A System and Method for Sleep Respiratory Event Recognition and Cardiopulmonary Load Quantification Based on Radiofrequency Chest and Abdominal Micro-motion
By using a radiofrequency chest and abdominal micro-motion recognition system, and combining sleep stages and body positions to establish an individual condition baseline, a three-stage evidence chain is constructed, which solves the problems of misjudgment and missed detection in non-contact sleep breathing monitoring, and achieves efficient identification and quantification of sleep breathing abnormalities.
Patent Information
- Application Number
- CN202611344420.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-09-01
- Publication Date
- 2026-11-10
AI Technical Summary
Existing non-contact sleep apnea monitoring technologies are prone to misjudgment or missed detection due to factors such as body position, sleep stage, and physical activity pollution. They also lack post-event verification, resulting in a large number of false events and making it impossible to conduct continuous multi-night monitoring in home settings.
A radio frequency chest and abdominal micromotion recognition system was adopted to collect chest and abdominal micromotion signals through radar. Individual conditional baselines were established by combining sleep stage and body position. A three-stage evidence chain of 'conditional baseline - respiratory effort during the event - cardiopulmonary recovery after the event' was constructed. Signal quality gating and modal quality assessment were performed to generate traceable event sequences and load indicators.
It effectively distinguishes the morphology of abnormal breathing events, reduces false alarms of spurious events, provides reliable quantification of sleep-disordered breathing load, and supports continuous home monitoring.
Smart Images

Figure CN123058724A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of biomedical information processing and non-contact sleep monitoring technology, and in particular, a system and method for sleep respiratory event recognition and cardiopulmonary load quantification based on radiofrequency chest and abdominal micro-movements. Background Technology
[0002] Conventional interpretation of sleep apnea events relies on multiple physiological signals, including nasal and oral airflow, chest and abdominal breathing effort, blood oxygen saturation, electrocardiogram (ECG), pulse rate, and EEG arousal. Polysomnography (PSG) can provide sufficient evidence for determining breathing events, but this technology has limitations such as complex equipment deployment, cumbersome sensor wearing procedures, and high manual interpretation costs, and does not support continuous multi-night monitoring of the same user in a home setting.
[0003] 24GHz or 60GHz linear frequency modulated continuous wave radar can detect minute body surface displacements caused by respiration and heartbeat without contact with the human body. However, it only acquires mechanical micro-motion signals from the chest and abdomen, and cannot directly obtain airflow or blood oxygenation signals from the mouth and nose. If respiratory abnormalities are judged solely based on the amplitude decrease or peak interval of a single respiratory displacement waveform, it is easy to misjudge events such as turning over, bedding movement, changes in body position, target distance gate switching, and normal respiratory amplitude fluctuations as respiratory abnormalities.
[0004] Existing non-contact sleep apnea monitoring methods mainly employ single-region respiratory waveform thresholding, peak-valley detection, or time-frequency classification models. Chinese patent CN108392186B discloses a non-contact sleep apnea detection method that acquires chest motion signals via radar and uses a heuristic peak-based algorithm to calculate the apnea-shallow-slow breathing index. However, this method relies solely on a single chest peak sequence for analysis and fails to construct a complete chain of evidence covering coordinated tracking of both chest and abdominal regions, conditional baselines, and post-event cardiopulmonary recovery verification.
[0005] Chinese patent CN102641125B discloses a device for determining the state of chest and abdominal movement by comparing the fundamental wave and odd harmonics of microwave Doppler signals. This device is used to distinguish between normal breathing, a state where the chest stops moving while the abdomen continues to move, and a state where both the chest and abdomen stop moving. However, this solution does not perform joint modeling for the range-angle dual-zone tracking of modern LFMCW radar, the sleep state conditional baseline, and the cardiopulmonary recovery characteristics after the event.
[0006] Chinese patent CN112686213A discloses a respiratory state classification system based on recurrent neural networks and millimeter waves, which classifies chest cavity motion signals into normal, obstructive, central, and mixed states. Chinese patent CN119700017A discloses a method for identifying apnea and hypoventilation by combining distance-orientation localization, variational mode extraction, and hybrid neural networks. These solutions focus on state classification within a single window or time-frequency graph, and their validation of the three-stage consistency of pre-event baseline conditions, intra-event respiratory effort, and post-event cardiopulmonary recovery remains insufficient.
[0007] Furthermore, sleep stage, body position, and nighttime duration all affect the relationship between normal respiratory amplitude and chest-abdomen phase. Using a uniform amplitude threshold throughout the night can easily misjudge normal amplitude decreases during lateral or REM sleep as hypoventilation, and may also miss true amplitude suppression even when baseline amplitude is high. When a scattering area is obscured by bedding or body position, failing to differentiate between signal modal availability in the whole-window scoring can also lead to the erroneous deletion of valid respiratory evidence or the erroneous retention of contamination evidence.
[0008] Therefore, there is an urgent need for a non-contact radiofrequency sleep apnea event identification scheme: this scheme can not only determine whether the respiratory amplitude has decreased, but also establish an individual local baseline based on sleep conditions on the basis of stable tracking in both chest and abdominal regions. At the same time, it can analyze respiratory effort during the event and cardiopulmonary recovery response after the event, and generate a traceable event sequence and load index through quality weighting. Summary of the Invention
[0009] The purpose of this invention is to overcome the problems of single-channel respiratory waveform thresholds being sensitive to body position, sleep stage, body movement contamination and distance gate switching, as well as the lack of post-event verification leading to a large number of false events. The invention provides a radiofrequency sleep breathing abnormality event identification and load quantification system and method consisting of a three-stage evidence chain of "conditional baseline - intra-event respiratory effort - post-event cardiopulmonary recovery".
[0010] The technical solution to achieve the objective of this invention is as follows: On one hand, a sleep respiratory event recognition and cardiopulmonary load quantification system based on radiofrequency chest and abdominal micro-movements is provided, the system comprising:
[0011] The radio frequency acquisition module is used to transmit radio frequency signals to the sleep area via radar and receive complex echo signals modulated by the target human body's chest and abdominal movements, cardiac micro-movements, and body movements.
[0012] The chest and abdomen dual-zone tracking and displacement reconstruction module is used to locate and track the chest scattering region and the abdomen scattering region in the spatial domain, respectively, and demodulate the chest respiratory displacement sequence, the abdomen respiratory displacement sequence and the cardiac response sequence based on the complex echoes of the chest scattering region and the abdomen scattering region.
[0013] The modal quality gating module is used to evaluate the signal quality and dynamic interference of each analysis window, and generate modal availability labels and quality weights for chest breathing mode, abdominal breathing mode and cardiac response mode;
[0014] The individual condition baseline module is used to combine physiological and environmental conditions to construct the chest respiratory amplitude baseline and abdominal respiratory amplitude baseline for the current analysis window from the qualified historical windows.
[0015] The candidate event generation module is used to jointly evaluate the current respiratory amplitude of the chest and abdomen relative to their respective corresponding respiratory amplitude baselines, and generate candidate intervals by combining duration and quality effective coverage constraints.
[0016] The intra-event respiratory effort characterization module is used to extract chest and abdominal coordinated motion features and residual motion energy within the candidate interval, and generate effort characterization parameters that characterize the degree and morphology of respiratory effort.
[0017] The post-event cardiopulmonary recovery response verification module is used to extract ventilation rebound, heart rate mutation, autonomic nervous system regulation changes and arousal / body movement response features after the candidate interval ends, based on the end time of the candidate interval, and to construct the post-event cardiopulmonary recovery response.
[0018] The event fusion and load output module is used to fuse the degree of ventilatory inhibition, the respiratory effort characterization parameters within the event, the cardiopulmonary recovery response after the event, and the quality weights to determine the event confidence and respiratory effort morphology type of the candidate interval, and output the quality-weighted sleep apnea load index.
[0019] Furthermore, the radio frequency acquisition module includes a linear frequency modulated continuous wave radar installed at the head or side of the bed;
[0020] The thoracic and abdominal dual-zone tracking and displacement reconstruction module calculates candidate distance-angle positions for the z-th anatomical region within the w-th analysis window. Tracking score :
[0021]
[0022] In the formula, For the anatomical region category index, where, Indicates the chest area. Indicates the abdominal region; To analyze the window index, indicating the first... One analysis window; As candidate distance units, For candidate azimuth angles, Represents candidate distance cells within the distance-angle plane. and candidate azimuth The jointly determined candidate positions; For the first Within the analysis window, the first anatomical regions at candidate locations Tracking score at the location; For the first Within the analysis window, the first anatomical regions at candidate locations The normalized echo energy at the location, where the superscript wavy line indicates that the corresponding parameter has been normalized; For the first Within the analysis window, the first anatomical regions at candidate locations Echo phase continuity score at the location, where the subscript Indicates the echo phase; For the first Within the analysis window, the first anatomical regions at candidate locations The tracking consistency score relative to the predicted position in the previous analysis window, subscript Indicates target tracking; For the first Within the analysis window, the first Candidate locations of anatomical regions Penalty for positional deviation relative to its predicted location or the pre-defined anatomical region range; For the first Within the analysis window, the first Candidate locations of anatomical regions Spatial overlap penalty with another candidate anatomical region location; The weights corresponding to the normalized echo energy. The weights corresponding to the echo phase continuity score. To track the weights corresponding to the consistency scores, The weight corresponding to the position deviation penalty. The weights corresponding to the spatial overlap penalty are, and to All are non-negative numbers; select respectively such that and The candidate position that achieves the maximum value and satisfies the spatial order constraint that the chest candidate position is above the abdomen candidate position is selected as the first candidate position. The chest scattering center and abdominal scattering center within each analysis window;
[0023] The complex echo of the z-th anatomical region is converted into radial displacement using the following formula:
[0024]
[0025] In the formula, For radar operating wavelength, Let be the radial displacement of the z-th anatomical region at time t. The complex echo of the z-th anatomical region at time t. It is a complex phase angle function. This is the phase unwrapping function.
[0026] Furthermore, when evaluating the analysis window, the modal quality gating module comprehensively considers target tracking stability, phase continuity, frequency band signal-to-noise ratio, volumetric contamination ratio, and data missing ratio.
[0027] Quality weight of the w-th analysis window Calculate according to the following formula:
[0028]
[0029] The Sigmoid mapping function is:
[0030]
[0031] The range of values for quality weights is:
[0032]
[0033] In the formula, For analyzing the window index; For the first The quality weights of each analysis window, and ; For Sigmoid mapping functions; The input value for the Sigmoid mapping function; It is an exponential function; Bias parameters for calculating quality weights; The weights corresponding to the signal-to-noise ratio in the respiratory band. The weights corresponding to the phase continuity score. The weights corresponding to the target tracking continuity score. The penalty weight corresponding to the proportion of physical pollution. The penalty weight corresponds to the proportion of missing data, and to All are non-negative numbers; For the first The ratio of respiratory band signal power to respiratory band noise power within each analysis window; For the first The echo phase continuity score for each analysis window, where the subscript... Indicates the echo phase; For the first The target tracking continuity score is analyzed in each window, where the subscript... Indicates target tracking; For the first The proportion of the sampling time for receptor dynamic contamination within a single analysis window to the total duration of that analysis window; For the first The proportion of missing data sampling points in an analysis window to the total number of sampling points in that analysis window;
[0034] When the chest or abdominal breathing modality is unavailable, the corresponding window is not included in the calculation of chest and abdominal effort representation; when the cardiac response modality is unavailable, breathing modality evidence is retained and event confidence is reduced.
[0035] Furthermore, in the individual condition baseline module, the condition status includes sleep stage, body position category, and nighttime period;
[0036] Conditional baseline amplitude for the z-th anatomical region in the w-th analysis window Calculate according to the following formula:
[0037]
[0038] In the formula, Let z be the respiratory amplitude of the z-th anatomical region in the u-th historical window. This refers to a set of historical windows that are consistent with or adjacent to the current w-th analysis window in terms of sleep stage, body position category, and nighttime time, and that meet quality standards. Let u be the quality weight of the u-th historical window. and These represent the center times of the w-th analysis window and the u-th historical window, respectively. is the baseline time decay constant, exp is the exponential function, and wmed is the weighted median calculation.
[0039] Furthermore, when the set When the number of valid windows is less than a preset threshold, the individual conditional baseline module expands the baseline candidate set in the order of adjacent sleep stages under the same body position, adjacent body positions under the same sleep stage, and the user's multi-night historical baselines, and assigns decreasing confidence weights to the expansion sources.
[0040] Furthermore, the candidate event generation module calculates the combined thoracic and abdominal amplitude suppression rate for the w-th analysis window. for:
[0041]
[0042] In the formula, and These represent the current respiratory amplitude of the chest and abdomen, respectively. and These are the baseline amplitudes for the chest and abdomen, respectively. To prevent positive numbers with a denominator of zero;
[0043] And calculate the effective support duration for the j-th candidate interval:
[0044]
[0045] In the formula, The window sliding step size, Let w be the window quality weight for the w-th analysis window. Let be the set of windows contained in the j-th candidate interval. For indicator functions; only if the calendar duration and quality of the candidate interval are valid support duration. When both the effective coverage rate and the quality reach the preset threshold, the candidate interval is retained.
[0046] Furthermore, the features extracted by the intra-event respiratory effort characterization module include thoracic-abdominal coherence and thoracic-abdominal phase anomalous index:
[0047]
[0048]
[0049] In the formula, For the chest and abdomen coherence of the j-th candidate interval, Let be the phase anomaly index of the j-th candidate interval. For the w-th analysis window, the minimum weights for chest and abdominal mass are... Let be the difference in the analytical phase between chest and abdominal respiration in the w-th analysis window, where i is the imaginary unit, exp is the exponential function, and cos is the cosine function. and To prevent positive numbers with a denominator of zero;
[0050] The intra-event respiratory effort characterization module further combines the chest-abdomen residual amplitude ratio and respiratory harmonic energy duration to construct a respiratory effort characterization value:
[0051]
[0052] In the formula, to The weights are non-negative and their sum is 1; Let j be the respiratory effort characterization value of the j-th candidate interval. These are the ratios of the residual amplitude of the chest and abdomen relative to their respective conditional baselines within the j-th candidate interval. The duration of the respiratory fundamental frequency and its harmonics in the j-th candidate interval.
[0053] Furthermore, the post-event cardiopulmonary recovery response verification module constructs a cardiopulmonary recovery response vector within the recovery observation window after the candidate interval ends:
[0054]
[0055] In the formula, Let j be the cardiopulmonary recovery response vector of the j-th candidate interval. This is the bounce ratio of the maximum post-event respiratory amplitude corresponding to the j-th candidate interval relative to the pre-event conditional baseline. Let be the difference between the maximum heart rate after the event and the median heart rate before the event corresponding to the j-th candidate interval. Let be the root mean square change in the difference between adjacent heartbeat intervals before and after the event corresponding to the j-th candidate interval. Let be the proximity response between the end of the j-th candidate interval and the moment when the most recent sleep stage became lighter or when body movement occurred. This represents the recovery time required for the respiratory amplitude or heart rate corresponding to the j-th candidate interval to return to the baseline allowable range.
[0056] Furthermore, the event fusion and load output module calculates the probability distribution of candidate intervals belonging to each morphological type based on the respiratory effort representation features within the event and the cardiopulmonary recovery response vector:
[0057]
[0058] In the formula, Let be the probability that the j-th candidate interval belongs to the k-th morphological type. The score is the discrimination score for the j-th candidate interval belonging to the k-th morphological type. The score is the discrimination score for the j-th candidate interval belonging to the h-th morphological type. The set of predefined morphological types includes at least the following: respiratory effort deficit, respiratory effort maintenance, mixed switching, and hypoventilation.
[0059] And calculate the comprehensive event confidence of the j-th candidate interval. :
[0060]
[0061] In the formula, σ is the Sigmoid mapping function. Bias parameters, to The fusion coefficient is... This represents the average value of the joint amplitude suppression rate within the j-th candidate interval. This represents the maximum probability of each morphological type in the j-th candidate interval. This represents the average quality weight within the j-th candidate interval. Let be the ratio of bodily contamination between the j-th candidate interval and the post-event recovery interval.
[0062] Furthermore, the quality-weighted sleep apnea load index output by the event fusion and load output module includes:
[0063] Quality-weighted AHI related load :
[0064]
[0065] In the formula, The effective sleep duration for quality is expressed in minutes, and subscript Indicates sleep period; This is the set of candidate interval indices for the confirmed event. For the first The overall event confidence of each candidate interval; To confirm the total number of events;
[0066] Quality-weighted low ventilation continuous load :
[0067]
[0068] In the formula, Let be the duration of the j-th candidate interval, expressed in minutes;
[0069] Post-event load recovery :
[0070]
[0071] In the formula, z(·) is the directional consistency normalization function completed using individual history or population reference distribution; to These are non-negative recovery feature weights;
[0072] Comprehensive load index :
[0073]
[0074] In the formula, to These are non-negative combination weights, and their sum is 1; Confidence-weighted proportions of events related to effort to maintain breathing and mixed transition events;
[0075] Each load indicator is associated with and outputs the start timestamp, end timestamp, morphology type, confidence level, and waveform evidence fragments of the corresponding event.
[0076] On the other hand, a method for sleep respiratory event identification and cardiopulmonary load quantification based on radiofrequency chest and abdominal micro-movements is provided, which is implemented based on the aforementioned system for sleep respiratory event identification and cardiopulmonary load quantification based on radiofrequency chest and abdominal micro-movements. The method includes the following steps:
[0077] Step 1: Transmit radio frequency signals to the sleep area via radar and collect complex echo signals modulated by the micro-movements of the human body;
[0078] Step 2: Locate and track the chest and abdominal scattering regions within the spatial domain;
[0079] Step 3: Based on the complex echoes from the chest and abdominal scattering regions, demodulate and reconstruct the chest respiratory displacement sequence, abdominal respiratory displacement sequence, and cardiac response sequence.
[0080] Step 4: Perform signal quality assessment on the analysis window to obtain the availability label and quality weight of each mode;
[0081] Step 5: Based on the current sleep stage, body position category, and time-related conditions in the analysis window, establish a baseline for individual conditional respiratory amplitude in the chest and abdomen.
[0082] Step 6: Generate candidate intervals based on the degree, duration, and effective quality of inhibition of thoracoabdominal breathing amplitude;
[0083] Step 7: Extract the characteristics of thoracic-abdominal coherence, phase anomaly, and residual respiratory effort within the candidate interval;
[0084] Step 8: Using the end time of the candidate interval as the alignment point, extract the post-event respiratory rebound, heart rate surge, changes in autonomic nervous regulation, and arousal or body movement response to construct the cardiopulmonary recovery response;
[0085] Step 9: Integrate evidence of ventilatory depression, intra-event respiratory effort characterization, post-event cardiopulmonary recovery response, and quality weights to output event confidence, morphological classification results, evidence time segments, and quality-weighted load indices.
[0086] On the other hand, a computer-readable storage medium is provided, on which a computer program is stored, which, when executed by a processor, implements the method for sleep breathing event recognition and cardiopulmonary load quantification based on radio frequency chest and abdominal micro-movements.
[0087] Compared with the prior art, the significant advantages of this invention are:
[0088] (1) By independently locating, tracking and reconstructing the two scattering regions of the chest and abdomen in the spatial domain, the chest-abdomen coordination relationship missing in the single-region respiratory waveform is preserved, overcoming the defect that traditional single-path respiratory waveforms cannot obtain the chest-abdomen coordination motion characteristics.
[0089] (2) By combining sleep stage, body position category and nighttime period, a dynamically weighted individual condition baseline is established, which reduces the misjudgment of normal physiological fluctuations by a uniform threshold throughout the night.
[0090] (3) By establishing respiratory effort characterization through intra-cardiac coherence, phase anomaly, residual amplitude and harmonic duration, the limitations of simply relying on the decrease in ventilation amplitude for event classification are overcome. It can effectively distinguish between respiratory effort loss type, maintenance type and other forms, so that event morphology analysis no longer depends solely on the decrease in amplitude.
[0091] (4) Incorporate respiratory rebound, heart rate surge, short-term HRV changes and sleep shallowing or body movement response after the event into the event verification to form a complete evidence chain of “pre-event conditional baseline - respiratory effort during the event - post-event cardiopulmonary recovery”, which greatly reduces false event false alarms.
[0092] (5) Introduce multimodal quality gating and event-level confidence to construct quality-weighted AHI related loads, low ventilation continuous loads, recovery loads and comprehensive loads to avoid equating low-quality candidate events with high-quality events.
[0093] (6) The start time, end time, chest and abdominal waveforms, cardiopulmonary recovery response, quality weight and main contribution features of each identified event are fully preserved, which facilitates long-term sleep phenotype modeling and subsequent manual review.
[0094] The present invention will now be described in further detail with reference to the accompanying drawings. Attached Figure Description
[0095] Figure 1 This is a block diagram of a sleep breathing event recognition and cardiopulmonary load quantification system based on radio frequency chest and abdominal micro-movements in one embodiment.
[0096] Figure 2 This is a schematic diagram of a three-stage evidence chain for a respiratory abnormality event, including the pre-event conditional baseline, the intra-event amplitude suppression and respiratory effort characterization, and the post-event cardiopulmonary recovery response.
[0097] Figure 3 This is a flowchart of a method for sleep respiratory event identification and cardiopulmonary load quantification based on radiofrequency chest and abdominal micro-movements in one embodiment. Detailed Implementation
[0098] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.
[0099] It should be noted that if the embodiments of the present invention involve descriptions such as "first" and "second," these descriptions are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined with "first" and "second" may explicitly or implicitly include at least one of those features. Furthermore, the technical solutions of the various embodiments can be combined with each other, but this must be based on the ability of those skilled in the art to implement them. When the combination of technical solutions is contradictory or impossible to implement, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection claimed by the present invention.
[0100] To achieve the objectives of this invention, the system provided by this invention includes a bedside radio frequency acquisition module, a chest and abdomen dual-zone tracking and displacement reconstruction module, a modal quality gating module, an individual conditional baseline module, a candidate event generation module, an intra-event respiratory effort characterization module, a post-event cardiopulmonary recovery response verification module, and an event fusion and load output module. Each module transmits window-level and event-level data according to timestamps.
[0101] The present invention also provides a corresponding method, which first tracks two scattering regions of the chest and abdomen in the distance-angle plane and establishes a submodal quality label; secondly, establishes an individual conditional baseline based on sleep stage, body position and time period, and generates amplitude inhibition candidate intervals; then, analyzes the intra-event morphology through chest-abdominal coherence, anomalous index and residual respiratory effort characterization; finally, aligns respiratory rebound, heart rate surge, HRV change and shallowing or body movement response at the end of the event, and outputs the event confidence and quality-weighted load.
[0102] In one embodiment, combined Figure 1 A sleep respiratory event recognition and cardiopulmonary load quantification system based on radiofrequency chest and abdominal micro-movements is provided, the system comprising:
[0103] The radio frequency acquisition module 100 is used to transmit radio frequency signals to the sleep area via radar and receive complex echo signals modulated by the target human body's chest and abdominal movements, cardiac micro-movements, and body movements.
[0104] The chest and abdomen dual-zone tracking and displacement reconstruction module 200 is used to locate and track the chest scattering region and the abdomen scattering region in the spatial domain respectively, and demodulate the chest respiratory displacement sequence, the abdomen respiratory displacement sequence and the cardiac response sequence based on the complex echoes of the chest scattering region and the abdomen scattering region.
[0105] The modal quality gating module 300 is used to evaluate the signal quality and dynamic interference of each analysis window, and generate modal availability labels and quality weights for chest breathing modality, abdominal breathing modality and cardiac response modality; here, a whole window label is not used to replace the quality status of different modalities;
[0106] Individual Condition Baseline Module 400 is used to combine physiological and environmental conditions to construct the chest respiratory amplitude baseline and abdominal respiratory amplitude baseline for the current analysis window from the qualified historical windows.
[0107] The candidate event generation module 500 is used to jointly suppress the current respiratory amplitude of the chest and abdomen (the amplitude can be obtained by peak-to-valley difference, Hilbert analytical envelope or periodic root mean square amplitude) relative to their respective corresponding respiratory amplitude baselines, and generate candidate intervals by combining duration and quality effective coverage constraints.
[0108] The intra-event respiratory effort characterization module 600 is used to extract chest and abdominal coordinated motion features and residual motion energy within the candidate interval, and generate effort characterization parameters that characterize the degree and morphology of respiratory effort.
[0109] The post-event cardiopulmonary recovery response verification module 700 is used to extract ventilation rebound, heart rate mutation, autonomic nervous system regulation changes and arousal / body movement response features after the end of the candidate interval based on the end time of the candidate interval, and construct the post-event cardiopulmonary recovery response.
[0110] The event fusion and load output module 800 is used to fuse the degree of ventilatory inhibition, the respiratory effort characterization parameters within the event, the cardiopulmonary recovery response after the event, and the quality weights, to determine the event confidence and respiratory effort morphology type of the candidate interval, and to output the quality-weighted sleep apnea load index.
[0111] Furthermore, in one embodiment, the radio frequency acquisition module includes a linear frequency modulated continuous wave radar mounted at the head or side of the bed;
[0112] The thoracic and abdominal dual-zone tracking and displacement reconstruction module calculates candidate distance-angle positions for the z-th anatomical region within the w-th analysis window. Tracking score :
[0113]
[0114] In the formula, For the anatomical region category index, where, Indicates the chest area. Indicates the abdominal region; To analyze the window index, indicating the first... One analysis window; As candidate distance units, For candidate azimuth angles, Represents candidate distance cells within the distance-angle plane. and candidate azimuth The jointly determined candidate positions; For the first Within the analysis window, the first anatomical regions at candidate locations Tracking score at the location; For the first Within the analysis window, the first anatomical regions at candidate locations The normalized echo energy at the location, where the superscript wavy line indicates that the corresponding parameter has been normalized; For the first Within the analysis window, the first anatomical regions at candidate locations Echo phase continuity score at the location, where the subscript Indicates the echo phase; For the first Within the analysis window, the first anatomical regions at candidate locations The tracking consistency score relative to the predicted position in the previous analysis window, subscript Indicates target tracking; For the first Within the analysis window, the first Candidate locations of anatomical regions Penalty for positional deviation relative to its predicted location or the pre-defined anatomical region range; For the first Within the analysis window, the first Candidate locations of anatomical regions Spatial overlap penalty with another candidate anatomical region location; The weights corresponding to the normalized echo energy. The weights corresponding to the echo phase continuity score. To track the weights corresponding to the consistency scores, The weight corresponding to the position deviation penalty. The weights corresponding to the spatial overlap penalty are, and to All are non-negative numbers; select respectively such that and The candidate position that achieves the maximum value and satisfies the spatial order constraint that the chest candidate position is above the abdomen candidate position is selected as the first candidate position. The analysis window includes the chest scattering center and the abdominal scattering center; the scores of the two regions are jointly maximized to reduce the occurrence of two tracking trajectories merging into the same scattering center;
[0115] The complex echo of the z-th anatomical region is converted into radial displacement using the following formula:
[0116]
[0117] In the formula, For radar operating wavelength, Let be the radial displacement of the z-th anatomical region at time t. The complex echo of the z-th anatomical region at time t. It is a complex phase angle function. This is the phase unwrapping function. Here, for... Baseline drift correction and bandpass filtering from 0.08Hz to 0.60Hz are performed to obtain respiratory displacement. Corresponding filtering is applied to the cardiac frequency band or the cardiopulmonary signal separation module is called to obtain the cardiac response sequence.
[0118] Preferably, in some embodiments, the radio frequency acquisition module is mounted in a fixed position at the head of the bed or on the side of the bed towards the head, and the radar's distance and angle coverage includes the user's upper body. Some embodiments use a 60GHz multiple-transmitter-multiple-receiver (LFMCW) radar, while others use a 24GHz LFMCW radar.
[0119] Furthermore, in one embodiment, the modal quality gating module comprehensively considers target tracking stability, phase continuity, frequency band signal-to-noise ratio, body motion contamination ratio, and data missing ratio when evaluating the analysis window;
[0120] Quality weight of the w-th analysis window Calculate according to the following formula:
[0121]
[0122] The Sigmoid mapping function is:
[0123]
[0124] The range of values for quality weights is:
[0125]
[0126] In the formula, For analyzing the window index; For the first The quality weights of each analysis window, and ; For Sigmoid mapping functions; The input value for the Sigmoid mapping function; It is an exponential function; Bias parameters for calculating quality weights; The weights corresponding to the signal-to-noise ratio in the respiratory band. The weights corresponding to the phase continuity score. The weights corresponding to the target tracking continuity score. The penalty weight corresponding to the proportion of physical pollution. The penalty weight corresponds to the proportion of missing data, and to All are non-negative numbers; For the first The ratio of respiratory band signal power to respiratory band noise power within each analysis window; For the first The echo phase continuity score for each analysis window, where the subscript... Indicates the echo phase; For the first The target tracking continuity score is analyzed in each window, where the subscript... Indicates target tracking; For the first The proportion of the sampling time for receptor dynamic contamination within a single analysis window to the total duration of that analysis window; For the first The proportion of missing data sampling points within an analysis window relative to the total number of sampling points in that window; here, SNR can be used for the chest and abdomen respectively. and Generate independent quality weights;
[0127] When the chest or abdominal breathing modality is unavailable, the corresponding window is not included in the calculation of chest and abdominal effort representation; when the cardiac response modality is unavailable, breathing modality evidence is retained and event confidence is reduced.
[0128] Furthermore, in one embodiment, the individual condition baseline module includes a condition state comprising a sleep stage, a body position category, and a nighttime period; the sleep stage may include wakefulness, N1, N2, N3, and REM sleep, or may be combined into wakefulness, non-rapid eye movement (NREM) sleep, and rapid eye movement (REM) sleep; the body position category may include supine, left lateral, right lateral, and prone; the nighttime period may be divided according to the time after falling asleep or the sleep cycle number.
[0129] Conditional baseline amplitude for the z-th anatomical region in the w-th analysis window Calculate according to the following formula:
[0130]
[0131] In the formula, Let z be the respiratory amplitude of the z-th anatomical region in the u-th historical window. This refers to a set of historical windows that are consistent with or adjacent to the current w-th analysis window in terms of sleep stage, body position category, and nighttime time, and that meet quality standards. Let u be the quality weight of the u-th historical window. and These represent the center times of the w-th analysis window and the u-th historical window, respectively. is the baseline time decay constant, exp is the exponential function, and wmed is the weighted median calculation.
[0132] Preferably, in some embodiments, when the set When the number of valid windows is less than a preset threshold, the individual conditional baseline module expands the baseline candidate set in the order of adjacent sleep stages under the same body position, adjacent body positions under the same sleep stage, and the user's multi-night historical baselines, and assigns decreasing confidence weights to the expansion sources.
[0133] Furthermore, in one embodiment, the candidate event generation module calculates the combined thoracic and abdominal amplitude suppression rate for the w-th analysis window. (To avoid false alarms caused by a localized decrease in amplitude in a certain region, the suppression rate relative to the conditional baseline using the geometric mean of the chest and abdomen is used.)
[0134]
[0135] In the formula, and These represent the current respiratory amplitude of the chest and abdomen, respectively. and These are the baseline amplitudes for the chest and abdomen, respectively. To prevent positive numbers with a denominator of zero; The larger the value, the more significant the inhibition of thoracoabdominal ventilation amplitude. When only one respiratory modality is available, single-modality inhibition candidates can be generated based on the amplitude ratio of the available modalities, but these candidates are marked as lacking evidence of thoracoabdominal synergy.
[0136] And calculate the effective support duration for the j-th candidate interval:
[0137]
[0138] In the formula, The window sliding step size, Let w be the window quality weight for the w-th analysis window. Let be the set of windows contained in the j-th candidate interval. The indicator function takes the value 1 if the condition is met and 0 otherwise; it is only valid if the calendar duration and quality support duration of the candidate interval are met. When both the effective coverage rate and the quality reach the preset threshold, the candidate interval is retained.
[0139] Here, by way of example, in one embodiment, the calendar duration of the candidate interval is 10 to 120 seconds, and the effective quality coverage is not less than 0.60; short fluctuations below the aforementioned range are not considered as breathing events, and intervals exceeding the range and accompanied by target loss are preferentially marked as low quality.
[0140] Furthermore, in one embodiment, the features extracted by the intra-event respiratory effort characterization module include thoracic-abdominal coherence and thoracic-abdominal phase anomalous index:
[0141]
[0142]
[0143] In the formula, For the chest and abdomen coherence of the j-th candidate interval, Let be the phase anomaly index of the j-th candidate interval. For the w-th analysis window, the minimum weights for chest and abdominal mass are... Let be the difference in the analytical phase between chest and abdominal respiration in the w-th analysis window, where i is the imaginary unit, exp is the exponential function, and cos is the cosine function. and To prevent positive numbers with a denominator of zero; here, A higher value indicates a more stable relative phase between the chest and abdomen within the candidate interval. However, this relative phase can be in-phase or out-of-phase, therefore the phase anomaly index also needs to be calculated. When the chest and abdomen movements tend to be in-phase... When the chest and abdominal movements approach 0, it indicates that the movement of the chest and abdomen is inversely related. Close to 1;
[0144] The intra-event respiratory effort characterization module further combines the chest-abdomen residual amplitude ratio and respiratory harmonic energy duration to construct a respiratory effort characterization value:
[0145]
[0146] In the formula, to The weights are non-negative and their sum is 1; Let j be the respiratory effort characterization value of the j-th candidate interval. These are the ratios of the residual amplitude of the chest and abdomen relative to their respective conditional baselines within the j-th candidate interval. The duration of the respiratory fundamental frequency and its harmonics within the j-th candidate interval. When the residual amplitude and harmonic duration of the chest and abdomen are low, it tends to indicate a lack of respiratory effort; when the residual effort is maintained and... When elevated, there is a tendency to maintain or exhibit abnormal breathing patterns.
[0147] Furthermore, in one embodiment, such as Figure 2As shown, each candidate interval is divided into a pre-event baseline interval, an intra-event interval, and a post-event recovery interval. The pre-event baseline interval preferentially selects a quality-compliant window that aligns with the sleep stage and body position of the candidate interval. The post-event recovery interval begins at the end of the candidate interval and continues until respiratory amplitude and heart rate return to the baseline tolerance range or the maximum recovery observation time is reached. The post-event cardiopulmonary recovery response verification module constructs a cardiopulmonary recovery response vector within the recovery observation window after the candidate interval ends.
[0148]
[0149] In the formula, Let j be the cardiopulmonary recovery response vector of the j-th candidate interval. This is the bounce ratio of the maximum post-event respiratory amplitude corresponding to the j-th candidate interval relative to the pre-event conditional baseline. Let be the difference between the maximum heart rate after the event and the median heart rate before the event corresponding to the j-th candidate interval. Let be the root mean square change in the difference between adjacent heartbeat intervals before and after the event corresponding to the j-th candidate interval. Let be the proximity response between the end of the j-th candidate interval and the moment when the most recent sleep stage became lighter or when body movement occurred. This represents the recovery time required for the respiratory amplitude or heart rate corresponding to the j-th candidate interval to return to the baseline allowable range.
[0150] Furthermore, in one embodiment, the event fusion and load output module calculates the probability distribution of candidate intervals belonging to each morphological type based on the intra-event respiratory effort characterization features and the cardiopulmonary recovery response vector:
[0151]
[0152] In the formula, Let be the probability that the j-th candidate interval belongs to the k-th morphological type. The score is the discrimination score for the j-th candidate interval belonging to the k-th morphological type. The score is the discrimination score for the j-th candidate interval belonging to the h-th morphological type. The predefined set of morphological types includes at least the respiratory effort deficit type, respiratory effort maintenance type, mixed switching type, and hypoventilation type; here, the morphological type is used to characterize the effort pattern in radiofrequency chest and abdominal micromovements and does not replace the clinical diagnosis obtained from oral and nasal airflow, blood oxygenation, and polysomnography.
[0153] And calculate the comprehensive event confidence of the j-th candidate interval. :
[0154]
[0155] In the formula, σ is the Sigmoid mapping function. Bias parameters, to The fusion coefficient is... This represents the average value of the joint amplitude suppression rate within the j-th candidate interval. This represents the maximum probability of each morphological type in the j-th candidate interval. This represents the average quality weight within the j-th candidate interval. This represents the proportion of motion contamination between the j-th candidate interval and the post-event recovery interval. When cardiac response is unavailable, it is replaced with a missing marker. And by reducing Or increase the penalty for missing items to lower the penalty. Missing values are not treated as zero responses.
[0156] Furthermore, in one embodiment, the quality-weighted sleep apnea load index output by the event fusion and load output module includes:
[0157] Quality-weighted AHI related load (Unit: times / hour):
[0158]
[0159] In the formula, The effective sleep duration for quality is expressed in minutes, and subscript Indicates sleep period; This is the set of candidate interval indices for the confirmed event. For the first The overall event confidence of each candidate interval; To confirm the total number of events; a constant of 60 is used to convert the load per minute to the load per hour. Compared to an equal-weighted count of all events, The contribution of low-confidence events is automatically downweighted;
[0160] Quality-weighted low ventilation continuous load :
[0161]
[0162] In the formula, Let be the duration of the j-th candidate interval, expressed in minutes; the constant 100 expresses the result as a percentage of effective sleep time.
[0163] Post-event load recovery :
[0164]
[0165] In the formula, z(·) is the directional consistency normalization function completed using individual history or population reference distribution; to These are non-negative recovery feature weights;
[0166] Comprehensive load index :
[0167]
[0168] In the formula, to These are non-negative combination weights, and their sum is 1; Confidence-weighted proportions of events related to effort to maintain breathing and mixed transition events;
[0169] Each load indicator is associated with and outputs the start timestamp, end timestamp, morphology type, confidence level, and waveform evidence fragments of the corresponding event.
[0170] In one specific embodiment, the original radar frame rate is set to 50 to 200 frames per second according to hardware settings; the respiratory displacement sampling rate after coherent merging is 20 Hz; and the cardiac response sequence sampling rate is 50 Hz. The analysis window length is 30 seconds, and the sliding step size is 2 seconds. The conditional baseline reference interval covers a quality-compliant window of 5 to 20 minutes prior to the current moment and can be supplemented from data from the same user over the past 3 to 14 nights. The post-event recovery observation interval is 30 to 120 seconds. The above values are implementation parameters and can be adjusted within a reasonable range according to the radar frame rate, user age, and monitoring scenario.
[0171] For example, for a certain candidate interval, the calculated average joint amplitude suppression rate was 0.67, the quality effective coverage rate was 0.91, the residual amplitude ratio of the chest was 0.38, the residual amplitude ratio of the abdomen was 0.64, the chest-abdomen coherence was 0.82, the phase anomalous index was 0.71, the post-event respiratory rebound ratio was 0.46, the heart rate surge was 8.4 beats / min, and a phase shallowing occurred 12 seconds after the event ended. This candidate interval simultaneously possesses amplitude suppression, respiratory effort maintenance, and post-event cardiopulmonary response, and is therefore output as a candidate event for respiratory effort maintenance morphology, while retaining the corresponding chest-abdomen waveform and heart rate evidence time segments.
[0172] In an exemplary dataset of 20 minutes of effective sleep per night, 62 candidate events were retained after event fusion, with a sum of event confidence of 57.4, derived from the formula... get The rate was 8.20 times per hour. This calculation preserves the confidence level and evidence time segment for each event, allowing subsequent long-term sleep phenotypic modeling to utilize event counts, event duration load, morphological composition, and recovery load simultaneously, rather than just obtaining a single nightly count.
[0173] On the other hand, a method for sleep respiratory event identification and cardiopulmonary load quantification based on radiofrequency chest and abdominal micro-movements is provided, which is implemented based on the aforementioned system for sleep respiratory event identification and cardiopulmonary load quantification based on radiofrequency chest and abdominal micro-movements. The method includes the following steps:
[0174] S101 transmits radio frequency signals to the sleep area via radar and collects complex echo signals modulated by the micro-movements of the human body.
[0175] S102, locates and tracks the chest scattering region and the abdominal scattering region in the spatial domain;
[0176] S103, based on the complex echoes from the chest scattering region and the abdominal scattering region, demodulate and reconstruct the chest respiratory displacement sequence, the abdominal respiratory displacement sequence, and the cardiac response sequence;
[0177] S104, perform signal quality assessment on the analysis window to obtain the availability label and quality weight of each mode;
[0178] S105, Based on the current analysis window's sleep stage, body position category, and time-related conditions, establish a baseline for individual conditional respiratory amplitude in the chest and abdomen.
[0179] S106, candidate intervals are generated based on the degree, duration and effective quality of inhibition of thoracoabdominal respiratory amplitude;
[0180] S107, extracting thoracic-abdominal coherence, phase anomalousness, and residual respiratory effort characteristics within the candidate interval;
[0181] S108 uses the end time of the candidate interval as the alignment point to extract post-event respiratory rebound, heart rate surge, changes in autonomic nervous regulation, and arousal or body movement response to construct a cardiopulmonary recovery response.
[0182] S109 integrates evidence of ventilatory depression, intra-event respiratory effort characteristics, post-event cardiopulmonary recovery response, and quality weights to output event confidence, morphological classification results, evidence time segments, and quality-weighted load indices.
[0183] For specific limitations on each step, please refer to the above limitations on the sleep respiratory event recognition and cardiopulmonary load quantification system based on radiofrequency chest and abdominal micro-movements, which will not be repeated here.
[0184] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the method for sleep breathing event recognition and cardiopulmonary load quantification based on radio frequency chest and abdominal micro-movements.
[0185] For specific limitations on each step, please refer to the above limitations on the sleep respiratory event recognition and cardiopulmonary load quantification system based on radiofrequency chest and abdominal micro-movements, which will not be repeated here.
[0186] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps of the method for sleep breathing event recognition and cardiopulmonary load quantification based on radio frequency chest and abdominal micro-movements.
[0187] For specific limitations on each step, please refer to the above limitations on the sleep respiratory event recognition and cardiopulmonary load quantification system based on radiofrequency chest and abdominal micro-movements, which will not be repeated here.
[0188] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention without departing from its spirit and scope should be included within the protection scope of the present invention.
Claims
1. A sleep respiratory event recognition and cardiopulmonary load quantification system based on radiofrequency chest and abdominal micro-movements, characterized in that, The system includes: The radio frequency acquisition module is used to transmit radio frequency signals to the sleep area via radar and receive complex echo signals modulated by the target human body's chest and abdominal movements, cardiac micro-movements, and body movements. The chest and abdomen dual-zone tracking and displacement reconstruction module is used to locate and track the chest scattering region and the abdomen scattering region in the spatial domain, respectively, and demodulate the chest respiratory displacement sequence, the abdomen respiratory displacement sequence and the cardiac response sequence based on the complex echoes of the chest scattering region and the abdomen scattering region. The modal quality gating module is used to evaluate the signal quality and dynamic interference of each analysis window, and generate modal availability labels and quality weights for chest breathing mode, abdominal breathing mode and cardiac response mode; The individual condition baseline module is used to combine physiological and environmental conditions to construct the chest respiratory amplitude baseline and abdominal respiratory amplitude baseline for the current analysis window from the qualified historical windows. The candidate event generation module is used to jointly evaluate the current respiratory amplitude of the chest and abdomen relative to their respective corresponding respiratory amplitude baselines, and generate candidate intervals by combining duration and quality effective coverage constraints. The intra-event respiratory effort characterization module is used to extract chest and abdominal coordinated motion features and residual motion energy within the candidate interval, and generate effort characterization parameters that characterize the degree and morphology of respiratory effort. The post-event cardiopulmonary recovery response verification module is used to extract ventilation rebound, heart rate mutation, autonomic nervous system regulation changes and arousal / body movement response features after the candidate interval ends, based on the end time of the candidate interval, and to construct the post-event cardiopulmonary recovery response. The event fusion and load output module is used to fuse the degree of ventilatory inhibition, the respiratory effort characterization parameters within the event, the cardiopulmonary recovery response after the event, and the quality weights to determine the event confidence and respiratory effort morphology type of the candidate interval, and output the quality-weighted sleep apnea load index.
2. The sleep respiratory event recognition and cardiopulmonary load quantification system based on radio frequency chest and abdominal micro-movements according to claim 1, characterized in that, The radio frequency acquisition module includes a linear frequency modulated continuous wave radar installed at the head or side of the bed; The thoracic and abdominal dual-zone tracking and displacement reconstruction module calculates candidate distance-angle positions for the z-th anatomical region within the w-th analysis window. Tracking score : In the formula, For the anatomical region category index, where, Indicates the chest area. Indicates the abdominal region; To analyze the window index, indicating the first... One analysis window; As candidate distance units, For candidate azimuth angles, Represents candidate distance cells within the distance-angle plane. and candidate azimuth The jointly determined candidate positions; For the first Within the analysis window, the first anatomical regions at candidate locations Tracking score at the location; For the first Within the analysis window, the first anatomical regions at candidate locations The normalized echo energy at the location, where the superscript wavy line indicates that the corresponding parameter has been normalized; For the first Within the analysis window, the first anatomical regions at candidate locations Echo phase continuity score at the location, where the subscript Indicates the echo phase; For the first Within the analysis window, the first anatomical regions at candidate locations The tracking consistency score relative to the predicted position in the previous analysis window, subscript Indicates target tracking; For the first Within the analysis window, the first Candidate locations of anatomical regions Penalty for positional deviation relative to its predicted location or the pre-defined anatomical region range; For the first Within the analysis window, the first Candidate locations of anatomical regions Spatial overlap penalty with another candidate anatomical region location; The weights corresponding to the normalized echo energy. The weights corresponding to the echo phase continuity score. To track the weights corresponding to the consistency scores, The weight corresponding to the position deviation penalty. The weights corresponding to the spatial overlap penalty are, and to All are non-negative numbers; select respectively such that and The candidate position that achieves the maximum value and satisfies the spatial order constraint that the chest candidate position is above the abdomen candidate position is selected as the first candidate position. The chest scattering center and abdominal scattering center within each analysis window; The complex echo of the z-th anatomical region is converted into radial displacement using the following formula: In the formula, For radar operating wavelength, Let be the radial displacement of the z-th anatomical region at time t. The complex echo of the z-th anatomical region at time t. It is a complex phase angle function. This is the phase unwrapping function.
3. The sleep respiratory event recognition and cardiopulmonary load quantification system based on radio frequency chest and abdominal micro-movements according to claim 1, characterized in that, When evaluating the analysis window, the modal quality gating module comprehensively considers target tracking stability, phase continuity, frequency band signal-to-noise ratio, volumetric contamination ratio, and data missing ratio. Quality weight of the w-th analysis window Calculate according to the following formula: The Sigmoid mapping function is: The range of values for quality weights is: In the formula, For analyzing the window index; For the first The quality weights of each analysis window, and ; For Sigmoid mapping functions; The input value for the Sigmoid mapping function; It is an exponential function; Bias parameters for calculating quality weights; The weights corresponding to the signal-to-noise ratio in the respiratory band. The weights corresponding to the phase continuity score. The weights corresponding to the target tracking continuity score. The penalty weight corresponding to the proportion of physical pollution. The penalty weight corresponds to the proportion of missing data, and to All are non-negative numbers; For the first The ratio of respiratory band signal power to respiratory band noise power within each analysis window; For the first The echo phase continuity score for each analysis window, where the subscript... Indicates the echo phase; For the first The target tracking continuity score is analyzed in each window, where the subscript... Indicates target tracking; For the first The proportion of the sampling time for receptor dynamic contamination within a single analysis window to the total duration of that analysis window; For the first The proportion of missing data sampling points in an analysis window to the total number of sampling points in that analysis window; When the chest or abdominal breathing modality is unavailable, the corresponding window is not included in the calculation of chest and abdominal effort representation; when the cardiac response modality is unavailable, breathing modality evidence is retained and event confidence is reduced.
4. The sleep respiratory event recognition and cardiopulmonary load quantification system based on radio frequency chest and abdominal micro-movements according to claim 1, characterized in that, In the individual condition baseline module, the condition status includes sleep stage, body position category, and nighttime period; Conditional baseline amplitude for the z-th anatomical region in the w-th analysis window Calculate according to the following formula: In the formula, Let z be the respiratory amplitude of the z-th anatomical region in the u-th historical window. This is a set of historical windows that are consistent with or adjacent to the current w-th analysis window in terms of sleep stage, body position category, and nighttime period, and that meet quality requirements. Let u be the quality weight of the u-th historical window. and These represent the center times of the w-th analysis window and the u-th historical window, respectively. is the baseline time decay constant, exp is the exponential function, and wmed is the weighted median calculation; When set When the number of valid windows is less than a preset threshold, the individual conditional baseline module expands the baseline candidate set in the order of adjacent sleep stages under the same body position, adjacent body positions under the same sleep stage, and the user's multi-night historical baselines, and assigns decreasing confidence weights to the expansion sources.
5. The sleep respiratory event recognition and cardiopulmonary load quantification system based on radio frequency chest and abdominal micro-movements according to claim 4, characterized in that, The candidate event generation module calculates the amplitude suppression rate of the thoracic and abdominal joint in the w-th analysis window. for: In the formula, and These represent the current respiratory amplitude of the chest and abdomen, respectively. and These are the baseline amplitudes for the chest and abdomen, respectively. To prevent positive numbers with a denominator of zero; And calculate the effective support duration for the j-th candidate interval: In the formula, The window sliding step size, Let w be the window quality weight for the w-th analysis window. Let be the set of windows contained in the j-th candidate interval. For indicative functions; only if the calendar duration and quality of the candidate interval are valid support durations. When both the effective coverage rate and the quality reach the preset threshold, the candidate interval is retained.
6. The sleep respiratory event recognition and cardiopulmonary load quantification system based on radio frequency chest and abdominal micro-movements according to claim 5, characterized in that, The features extracted by the intra-event respiratory effort characterization module include thoracic-abdominal coherence and thoracic-abdominal phase anomalous index: In the formula, For the chest and abdomen coherence of the j-th candidate interval, Let be the phase anomaly index of the j-th candidate interval. For the w-th analysis window, the minimum weights for chest and abdominal mass are... Let be the difference in the analytical phase between chest and abdominal respiration in the w-th analysis window, where i is the imaginary unit, exp is the exponential function, and cos is the cosine function. and To prevent positive numbers with a denominator of zero; The intra-event respiratory effort characterization module further combines the chest-abdomen residual amplitude ratio and respiratory harmonic energy duration to construct a respiratory effort characterization value: In the formula, to The weights are non-negative and their sum is 1; Let j be the respiratory effort characterization value of the j-th candidate interval. These are the ratios of the residual amplitude of the chest and abdomen relative to their respective conditional baselines within the j-th candidate interval. Let be the duration of the fundamental frequency of breathing and its harmonics in the j-th candidate interval.
7. The sleep respiratory event recognition and cardiopulmonary load quantification system based on radio frequency chest and abdominal micro-movements according to claim 1, characterized in that, The post-event cardiopulmonary recovery response verification module constructs a cardiopulmonary recovery response vector within the recovery observation window after the candidate interval ends: In the formula, Let j be the cardiopulmonary recovery response vector of the j-th candidate interval. This is the bounce ratio of the maximum post-event respiratory amplitude corresponding to the j-th candidate interval relative to the pre-event conditional baseline. Let be the difference between the maximum heart rate after the event and the median heart rate before the event corresponding to the j-th candidate interval. Let be the root mean square change in the difference between adjacent heartbeat intervals before and after the event corresponding to the j-th candidate interval. Let be the proximity response between the end of the j-th candidate interval and the moment when the most recent sleep stage became lighter or when body movement occurred. The recovery time required for the respiratory amplitude or heart rate corresponding to the j-th candidate interval to return to the baseline tolerance range; The event fusion and load output module calculates the probability distribution of candidate intervals belonging to each morphological type based on the intra-event respiratory effort characterization features and the cardiopulmonary recovery response vector. In the formula, Let be the probability that the j-th candidate interval belongs to the k-th morphological type. The score is the discrimination score for the j-th candidate interval belonging to the k-th morphological type. The score is the discrimination score for the j-th candidate interval belonging to the h-th morphological type. The set of predefined morphological types includes at least the following: respiratory effort deficit, respiratory effort maintenance, mixed switching, and hypoventilation. And calculate the comprehensive event confidence of the j-th candidate interval. : In the formula, σ is the Sigmoid mapping function. Bias parameters, to The fusion coefficient is... This represents the average value of the joint amplitude suppression rate within the j-th candidate interval. This represents the maximum probability of each morphological type in the j-th candidate interval. This represents the average quality weight within the j-th candidate interval. The ratio of bodily contamination between the j-th candidate interval and the post-event recovery interval; Candidate intervals whose overall event confidence reaches a preset event confirmation confidence threshold will be included in the confirmation event index set: In the formula, This is the set of candidate interval indices for the confirmed event; Set a preset event confirmation confidence threshold; This represents the total number of candidate intervals. To confirm the total number of events; Represents a set The number of elements contained therein.
8. The sleep respiratory event recognition and cardiopulmonary load quantification system based on radio frequency chest and abdominal micro-movements according to claim 7, characterized in that, The quality-weighted sleep apnea load index output by the event fusion and load output module includes: Quality-weighted AHI related load : In the formula, The effective sleep duration for quality is expressed in minutes, and subscript Indicates sleep period; This is the set of candidate interval indices for the confirmed event; For the first The overall event confidence of each candidate interval; To confirm the total number of events; Quality-weighted low ventilation continuous load : In the formula, Let be the duration of the j-th candidate interval, expressed in minutes; Post-event load recovery : In the formula, z(·) is the directional consistency normalization function completed using individual history or population reference distribution; to These are non-negative recovery feature weights; Comprehensive load index : In the formula, to These are non-negative combination weights, and their sum is 1; Confidence-weighted proportions of events related to effort to maintain breathing and mixed transition events; Each load indicator is associated with and outputs the start timestamp, end timestamp, morphology type, confidence level, and waveform evidence fragments of the corresponding event.
9. A method for sleep respiratory event identification and cardiopulmonary load quantification based on radiofrequency chest and abdominal micro-movements, implemented based on the system described in any one of claims 1 to 8, characterized in that, The method includes the following steps: Step 1: Transmit radio frequency signals to the sleep area via radar and collect complex echo signals modulated by the micro-movements of the human body; Step 2: Locate and track the chest and abdominal scattering regions within the spatial domain; Step 3: Based on the complex echoes from the chest and abdominal scattering regions, demodulate and reconstruct the chest respiratory displacement sequence, abdominal respiratory displacement sequence, and cardiac response sequence. Step 4: Perform signal quality assessment on the analysis window to obtain the availability label and quality weight of each mode; Step 5: Based on the current sleep stage, body position category, and time-related conditions in the analysis window, establish a baseline for individual conditional respiratory amplitude in the chest and abdomen. Step 6: Generate candidate intervals based on the degree, duration, and effective quality of inhibition of thoracoabdominal breathing amplitude; Step 7: Extract the characteristics of thoracic-abdominal coherence, phase anomaly, and residual respiratory effort within the candidate interval; Step 8: Using the end time of the candidate interval as the alignment point, extract the post-event respiratory rebound, heart rate surge, changes in autonomic nervous regulation, and arousal or body movement response to construct the cardiopulmonary recovery response; Step 9: Integrate evidence of ventilatory depression, intra-event respiratory effort characterization, post-event cardiopulmonary recovery response, and quality weights to output event confidence, morphological classification results, evidence time segments, and quality-weighted load indices.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the method of claim 9.
Citation Information
Patent Citations
Sleep apnea detection device
CN102641125B
A non-contact method and system for detecting sleep apnea
CN108392186B
OSAHS diagnosis system based on recurrent neural network and millimeter waves
CN112686213A
Sleep apnea intelligent identification method based on hybrid neural network
CN119700017A
