuous heat-sealing ribs is 3mm and the welding depth is 0.15mm.

[0008] Optionally, the upper surface of the double-layer passive quilted transverse pleated airbag is uniformly provided with 18 transverse accordion pleats, the spacing of the transverse accordion pleats is 31.1mm and the pleat depth is 2mm.

[0009] Optionally, the signal monitoring and coupling module includes: The original monitoring unit is used to transmit signals using a millimeter-wave radar monitoring module and receive echo signals transmitted from the human head and neck and the surface of the airbag. The transmitted signal and the echo signal are mixed to obtain an intermediate frequency signal. The intermediate frequency signal is subjected to a fast Fourier transform to obtain a range dimension signal. Then, the range dimension signal is subjected to phase information extraction and micro-displacement signal conversion to obtain the original monitoring data. The static background unit is used to construct a baseline background model using echo signals continuously collected by the millimeter-wave radar monitoring module in the absence of human targets, so as to perform static background elimination. The micro-motion enhancement unit is used to divide the echo signal into airbag region signal, human head and neck region signal and environmental region signal according to the distance distribution based on the original monitoring data and the baseline background model. Based on the signals of each region, phase unfolding and micro-displacement transformation are performed to complete the micro-motion signal enhancement and obtain the micro-motion sequence. The hierarchical analysis unit is used to define the micro-motion sequence of the human head and neck region as direct human micro-motion data, define the micro-motion sequence of the air sac region as air sac deformation coupling data, and remove body motion interference to obtain the direct human respiratory signal and the air sac coupled amplified signal.

[0010] Optionally, the state recognition and inversion module includes: The signal fusion unit is used to perform time synchronization, constraint within the respiratory dominance frequency range, correlation calculation, and amplitude normalization processing based on the direct respiratory measurement signal and the airbag coupled amplified signal, and then perform signal fusion to obtain dual-domain fused respiratory candidate data; wherein, the correlation calculation includes waveform period consistency, dominance frequency consistency, phase coupling degree, and envelope similarity; The in-situ state identification unit is used to determine, based on the dual-domain fused respiratory candidate data, whether there is stable echo energy in the head and neck region, whether there is a continuous pressure deformation signal in the air sac region, and whether the fused respiratory candidate waveform is continuous, to obtain the judgment result, and then obtain the in-situ state identification result based on the judgment result. The sleeping posture recognition unit is used to extract the energy distribution difference between the left and right sides, the coupling ratio between the central area and the edge area, the offset of the head and neck echo center, and the asymmetry of respiratory wave conduction based on the dual-domain fused respiratory candidate data to obtain sleeping posture features. According to the sleeping posture features, the posture is divided into supine, left lateral, right lateral, prone mixed position and head offset position to obtain the sleeping posture state recognition result. The airbag state inversion unit is used to compare the direct respiratory measurement signal and the airbag coupled amplified signal to calculate the signal amplitude ratio, average phase difference and airbag signal fluctuation degree to obtain inversion state parameters, and then determine the airbag state based on the inversion state parameters; the airbag state includes normal state, under-pressure state, collapsed or folded state and contact offset state. The state fusion unit is used to integrate the in-place state recognition results, sleeping posture state recognition results, and airbag state to obtain state constraint data.

[0011] Optionally, the respiratory reconstruction and feature extraction module includes: The respiratory reconstruction unit is used to remove signals from the bed-out stage and strong body movement stage in the dual-domain fused respiratory candidate data according to the state constraint data to obtain effective respiratory segments, perform peak and valley detection on the effective respiratory segments to extract the inspiratory peak, expiratory valley and cycle length, and use linear interpolation to constrain and compensate for short-term missing segments in the effective respiratory segments to complete the respiratory cycle reconstruction and form continuous respiratory characterization data. The respiratory feature unit is used to extract basic physiological features, stability features, anomaly sensitivity features, and coupling quality features based on the continuous respiratory characterization data to obtain respiratory feature data. The basic physiological features include respiratory rate, respiratory cycle, waveform amplitude, inspiratory-to-expiratory time ratio, and difference between adjacent respiratory cycle events. The stability features include the difference between adjacent respiratory cycles, the overall fluctuation of respiratory amplitude, the continuity of respiratory rhythm, and the proportion of effective signal. The anomaly sensitivity features include the degree of gradual decrease in respiratory amplitude, the intensity of restart before and after a pause, the duration of hypoventilation, the degree of irregular breathing, and the recovery time after an anomaly. The coupling quality features include the correlation calculation, airway conduction gain score, current sleep position matching score, and current segment monitoring confidence score.

[0012] Optionally, the anomaly identification and confirmation module includes: The apnea identification unit is used to combine the state constraint data and the respiratory feature data to determine apnea candidate events when the respiratory amplitude is lower than the baseline, the apnea duration exceeds the apnea time threshold, the dual-domain fused respiratory candidate data shows synchronous signal attenuation, or a non-bed-out state and non-body movement state are detected. Other anomaly identification units are used to combine the state constraint data and the respiratory feature data to determine a hypoventilation abnormal event when the respiratory amplitude continues to decrease relative to the baseline and the cycle still exists; to determine a respiratory rhythm abnormal event when the cycle is irregular and the respiratory rate is abnormal; and to determine a device abnormal event when the airbag status is under-pressure or deviated. An anomaly confirmation unit is used to integrate the candidate sleep apnea events and various abnormal events into anomaly candidate data, and to review and confirm the anomaly candidate data in a short time window and a long time window respectively, to obtain an anomaly determination result. An anomaly alarm unit is used to classify the anomaly determination results into different anomaly levels, generate different alarm information according to different anomaly levels, and push the alarm information to mobile terminals or computer terminals in real time.

[0013] This invention also provides a radar-based non-intrusive respiratory monitoring method based on a passive pleated airbag, comprising: Raw monitoring data is obtained through a sleep monitoring pillow. Based on the raw monitoring data, static background elimination, micro-motion signal enhancement, and regional signal layering analysis are performed to obtain direct human breathing signals and airbag coupled amplified signals. The direct respiratory measurement signal of the body and the coupled amplified signal of the airbag are coupled to obtain dual-domain fused respiratory candidate data. Based on the dual-domain fused respiratory candidate data, in-situ state recognition, sleeping posture recognition and airbag state inversion are performed to obtain state constraint data. Based on the state constraint data, the dual-domain fused respiratory candidate data is filtered and features are extracted to obtain respiratory feature data; By combining the state constraint data and the respiratory feature data, candidate events and abnormal events of sleep apnea are identified, abnormal candidate data is obtained, the abnormal candidate data is reconfirmed to obtain the abnormal judgment result, and then a graded alarm is pushed according to the abnormal judgment result.

[0014] This invention discloses the following technical advantages by providing a radar-based non-intrusive respiratory monitoring system and method based on a passive pleated airbag: 1. By designing a sleep monitoring pillow, the airbag structure achieves a high sensitivity response to weak respiratory vibrations. The independent deformation chamber and accordion pleat design enhance local deformation capabilities and improve signal coupling efficiency. The full-length horizontal pleats are perpendicular to the cervical spine, linearly amplifying weak respiratory signals by 5-10 times, solving the core pain point of inaccurate and missed detections in civilian radar monitoring of weak respiratory signals. The passive design (no external air source required) reduces system complexity and energy consumption, serving as the physical sensing foundation of the entire system, improving signal acquisition quality, and solving the interference problem to users caused by traditional respiratory monitoring equipment.

[0015] 2. By coordinating millimeter-wave radar and airbags to acquire direct human signals and airbag-coupled amplified signals, weak respiratory signals are extracted using techniques such as static background removal, micro-motion enhancement, and layered analysis. This improves the signal-to-noise ratio of respiratory signals, especially the signal stability under body movement interference. Furthermore, the dual-signal acquisition provides redundant and complementary information for subsequent fusion and state recognition, thereby enhancing the identifiability and reliability of respiratory signals.

[0016] 3. By fusing dual-domain signals, jointly identifying and retrieving in-situ state, sleeping posture, and airbag status, as well as multi-dimensional correlation indicators, the ability to adapt to different sleeping postures, body movements, and abnormal airbag status can be improved. Furthermore, state constraint data provides contextual information for subsequent respiratory reconstruction, reducing misjudgments and improving the perception of the environment and user status.

[0017] 4. By reconstructing the respiratory cycle and extracting four categories of features—basic physiology, stability, abnormal sensitivity, and coupling quality—it is possible to achieve continuous and complete representation of respiratory signals, enrich the dimensions of feature extraction, support the accuracy of subsequent anomaly identification, and provide structured input for anomaly identification.

[0018] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments 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 drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a schematic diagram of the system architecture provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the pillow core structure provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of the sleep monitoring pillow structure provided in an embodiment of the present invention; Figure 4 This is a schematic diagram of the respiratory recognition and abnormality monitoring subsystem architecture provided in an embodiment of the present invention; Figure 5 This is a schematic diagram of the method flow provided in an embodiment of the present invention; Explanation of reference numerals in the attached diagram: 1. Pillow core; 2. First central groove; 3. Second central groove; 4. Double-layer passive simulated quilted transverse pleated airbag; 5. Continuous heat-sealed rib; 6. Independent deformation chamber. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0023] like Figure 1 As shown, the present invention provides a radar-based non-intrusive respiratory monitoring system based on a passive pleated airbag, including a sleep monitoring pillow and a respiratory recognition and abnormality monitoring subsystem connected to the sleep monitoring pillow.

[0024] I. Sleep Monitoring Pillow like Figure 2 , Figure 3 As shown, the sleep monitoring pillow includes a pillow core 1, a first central groove 2 disposed on the upper surface of the pillow core 1, a second central groove 3 disposed in the upper part of the first central groove 2, a double-layer passive quilted transverse pleated airbag 4 disposed on the first central groove 2, and a millimeter-wave radar monitoring module disposed on the second central groove 3.

[0025] The double-layer passive simulated quilted transverse pleated airbag 4 includes an inner airtight layer, an outer protective layer disposed on the outer surface of the inner airtight layer, and an air chamber disposed inside the inner airtight layer. A 15mm wide non-inflatable sealing margin is reserved around the airbag. The Velcro surface of the bottom of the airbag (the margin area) is omnidirectionally fitted and fixed to the Velcro hook surface pasted in the groove of the pillow core 1, without needles, threads, or glue, and is detachable and does not shift.

[0026] The inner airtight layer is made of a 0.12mm high-barrier polyester TPU film to ensure an extremely low nitrogen leakage rate; the outer protective layer is made of 0.08mm TPU composite 40D nylon knitted fabric, which is tear-resistant and puncture-resistant, while not restricting wrinkle deformation; the inner airtight layer and the outer protective layer are welded together by high-frequency heat sealing, and the air chamber is filled with 99.9% dry nitrogen gas at a pre-filling pressure of 10kPa.

[0027] The double-layer passive quilted transverse pleated airbag 4 divides the air chamber into 12 independent deformation chambers 6 through 3 horizontal and 2 vertical continuous heat-sealing ribs 5, which disperses the pressure stress and prevents air leakage. The width of the continuous heat-sealing rib 5 is 3mm and the welding depth is 0.15mm.

[0028] The upper surface of the double-layer passive quilted transverse pleated airbag 4 is uniformly provided with 18 transverse accordion pleats, the spacing of which is 31.1 mm and the depth of which is 2 mm. The pleats are parallel to the long side of the pillow (left-right direction) and perpendicular to the direction of the human cervical spine (front-back direction), which can linearly amplify the micro-movement of breathing (0.05-0.2 mm) by 5-10 times.

[0029] II. Figure 4 As shown, the respiratory recognition and abnormality monitoring subsystem includes: 1. A signal monitoring and coupling module, used to acquire raw monitoring data through a sleep monitoring pillow, and based on the raw monitoring data, perform static background elimination, micro-motion signal enhancement, and regional signal layering analysis to obtain direct human respiratory measurement signals and airbag coupled amplified signals. The signal monitoring and coupling module includes: 1.1 The original monitoring unit is used to transmit signals using a millimeter-wave radar monitoring module and receive echo signals transmitted from the human head and neck and the surface of the airbag. The transmitted signal and the echo signal are mixed to obtain an intermediate frequency signal. The intermediate frequency signal is subjected to a fast Fourier transform to obtain a range-dimensional signal. The range-dimensional signal is then subjected to phase information extraction and micro-displacement signal conversion to obtain the original monitoring data. The original monitoring data is a range-time matrix, which represents the change of echo intensity at different distance positions over time.

[0030] 1.2 Static background unit, used to eliminate static background even when there is no human target, as stationary objects such as beds, walls, and tables will reflect radar waves. The baseline background model is constructed by continuously collecting echo signals from the millimeter-wave radar monitoring module.

[0031] 1.3 Micro-motion enhancement unit, used to divide the echo signal into airbag region signal, human head and neck region signal and environmental region signal according to the distance distribution based on the original monitoring data and the baseline background model.

[0032] Based on the signals from each region, phase unrolling and micro-displacement transformation are performed to enhance the micro-motion signal and obtain the micro-motion sequence. Radar can not only measure distance but also detect extremely small movements of objects at the sub-millimeter level. By calculating the changes in the echo phase, the waveform of the object's vibration over time at each distance point is obtained, i.e., the micro-motion sequence.

[0033] The 1.4 hierarchical analysis unit is used to define the micro-motion sequence of the human head and neck region as direct human micro-motion data, define the micro-motion sequence of the air sac region as air sac deformation coupling data, and remove body motion interference to obtain the direct human respiratory signal and the air sac coupled amplified signal.

[0034] 2. A state recognition and inversion module, used to couple the direct respiratory measurement signal and the airbag coupled amplified signal to obtain dual-domain fused respiratory candidate data, and perform in-situ state recognition, sleeping posture recognition, and airbag state inversion based on the dual-domain fused respiratory candidate data to obtain state constraint data. The state recognition and inversion module includes: 2.1 Signal fusion unit, used to perform time synchronization, constraint within the respiratory dominance frequency range, correlation calculation and amplitude normalization processing based on the direct respiratory measurement signal and the airbag coupled amplified signal, and then perform signal fusion to obtain dual-domain fused respiratory candidate data; wherein, the correlation calculation includes waveform period consistency, dominance frequency consistency, phase coupling degree and envelope similarity.

[0035] 2.2 In-situ state identification unit, based on the dual-domain fused respiratory candidate data: Determining whether there is stable echo energy in the head and neck region can indicate the presence of an object in that region. Determining whether there is a continuous pressure deformation signal in the airbag area can indicate that there is weight pressing on the pillow; Determining whether the candidate waveforms of fused breathing are continuous can indicate regular breathing. Based on the above judgment, the judgment result is obtained, and then the presence status recognition result is obtained based on the judgment result; if all three conditions are negative, it is determined that the person has left the bed.

[0036] 2.3 Sleep posture recognition unit, used to extract the energy distribution difference between the left and right sides, the coupling ratio between the central area and the edge area, the offset of the head and neck echo center and the asymmetry of respiratory wave conduction based on the dual-domain fused respiratory candidate data, to obtain sleep posture features, and to divide the posture into supine, left lateral, right lateral, prone mixed position and head offset position according to the sleep posture features, to obtain the sleep posture state recognition result.

[0037] Difference in energy distribution between the left and right sides: If there is more energy on the left side and less on the right side, it may be due to lying on the right side with the body turned to one side. Coupling ratio between central and peripheral regions, and offset of head and neck echo center: whether the position of the head and neck detected by radar is to the left or right relative to the center of the pillow; Asymmetry in respiratory wave conduction: Whether the deformation amplitude on the left and right sides of the air sac is consistent.

[0038] 2.4 The airbag state inversion unit is used to compare the direct respiratory measurement signal and the airbag coupled amplified signal to calculate the signal amplitude ratio, average phase difference and airbag signal fluctuation degree to obtain inversion state parameters, and then determine the airbag state based on the inversion state parameters; the airbag state includes normal state, underpressure state, collapsed or folded state and contact offset state.

[0039] 2.5 State fusion unit, used to integrate the in-place state recognition results, sleeping posture state recognition results and airbag state to obtain state constraint data.

[0040] 3. A respiratory reconstruction and feature extraction module, used to filter and extract features from the dual-domain fused respiratory candidate data based on the state constraint data to obtain respiratory feature data. The respiratory reconstruction and feature extraction module includes: 3.1 The respiratory reconstruction unit is used to remove signals from the bed-out stage and strong body movement stage in the dual-domain fused respiratory candidate data according to the state constraint data to obtain effective respiratory segments, perform peak and valley detection on the effective respiratory segments to extract the inspiratory peak, expiratory valley and cycle length, and use linear interpolation to constrain and compensate for short-term missing segments in the effective respiratory segments to complete the respiratory cycle reconstruction and form continuous respiratory characterization data.

[0041] Constraint compensation for short-term missing segments: 1) If the missing time is short and the waveforms before and after are stable, use a local model to complete it: If the signal is poor for a short time, such as 2-3 seconds, and the breathing before and after is very regular, use interpolation or a simple prediction model to fill in the gap and maintain the continuity of the waveform. For example, if the interval between the previous inhalation and the next inhalation is 4 seconds, and there is a 1-second gap in the middle, it can be filled in according to a sine curve.

[0042] 2) If the missing time is long, leave the segment open and do not force a reconstruction: If the missing time is very long, such as more than 10 seconds, it may indicate that there is real apnea or that the person has left the bed. Do not fabricate a breathing waveform. In this case, leave the segment open, mark it as invalid, and do not force a connection.

[0043] 3.2 Respiratory feature unit, used to extract basic physiological features, stability features, abnormal sensitivity features and coupling quality features based on the continuous respiratory characterization data to obtain respiratory feature data.

[0044] 1) Basic physiological characteristics include: Respiratory rate: How many times per minute an adult breathes. A normal adult at rest breathes 12-20 times per minute. This is used to determine whether breathing is too fast or too slow. Respiratory cycle: the time of one complete inhalation plus exhalation, the basis for calculating frequency; Waveform amplitude: The depth of each breath, representing the signal strength, used to determine whether the breath is deep or shallow; Inspiratory-to-expiratory time ratio: Normal is about 1:1.5 to 1:2, used to determine whether there is airway obstruction; Difference between adjacent respiratory cycle events: The degree of time difference between adjacent respiratory cycles, used to determine whether breathing is regular.

[0045] 2) Stability characteristics include: The difference between the current respiratory cycle and the previous one is used to determine whether breathing has suddenly become faster or slower. Overall fluctuation in respiratory amplitude: The overall fluctuation in respiratory amplitude is used to determine whether the depth of breathing fluctuates greatly. Continuity of respiratory rhythm: Whether the respiratory waveform remains regular for a long time, used to determine whether there are frequent interruptions; Effective signal percentage: The proportion of clean breathing time to total time, used to assess monitoring quality. A low percentage indicates frequent turning or poor air bag contact.

[0046] 3) Abnormally sensitive characteristics include: The degree of gradual decrease in respiratory amplitude: whether breathing becomes increasingly shallow, such as a precursor to central sleep apnea, used to provide early warning of sleep apnea; Intensity of restart before and after pause: The depth of the first breath after the pause ends, used to determine whether the recovery is normal; Duration of hypoventilation: The length of time during which breathing becomes shallow but does not completely stop, used to determine the severity of a hypoventilation event; Irregular breathing degree: The degree of disorder in breathing rhythm, such as Cheyne-Stokes respiration, which is used to help identify diseases such as heart failure; Post-abnormality recovery time: The time from the end of the abnormal event to the return of normal breathing, used to assess recovery capacity.

[0047] 4) Coupling quality characteristics include: Correlation calculation: The degree of agreement between radar direct measurement signal and airbag signal to determine whether the two signal sources corroborate each other; Airbag conduction gain score: whether the amplification effect of the airbag on breathing is normal, used to determine whether the airbag is aging or leaking; Current sleeping posture matching score: Whether the current waveform matches the sleeping posture template, used to determine whether the sleeping posture recognition is correct; Current segment monitoring confidence score: comprehensively judges whether this segment of data is reliable, and is used for subsequent alarm decisions.

[0048] 4. An anomaly identification and confirmation module, used to combine the state constraint data and the respiratory feature data to identify candidate events and abnormal events of sleep apnea, obtain candidate abnormal data, perform secondary confirmation on the candidate abnormal data to obtain an anomaly determination result, and then push tiered alarms based on the anomaly determination result. The anomaly identification and confirmation module includes: 4.1 The apnea identification unit is used to combine the state constraint data and the respiratory feature data to determine the apnea candidate event when the respiratory amplitude is lower than the baseline, the apnea duration exceeds the apnea time threshold, the dual-domain fused respiratory candidate data shows synchronous signal attenuation, or the non-bed-out state and non-body movement state are detected.

[0049] 4.2 Other anomaly identification units are used to combine the state constraint data and the respiratory feature data to determine a hypoventilation abnormal event when the respiratory amplitude continues to decrease relative to the baseline and the cycle still exists; to determine a respiratory rhythm abnormal event when the cycle is irregular and the respiratory rate is abnormal; and to determine a device abnormal event when the airbag status is under-pressure or deviated.

[0050] Hypopnea is an abnormal event characterized by shallow breathing, such as only half the normal level, even though breathing has not completely stopped. The breathing cycle still exists, with inhalation and exhalation continuing, albeit very shallowly. This condition is called hypoventilation, a manifestation of sleep apnea syndrome, but milder than complete apnea.

[0051] Abnormal respiratory rhythm events: breathing is sometimes fast and sometimes slow, irregular, with an irregular cycle, and abnormal breathing frequency (>30 breaths / min) or (<8 breaths / min). All of the above are rhythm abnormalities and may indicate heart failure, central sleep apnea, or other diseases.

[0052] 4.3 Anomaly Confirmation Unit, used to integrate the candidate sleep apnea events and various abnormal events into anomaly candidate data, and to review and confirm the anomaly candidate data within a short time window and a long time window respectively, to obtain anomaly judgment results.

[0053] 1) Short window review: Candidate events are re-verified within a short time window to determine whether they were caused by transient disturbances such as turning over, coughing, talking, or localized compression.

[0054] When a candidate abnormal event occurs, such as a candidate for apnea lasting 10 seconds, an alarm is not immediately triggered. Instead, a short time window is observed, such as 5 seconds before and after, for a total of 10 seconds. This window is checked for transient disturbances such as turning over, coughing, talking, or pressing a hand on a pillow. If a disturbance is found, the candidate event is rejected, and no confirmed abnormality is output, significantly reducing false alarms caused by actions such as turning over.

[0055] 2) Long window confirmation: If a candidate event occurs consecutively or repeatedly within a long time window, its anomaly level is increased.

[0056] If the same candidate event occurs repeatedly, such as five apnea candidates within 10 minutes, or persists for a long period, such as a single instance of hypopnea lasting two minutes, it indicates that it is not a random disturbance. The abnormality level should be escalated, for example, from a Level 1 alert to a Level 2 warning or Level 3 alarm. Genuine pathological events should receive higher priority alerts.

[0057] 4.4 Anomaly alarm unit, used to classify the anomaly judgment result into anomaly levels, generate different alarm information according to different anomaly levels, and push the alarm information to mobile terminal or computer terminal in real time.

[0058] For example, abnormal confirmation results can be categorized as follows: Level 1 alert: Minor abnormalities or events requiring observation; such as occasional brief episodes of hypoventilation, which are only recorded in the app and do not disturb the user. Level 2 warning: Clearly abnormal but with moderate risk; if the respiratory rate remains high, a notification will be sent, but no audible or visual alarm will be triggered. Level 3 alarm: High risk of apnea or prolonged absence from bed; if breathing apnea lasts more than 15 seconds, immediately activate the local buzzer, send an app notification, or make a phone call; Level 4 maintenance prompt: If the equipment status is abnormal, such as undervoltage, offset, or persistent low confidence, the user is prompted to check the equipment.

[0059] like Figure 5 As shown, the present invention also provides a radar-based non-intrusive respiratory monitoring method based on a passive pleated airbag, comprising: Raw monitoring data is obtained through a sleep monitoring pillow. Based on the raw monitoring data, static background elimination, micro-motion signal enhancement, and regional signal layering analysis are performed to obtain direct human breathing signals and airbag coupled amplified signals. The direct respiratory measurement signal of the body and the coupled amplified signal of the airbag are coupled to obtain dual-domain fused respiratory candidate data. Based on the dual-domain fused respiratory candidate data, in-situ state recognition, sleeping posture recognition and airbag state inversion are performed to obtain state constraint data. Based on the state constraint data, the dual-domain fused respiratory candidate data is filtered and features are extracted to obtain respiratory feature data; By combining the state constraint data and the respiratory feature data, candidate events and abnormal events of sleep apnea are identified, abnormal candidate data is obtained, the abnormal candidate data is reconfirmed to obtain the abnormal judgment result, and then a graded alarm is pushed according to the abnormal judgment result.

[0060] Therefore, this invention provides a radar-based non-intrusive respiratory monitoring system and method based on a passive pleated airbag, which can achieve intelligent monitoring of respiratory abnormalities with high robustness and low false alarm rate through non-intrusive pillow structure and radar-airbag dual-domain fusion perception.

[0061] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0062] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. Furthermore, those skilled in the art will recognize that, based on the ideas of the present invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A radar-based non-intrusive respiratory monitoring system based on a passive pleated airbag, characterized in that, The system includes a sleep monitoring pillow and a breathing recognition and anomaly monitoring subsystem connected to the sleep monitoring pillow, wherein the breathing recognition and anomaly monitoring subsystem includes: The signal monitoring and coupling module is used to acquire raw monitoring data through the sleep monitoring pillow, and based on the raw monitoring data, to perform static background elimination, micro-motion signal enhancement and regional signal layering analysis to obtain the human respiratory direct measurement signal and the airbag coupling amplification signal. The state recognition and inversion module is used to couple the body breathing direct measurement signal and the airbag coupled amplification signal to obtain dual-domain fused breathing candidate data. Based on the dual-domain fused breathing candidate data, in-situ state recognition, sleeping posture recognition and airbag state inversion are performed to obtain state constraint data. The respiratory reconstruction and feature extraction module is used to filter and extract features from the dual-domain fused respiratory candidate data based on the state constraint data to obtain respiratory feature data. The anomaly identification and confirmation module is used to combine the state constraint data and the respiratory feature data to identify candidate events and abnormal events of sleep apnea, obtain abnormal candidate data, perform secondary confirmation on the abnormal candidate data to obtain an anomaly judgment result, and then push graded alarms based on the anomaly judgment result. The signal monitoring and coupling module, the state recognition and inversion module, the respiratory reconstruction and feature extraction module, and the anomaly recognition and confirmation module are interconnected.

2. The radar-based non-intrusive respiratory monitoring system based on a passive pleated airbag according to claim 1, characterized in that, The sleep monitoring pillow includes a pillow core, a first central groove on the upper surface of the pillow core, a second central groove in the upper part of the first central groove, a double-layer passive quilted transverse pleated airbag on the first central groove, and a millimeter-wave radar monitoring module on the second central groove.

3. The radar-based non-intrusive respiratory monitoring system based on a passive pleated airbag according to claim 2, characterized in that, The double-layer passive simulated quilted transverse pleated airbag includes an inner airtight layer, an outer protective layer disposed on the outer surface of the inner airtight layer, and an air chamber disposed inside the inner airtight layer. The inner airtight layer is a 0.12mm high-barrier polyester TPU film, and the outer protective layer is a 0.08mm TPU composite 40D nylon knitted fabric. The inner airtight layer and the outer protective layer are welded together by high-frequency heat sealing. The air chamber is filled with 99.9% dry nitrogen.

4. The radar-based non-intrusive respiratory monitoring system based on a passive pleated airbag according to claim 3, characterized in that, The double-layer passive quilted transverse pleated airbag is divided into 12 independent deformation chambers by 3 horizontal and 2 vertical continuous heat-sealing ribs. The width of the continuous heat-sealing ribs is 3mm and the welding depth is 0.15mm.

5. The radar-based non-intrusive respiratory monitoring system based on a passive pleated airbag according to claim 4, characterized in that, The upper surface of the double-layer passive quilted transverse pleated airbag is uniformly provided with 18 transverse accordion pleats, the spacing of the transverse accordion pleats is 31.1mm and the pleat depth is 2mm.

6. The radar-based non-intrusive respiratory monitoring system based on a passive pleated airbag according to claim 5, characterized in that, The signal monitoring and coupling module includes: The original monitoring unit is used to transmit signals using a millimeter-wave radar monitoring module and receive echo signals transmitted from the human head and neck and the surface of the airbag. The transmitted signal and the echo signal are mixed to obtain an intermediate frequency signal. The intermediate frequency signal is subjected to a fast Fourier transform to obtain a range dimension signal. Then, the range dimension signal is subjected to phase information extraction and micro-displacement signal conversion to obtain the original monitoring data. The static background unit is used to construct a baseline background model using echo signals continuously collected by the millimeter-wave radar monitoring module in the absence of human targets, so as to perform static background elimination. The micro-motion enhancement unit is used to divide the echo signal into airbag region signal, human head and neck region signal and environmental region signal according to the distance distribution based on the original monitoring data and the baseline background model. Based on the signals of each region, phase unfolding and micro-displacement transformation are performed to complete the micro-motion signal enhancement and obtain the micro-motion sequence. The hierarchical analysis unit is used to define the micro-motion sequence of the human head and neck region as direct human micro-motion data, define the micro-motion sequence of the air sac region as air sac deformation coupling data, and remove body motion interference to obtain the direct human respiratory signal and the air sac coupled amplified signal.

7. The radar-based non-intrusive respiratory monitoring system based on a passive pleated airbag according to claim 6, characterized in that, The state recognition and inversion module includes: The signal fusion unit is used to perform time synchronization, constraint within the respiratory dominance frequency range, correlation calculation, and amplitude normalization processing based on the direct respiratory measurement signal and the airbag coupled amplified signal, and then perform signal fusion to obtain dual-domain fused respiratory candidate data; wherein, the correlation calculation includes waveform period consistency, dominance frequency consistency, phase coupling degree, and envelope similarity; The in-situ state identification unit is used to determine, based on the dual-domain fused respiratory candidate data, whether there is stable echo energy in the head and neck region, whether there is a continuous pressure deformation signal in the air sac region, and whether the fused respiratory candidate waveform is continuous, to obtain the judgment result, and then obtain the in-situ state identification result based on the judgment result. The sleeping posture recognition unit is used to extract the energy distribution difference between the left and right sides, the coupling ratio between the central area and the edge area, the offset of the head and neck echo center, and the asymmetry of respiratory wave conduction based on the dual-domain fused respiratory candidate data to obtain sleeping posture features. According to the sleeping posture features, the posture is divided into supine, left lateral, right lateral, prone mixed position and head offset position to obtain the sleeping posture state recognition result. The airbag state inversion unit is used to compare the direct respiratory measurement signal and the airbag coupled amplified signal to calculate the signal amplitude ratio, average phase difference and airbag signal fluctuation degree to obtain inversion state parameters, and then determine the airbag state based on the inversion state parameters; the airbag state includes normal state, under-pressure state, collapsed or folded state and contact offset state. The state fusion unit is used to integrate the in-place state recognition results, sleeping posture state recognition results, and airbag state to obtain state constraint data.

8. The radar-based non-intrusive respiratory monitoring system based on a passive pleated airbag according to claim 7, characterized in that, The respiratory reconstruction and feature extraction module includes: The respiratory reconstruction unit is used to remove signals from the bed-out stage and strong body movement stage in the dual-domain fused respiratory candidate data according to the state constraint data to obtain effective respiratory segments, perform peak and valley detection on the effective respiratory segments to extract the inspiratory peak, expiratory valley and cycle length, and use linear interpolation to constrain and compensate for short-term missing segments in the effective respiratory segments to complete the respiratory cycle reconstruction and form continuous respiratory characterization data. The respiratory feature unit is used to extract basic physiological features, stability features, anomaly sensitivity features, and coupling quality features based on the continuous respiratory characterization data to obtain respiratory feature data. The basic physiological features include respiratory rate, respiratory cycle, waveform amplitude, inspiratory-to-expiratory time ratio, and difference between adjacent respiratory cycle events. The stability features include the difference between adjacent respiratory cycles, the overall fluctuation of respiratory amplitude, the continuity of respiratory rhythm, and the proportion of effective signal. The anomaly sensitivity features include the degree of gradual decrease in respiratory amplitude, the intensity of restart before and after a pause, the duration of hypoventilation, the degree of irregular breathing, and the recovery time after an anomaly. The coupling quality features include the correlation calculation, airway conduction gain score, current sleep position matching score, and current segment monitoring confidence score.

9. A radar-based non-intrusive respiratory monitoring system based on a passive pleated airbag according to claim 8, characterized in that, The anomaly identification and confirmation module includes: The apnea identification unit is used to combine the state constraint data and the respiratory feature data to determine apnea candidate events when the respiratory amplitude is lower than the baseline, the apnea duration exceeds the apnea time threshold, the dual-domain fused respiratory candidate data shows synchronous signal attenuation, or a non-bed-out state and non-body movement state are detected. Other anomaly identification units are used to combine the state constraint data and the respiratory feature data to determine a hypoventilation abnormal event when the respiratory amplitude continues to decrease relative to the baseline and the cycle still exists; to determine a respiratory rhythm abnormal event when the cycle is irregular and the respiratory rate is abnormal; and to determine a device abnormal event when the airbag status is under-pressure or deviated. An anomaly confirmation unit is used to integrate the candidate sleep apnea events and various abnormal events into anomaly candidate data, and to review and confirm the anomaly candidate data in a short time window and a long time window respectively, to obtain an anomaly determination result. An anomaly alarm unit is used to classify the anomaly determination results into different anomaly levels, generate different alarm information according to different anomaly levels, and push the alarm information to mobile terminals or computer terminals in real time.

10. A radar-based non-intrusive respiratory monitoring method based on a passive pleated airbag, characterized in that, include: Raw monitoring data is obtained through a sleep monitoring pillow. Based on the raw monitoring data, static background elimination, micro-motion signal enhancement, and regional signal layering analysis are performed to obtain direct human breathing signals and airbag coupled amplified signals. The direct respiratory measurement signal of the body and the coupled amplified signal of the airbag are coupled to obtain dual-domain fused respiratory candidate data. Based on the dual-domain fused respiratory candidate data, in-situ state recognition, sleeping posture recognition and airbag state inversion are performed to obtain state constraint data. Based on the state constraint data, the dual-domain fused respiratory candidate data is filtered and features are extracted to obtain respiratory feature data; By combining the state constraint data and the respiratory feature data, candidate events and abnormal events of sleep apnea are identified, abnormal candidate data is obtained, the abnormal candidate data is reconfirmed to obtain the abnormal judgment result, and then a graded alarm is pushed according to the abnormal judgment result.

Images