Non-invasive sleep monitoring and intervention system

Abstract

The invention discloses a non-invasive sleep monitoring and intervening system, which constructs a thoracic and abdominal ventilation displacement model through human body micro-motion phase calculation of millimeter wave radar, and realizes continuous and non-contact monitoring of respiratory rhythm, ventilation change and body motion state. The system further utilizes a time-frequency fusion feature and trend analysis method to identify and predict sleep disorder events such as snoring, apnea and hypopnea in real time, and generates acoustic disturbance signals meeting sound pressure and frequency constraints before and after the events are formed, so that the ventilation mode or sleeping posture of a user is promoted to be controllably and finely adjusted on the premise that the user is not awakened, and the sleep disorder events are prevented from being disturbed. Therefore, the abnormal event duration is shortened and the sleep quality is improved. The method is suitable for sleep monitoring equipment, bedside lamps, intelligent sound boxes or desktop type health equipment, and algorithms and intervention strategies can be further optimized through software upgrading.

Description

Technical Field

[0001] This invention relates to the field of sleep health monitoring and intelligent intervention technology, and in particular to a non-invasive sleep monitoring and intervention system. Background Technology

[0002] With the increasing popularity of home health monitoring devices, issues such as sleep apnea, snoring, and sleep apnea are receiving more attention from users. To capture breathing depth, ventilation stability, and changes in body movement, existing products can be broadly categorized into three types: wearable devices, mattress-based sensors, and monitoring devices based on audio or cameras.

[0003] However, these solutions all have their own unavoidable technical limitations in practical use.

[0004] Wearable devices rely on contact sensors to acquire information about chest and abdominal movements or heart rate, requiring high user compliance. They are prone to slipping or falling off at night, and signal stability cannot be guaranteed over long periods. Furthermore, wearable devices struggle to capture accurate sleep posture and ventilation changes without the user noticing. While mattress sensors can record pressure changes, they are significantly affected by mattress firmness, user posture, and the presence of multiple people sharing the bed. They cannot effectively separate the actual displacement of the chest and abdominal cavities, resulting in accuracy limitations in recognizing events such as sleep apnea and shallow breathing.

[0005] While monitoring methods based on audio or cameras achieve some non-contact monitoring, audio is highly susceptible to environmental noise interference, and cameras raise issues of user privacy and low-light imaging, making them unable to meet the requirements for stable monitoring around the clock. In real-world scenarios, user actions such as covering their eyes at night, turning over, or being covered by bedding can further degrade the signal quality of these solutions, leading to monitoring interruptions or misjudgments.

[0006] Regarding interventions for sleep disorders, most existing technologies rely on fixed-intensity audio-visual stimulation, which cannot be dynamically controlled based on event type, severity, or user sensitivity. This often results in situations where "intervention is too strong, leading to awakening" or "intervention is too weak, failing to change the state," lacking truly flexible and controllable intervention strategies. Furthermore, existing solutions mostly analyze raw data by uploading it to the cloud, posing privacy risks and depending on the network environment, making real-time judgment and response impossible locally.

[0007] Therefore, how to stably and continuously acquire micro-motion information related to chest and abdominal ventilation under completely non-contact conditions, how to establish a reliable identification and prediction mechanism before and after the event occurs, and how to generate a minimal intervention constraint intervention strategy that is both effective and does not disrupt sleep remain key issues that existing technologies have long failed to solve.

[0008] Therefore, existing technologies still need to be improved. Summary of the Invention

[0009] Given that the existing technologies mentioned above still have significant shortcomings in terms of inaccurate acquisition of ventilation micro-movements, unstable event recognition, rigid intervention methods that are easy to disturb users, and data processing that relies on the cloud, there is an urgent need for an overall solution that can stably reconstruct breathing and ventilation states under completely non-contact conditions, make timely judgments and implement flexible interventions before and after events occur, and complete privacy and security processing on the device side.

[0010] The technical solution of the present invention is as follows: This invention provides a non-invasive sleep monitoring and intervention system, comprising a continuous non-contact monitoring module for respiratory cycles, ventilation status, and body movement changes during sleep, a sleep disorder event determination module, and a flexible intervention module; wherein: The monitoring module models the ventilation-related displacements of the human chest and abdomen based on the phase micro-motion calculation of millimeter-wave radar to obtain respiratory rhythm, heart rate micro-motion, and body movement intensity. The event determination module performs time-frequency fusion, trend analysis, and pattern matching on the data output by the monitoring module to identify or predict sleep disorder events such as snoring, sleep apnea, and hypopnea. The flexible intervention module generates an acoustic disturbance signal that meets the sound pressure threshold, frequency range and minimum intervention constraint when the determination module is triggered, so as to cause the user's posture or breathing mode to be finely adjusted without waking the user. The system further includes a localized data processing module, which is used to complete feature extraction, event determination and intervention decision-making on the device side, and output desensitized data to the terminal device; thereby realizing closed-loop control of non-contact detection, real-time identification and subthreshold intervention for sleep abnormalities.

[0011] In one embodiment, the millimeter-wave radar uses an FMCW waveform, estimates the echo angle through a multi-antenna array, and constructs a thoracic-abdominal displacement curve using phase demodulation to reconstruct the user's respiratory depth, rhythm changes, and ventilation discontinuities.

[0012] In one embodiment, the time-frequency fusion includes short-time Fourier transform, phase difference demodulation features, body kinetic energy envelope, snoring vibration spectrum, and heart rate variability, and achieves stable feature extraction under low signal-to-noise ratio conditions through Kalman filtering or adaptive weighted filtering.

[0013] In one embodiment, the event determination module employs a trend-based sleep disorder event prediction model. This model establishes joint temporal judgment rules for prolonged respiratory cycles, decreased amplitude, and increased snoring energy, and triggers intervention before the event occurs to shorten the event duration.

[0014] In one embodiment, the acoustic disturbance signal generated by the flexible intervention module includes: Short pulse signals, pulse sequences, or frequency-modulated signals with frequencies in the range of 0.5kHz to 4kHz and sound pressure levels in the range of 20 to 30dB are used, and a minimum intervention strategy with progressively increasing intensity is adopted according to the severity of the event to ensure that the user is not awakened as much as possible.

[0015] In one embodiment, the localized data processing module only performs feature calculation and model inference, does not store the original radar echo data, original audio or image information, and only outputs the desensitized event statistics and sleep parameters, which are then uploaded to the mobile terminal or the cloud.

[0016] In one embodiment, the system supports multi-user differentiation, performs target demixing through the spatial distribution of millimeter-wave echoes, respiratory rate differences, and micro-motion trajectories, and independently constructs respiratory characteristic baselines and intervention parameter models for each user.

[0017] In one embodiment, the system further includes an alternative sensing module comprising at least one of an ultrasonic array, an infrared array, or a mattress pressure sensing layer, for maintaining continuity of respiratory and body movement monitoring when millimeter-wave echoes are blocked or fail.

[0018] In one embodiment, the system receives data from environmental sensors or a mattress micro-vibration module, and dynamically adjusts the event determination threshold and intervention intensity based on environmental noise, temperature and humidity, and sleep stage to improve the accuracy and stability of the intervention.

[0019] In one embodiment, the system adopts the structure of a bedside lamp, speaker, or desktop device, and supports encrypted log storage and remote upgrades for algorithm optimization and intervention strategy updates.

[0020] In summary, this invention, focusing on the four core requirements of "non-invasiveness, continuity, real-time performance, and minimal intervention," establishes a complete closed-loop technical path from acquiring ventilation micro-movements and identifying sleep events to implementing interventions. By reconstructing chest and abdominal ventilation displacement through millimeter-wave phase calculation, this invention can maintain stable monitoring even under varying bedding conditions, posture changes, and multi-user environments. Combined with time-frequency fusion and trend analysis, the system not only identifies existing sleep disorder events but also has the ability to predict event trends in advance, thereby enabling more proactive treatment strategies.

[0021] In the intervention section, this invention employs a flexible acoustic perturbation chain that satisfies sound pressure and frequency constraints. This allows intervention actions to guide users to make slight posture or ventilation adjustments without disrupting sleep, improving the success rate of interventions and reducing the risk of awakening. Simultaneously, the system moves data processing forward to the device, outputting only anonymized statistical information, significantly reducing the storage and transmission of raw physiological signals, thus balancing practicality and privacy.

[0022] Overall, this invention addresses the core issues that have long existed in existing non-contact sleep monitoring and intervention technologies, such as unstable accuracy, delayed recognition, rough intervention, and privacy risks, through feasible engineering methods. It provides an effective technical solution for building smart sleep health devices that are more reliable, lighter, and more suitable for home scenarios.

[0023] Compared to existing solutions that rely on wearable sensors, mattress pressure layers, or audio / video monitoring, this invention directly reconstructs chest and abdominal ventilation displacement through millimeter-wave micro-motion phase calculation under completely non-contact conditions. This not only provides a signal source that more closely approximates real respiratory dynamics but also maintains signal continuity even under conditions of bedding obstruction, posture changes, and multiple people in the room—a feat difficult to achieve with traditional solutions. The trend-based event determination mechanism of this invention can identify transitional signs before respiratory abnormalities fully develop, shifting intervention from "post-event correction" to "pre-event adjustment," significantly shortening the duration of events—a capability lacking in existing products based on static threshold determination.

[0024] In terms of intervention methods, this invention employs a flexible acoustic perturbation chain with low sound pressure and narrow frequency. Through a step-by-step enhancement strategy, the intervention actions are made closer to the natural response patterns of the human body, enabling users to make slight posture or ventilation adjustments without interrupting sleep. Compared with existing solutions that use fixed-intensity sound and light stimulation, which easily lead to awakening, this invention achieves a difficult balance between intervention effectiveness and sleep continuity, representing an improvement beyond the expectations of conventional technologies.

[0025] Furthermore, this invention implements end-to-end processing of all data in its system architecture, outputting only desensitized features without storing the original echo or audio, thus significantly reducing privacy risks while maintaining real-time performance. This overall architecture of "reconstructing ventilation—determining trends—implementing intervention—local closed-loop" does not exist in existing technologies, and it forms a comprehensive and unexpected technical advantage in terms of monitoring stability, intervention accuracy, and privacy security. Attached Figure Description

[0026] The present invention will be further described below with reference to the accompanying drawings and embodiments. In the accompanying drawings: Figure 1 The present invention provides a system structure block diagram of a non-invasive sleep monitoring and intervention system. Detailed Implementation

[0027] To make the objectives, technical solutions, and effects of this invention clearer and more explicit, the invention is further described in detail below. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention. The embodiments of the invention are described below in conjunction with the accompanying drawings.

[0028] This invention provides a non-invasive sleep monitoring and intervention system; please refer to [link to relevant documentation]. Figure 1 The system includes a continuous non-contact monitoring module for respiratory cycles, ventilation status, and body movement changes during sleep; a sleep disorder event determination module; and a flexible intervention module; wherein: The monitoring module models the ventilation-related displacements of the human chest and abdomen based on the phase micro-motion calculation of millimeter-wave radar to obtain respiratory rhythm, heart rate micro-motion, and body movement intensity. In an embodiment that can be practically deployed in a bedside lamp or desktop health device, this system uses millimeter-wave radar as the core non-contact sensing component. It transmits frequency-modulated continuous waves into the sleep area via an FMCW (Frequency Modulated Wave) system, utilizing multi-antenna echoes to construct a respiratory-related micro-motion model of the human chest and abdomen. During device operation, the processor first performs range-gating and phase-stabilizing processing on the echo signals to obtain the subtle displacement changes in the thoracic and abdominal cavities during the respiratory cycle, thereby reconstructing the respiratory rhythm, ventilation depth, and heart rate-related micro-motion characteristics. The entire monitoring chain remains continuously operational, maintaining stable extraction of the basal ventilation curve even if the user turns over or is covered by bedding.

[0029] The event determination module performs time-frequency fusion, trend analysis, and pattern matching on the data output by the monitoring module to identify or predict sleep disorder events such as snoring, sleep apnea, and hypopnea. After acquiring the real-time respiratory displacement curve, the processor performs time-frequency feature fusion and trend judgment. By comprehensively identifying snoring, hypopnea, or sleep apnea events through indicators such as amplitude changes, period lengthening, energy attenuation, and energy mutations in snoring-related frequency bands within a short time window, the processor determines the events. When the decision chain detects that an event has occurred or is about to occur, it immediately enters the intervention decision-making stage.

[0030] The flexible intervention module generates an acoustic disturbance signal that meets the sound pressure threshold, frequency range and minimum intervention constraint when the determination module is triggered, so as to cause the user's posture or breathing mode to be finely adjusted without waking the user. At this time, the system selects appropriate acoustic disturbance parameters based on the current event type and intensity. For example, it can trigger the user to make subconscious posture adjustments with a short pulse or a slightly frequency-modulated signal with a low sound pressure. The intervention is kept below the sound pressure threshold that the user is not easily awakened, thereby achieving the purpose of adjusting the ventilation state.

[0031] The system further includes a localized data processing module, which is used to complete feature extraction, event determination and intervention decision-making on the device side, and output desensitized data to the terminal device; thereby realizing closed-loop control of non-contact detection, real-time identification and subthreshold intervention for sleep abnormalities.

[0032] To ensure privacy and security, this embodiment employs a localized processing architecture. Raw millimeter-wave echoes, audio, and image data are not stored or uploaded. The device only performs feature extraction, event determination, and intervention triggering within its processor. The final output consists only of desensitized sleep parameters, such as respiratory stability index, event frequency statistics, and intervention execution status. This data can be sent to the user's mobile device via an encrypted channel for subsequent viewing or trend analysis.

[0033] Furthermore, this embodiment supports multi-person bed-sharing environments. When multiple spatial scattering centers exist for millimeter-wave echoes, the system distinguishes users based on differences in respiratory rate, consistency of displacement waveform, and spatial angle distribution among different targets, establishing a monitoring chain and intervention strategy for each user. Through the above structural configuration, this embodiment achieves continuous monitoring of ventilation status during sleep, real-time identification of abnormal events, and flexible adjustment based on "minimal intervention," forming the complete monitoring-judgment-intervention closed loop described above.

[0034] In a further embodiment, the millimeter-wave radar uses FMCW waveform, estimates the echo angle through a multi-antenna array, and constructs a thoracic-abdominal displacement curve using phase demodulation to reconstruct the user's respiratory depth, rhythm changes, and ventilation discontinuities.

[0035] Specifically, the system uses a 60GHz FMCW millimeter-wave radar as the primary sensing unit. The radar module incorporates a multi-transmitter, multi-receiver antenna array. After startup, the controller periodically transmits continuously modulated signals according to preset frequency modulation parameters and captures the echoes reflected from the human body surface through a multi-channel receiver. The system first uses range-gate filtering to eliminate reflections from other objects on the bed, and then determines the approximate location of the human target using the antenna array's angle estimation function. Based on this, the processor performs phase demodulation on the echoes from the selected range gate, extracting the micro-motion curves generated by breathing in the chest and abdomen. Since phase changes are linearly correlated with the periodic displacements of the chest and abdominal cavities, by performing drift-reduction and phase unrolling processing on the phase sequence, the system can reconstruct the true shape of the respiratory waveform, including fluctuations in respiratory depth and slight rhythmic drifts. This process is completed entirely without contact and maintains continuous tracking of the respiratory signal even when the user turns over or is covered by bedding, ensuring high-quality input data for subsequent event recognition chains.

[0036] In a further embodiment, the time-frequency fusion includes short-time Fourier transform, phase difference demodulation features, body kinetic energy envelope, snoring vibration spectrum and heart rate variability, and stable feature extraction under low signal-to-noise ratio conditions is achieved through Kalman filtering or adaptive weighted filtering.

[0037] Specifically, in this embodiment, the system employs a time-frequency fusion strategy for the respiratory signal obtained from phase demodulation to enhance the stability of sleep event detection. The processor first performs a short-time Fourier transform on the signal to obtain the dominant respiratory frequency, subharmonic components, and energy distribution; simultaneously, it extracts auxiliary features such as the rate of change of phase difference, body motion energy envelope, and heart rate micro-fluctuations. To address the low signal-to-noise ratio problem at night, this embodiment uses Kalman filtering to smooth the dominant frequency fluctuations and adaptive weighted filtering to suppress high-frequency noise caused by sudden body movements, ensuring the continuity and stability of the respiratory waveform even in the presence of interference. After the above feature fusion processing, the system can accurately identify significant ventilation abnormalities in noisy sleep scenarios with frequent posture changes, providing more reliable input data for subsequent event determination.

[0038] In a further embodiment, the event determination module adopts a sleep disorder event prediction model based on trend changes. This model establishes a joint temporal judgment rule for prolonged breathing cycle, decreased amplitude, and increased snoring energy, and triggers intervention before the event occurs to shorten the event duration.

[0039] Specifically, the system constructs a sleep event prediction model based on trend changes to identify potential hypoventilation or apnea in advance. The processor records the trends in respiratory cycle, amplitude, and energy during continuous monitoring, and combines this with the energy rise curve of the snoring frequency band for comprehensive judgment. When the respiratory cycle gradually lengthens, the amplitude decreases, and is accompanied by a slight increase in sound energy, the system considers it to be in a transitional zone before the event occurs, according to preset rules. At this time, the event judgment chain triggers a mild intervention signal in advance to intervene in the user's posture or ventilation mode as early as possible before the event fully develops, thereby shortening the event duration and reducing the impact on sleep stability. This predictive judgment strategy is more forward-looking than the traditional approach of only dealing with events after they occur, and it aligns with the "early intervention" design goal proposed in the technical disclosure.

[0040] In a further embodiment, the acoustic disturbance signal generated by the flexible intervention module includes: Short pulse signals, pulse sequences, or frequency-modulated signals with frequencies in the range of 0.5kHz to 4kHz and sound pressure levels in the range of 20 to 30dB are used, and a minimum intervention strategy with progressively increasing intensity is adopted according to the severity of the event to ensure that the user is not awakened as much as possible.

[0041] More specifically, in this embodiment, after the event determination chain detects a trend of hypoventilation, snoring, or apnea, the intervention chain generates an acoustic disturbance signal that meets the constraints of sound pressure and frequency. The processor first selects the intervention level according to the event type; for example, it outputs only short, low sound pressure pulses for mild snoring, and uses a more continuous frequency modulation sequence for apnea tendencies. The frequency of the acoustic disturbance signal is limited to the range of 0.5kHz to 4kHz to balance the propagation efficiency of the device structure and the low sensitivity of the human body to high-frequency sounds. The sound pressure level is controlled between 20 and 30dB. This range has been tested in actual scenarios and can avoid waking the user while still being sufficient to guide them to make subtle, imperceptible posture adjustments. The entire intervention chain adopts a step-by-step enhancement strategy; that is, if the initial intervention fails to restore the breathing waveform, the system will slightly increase the signal pulse width or modulation depth, but still within a range that does not affect the continuity of the user's sleep, making the intervention closer to the changing rhythm of the natural sleep environment.

[0042] In a further embodiment, the localized data processing module only performs feature calculation and model inference, does not store the original radar echo data, original audio or image information, and only outputs the desensitized event statistics and sleep parameters, which are then uploaded to the mobile terminal or the cloud.

[0043] Specifically, in this embodiment, the system's data processing architecture is entirely device-side based to avoid privacy risks associated with the external transmission of raw physiological data. The raw echo data from the millimeter-wave radar is temporarily cached within the processor after phase demodulation, filtering, and feature extraction, used only for model inference in the current cycle, and then immediately overwritten, not retained in memory long-term. The system also does not record audio or image information that may involve user privacy, retaining only desensitized statistical parameters processed by the algorithm, such as summary information like respiratory stability, event frequency, and intervention effectiveness. The final output data is rigorously compressed and transmitted to the user's mobile terminal in an encrypted manner. Under this structure, even with long-term device operation, there is no risk of the raw respiratory micro-motion waveform or raw acoustic data being extracted, making this embodiment significantly superior to traditional cloud analytics solutions in terms of privacy protection.

[0044] In a further embodiment, the system supports multi-user differentiation, performs target demixing through the spatial distribution of millimeter-wave echoes, respiratory rate differences, and micro-motion trajectories, and independently constructs respiratory characteristic baselines and intervention parameter models for each user.

[0045] Specifically, the system supports monitoring and intervention in multi-user bedside environments. Millimeter-wave radar can simultaneously capture echoes from multiple scattering centers in space. The system first makes preliminary distinctions based on the distribution of each target in the range-angle domain; then, it further confirms the identity by analyzing differences in respiratory rates, rhythm consistency, and micro-motion trajectory stability among different targets. For each identified user, the system maintains a separate monitoring chain, including independent respiratory baselines, ventilation waveform characteristics, and event judgment thresholds, thus avoiding confusion when two users have similar respiratory rhythms. When triggering intervention, the system combines the directivity of the sound source with model weights to ensure the intervention signal is effective for the target user while minimizing the impact on the other user. For example, when the spatial angular positions of two users from the device differ significantly, the system uses a main radiation direction control strategy to bring the intervention signal closer to the target user's location, thereby improving the effectiveness and privacy of the intervention.

[0046] In a further embodiment, the system further includes an alternative sensing module, which is composed of at least one of an ultrasonic array, an infrared array, or a mattress pressure sensing layer, for maintaining the continuity of respiratory and body movement monitoring when the millimeter wave echo is blocked or fails.

[0047] In a practical deployment scenario, to avoid signal attenuation of the millimeter-wave radar due to obstruction by bedding or specific sleeping positions, this embodiment employs an alternative sensing chain. An ultrasonic array is integrated within the device to take over monitoring breathing and body movement when the millimeter-wave signal quality deteriorates. The ultrasonic array operates in a low-power pulse mode, acquiring the chest and abdominal undulation trend through echo amplitude changes. The system evaluates the millimeter-wave echo quality in each monitoring cycle, determining its usability based on factors such as the continuity of the phase trajectory, signal-to-noise ratio, and changes in body kinetic energy. When a significant waveform discontinuity or abnormal fluctuation is detected in the millimeter-wave signal, the processor automatically switches to the ultrasonic data channel and uses its output breathing trend curve in the event determination chain. If the device is also equipped with an infrared array or a mattress pressure layer, the same switching logic is employed to ensure continuous tracking of ventilation status regardless of whether the user is completely covered by blankets, improving overall monitoring stability and availability.

[0048] In a further embodiment, the system receives data from environmental sensors or a mattress micro-vibration module, and dynamically adjusts the event determination threshold and intervention intensity based on environmental noise, temperature and humidity, and sleep stage to improve the accuracy and stability of the intervention.

[0049] Specifically, the system dynamically adjusts the judgment threshold and intervention intensity by combining inputs from environmental sensors and the mattress micro-vibration module. The device integrates temperature and humidity sensors and environmental noise sensors. The processor records the average level of ambient noise at night and adjusts the recognition threshold for snoring vibration energy accordingly. For example, in a noisy room, to avoid misidentifying ambient sound as snoring, the system raises the judgment threshold, making event recognition more robust. Furthermore, if the user's mattress has a micro-vibration sensing layer, the system cross-validates its output body movement information with millimeter-wave body movement characteristics. When the mattress detects a slight tendency to turn over, the system reduces the intervention intensity in advance, avoiding over-intervention when the user has already made natural adjustments. Through this combined sensing strategy, this embodiment can adjust parameters in real time according to the actual sleep environment, making event judgment more accurate and the intervention strategy closer to the user's actual sleep state.

[0050] In a further embodiment, the system adopts the structure of a bedside lamp, speaker, or desktop device, and supports encrypted log storage and remote upgrades for algorithm optimization and intervention strategy updates.

[0051] Specifically, the system is designed with a structure compatible with common household appliances, such as bedside lamps or desktop speakers, for ease of use in daily life. In its hardware design, the device's casing provides fixed mounting positions for the millimeter-wave radar, speaker module, environmental sensors, and processor, ensuring good signal coverage in ordinary living spaces. Log data generated during system operation is stored in encrypted format on local storage and will not be transmitted without user authorization. To facilitate subsequent algorithm upgrades, the device supports remote firmware updates; users can trigger updates via a mobile application to obtain improved event judgment models and intervention strategies. This embodiment enables the system to be implemented as an everyday item while possessing continuous upgrade capabilities, resulting in more reliable monitoring performance over long-term use.

[0052] In summary, this invention integrates millimeter-wave micro-motion sensing, trend-based event determination, and low-impact intervention strategies into a unified sleep monitoring and regulation system, enabling the device to operate stably and continuously in a typical bedroom environment. As can be seen from multiple embodiments, the core idea of ​​this invention is not simply to add sensors or single functions, but rather to re-examine the technical path of non-invasive sleep monitoring around the themes of "realistic reconstruction of ventilation status," "early identification of event formation," and "minimization of intervention actions." This allows the system to more closely approximate clinical monitoring logic while maintaining a low-invasive experience for home use.

[0053] In millimeter-wave ventilation modeling, the phase demodulation and feature processing flow provided by this invention enables the system to maintain high signal utilization even under conditions of bedding occlusion, posture changes, and multiple users. Regarding event determination, the system uses a joint judgment of temporal trends and energy characteristics to ensure that intervention timing no longer depends on the event having fully occurred, but rather on responding promptly during the "transition phase before the event." In terms of intervention execution, the acoustic perturbation chain proposed in this invention does not rely on fixed stimulus intensity, but dynamically adjusts according to the severity of the event and the user's current state, achieving gentle intervention rather than abrupt awakening. Combined with a localized data processing structure, the entire system minimizes the exposure of raw physiological data while ensuring real-time performance, demonstrating high feasibility from both engineering and privacy protection perspectives.

[0054] It should be noted that the specific embodiments are technical representations of the present invention and are used to help understand the structure and process of the present invention. Those skilled in the art can make appropriate adjustments to the parameter configuration, signal processing details, and intervention methods in the system according to actual hardware conditions, algorithm capabilities, or device form. These adjustments do not change the core idea of ​​the present invention and should be considered to be within the protection scope of the present invention.

Claims

1. A non-invasive sleep monitoring and intervention system, characterized in that, The system includes a continuous non-contact monitoring module for respiratory cycles, ventilation status, and body movement changes during sleep; a sleep disorder event determination module; and a flexible intervention module; wherein: The monitoring module models the ventilation-related displacements of the human chest and abdomen based on the phase micro-motion calculation of millimeter-wave radar to obtain respiratory rhythm, heart rate micro-motion, and body movement intensity. The event determination module performs time-frequency fusion, trend analysis, and pattern matching on the data output by the monitoring module to identify or predict sleep disorder events such as snoring, sleep apnea, and hypopnea. The flexible intervention module generates an acoustic disturbance signal that meets the sound pressure threshold, frequency range and minimum intervention constraint when the determination module is triggered, so as to cause the user's posture or breathing mode to be finely adjusted without waking the user. The system further includes a localized data processing module, which is used to complete feature extraction, event determination and intervention decision-making on the device side, and output desensitized data to the terminal device; thereby realizing closed-loop control of non-contact detection, real-time identification and subthreshold intervention for sleep abnormalities.

2. The non-invasive sleep monitoring and intervention system according to claim 1, characterized in that, The millimeter-wave radar uses FMCW waveforms, estimates echo angles through a multi-antenna array, and constructs a thoracic-abdominal displacement curve using phase demodulation to reconstruct the user's respiratory depth, rhythm changes, and ventilation discontinuities.

3. The non-invasive sleep monitoring and intervention system according to claim 1 or 2, characterized in that, The time-frequency fusion includes short-time Fourier transform, phase difference demodulation features, body kinetic energy envelope, snoring vibration spectrum and heart rate variability, and achieves stable feature extraction under low signal-to-noise ratio conditions through Kalman filtering or adaptive weighted filtering.

4. The non-invasive sleep monitoring and intervention system according to any one of claims 1 to 3, characterized in that, The event determination module adopts a sleep disorder event prediction model based on trend changes. This model establishes a joint time-series judgment rule for prolonged breathing cycle, decreased amplitude, and increased snoring energy, and triggers intervention before the event occurs to shorten the event duration.

5. The non-invasive sleep monitoring and intervention system according to any one of claims 1 to 4, characterized in that, The acoustic disturbance signal generated by the flexible intervention module includes: Short pulse signals, pulse sequences, or frequency-modulated signals with frequencies in the range of 0.5kHz to 4kHz and sound pressure levels in the range of 20 to 30dB are used, and a minimum intervention strategy with progressively increasing intensity is adopted according to the severity of the event to ensure that the user is not awakened as much as possible.

6. The non-invasive sleep monitoring and intervention system according to any one of claims 1 to 5, characterized in that, The localized data processing module only performs feature calculations and model inferences, and does not store the original radar echo data, original audio or image information. It only outputs the desensitized event statistics and sleep parameters and uploads them to the mobile terminal or cloud.

7. The non-invasive sleep monitoring and intervention system according to any one of claims 1 to 6, characterized in that, The system supports multi-user differentiation, performs target demixing through the spatial distribution of millimeter wave echoes, respiratory rate differences, and micro-motion trajectories, and independently constructs respiratory characteristic baselines and intervention parameter models for each user.

8. The non-invasive sleep monitoring and intervention system according to any one of claims 1 to 7, characterized in that, The system further includes an alternative sensing module, which is composed of at least one of an ultrasonic array, an infrared array, or a mattress pressure sensing layer, for maintaining the continuity of respiratory and body movement monitoring when the millimeter wave echo is blocked or fails.

9. The non-invasive sleep monitoring and intervention system according to any one of claims 1 to 8, characterized in that, The system receives data from environmental sensors or mattress micro-vibration modules, and dynamically adjusts the event judgment threshold and intervention intensity based on environmental noise, temperature and humidity, and sleep stages to improve the accuracy and stability of the intervention.

10. The non-invasive sleep monitoring and intervention system according to any one of claims 1 to 8, characterized in that, The system adopts the structure of a bedside lamp, speaker, or desktop device, and supports encrypted log storage and remote upgrades for algorithm optimization and intervention strategy updates.

Images