A non-contact respiratory and cardiac monitoring and fall detection system and method
By combining millimeter-wave radar and a signal processing board, non-contact monitoring of breathing, heartbeat, and falls has been achieved, solving the problems of discomfort when wearing the device and high misjudgment rate, and providing a privacy-friendly home-based elderly care monitoring solution.
Patent Information
- Application Number
- CN202611219787.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-12
- Publication Date
- 2026-11-10
AI Technical Summary
Existing home-based elderly care monitoring devices suffer from problems such as discomfort when worn, high misjudgment rate, privacy concerns, and inability to detect abnormal conditions in the elderly in a timely manner. In particular, they are difficult to achieve continuous, unobtrusive, and privacy-friendly respiratory and heart rate monitoring and fall detection in complex home environments.
It employs millimeter-wave radar for non-contact monitoring, combined with adaptive background suppression, respiratory and heart rate co-calculation, and multi-stage causal chain recognition for falls. Through the main control and signal processing board, it achieves coordinated monitoring of respiratory status, heart rate, and body movement, and constructs an alarm link to adapt to changes in the home environment.
It enables stable monitoring of breathing, heart rate, and falls in complex home environments, reduces false alarm rates, provides privacy-friendly long-term monitoring, and is simple in structure and easy to deploy, making it suitable for home-based elderly care safety monitoring.
Smart Images

Figure CN123058684A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of home-based elderly care technology, specifically relating to a non-contact respiratory and heart rate monitoring and fall detection system and method. Background Technology
[0002] With an aging population and increasing demand for home-based elder care, there is a significant increase in demand for continuous, discreet, and privacy-friendly respiratory and heart rate monitoring and fall risk management among elderly people living alone, empty-nest elderly, disabled or semi-disabled elderly, and those recovering from chronic diseases. Traditional monitoring devices mostly use contact-based data collection methods, such as heart rate belts, ECG patches, finger-clip pulse oximeters, wristwatches, or mattress pressure sensors. While these devices can obtain data under specific conditions, they typically need to be worn close to the body, adhered to the skin, or pressed under the body. Long-term use can easily lead to problems such as a feeling of constriction, skin discomfort, falling off during sleep, forgetting to wear them, or interruption of charging, making them unsuitable for the discreet monitoring needs of elderly people who spend long periods of time alone at home.
[0003] Camera-based monitoring devices can perform some posture recognition and fall detection, but they rely on lighting conditions and are easily affected by obstructions. Furthermore, the images captured by cameras pose significant privacy concerns in private home settings such as bedrooms and living areas. Manual phone calls, home visits, or scheduled checks by family members are insufficient to promptly detect risk events such as abnormal breathing and heartbeats at night, prolonged periods of stillness, failure to return to bed, or inability to seek help independently after a fall.
[0004] Millimeter-wave radar can detect minute displacements in the human chest or abdomen using electromagnetic waves. It is non-contact, can penetrate ordinary clothing and bedding, does not rely on light, and does not require the acquisition of human images, making it suitable for continuous monitoring in home-based elderly care. However, the home environment is complex, with bed frames, walls, bedside tables, curtains, furniture, fans, and the activities of caregivers generating complex reflections. Changes in the elderly's sleeping posture, turning over, getting up, and differences in day and night activity can also cause fluctuations in respiratory and heart rate calculations. Existing solutions, if using only fixed filters or a single threshold, are prone to misjudging short-term large-amplitude movements, sitting up, getting out of bed, or family members entering as falls, or struggling to stably extract heart rate information under strong respiratory harmonic interference. Summary of the Invention
[0005] This invention addresses the shortcomings of existing technologies by providing a non-contact respiratory and heart rate monitoring and fall detection system and method. It combines adaptive home environment sensing, coordinated respiratory and heart rate calculation, multi-stage fall confirmation, day and night behavior baseline, audible and visual alarms, and remote early warning from family members / communities to achieve home-based elderly care safety monitoring that is privacy-friendly, long-term usable, and has controllable false alarms.
[0006] This invention provides the following technical solution:
[0007] Firstly, a non-contact respiratory and cardiac monitoring and fall detection system is provided, including: The millimeter-wave radar sensing module is fixed in the area to be monitored and is used to transmit detection signals and receive echo signals according to a preset sampling strategy. The main control and signal processing board includes a preprocessing module, a distance gate selection module, a respiratory and heart rate monitoring module, and a fall multi-stage causal chain recognition module; The preprocessing module is used to perform time-distance processing on the received echo signals and filter them to obtain an effective distance gate. The range gate selection module is used to calculate the joint confidence of each effective range gate and select the main range gate or the fused range gate; the fused range gate is the fused echo of multiple consecutive high-confidence effective range gates; The respiratory and cardiac coordinated monitoring module is used to synchronously extract phase micro-displacement and amplitude envelope sequences from the echoes of the main distance gate or fusion distance gate, and estimate the time-varying trajectory of the respiratory dominant frequency and output the respiratory rate after joint spectral and time-domain analysis. Based on the time-varying trajectory of the respiratory dominant frequency, a prediction reference signal including the respiratory fundamental wave and at least one harmonic is generated. The prediction reference signal is used to adaptively cancel the phase micro-displacement sequence and amplitude envelope sequence to obtain the candidate residual signal of the heartbeat. Then, the joint confidence of the heartbeat is calculated, and the heart rate estimate is output based on the joint confidence of the heartbeat. The fall multi-stage causal chain recognition module is used to construct a body movement detection area, extract body movement and fall features, identify the current body movement state, and output a confirmation of fall when body movement states occur in a preset order within a continuous stage.
[0008] Optionally, the preprocessing module is used to perform time-distance processing on the received echo signal and filter to obtain an effective distance gate, specifically: The received echo signal is processed by time-distance processing to obtain the echo energy of different distance units; Extract the phase abrupt change rate, amplitude stability and short-time signal-to-noise ratio characteristics of each distance cell, and simultaneously quantify the target's body motion strength, current day-night pattern and environmental changes; Based on amplitude stability and environmental changes, the background suppression coefficient is dynamically mapped, and combined with phase mutation rate, short-time signal-to-noise ratio characteristics, body motion strength and current day-night pattern, the outlier removal threshold is dynamically mapped. Based on the pre-acquired background template and the mapped background suppression coefficient, background removal is performed on the echo energy of each distance cell; Based on the outlier removal threshold, distance cells with outlier scores exceeding the threshold are removed. After spatial clustering and inter-frame continuity verification of the remaining distance cells, all valid distance gates are selected.
[0009] Optionally, the distance gate selection module is used to calculate the joint confidence of each effective distance gate and select the main distance gate or the fused distance gate, specifically: The joint confidence level of each effective distance gate is calculated based on echo energy, phase stability, amplitude coherence, respiratory cycle continuity, and body motion interference. Select the effective range gate with the highest joint confidence as the primary range gate, or select the fused range gate using the following methods: Multiple effective range gates with joint confidence exceeding a set threshold and adjacent to each other are selected. An adaptive weight for each effective range gate is determined based on the normalized joint confidence of each effective range gate. The complex echoes of the selected effective range gates are then weighted and fused to obtain the fused range gate.
[0010] Optionally, the respiratory and heart rate co-monitoring module synchronously extracts phase micro-shift and amplitude envelope sequences from the echoes of the main distance gate or the fusion distance gate, and estimates the time-varying trajectory of the respiratory dominant frequency after joint spectral and time-domain analysis, specifically as follows: Phase unwrapping is performed on the echoes from the main range gate or the fused range gate to obtain the phase micro-shift sequence. and amplitude envelope sequence Subsequently, a dual-channel micro-motion characterization was formed. ; Characterization of dual-channel micro-motion using a sliding window with a set length and a set step size. Perform joint spectral and time-domain analysis to extract the respiratory spectral peak energy of each sliding window within the respiratory frequency band. Main peak width ratio Periodic autocorrelation coefficient and continuity of adjacent windows , constitute respiratory feature vector ; The respiratory fit score of each sliding window is calculated based on the respiratory feature vector, and the time-varying trajectory of the main respiratory frequency is found according to the following formula, and the output respiratory rate is calculated simultaneously. ; ; in, For the first The time-varying trajectory of the breathing frequency of a sliding window. For any breathing frequency within the breathing frequency band, For the first The breathing adaptation score for each sliding window. For the breathing frequency band, This is the frequency jump penalty coefficient. For the output of the first The breathing rate of each sliding window.
[0011] Optionally, the respiratory and cardiac coordinated monitoring module generates a predictive reference signal including the respiratory fundamental wave and at least one harmonic based on the time-varying trajectory of the respiratory dominant frequency. The predictive reference signal is then used to adaptively cancel the phase micro-shift sequence and the amplitude envelope sequence to obtain a candidate residual signal for the heartbeat, specifically: The prediction reference signal is generated according to the following formula: ; Breathing prediction cancellation is performed on the phase micro-shift sequence and amplitude envelope sequence to obtain the phase residual sequence and amplitude residual sequence; ; in, To predict the reference signal, The harmonic order is... and The first First-order respiratory component in the first The amplitude and phase of each sliding window, For sampling points, For the first The time-varying trajectory of the breathing frequency of a sliding window. The sampling period is and These are phase micro-displacement sequences and amplitude envelope sequences, respectively. It is a phase residual sequence. For amplitude residual sequences, and These are the adaptive cancellation coefficients for phase and amplitude, respectively.
[0012] Optionally, the respiratory and cardiac coordinated monitoring module, after obtaining the candidate residual signal of the heartbeat, calculates the joint confidence of the heartbeat and outputs the estimated heart rate value based on the joint confidence of the heartbeat, specifically as follows: Calculate the phase residual sequence peak score, amplitude residual sequence peak score, frequency continuity score, physiological range score, and respiratory harmonic overlap penalty for each sliding window in the cardiac frequency band to form the cardiac joint confidence score. Determine whether the combined confidence level of the heartbeat in the current sliding window is not less than a set threshold. If not, prevent the triggering of a high-level heartbeat abnormality alarm in the current time window. If so, output the heart rate estimate based on the combined spectral peak of the phase residual sequence and the amplitude residual sequence.
[0013] Optionally, the fall multi-stage causal chain recognition module is used to construct a body movement detection region, extract body movement and fall features, identify the current body movement state, and output a confirmation of fall when body movement states occur in a preset order within consecutive stages. Specifically: The area to be monitored was divided into the bedside area, chest and abdominal micro-movement area, limb movement area and landing candidate area, and the body movement and fall feature groups of each sub-region were extracted in the form of a sliding window. The entropy weight method is used to calculate the contribution weight of each feature in the body movement and fall feature group, and the state fit score of each sub-region in the current sliding window is calculated. Based on the state adaptation scores of each sub-region in several sliding windows and the human activity posture feature vector, it is determined whether a complete state chain of falls is triggered under the set time constraints. If so, the fall is confirmed; otherwise, monitoring continues.
[0014] Optionally, determine whether the complete state chain of the fall is triggered under the set time constraint, specifically: Determine whether a pre-fall instability has been triggered and record the trigger time. At the trigger time The set time afterwards Within the specified timeframe, confirm the impact landing. If the impact landing is confirmed, record the trigger time. At the trigger time The set time after Inside, identify whether it is a post-fall state. If the above process is met, the fall is confirmed; otherwise, continue monitoring.
[0015] Optionally, it also includes an alarm module, a wireless communication module, a power supply component, and a housing that encloses all modules. The alarm module is used to issue an alarm based on the output respiratory rate, heart rate estimate, and whether a fall has occurred. The wireless communication module is used to push the respiratory rate, heart rate estimate, and whether a fall has occurred to a designated terminal.
[0016] Secondly, a non-contact method for monitoring respiratory and cardiac function and for detecting falls is provided, including: The system transmits detection signals and receives echo signals according to a preset sampling strategy. The received echo signal is processed for time and distance, and then filtered to obtain an effective distance gate; Calculate the joint confidence of each effective range gate, and select the main range gate or the fused range gate; the fused range gate is the fused echo of multiple consecutive high-confidence effective range gates; Phase micro-displacement and amplitude envelope sequences are synchronously extracted from the echoes of the main distance gate or fusion distance gate. After joint spectral and temporal analysis, the time-varying trajectory of the respiratory dominant frequency and the output respiratory rate are estimated. A prediction reference signal including the respiratory fundamental wave and at least one harmonic is generated based on the time-varying trajectory of the respiratory dominant frequency. The phase micro-displacement sequence and amplitude envelope sequence are adaptively canceled using the prediction reference signal to obtain the candidate residual signal of the heartbeat. The joint confidence of the heartbeat is calculated, and the heart rate estimate is output based on the joint confidence of the heartbeat. After constructing a body movement detection region and extracting body movement and fall features, the current body movement state is identified, and when a body movement state occurs in a preset order within a continuous phase, a fall confirmation is output.
[0017] Compared with the prior art, the beneficial effects of the present invention are: (1) This invention uses millimeter-wave radar for non-contact respiratory and heart rate monitoring. The monitored elderly do not need to wear heart rate belts, patches, finger clips, or wristwatches, which can reduce discomfort, forgetting to wear them, and compliance issues in long-term home monitoring. It does not rely on cameras to collect human images and can be used in privacy-sensitive scenarios such as the elderly’s bedroom, bedside, and living area, achieving continuous monitoring without exposing home life. This invention integrates the millimeter-wave radar sensing module with the main control signal processing board to achieve coordinated monitoring of respiratory status, heart rate, body movement, getting out of bed, and fall status, rather than just monitoring the presence of the human body or single respiratory status. This is more in line with the actual needs of “daily care + abnormal early warning” in home-based elderly care. This invention improves monitoring stability under conditions of bed frame, bedding, furniture reflection, turning over, and environmental vibration by suppressing static backgrounds at home, multi-distance gate phase-amplitude quality assessment, collaborative fusion, low-confidence isolation, and personalized health baseline updates during both sleep and daytime periods. The invention integrates local display, audible and visual alarms, wireless communication, family member mobile terminals, community elderly care service terminals, and preset emergency contact terminals into a home-based elderly care alarm chain, enabling abnormal events to quickly reach the relevant personnel through general reminders, key reminders, or emergency alarms. The invention has a simple structure and is easy to deploy; it can be wall-mounted, fixed to the headboard, bedside, or placed on a table. It can also be temporarily deployed between different rooms or resting locations in the home via magnetic attachment and a portable strap, making it suitable for long-term safety monitoring of home-based elderly care.
[0018] (2) By adaptively adjusting the background suppression coefficient and the outlier removal threshold, the present invention can make the selection of the effective distance gate simultaneously have the evaluation criteria of phase stability, amplitude coherence and continuity, and can reduce the fluctuations caused by changes in sleeping posture, bedding obstruction, bedside furniture and wall reflection.
[0019] (3) The method of combining respiratory time-varying harmonic prediction cancellation and cardiac beat dual evidence residual reconstruction is adopted: the respiratory fundamental frequency and harmonic prediction signals are generated according to the evolution trend of the respiratory dominant frequency in adjacent time windows, which suppresses strong respiratory leakage while retaining weak cardiac beat residuals; then the consistency of phase residual spectrum peak and amplitude envelope modulation is used to jointly confirm the heartbeat characteristics, thereby improving the stability and usability of heart rate calculation under low signal-to-noise ratio conditions at home.
[0020] (4) The present invention adopts a multi-stage causal chain identification mechanism of pre-fall instability - impact landing - post-fall recovery / continuous stillness, and uses motion changes, distance migration, echo energy mutation, micro-Doppler evolution, respiratory and heartbeat continuity and home day and night behavior baseline for time-series verification; compared with the fall detection method that only relies on a single threshold trigger, it can effectively distinguish between rolling over, sitting up, getting out of bed, family members passing by, short-term large body movements and real fall events, and can implement graded early warning by combining impact intensity and recovery status. Attached Figure Description
[0021] Figure 1 This is an overall block diagram of the non-contact respiratory and heart rate monitoring and fall detection system of the present invention; Figure 2 This is a flowchart of the steps of the non-contact respiratory and heart rate monitoring and fall detection method of the present invention. Detailed Implementation
[0022] The present invention will be further described below with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solutions of the present invention and should not be used to limit the scope of protection of the present invention. It should be noted that the term "comprising" and any variations thereof in the specification, claims and the above-mentioned drawings of the present invention are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to these processes, methods, products or devices.
[0023] Example 1: like Figure 1 As shown, a non-contact respiratory and heart rate monitoring and fall detection system includes: a millimeter-wave radar sensing module, a main control and signal processing board, an alarm module, a local display module, a wireless communication module, a power supply component, and a housing that encloses all modules.
[0024] I. Casing, power supply components, and local display module.
[0025] The housing is used to enclose all modules. It consists of an upper housing and a lower housing, made of PC+ABS material and secured with clips, screws, or ultrasonic welding. A radar-transparent front cover or sensor panel can be installed on the upper housing at the radar sensor window. Preferably, it uses plastic or composite materials with low millimeter-wave loss, and its outer surface is rounded, dustproof, and scratch-resistant to suit home installation environments such as beside a bed, on a wall, or on a desktop.
[0026] The power supply components include a USB-C charging and power supply module, a charging and battery management module, and a rechargeable lithium battery. The USB-C charging and power supply module is located on the side or bottom of the casing. The charging and battery management module connects to the rechargeable lithium battery and provides charging protection, overvoltage protection, overcurrent protection, over-discharge protection, and temperature protection. The power supply components provide a stable operating voltage for the millimeter-wave radar sensing module, main control and signal processing board, wireless communication module, local display module, and alarm module. During stable rest at night, it reduces the power consumption of unnecessary modules and increases the sampling and reporting frequency in cases of fall detection or abnormal breathing and heartbeat.
[0027] The casing is usually equipped with a local display module, which is used to display the current respiratory rate, heart rate estimate and whether a fall has occurred based on the trigger status (such as display triggered by a button), or to automatically display the respiratory rate, heart rate estimate and whether a fall has occurred locally when an alarm is triggered.
[0028] The back of the lower cover features a magnetic backplate or wall mount, allowing it to be secured to walls, headboards, bedside tables, or other surfaces via magnetic clips, mounting holes, adhesive, or screws. A portable handle assembly can also be installed on the cover, connecting to it for easy movement between different areas of the home, such as bedrooms, living rooms, and relaxation areas. For families requiring coverage of both the bedside and frequently used activity areas, multiple devices can be deployed, or calibration of the respective monitoring areas can be performed separately through mobile installation.
[0029] II. Millimeter-wave radar sensing module.
[0030] It is fixed in the area to be monitored and is used to transmit detection signals and receive echo signals according to a preset sampling strategy.
[0031] The millimeter-wave radar sensing module can employ a millimeter-wave radar chip, radar module, or equivalent combination thereof with human body echo acquisition capabilities, without limitation on specific chip models, board forms, or operating frequency bands. The millimeter-wave radar sensing module is located inside the radar-transparent front cover or sensing panel of the casing, with its antenna array area facing the home monitoring area. The millimeter-wave radar sensing module includes a communication interface, a debugging or configuration interface, a power interface, a millimeter-wave radar transceiver unit, and an antenna array area. The communication interface connects to the main control and signal processing board for outputting sampled data or intermediate processed data; the debugging or configuration interface is used for reset, startup, or parameter configuration; and the power interface connects to the power supply components.
[0032] III. Main Control and Signal Processing Board.
[0033] The main control and signal processing board is located below the millimeter-wave radar sensing module or within the same housing (inside the outer casing). The main control and signal processing board can be an MCU, DSP, FPGA, embedded SoC, or edge computing module. After receiving data from the millimeter-wave radar sensing module, the main control and signal processing board performs home static background suppression, multi-distance gate phase-amplitude collaborative fusion, respiratory / heartbeat collaborative decoupling calculation, fall multi-stage causal chain identification, home day and night behavior baseline update, alarm strategy judgment, and communication control.
[0034] IV. Wireless communication module.
[0035] It can be detachably or welded to the vicinity of the main control and signal processing board, and can choose between Wi-Fi, Bluetooth, cellular communication, or low-power wide-area communication depending on the home network conditions. Home users can connect to family members' mobile phone apps via Wi-Fi or Bluetooth; when network conditions are limited or remote monitoring is required, the device can send monitoring status and alarm information to community elderly care service terminals, home care terminals, or preset emergency contacts via cellular communication or low-power wide-area communication.
[0036] V. Alarm Module.
[0037] The alarm module connects to the main control and signal processing board. When abnormal breathing, abnormal heartbeat, prolonged absence from bed, suspected fall, confirmed fall, low battery, device offline, or sensor malfunction occurs, the main control and signal processing board controls the alarm module to output a buzzer, flashing indicator light, or screen alarm notification. The local display module shows respiratory rate, heart rate, battery level, network connection status, and alarm status. The elderly person, their family members, or home caregivers can use buttons to mute, reset, confirm alarms, or configure network settings.
[0038] The main control and signal processing board includes: a preprocessing module, a distance gate selection module, a respiratory and heart rate monitoring module, and a fall multi-stage causal chain recognition module.
[0039] VI. Preprocessing Module.
[0040] Used to process the received echo signal in time and distance and filter to obtain an effective distance gate.
[0041] The specific method for obtaining the effective distance door is as follows: Step S1: Perform time-distance processing on the received echo signal to obtain the echo energy of different distance cells.
[0042] The time-spacing processing method refers to existing technologies, such as Fast Fourier Transform, etc. Of course, some other conventional preprocessing methods can also be performed before time-spacing processing.
[0043] Step S2: Extract the phase abrupt change rate, amplitude stability and short-time signal-to-noise ratio features of each distance cell, and simultaneously quantify the target's body motion strength, current day / night pattern and environmental changes.
[0044] The current day and night mode can be mapped based on clock signals, etc. The strength of the target's motion can be quantified by analyzing the micro-Doppler characteristics or energy accumulation of the echo. Environmental changes can be calculated by calculating several environmental features and then mapping based on all environmental features, such as weighted fusion. Environmental features include: electromagnetic background drift, physical structure differences, air medium shift and / or airflow disturbance index, etc., which can be selected and configured based on expert experience.
[0045] Step S3: Dynamically map the background suppression coefficient based on amplitude stability and environmental changes; combine phase mutation rate, short-time signal-to-noise ratio characteristics, body motion strength and current day / night mode to dynamically map the outlier removal threshold.
[0046] That is, the background suppression coefficient is adaptively adjusted based on amplitude stability and environmental changes, combined with a predefined mapping function. The outlier removal threshold is adaptively adjusted based on phase abrupt change rate, short-time signal-to-noise ratio characteristics, body motion intensity, and current day / night pattern, combined with a predefined mapping function. The specific selection method for the mapping function can refer to existing technologies, such as lookup table mapping or fuzzy logic.
[0047] Step S4: Based on the pre-acquired background template and the mapped background suppression coefficient, perform background removal on the echo energy of each distance cell.
[0048] By establishing an empty field or a long-term static background reflection template during the installation and calibration period, the background suppression coefficient can be used to perform template subtraction on fixed reflections such as walls, bed frames, bedside tables, bedding, and furniture.
[0049] Step S5: Based on the outlier removal threshold, remove distance cells whose outlier scores exceed the outlier removal threshold, and after performing spatial clustering and inter-frame continuity verification on the remaining distance cells, select all valid distance gates.
[0050] The anomaly score for the distance cell can be obtained by referring to existing technologies, such as calculating the anomaly score based on phase abrupt change rate, amplitude stability, and normalized signal-to-noise ratio. Spatial clustering and inter-frame continuity verification can also refer to existing technologies.
[0051] VII. Distance Door Selection Module.
[0052] Used to calculate the joint confidence of each effective range gate, and select the main range gate or the fused range gate.
[0053] The fusion range gate is the fusion echo of multiple consecutive high-confidence effective range gates. The specific steps for selecting the main range gate or the fusion range gate are as follows: Step E1: Calculate the joint confidence of each effective distance gate based on echo energy, phase stability, amplitude coherence, respiratory cycle continuity, and body motion interference.
[0054] Calculate the joint confidence level using the following formula; ; in, For the joint confidence level of the effective distance gate, To score the normalized echo energy, To score phase stability, For amplitude coherence scoring, For respiratory cycle continuity scoring, For scoring body motion interference, the scores of each item in the joint confidence formula can be determined with reference to existing techniques. , , , and These are the preset or adaptive weights corresponding to each score.
[0055] Step E2: Select the effective range gate with the highest joint confidence as the main range gate, or select the fused range gate using the following methods: Multiple adjacent effective range gates with a joint confidence score exceeding a set threshold are selected. An adaptive weight for each effective range gate is determined based on its normalized joint confidence score. The complex echoes of the selected effective range gates are then weighted and fused to obtain the fused range gate. The specific formula is as follows: ; ; in, For the complex echoes of each adjacent effective range gate with a joint confidence level exceeding a set threshold, and The first The effective distance gate is at the... In-phase and quadrature components at each sampling point For the first The weight of each effective distance gate, For the first Joint confidence of each effective distance gate, For multiple valid distance gate sets whose joint confidence exceeds a set threshold and are adjacent Any one of them, To prevent tiny constants with a denominator of zero, This is the fused echo sequence corresponding to the fused range gate.
[0056] 8. Respiratory and cardiac monitoring module.
[0057] The specific steps for performing respiratory and heart rate coordination calculations are as follows: Step D1: Synchronously extract the phase micro-shift and amplitude envelope sequences from the echoes of the main distance gate or the fusion distance gate, and estimate the time-varying trajectory of the respiratory main frequency and the output respiratory rate after joint spectral and time domain analysis.
[0058] Step D1 is as follows: Step D11: Perform phase unwrapping on the echoes from the main range gate or the fused range gate to obtain the phase micro-shift sequence. and amplitude envelope sequence Subsequently, a dual-channel micro-motion characterization was formed. .
[0059] ; in, This indicates the fused echo sequence corresponding to the fused range gate or the echo sequence corresponding to the main range gate. The sequence after phase unwrapping. The argument is expressed in radians. This refers to the wavelength of millimeter-wave radar signals.
[0060] Step D12: Characterize the dual-channel micro-motion using a sliding window with a set length and a set step size. Perform joint spectral and time-domain analysis to extract the respiratory spectral peak energy of each sliding window in the respiratory frequency band. Main peak width ratio Periodic autocorrelation coefficient and continuity of adjacent windows , constitute respiratory feature vector .
[0061] The joint spectral and time-domain analysis method references existing techniques, such as Fourier transform. (Respiratory frequency band) , and These are the preset minimum and maximum respiratory rates, respectively.
[0062] Step D13: Calculate the respiratory adaptation score of each sliding window based on the respiratory feature vector, find the time-varying trajectory of the main respiratory frequency according to the following formula, and simultaneously calculate the output respiratory rate.
[0063] ; ; in, For the first The time-varying trajectory of the breathing frequency of a sliding window. For any candidate frequency in the breathing frequency band, For the first The breathing fit score for each sliding window is obtained either by weighting each breathing feature or by mapping each breathing feature to a lookup table. For the breathing frequency band, This is the frequency jump penalty coefficient. For the output of the first The breathing rate of each sliding window.
[0064] Step D2: Generate a predictive reference signal including the respiratory fundamental wave and at least one harmonic based on the time-varying trajectory of the respiratory dominant frequency. Use the predictive reference signal to adaptively cancel the phase micro-displacement sequence and amplitude envelope sequence to obtain the candidate residual signal of the heartbeat.
[0065] Step D2: Specifically: Step D21: Generate a predictive reference signal according to the following formula to describe the leakage of the strong respiratory component in the candidate frequency band of the heartbeat.
[0066] ; Step D22: Perform respiratory prediction cancellation on the phase micro-shift sequence and amplitude envelope sequence to obtain the phase residual sequence and amplitude residual sequence. The cancellation only applies to the narrowband components associated with the respiratory fundamental wave and its harmonics to preserve the weak residuals of adjacent heartbeats.
[0067] ; in, To predict the reference signal, The harmonic order is... and The first First-order respiratory component in the first The amplitude and phase of each sliding window, For sampling points, For the first The time-varying trajectory of the breathing frequency of a sliding window. The sampling period is and These are phase micro-displacement sequences and amplitude envelope sequences, respectively. It is a phase residual sequence. For amplitude residual sequences, and These are the adaptive cancellation coefficients for phase and amplitude, respectively.
[0068] Step D3: Calculate the confidence level of the combined heartbeat and output the heart rate estimate based on the confidence level of the combined heartbeat.
[0069] Step D3: Specifically: Step D31: Calculate the phase residual sequence peak score, amplitude residual sequence peak score, frequency continuity score, physiological range score, and respiratory harmonic overlap penalty for each sliding window in the heartbeat frequency band (usually referring to the main heartbeat frequency of the current sliding window) to form the joint confidence score of the heartbeat.
[0070] ; in, The joint confidence level of heartbeats in the current sliding window. , , , and They are heart rate bands The phase residual sequence peak score, amplitude residual sequence peak score, frequency continuity score, physiological range score, and respiratory harmonic overlap penalty are included. , , , and They are respectively , , , and The corresponding weights.
[0071] Step D32: Determine whether the combined confidence level of the heartbeat in the current sliding window is not less than a set threshold. If not, then the triggering of a high-level heart rate abnormality alarm in the current time window is prohibited; if yes, then the heart rate estimate is output based on the joint spectral peak of the phase residual sequence and the amplitude residual sequence.
[0072] ; in, This is a heart rate estimate. Heart rate Any candidate frequency, For the first Phase residual sequence within a sliding window At candidate frequency Peak score at the location, For the first Amplitude residual sequence within a sliding window At candidate frequency Peak score at the location, Candidate frequencies Penalty term for overlap with respiratory fundamental frequency or respiratory harmonics. , and They are respectively , and Corresponding weights.
[0073] 9. Fall multi-stage causal chain identification module.
[0074] It is used to construct a body movement detection region, extract body movement and fall features, identify the current body movement state, and output a confirmation fall when body movement states occur in a preset order within a continuous phase.
[0075] The specific steps for body movement recognition and multi-stage causal chain recognition of falls using the fall multi-stage causal chain recognition module are as follows: Step i: Divide the area to be monitored into the bedside area, chest and abdominal micro-movement area, limb movement area and landing candidate area, and extract the body movement and fall feature groups of each sub-region in the form of a sliding window.
[0076] Based on the installation location of the millimeter-wave radar sensing module, or by conducting prior sensing, the area to be monitored is divided into the bedside area, the chest and abdomen micro-movement area, the limb movement area, and the landing candidate area.
[0077] Extract the peak phase change rate for each sub-region Energy mutation amount Distance from center migration Velocity spectrum broadening MicroDoppler transient intensity and static duration This forms candidate feature vectors for body movement and falls. .
[0078] ; ; in, For the first The sub-regions in the first The phase change between each sampling point and its adjacent sampling points The sampling time interval between adjacent sampling points. For the first The sub-regions in the first Echo energy or accumulated echo energy within a sliding window For the first The target distance from the center within each sliding window is used to calculate the peak phase change rate, energy mutation amount, and distance from the center migration amount, respectively.
[0079] Step ii: Calculate the contribution weight of each feature in the body movement and fall feature group using the entropy weight method, and score the state fit of each sub-region in the current sliding window.
[0080] ; ; ; in, and These are the first and second normalized values, respectively. The first sub-region One characteristic, and The first in each of all regions The maximum and minimum values of each feature; For the first The first sub-region The proportion of each normalized feature across all sub-regions. For the first Information entropy of each feature The number of sub-regions. For the first The difference coefficient of each feature For the first The contribution weight of each feature.
[0081] Membership functions were set for normal rest, turning over / short-term body movement, getting up, getting out of bed, pre-fall instability, impact landing, post-fall recovery, and post-fall static states, and a spatial discrete penalty term was introduced to obtain the fitness score of each sub-region for each state.
[0082] ; in, Indicates the state to be identified. For the first Each feature pairs the state membership function, For spatial discrete penalty terms, This is the penalty coefficient.
[0083] Step iii: Based on the state adaptation scores of each sub-region in several sliding windows and the human activity posture feature vector, determine whether a complete state chain of falls is triggered under the set time constraints. If yes, confirm the fall; otherwise, continue monitoring.
[0084] Determine whether a pre-fall instability has been triggered and record the trigger time. At the trigger time The set time after Within the specified timeframe, confirm the impact landing. If the impact landing is confirmed, record the trigger time. At the trigger time The set time after Inside, identify whether it is a post-fall state. If the above process is met, the fall is confirmed; otherwise, continue monitoring.
[0085] The specific criteria for determining whether a pre-fall instability has been triggered are: continuous A time window satisfies ; The scoring threshold is the corresponding score for the pre-fall instability state. For the first Each sub-region represents the fit score for the pre-fall instability state, and includes signs such as increased speed changes, abrupt changes in movement direction, broadening of the speed spectrum, or deviation of the target trajectory from the individual's day-night behavioral baseline.
[0086] Impact landing confirmation specifically refers to: at the trigger time The set time after Within, if the candidate area or adjacent sub-area meets the requirements... If at least two of the following conditions are met simultaneously: rapid migration from the center, sudden change in echo energy, and micro-Doppler transient enhancement, then the impact landing phase is confirmed. For the first Each sub-region represents the fit score for the impact landing state. The scoring threshold corresponds to the impact landing state.
[0087] Identifying whether it is a post-fall state specifically involves: at the trigger time The set time after If the target is located in the current event area and the duration of stillness exceeds the set threshold, while the continuity of breathing and heartbeat or low-amplitude micro-movements can still be detected, it is determined to be a continuous stillness after a fall. If there is a reverse getting-up trajectory, a return to the normal activity area from the center, or recovery of activity energy, it is determined to be a recovery after a fall.
[0088] For this invention, a confirmed fall is output only when the pre-fall instability, impact landing, and post-fall state meet the time sequence constraints; if only a single velocity threshold, a single energy mutation, or no sustained low activity after the fall is met, candidates for turning over, short-term body movement, getting up, or getting out of bed are output, without directly triggering an emergency alarm.
[0089] The time order constraint is: .
[0090] In this invention, the respiratory rate, heart rate, frequency of body movements, common times of getting up / getting out of bed, activity range, and alarm events of the user can also be recorded during sleep and daytime activity periods to form a personalized health and activity baseline. When the current data deviates from the health baseline and exceeds a threshold, the alarm abnormality level is increased. If the system detects the entry of family members or visitors, violent body movements, loss of target, or insufficient signal-to-noise ratio, the corresponding time window is marked as low-confidence data, which can prevent the triggering of high-level heart rate abnormality alarms or prevent participation in personalized health baseline updates.
[0091] Example 2: like Figure 2 As shown, a non-contact respiratory and cardiac monitoring and fall detection method includes: The system transmits detection signals and receives echo signals according to a preset sampling strategy. The received echo signal is processed for time and distance, and then filtered to obtain an effective distance gate; Calculate the joint confidence of each effective range gate, and select the main range gate or the fused range gate; the fused range gate is the fused echo of multiple consecutive high-confidence effective range gates; Phase micro-displacement and amplitude envelope sequences are synchronously extracted from the echoes of the main distance gate or fusion distance gate. After joint spectral and temporal analysis, the time-varying trajectory of the respiratory dominant frequency and the output respiratory rate are estimated. A prediction reference signal including the respiratory fundamental wave and at least one harmonic is generated based on the time-varying trajectory of the respiratory dominant frequency. The phase micro-displacement sequence and amplitude envelope sequence are adaptively canceled using the prediction reference signal to obtain the candidate residual signal of the heartbeat. The joint confidence of the heartbeat is calculated, and the heart rate estimate is output based on the joint confidence of the heartbeat. After constructing a body movement detection region and extracting body movement and fall features, the current body movement state is identified, and when a body movement state occurs in a preset order within a continuous phase, a fall confirmation is output.
[0092] For a more detailed explanation of the above method, please refer to the relevant content disclosed in the foregoing embodiments, which will not be repeated here.
[0093] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the systems disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple; relevant parts can be referred to the method section.
[0094] Those skilled in the art will clearly understand that the techniques in the embodiments of the present invention can be implemented using software plus necessary general-purpose hardware platforms. Based on this understanding, the technical solutions in the embodiments of the present invention, or the parts that contribute to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in various embodiments or certain parts of the embodiments of the present invention.
[0095] The above are merely preferred embodiments of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should be considered within the scope of protection of the present invention.
Claims
1. A non-contact respiratory and cardiac monitoring and fall detection system, characterized in that, include: The millimeter-wave radar sensing module is fixed in the area to be monitored and is used to transmit detection signals and receive echo signals according to a preset sampling strategy. The main control and signal processing board includes a preprocessing module, a distance gate selection module, a respiratory and heart rate monitoring module, and a fall multi-stage causal chain recognition module; The preprocessing module is used to perform time-distance processing on the received echo signals and filter them to obtain an effective distance gate. The distance gate selection module is used to calculate the joint confidence of each effective distance gate and select the main distance gate or the fused distance gate. The fusion range gate is a fusion echo of multiple consecutive high-confidence effective range gates; The respiratory and cardiac coordinated monitoring module is used to synchronously extract phase micro-displacement and amplitude envelope sequences from the echoes of the main distance gate or fusion distance gate, and estimate the time-varying trajectory of the respiratory dominant frequency and output the respiratory rate after joint spectral and time-domain analysis. Based on the time-varying trajectory of the respiratory dominant frequency, a prediction reference signal including the respiratory fundamental wave and at least one harmonic is generated. The prediction reference signal is used to adaptively cancel the phase micro-displacement sequence and amplitude envelope sequence to obtain the candidate residual signal of the heartbeat. Then, the joint confidence of the heartbeat is calculated, and the heart rate estimate is output based on the joint confidence of the heartbeat. The fall multi-stage causal chain recognition module is used to construct a body movement detection area, extract body movement and fall features, identify the current body movement state, and output a confirmation of fall when body movement states occur in a preset order within a continuous stage.
2. The non-contact respiratory and cardiac monitoring and fall detection system according to claim 1, characterized in that, The preprocessing module is used to perform time-distance processing on the received echo signal and filter to obtain an effective distance gate, specifically: The received echo signal is processed by time-distance processing to obtain the echo energy of different distance units; Extract the phase abrupt change rate, amplitude stability and short-time signal-to-noise ratio features of each distance cell, and simultaneously quantify the target's body motion intensity, current day-night pattern and environmental changes; Based on amplitude stability and environmental changes, the background suppression coefficient is dynamically mapped, and combined with phase mutation rate, short-time signal-to-noise ratio characteristics, body motion strength and current day-night pattern, the outlier removal threshold is dynamically mapped. Based on the pre-acquired background template and the mapped background suppression coefficient, background removal is performed on the echo energy of each distance cell; Based on the outlier removal threshold, distance cells with outlier scores exceeding the threshold are removed. After spatial clustering and inter-frame continuity verification of the remaining distance cells, all valid distance gates are selected.
3. The non-contact respiratory and cardiac monitoring and fall detection system according to claim 1, characterized in that, The distance gate selection module is used to calculate the joint confidence of each effective distance gate and select the main distance gate or the fused distance gate, specifically as follows: The joint confidence level of each effective distance gate is calculated based on echo energy, phase stability, amplitude coherence, respiratory cycle continuity, and body motion interference. Select the effective range gate with the highest joint confidence as the primary range gate, or select the fused range gate using the following methods: Multiple effective range gates with joint confidence exceeding a set threshold and adjacent to each other are selected. An adaptive weight for each effective range gate is determined based on the normalized joint confidence of each effective range gate. The complex echoes of the selected effective range gates are then weighted and fused to obtain the fused range gate.
4. The non-contact respiratory and cardiac monitoring and fall detection system according to claim 1, characterized in that, The respiratory and heart rate co-monitoring module synchronously extracts phase micro-shift and amplitude envelope sequences from the echoes of the main distance gate or fusion distance gate, and estimates the time-varying trajectory of the respiratory dominant frequency after joint spectral and time-domain analysis, specifically: Phase unwrapping is performed on the echoes from the main range gate or the fused range gate to obtain the phase micro-shift sequence. and amplitude envelope sequence Subsequently, a dual-channel micro-motion characterization was formed. ; Characterization of dual-channel micro-motion using a sliding window with a set length and a set step size. Perform joint spectral and time-domain analysis to extract the respiratory spectral peak energy of each sliding window within the respiratory frequency band. Main peak width ratio Periodic autocorrelation coefficient and continuity of adjacent windows , constitute respiratory feature vector ; The respiratory fit score of each sliding window is calculated based on the respiratory feature vector, and the time-varying trajectory of the main respiratory frequency is found according to the following formula, and the output respiratory rate is calculated simultaneously. ; ; in, For the first The time-varying trajectory of the breathing frequency of a sliding window. For any breathing frequency within the breathing frequency band, For the first The breathing adaptation score for each sliding window. For the breathing frequency band, This is the frequency jump penalty coefficient. For the output of the first The breathing rate of each sliding window.
5. The non-contact respiratory and cardiac monitoring and fall detection system according to claim 1, characterized in that, The respiratory and heart rate co-monitoring module generates a predictive reference signal including the respiratory fundamental wave and at least one harmonic based on the time-varying trajectory of the respiratory dominant frequency. The predictive reference signal is then used to adaptively cancel the phase micro-shift sequence and amplitude envelope sequence to obtain a candidate residual signal for the heartbeat, specifically: The prediction reference signal is generated according to the following formula: ; Breathing prediction cancellation is performed on the phase micro-shift sequence and amplitude envelope sequence to obtain the phase residual sequence and amplitude residual sequence; ; in, To predict the reference signal, The harmonic order is... and The first First-order respiratory component in the first The amplitude and phase of each sliding window, For sampling points, For the first The time-varying trajectory of the breathing frequency of a sliding window. The sampling period is and These are phase micro-displacement sequences and amplitude envelope sequences, respectively. It is a phase residual sequence. For amplitude residual sequences, and These are the adaptive cancellation coefficients for phase and amplitude, respectively.
6. The non-contact respiratory and cardiac monitoring and fall detection system according to claim 5, characterized in that, The respiratory and cardiac coordinated monitoring module, after obtaining the candidate residual signal of the heartbeat, calculates the joint confidence of the heartbeat and outputs the estimated heart rate value based on the joint confidence of the heartbeat, specifically as follows: Calculate the phase residual sequence peak score, amplitude residual sequence peak score, frequency continuity score, physiological range score, and respiratory harmonic overlap penalty for each sliding window in the cardiac frequency band to form the cardiac joint confidence score. Determine whether the combined confidence level of the heartbeat in the current sliding window is not less than a set threshold. If not, prevent the triggering of a high-level heartbeat abnormality alarm in the current time window. If so, output the heart rate estimate based on the combined spectral peak of the phase residual sequence and the amplitude residual sequence.
7. The non-contact respiratory and cardiac monitoring and fall detection system according to claim 1, characterized in that, The fall multi-stage causal chain recognition module is used to construct a body movement detection region, extract body movement and fall features, identify the current body movement state, and output a confirmation of fall when body movement states occur in a preset order within consecutive stages. Specifically: The area to be monitored was divided into the bedside area, chest and abdominal micro-movement area, limb movement area and landing candidate area, and the body movement and fall feature groups of each sub-region were extracted in the form of a sliding window. The entropy weight method is used to calculate the contribution weight of each feature in the body movement and fall feature group, and the state fit score of each sub-region in the current sliding window is calculated. Based on the state adaptation scores of each sub-region in several sliding windows and the human activity posture feature vector, it is determined whether a complete state chain of falls is triggered under the set time constraints. If so, the fall is confirmed; otherwise, monitoring continues.
8. The non-contact respiratory and cardiac monitoring and fall detection system according to claim 7, characterized in that, To determine whether a complete state chain triggers a fall under the set time constraints, the following steps are taken: Determine whether a pre-fall instability has been triggered and record the trigger time. At the trigger time The set time after Within the specified timeframe, confirm the impact landing. If the impact landing is confirmed, record the trigger time. At the trigger time The set time after Inside, identify whether it is a post-fall state. If the above process is met, the fall is confirmed; otherwise, continue monitoring.
9. The non-contact respiratory and cardiac monitoring and fall detection system according to claim 1, characterized in that, It also includes an alarm module, a wireless communication module, a power supply component, a local display module, and a housing that encloses all modules. The alarm module is used to issue an alarm based on the output respiratory rate, heart rate estimate, and whether a fall has occurred. The wireless communication module is used to push the respiratory rate, heart rate estimate, and whether a fall has occurred to a designated terminal. The local display module is used to display the current respiratory rate, heart rate estimate, and whether a fall has occurred based on the trigger status, or to automatically display the respiratory rate, heart rate estimate, and whether a fall has occurred locally when an alarm occurs.
10. A non-contact method for monitoring respiratory and cardiac activity and for detecting falls, characterized in that, include: The system transmits detection signals and receives echo signals according to a preset sampling strategy. The received echo signal is processed for time and distance, and then filtered to obtain an effective distance gate; Calculate the joint confidence of each effective range gate and select the main range gate or the fused range gate; The fusion range gate is a fusion echo of multiple consecutive high-confidence effective range gates; Phase micro-displacement and amplitude envelope sequences are synchronously extracted from the echoes of the main distance gate or fusion distance gate. After joint spectral and temporal analysis, the time-varying trajectory of the respiratory dominant frequency and the output respiratory rate are estimated. A prediction reference signal including the respiratory fundamental wave and at least one harmonic is generated based on the time-varying trajectory of the respiratory dominant frequency. The phase micro-displacement sequence and amplitude envelope sequence are adaptively canceled using the prediction reference signal to obtain the candidate residual signal of the heartbeat. The joint confidence of the heartbeat is calculated, and the heart rate estimate is output based on the joint confidence of the heartbeat. After constructing a body movement detection region and extracting body movement and fall features, the current body movement state is identified, and when a body movement state occurs in a preset order within a continuous phase, a fall confirmation is output.
