Vital sign and sleep monitoring method and system based on millimeter wave radar

Abstract

The invention discloses a vital sign and sleep monitoring method and system based on a millimeter-wave radar, the method utilizes the millimeter-wave radar to realize non-contact sleep monitoring, and the millimeter-wave radar is firstly used to collect radar echo signals obtained by vibration of a chest wall in a sleep state; determining the position of a monitored object and extracting a vital sign signal of the target; then accurately detecting a first amplitude body movement and a second amplitude body movement of the target; whether an apnea event happens to the target is detected in real time, and alarm information is prompted under the condition that the event happens for a long time; three sleeping postures of supine, side-lying and prostrate are detected; and finally, sleep state monitoring is formed, a user sleep quality analysis report is formed in combination with the monitored information, and health suggestions are given. The method is simple, convenient, accurate, economical and efficient, is suitable for various application scenes such as families, hospitals and nursing homes, and provides all-weather health monitoring services for users.

Description

Technical Field

[0001] The present invention relates to the field of biomedical engineering and wireless sensor equipment technology, and in particular to a method and system for non-contact monitoring of human vital signs, sleeping posture, sleep stages and apnea alarm using millimeter wave radar. Background Art

[0002] In recent years, with the advancement of millimeter-wave radar technology, its application in non-contact human monitoring has received widespread attention. Due to its characteristics of working in the high-frequency band, millimeter-wave radar has the ability to penetrate non-metallic materials and can detect weak vital signs of the human body, such as breathing and heartbeat, within a certain distance. The advantage of this technology is that it does not require direct contact with the human body, thus avoiding the inconvenience and discomfort caused by traditional monitoring methods. Although traditional monitoring equipment such as electrocardiograms (ECGs) and pulse oximeters can provide accurate vital signs data, they may cause discomfort to users during long-term wear, especially during nighttime sleep monitoring, which can affect sleep quality.

[0003] At the same time, with the improvement of people's health awareness and the growth of demand for high-quality sleep, non-contact health monitoring technology has gradually become a research hotspot. As an emerging non-contact monitoring tool, millimeter wave radar can not only monitor basic vital signs such as human breathing rate and heart rate, but also further identify the user's sleeping posture, divide the sleep stage, and monitor sleep disorder symptoms such as apnea. Compared with other non-contact technologies, such as infrared sensors and ultrasonic sensors, millimeter wave radar has higher resolution and stronger anti-interference ability, and can work stably in complex environments.

[0004] Therefore, a method and system for monitoring vital signs and sleep based on millimeter-wave radar is needed. Summary of the invention

[0005] In view of this, the purpose of the present invention is to provide a vital sign and sleep monitoring method and system based on millimeter wave radar, which can not only improve the accuracy of vital sign monitoring, but also accurately identify sleeping posture, divide sleep stages, and effectively monitor the occurrence of sleep apnea.

[0006] In order to achieve the above object, the present invention provides the following technical solutions:

[0007] The millimeter wave radar-based vital sign and sleep monitoring method provided by the present invention comprises the following steps:

[0008] Use millimeter wave radar to collect radar echo signals obtained by chest wall vibration during sleep;

[0009] Determine the location of the monitoring target through the bed area positioning algorithm;

[0010] The adaptive noise extraction algorithm can effectively eliminate interference and extract the target's vital signs.

[0011] The first amplitude body movement and the second amplitude body movement of the target are accurately detected through a multi-feature body movement detection algorithm;

[0012] Extract breathing and heartbeat signals, detect whether the target has apnea events in real time through a multi-feature apnea judgment algorithm, and prompt an alarm message when the event occurs for a time exceeding the threshold time;

[0013] Effectively detect different sleeping positions through a multi-feature sleeping position judgment algorithm;

[0014] Different sleeping positions are monitored and analyzed through the sleep monitoring algorithm, and the monitoring information is combined to form a user sleep quality analysis report.

[0015] Further, the position of the monitoring target is carried out according to the following steps:

[0016] Determine the installation location of the millimeter-wave radar and the bed parameters, and determine the bed monitoring area;

[0017] By detecting whether there is a target exceeding the threshold in the window monitoring area, if a target appears, continue to detect whether there is a signal containing vital signs. If a vital sign signal is detected, it is determined that there is a monitoring target in the bed monitoring area.

[0018] Further, the multi-feature body motion detection algorithm is performed according to the following steps:

[0019] Calculate the range time second-order derivative information, range image spectrum amplitude information, and range gate diffusion information based on the range gate data of the target;

[0020] When it is detected that: the second-order derivative information of distance time is less than the first body motion threshold, the amplitude information of the range image spectrum is less than the second body motion threshold, and the range gate diffusion information is less than the third body motion threshold, it is judged that there is no body motion and the target is in a resting state;

[0021] When it is detected that: the distance time second-order derivative information is less than the first body motion threshold, the range image spectrum amplitude information is greater than the second body motion threshold and less than the fourth body motion threshold, and the range gate diffusion information is greater than the third body motion threshold, it is judged as the second amplitude body motion and the target is in the second amplitude body motion state;

[0022] When it is detected that: the second-order derivative information of distance time is greater than the first body motion threshold, the distance image spectrum amplitude information is greater than the fourth body motion threshold, and the distance gate diffusion information is greater than the third body motion threshold, it is judged as first-amplitude body motion, and the first-amplitude body motion information is input into the sleep monitoring module as a sleep feature, and the data of the body movement occurrence time is separated from the data entering the vital signs monitoring.

[0023] Further, the extraction of breathing and heartbeat signals is performed according to the following steps:

[0024] Phase information is extracted from resting data without body movement, and the original phase information is obtained by phase unwrapping;

[0025] Performing Fourier transform on the original phase information to obtain phase change period information, and confirming the period information of the Fourier transformed phase period information that is higher than the set signal-to-noise ratio as the target breathing signal period;

[0026] Perform IIR filtering on the original phase information to extract the heartbeat signal, perform spectrum detection on the heartbeat signal, calculate the Fourier transform of the heartbeat signal, and find out the period information of the detected heartbeat signal;

[0027] Calculate real-time heart rate and breathing.

[0028] Further, the multi-feature apnea determination algorithm is performed according to the following steps:

[0029] When apnea occurs, the intensity, standard score, and range of the respiratory signal are used as features to determine apnea:

[0030] When the intensity of the breathing signal is less than the first breathing feature threshold, the intensity feature of the breathing signal is valid;

[0031] When the standard score of the respiratory signal is less than the second respiratory feature threshold, the standard score feature of the respiratory signal is valid;

[0032] When the range of the breathing signal is less than the third breathing feature threshold, the range feature of the breathing signal is valid;

[0033] When the three characteristics are simultaneously effective for a period exceeding the set apnea judgment time, an apnea event is confirmed to have occurred.

[0034] Further, the multi-feature sleeping posture judgment algorithm is performed according to the following steps:

[0035] The sleeping position is determined according to the one-dimensional range image spectrum, the two-dimensional range speed spectrum, the heart rate change, and the breathing rate change; the sleeping position includes three sleeping positions: supine, side-lying, and prone.

[0036] Further, the sleep monitoring algorithm is performed according to the following steps:

[0037] A sleep status analysis report is obtained by monitoring and analyzing different sleeping positions, wherein the sleep status includes wakefulness, light sleep, deep sleep, and deep sleep REM period.

[0038] Further, the awake period, light sleep period, deep sleep period, and deep sleep REM period are determined according to the following steps:

[0039] When the number of body movements within the recording window T3 is greater than the threshold Thod8, it is in the awake period;

[0040] When the number of body movements within the recording window T3 is less than the threshold Thod8 but greater than Thod9, it is in the light sleep stage;

[0041] When the number of body movements within the recording window T3 is less than the ninth threshold value Thod9 but greater than the threshold value Thod10, the body enters the deep sleep period and the deep sleep REM period.

[0042] The millimeter wave radar-based vital sign and sleep monitoring system provided by the present invention includes a memory, a processor, and a computer program stored in the memory and executable on the processor. The processor implements the above method when executing the program.

[0043] The beneficial effects of the present invention are:

[0044] The present invention provides a vital sign and sleep monitoring method and system based on millimeter wave radar. The method uses millimeter wave radar to realize non-contact sleep monitoring. First, the millimeter wave radar is used to collect radar echo signals obtained by chest wall vibration in the sleeping state; the position of the monitored object is determined by a bed area positioning algorithm; the interference is effectively eliminated by an adaptive noise extraction algorithm to extract the vital sign signals of the target; the first amplitude body movement and the second amplitude body movement of the target are accurately detected by a multi-feature body motion detection algorithm; then, a multi-feature apnea judgment algorithm is used to detect in real time whether the target has an apnea event and prompt an alarm message if the event occurs for a long time; a multi-feature sleeping posture judgment algorithm is used to effectively detect three sleeping postures: supine, side-lying, and prone; finally, a sleep monitoring algorithm is used to monitor four sleep states of awake period, light sleep period, deep sleep period, and deep REM sleep period, and a user sleep quality analysis report is formed in combination with the above monitored information, and health advice is given.

[0045] The present invention also provides an implementation system or WeChat applet that can generate detailed sleep reports through all-night vital sign monitoring, effectively detect sleep apnea events, and is suitable for long-term monitoring and analysis of personal sleep conditions in a home environment.

[0046] In addition, the present invention also combines advanced signal processing technology and algorithms to make the monitoring results more accurate and reliable. The method is simple, accurate, economical and efficient, and is suitable for a variety of application scenarios such as homes, hospitals and nursing homes, providing users with round-the-clock health monitoring services.

[0047] Other advantages, objectives and features of the present invention will be described in the following description to some extent, and to some extent, will be obvious to those skilled in the art based on the following examination and study, or can be taught from the practice of the present invention. The objectives and other advantages of the present invention can be realized and obtained through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to make the purpose, technical solution and beneficial effects of the present invention clearer, the present invention provides the following drawings for illustration:

[0049] Figure 1 This is an actual scene installation diagram of this embodiment.

[0050] Figure 2 This is a diagram of the bed area positioning algorithm of this embodiment.

[0051] Figure 3 Schematic diagram of the adaptive noise extraction algorithm of this embodiment.

[0052] Figure 4 This is a diagram of the multi-feature body motion detection algorithm of this embodiment.

[0053] Figure 5 This is a diagram of the multi-feature apnea determination algorithm of this embodiment.

[0054] Figure 6 This is a diagram of the multi-feature sleeping posture judgment algorithm of this embodiment.

[0055] Figure 7 Schematic diagram of the sleep monitoring algorithm of this embodiment.

[0056] Figure 8 This is a vital sign monitoring method based on millimeter wave radar and a sleep monitoring and analysis flow chart.

[0057] Fig. 9 This is a schematic diagram of the effect of judging based on the body movement threshold.

[0058] Fig.10 This is a schematic diagram of the effect of judging based on the time breathing feature threshold.

[0059] Fig.11 This is a schematic diagram of the sleeping posture judgment effect. DETAILED DESCRIPTION

[0060] The present invention is further described below in conjunction with the accompanying drawings and specific embodiments so that those skilled in the art can better understand the present invention and implement it, but the embodiments are not intended to limit the present invention.

[0061] Example 1

[0062] The high-precision human sleep breathing monitoring method based on millimeter-wave radar provided in this embodiment can accurately monitor the vital signs of the human body in the sleeping state without direct contact with the human body, and innovatively use radar spectrum data and innovative signal processing algorithms to accurately identify human apnea and sleeping posture. In addition, sleep monitoring is realized. After sleep, users can view their daily sleep status report and health analysis in the mini program provided on WeChat.

[0063] The millimeter wave radar-based vital sign monitoring method and sleep monitoring method provided in this embodiment mainly include the following steps:

[0064] Step 1: Use millimeter-wave radar to collect radar echo signals obtained by chest wall vibration during sleep; determine the location of the monitored object through the bed area positioning algorithm;

[0065] The millimeter-wave radar stand is placed at the head of the bed. After being powered on, the millimeter-wave radar will continuously emit electromagnetic waves to the bed area, and receive the echo from the bed area for signal processing. After receiving the echo signal, the device performs a distance dimension Fourier transform on the echo information to extract the bed area distance image data, and uses the bed area positioning algorithm to determine whether the human body is in bed. If it is determined that the person is not in bed, vital signs monitoring will not be performed. If it is detected that the person is in bed, vital signs monitoring and sleep monitoring will be performed.

[0066] Step 2: After the human body is located in the bed area and judged as being in bed in step 1, interference is effectively eliminated through an adaptive noise extraction algorithm to extract the target's vital sign signals;

[0067] This embodiment uses an adaptive noise extraction algorithm to extract the background noise of the current scene, and then performs a signal-to-noise ratio analysis on the range image data, and performs moving target detection on the range image to extract the range gate data where the target with a high signal-to-noise ratio is located.

[0068] Step 3: Use the target range gate data obtained in step 2 to perform a multi-feature body motion detection algorithm. The multi-feature body motion detection algorithm accurately detects the first amplitude body motion and the second amplitude body motion of the target;

[0069] Among them, the features mainly include the second-order derivative information of distance time, the amplitude information of the distance image spectrum, and the diffusion information of the distance gate. Through multi-feature detection and identification, whether the target has first-amplitude body movement such as turning over, as well as second-amplitude body movement such as coughing and swiping the phone, etc. If the first-amplitude body movement is detected, the data of the time when the body movement occurs needs to be separated from the data entering the vital signs monitoring.

[0070] In this embodiment, the first amplitude of body movement is set as a large amplitude of body movement, and the second amplitude of body movement is set as a small amplitude of body movement. The amplitude of body movement can be determined according to actual conditions.

[0071] Step 4: Extract the breathing and heartbeat signals. Extract the phase information from the resting data without body movement in step 3, perform Fourier transform on the phase to obtain its change cycle information, and use the adaptive noise extraction algorithm used in step 1 to extract the highest intensity breathing signal within the breathing frequency range, and use the IIR filter to extract the heartbeat signal. Perform frequency detection on the extracted breathing and heartbeat signals, calculate the real-time heart rate and breathing using the formula, and report the abnormalities such as slow breathing, fast breathing, and heartbeat variability in the sleep report.

[0072] Step 5: Use a multi-feature apnea judgment algorithm to detect in real time whether the target has a apnea event and prompt an alarm message if the event lasts for a long time (i.e., exceeds a preset time);

[0073] In this embodiment, a multi-feature apnea judgment algorithm is used to detect in real time whether apnea has occurred. When apnea occurs, in addition to the weakening of the frequency peak of the corresponding respiratory frequency band, the intensity, standard score, and range of the respiratory signal are used as features to judge apnea. Multi-feature fusion signal detection greatly improves detection accuracy and real-time performance.

[0074] Step 6: Use a multi-feature sleeping posture judgment algorithm to detect the target sleeping posture in real time, and effectively detect the three sleeping postures of supine, side, and prone;

[0075] This embodiment performs data statistical processing on the one-dimensional range image spectrum, the two-dimensional range speed spectrum, the heart rate change and the respiratory rate change under different sleeping postures as the sleep posture characteristic information, and jointly judges the three kinds of characteristic information to detect the current detection target sleep posture.

[0076] Step 7: Use the sleep monitoring algorithm to monitor the four sleep states of awake, light sleep, deep sleep, and REM sleep, and combine the previously monitored information to form a user sleep quality analysis report and give health advice;

[0077] The data on the bed status, heartbeat, breathing, body movement, breathing status (pause, bradyphrenia, tachypnea), and sleeping posture judgment obtained through the above steps in this embodiment will be used for target sleep monitoring and analysis in this step. The sleep monitoring part uses the above characteristic data information, body movement recording and cardiopulmonary coupling method to evaluate and analyze the target person's wakefulness time, light sleep time, deep sleep time, and REM deep sleep time, and forms a sleep quality analysis report for the target person based on the monitored heartbeat, breathing, body movement, breathing status, and sleeping posture.

[0078] Furthermore, step 1 has the following sub-steps:

[0079] Step 1.1: The echo received by the millimeter wave radar R(t) = A r cos(2π(f c +Δf)t+φ) and the local oscillator signal S(t)=A s cos(2πf c t) is mixed to obtain the intermediate frequency signal IF(t), the formula is as follows:

[0080]

[0081] Where R(t) represents the echo received by the millimeter-wave radar; A r Indicates the signal strength of the received echo; f c represents the signal frequency of the local oscillator signal; Δf represents the frequency offset of the echo signal relative to the local oscillator signal; t represents the time variable; φ represents the phase of the echo signal; S(t) represents the local oscillator signal; A s represents the local oscillator signal strength; IF(t) represents the intermediate frequency signal;

[0082] Step 1.2: Perform Fourier transform on the intermediate frequency signal in the range image dimension to obtain a one-dimensional range image spectrum.

[0083] Step 1.3: The bed area positioning algorithm determines whether a person is in bed: First, the millimeter wave radar can adaptively update the bed monitoring range according to the installation height and bed width input in the user interface; if there is no input in the user interface, the default installation height and bed width are set; and the CACFAR algorithm is used to detect whether there is a target exceeding the threshold. Assuming that a suspected target is detected, it will continue to detect whether there is a signal containing vital signs. If no signal is detected, it is only suspected that there is a person but not judged as being in bed, thereby avoiding the detection of false human targets and improving anti-interference performance.

[0084] Furthermore, the adaptive noise extraction algorithm in step 2 is described in detail as follows:

[0085] The one-dimensional range image spectrum obtained in step 1 is used for the adaptive noise extraction algorithm to find the range gate of the human target: first, the range image signals of the bed area and the non-bed area are processed. After the moving target processing, the targets with motion information will be screened out. There are clutter and background noise in the screened-out static signals, and the background noise can be filtered out by smoothing filtering. When the range gate signal strength of the suspected target is greater than the set signal-to-noise ratio, it is confirmed as the target, and the corresponding range gate signal is moved to the next step of processing.

[0086] Furthermore, the multi-feature body motion detection algorithm in step 3 is described in detail as follows:

[0087] A multi-feature body motion detection algorithm is used for the target's range gate data: the three features of the range time second-order derivative information, the range image spectrum amplitude information, and the range gate diffusion information are calculated for the target's range gate data. The first amplitude body motion mainly refers to the more violent body movement, such as turning over, kicking, etc.; the second amplitude body motion mainly refers to slight body movement, such as slight head shaking, small movement of hands or feet.

[0088] When it is detected that: the second-order derivative information of distance time is less than the body motion threshold 1, the range image spectrum amplitude information is less than the body motion threshold 2, and the range gate diffusion information is less than the body motion threshold 3, it is judged that there is no body motion and the target is in a resting state;

[0089] When it is detected that: the second-order derivative information of the distance time is less than the body motion threshold 1, the amplitude information of the distance image spectrum is greater than the body motion threshold 2 and less than the body motion threshold 4, and the range gate diffusion information is greater than the body motion threshold 3, it is judged as the second amplitude body motion, and the target is in the second amplitude body motion state;

[0090] When it is detected that: the distance time second-order derivative information is greater than the body motion threshold 1, the distance image spectrum amplitude information is greater than the body motion threshold 4, and the distance gate diffusion information is greater than the body motion threshold 3, it is judged as the first amplitude body motion, and the target is in the first amplitude body motion state. At this time, the target's sleep stage may change, and the first amplitude body motion information will be input into the sleep monitoring module as a sleep feature, and the data of the body movement occurrence time needs to be separated from the data entering the vital signs monitoring.

[0091] In this embodiment, the first body movement threshold 1, the second body movement threshold 2, the third body movement threshold 3, and the fourth body movement threshold 4 are determined as follows:

[0092] After data collection and analysis of 10 test persons (7 males, 3 females, 5 obese persons, and 5 normal persons), when body movement occurs, the value of the second-order derivative information of distance time is significantly lower than when there is no body movement. Therefore, the first body movement threshold 1 is determined to be -5 times the second-order derivative information of distance time in a stable state. When body movement occurs, the value of the distance image spectrum amplitude information is significantly larger than when there is no body movement, but the value difference between the second amplitude body movement and the first amplitude body movement is different. Therefore, the second body movement threshold 2 is determined to be 5 times the distance image spectrum amplitude information in a stable state, and the third body movement threshold 3 is determined to be 10 times the distance image spectrum amplitude information in a stable state. In order to better distinguish the second amplitude body movement from the first amplitude body movement, the distance gate diffusion information is also introduced, such as Fig. 9 As shown, Fig. 9 The fourth body motion threshold 4 is determined according to the difference in actual displacement caused by small and large movements of a normal body. However, due to the different body shapes of the test subjects, it can be seen that any one of the indicators alone may not be completely accurate, so the introduction of multiple indicators for joint judgment greatly improves the accuracy and reduces the misjudgment rate.

[0093] Furthermore, the frequency detection process of the extracted breathing and heartbeat signals in step 4 is performed according to the following sub-steps:

[0094] Step 4.1: Extract phase information from the resting data without body movement in step 3, and obtain original phase information through phase unwrapping.

[0095] Step 4.2: Perform Fourier transform on the original phase information to obtain the phase change period information. Use the adaptive noise extraction algorithm to filter out the background noise through smoothing filtering, and confirm the period information of the Fourier transformed phase period information that is higher than the set signal-to-noise ratio as the target respiratory signal period.

[0096] Step 4.3: Perform IIR filtering on the original phase information to extract the heartbeat signal, perform spectrum detection on the heartbeat signal, calculate the Fourier transform of the heartbeat signal, and find out the period information of the detected heartbeat signal.

[0097] Step 4.4: Use the formula to calculate the real-time heart rate and respiration. The formula is as follows:

[0098] Respiratory rate = 60*respiratory cycle

[0099] Slow breathing, fast breathing, and heart rate variability are reported as abnormalities in the sleep report.

[0100] Furthermore, the multi-feature apnea determination algorithm in step 5 is described in detail as follows:

[0101] When apnea occurs, in addition to the weakening of the frequency peak of the corresponding respiratory frequency band, the intensity, standard score and range of the respiratory signal are also used as features to judge apnea.

[0102] Perform apnea detection on the intensity, standard score, and range of the original phase breathing signal processed in step 4:

[0103] When the intensity of the breathing signal is less than the first breathing feature threshold 1, the intensity feature of the breathing signal is valid.

[0104] When the standard score of the breathing signal is less than the second breathing feature threshold 2, the standard score feature of the breathing signal is valid.

[0105] When the range of the breathing signal is less than the third breathing feature threshold 3, the range feature of the breathing signal is valid.

[0106] When the three characteristics are simultaneously effective for a period exceeding the set apnea judgment time, an apnea event is confirmed to have occurred.

[0107] Multi-feature fusion signal detection greatly improves detection accuracy and real-time performance.

[0108] The first breathing feature threshold 1, the second breathing feature threshold 2, the third breathing feature threshold 3, and the set apnea judgment time provided in this embodiment are determined as follows:

[0109] Data was collected from 10 test subjects (7 males, 3 females, 5 obese people, and 5 normal people). After breathing normally for a period of time, they held their breath for 30 seconds to simulate apnea. After breathing normally for a period of time, they held their breath for 30 seconds to simulate apnea. In one experiment, 3 apneas were performed, and one experiment lasted about 6 minutes. Fig.10 As shown, Fig.10 This is a schematic diagram of the effect of judging based on the time breathing feature threshold; the signal strength, extreme difference, and standard score will have obvious troughs. Therefore, after statistically analyzing the data of 10 testers, it was found that the data basically showed the same trend, and a unified threshold could be used to distinguish between normal breathing and apnea. It can also be seen that any one of the indicators alone may not be completely accurate, so the introduction of multiple indicators for joint judgment greatly improves the accuracy and reduces the misjudgment rate. The apnea judgment time is mainly set based on sleep disorders. If an adult has 5 or more apnea events (each lasting at least 10 seconds) within an hour, and is accompanied by symptoms such as daytime sleepiness, then further examination may be required to determine whether there is sleep apnea syndrome. Therefore, the apnea judgment time is set to 10 seconds.

[0110] Furthermore, the multi-feature sleeping posture judgment algorithm in step 6 is described in detail as follows:

[0111] The sleeping posture judgment mainly judges the three sleeping postures of supine, side and prone. The radar echo signals of the human body in different sleeping postures will be different, because the sleeping posture affects the body's reflection characteristics of radar waves.

[0112] When lying on your back, your chest and abdomen face the radar, your body is relatively flat, and because the main organs of the body (such as the lungs and heart) face the radar directly, a stronger echo signal will be generated. The rise and fall of the chest caused by breathing and heartbeat will form a more obvious periodic change in the signal.

[0113] When lying on your side, the rise and fall of one side of the body (especially the side facing the radar) will produce a more significant signal change compared to lying on your back. The signal strength and pattern will also vary depending on the part of the body facing the radar (such as the left chest facing the radar when lying on the left side).

[0114] When lying prone, the radar waves mainly come into contact with the back. Although the ups and downs of the back caused by breathing can still be detected, the signal is not as strong as when in frontal contact.

[0115] Therefore, after a large number of experimental tests, we selected one-dimensional range image spectrum, two-dimensional range speed spectrum, heart rate changes, and breathing rate changes as the characteristic information of sleeping posture.

[0116] The present invention also provides a WeChat applet that supports target users to view the generated sleep report every day. Through all-night vital sign monitoring, a detailed sleep report can be generated, and sleep apnea events can be effectively detected. It is suitable for long-term monitoring and analysis of personal sleep conditions in a home environment.

[0117] Example 2

[0118] In order to more clearly explain the objectives, technical solutions and advantages of the embodiments of the present application, the following will provide a detailed and systematic description of the technical solutions involved in conjunction with the drawings provided in the embodiments of the present application. We will strive to ensure the accuracy and completeness of the description so that the technical details of the embodiments and their application value can be fully understood.

[0119] like Figures 1 to 8 As shown, a method for realizing non-contact sleep monitoring using millimeter wave radar is adopted in an example of the present invention, and the steps are as follows:

[0120] Step 1: Power on the 77GHz frequency modulated continuous millimeter wave radar and hang it in the middle of the bed about 1.2 meters above the bed. The radar continuously receives the following echoes:

[0121] R(t)=A r cos(2π(fc +Δf)t+φ)

[0122] Among them, A r is the echo signal strength, f c is the local oscillator frequency of the transmitted signal, φ is the original phase of the transmitted signal, and Δf is the Doppler frequency change caused by the object containing motion information collected by the echo. After receiving the echo, it will be combined with the local oscillator signal S(t) = A s cos(2πf c t), where A s is the frequency of the local oscillator signal, and the intermediate frequency signal is obtained by mixing:

[0123]

[0124] The obtained intermediate frequency signal is subjected to a one-dimensional distance image Fourier transform, and then the bed area positioning algorithm is used to determine whether the human body is in the bed. The bed area reflected on the distance image is calculated by the following steps: the radar height from the bed surface is H, which is equal to 1.2 meters in this example; the radar distance resolution is RES, which is 0.025 meters in this example; the bed length is L, which is 1.8 meters in this example; the one-dimensional distance image Fourier transform point number is NumFFT, and in this example, the one-dimensional distance image Fourier transform point number is 512 points; then for the radar distance image, the area range of the bed is:

[0125] Where x is the detected range gate position; H is the height of the radar from the bed surface; RES is the radar range resolution; L is the length of the bed;

[0126] The one-dimensional range image is used for bed detection. If the target's range gate is x and is within the bed area, the suspected target is detected. The data at x is extracted to check whether the frequency component of the vital sign signal appears in its spectrum data. If so, it is considered to be in bed.

[0127] Step 2: If a human target is identified as being in bed, the target with motion information will be screened out after processing the one-dimensional range image moving target. There are clutter and background noise in the screened static signal. The background noise n1 can be filtered out through smoothing filtering. When the range gate signal strength of the suspected target is greater than the set signal-to-noise ratio SNR, it is confirmed as the target. At this time, the range gate signal X(t) corresponding to the maximum signal-to-noise ratio is extracted, which is the reflection signal at the target chest.

[0128] Step 3: Perform multi-feature body motion detection algorithm on X(t) to calculate three features: distance time second-order derivative information A, distance image spectrum amplitude information Po, and distance gate diffusion information Diff. Make a joint judgment on whether the three features are effective using four thresholds: Thod1, Thod2, Thod3, and Thod4. Among them, Thod1 represents the first body motion threshold; Thod2 represents the second body motion threshold; Thod3 represents the third body motion threshold; Thod4 represents the fourth body motion threshold.

[0129] First, calculate the distance-time second-order derivative information. The second-order derivative is the derivative of the first-order derivative of the function, which can reflect the concavity information of the signal. In this example, it is necessary to calculate the distance-time second-order derivative information of the target chest reflection signal X(t), and the calculation of the first-order derivative is recorded as Calculating the second-order derivative is to calculate the first-order derivative again, so the formula is:

[0130]

[0131] Where T is the observation window length.

[0132] In this example, it is necessary to calculate the range profile spectrum amplitude information Po of the target chest reflection signal X(t), and the calculation formula is:

[0133] Po=max(X(t)), t=1,2...T

[0134] In this example, it is necessary to calculate the range gate diffusion information Diff of the target chest reflection signal X(t), and the calculation formula is:

[0135] Diff=max(X(t))-min(X(t)),t=1,2...T

[0136] This example divides body movements into first-amplitude body movements and second-amplitude body movements based on the distance, intensity, and speed of the body movements. The first-amplitude body movements are mainly more violent body movements, such as turning over, kicking, etc., so the characteristics of the first-amplitude body movements are as follows:

[0137] When the distance time second-order derivative information A is less than the first body motion threshold Thod1, the range image spectrum amplitude information Po is less than the second body motion threshold Thod2, and the range gate diffusion information Diff is less than the third body motion threshold Thod3, it is judged that there is no body motion and the target is in a resting state;

[0138] When it is detected that: the distance time second-order derivative information A is less than the first body motion threshold Thod1, the distance image spectrum amplitude information Po is greater than the second body motion threshold Thod2 and less than the fourth body motion threshold Thod4, and the range gate diffusion information Diff is greater than the third body motion threshold Thod3, it is determined to be the second amplitude body motion, and the target is in the second amplitude body motion state;

[0139] When it is detected that: the distance time second-order derivative information A is greater than the first body movement threshold Thod1, the distance image spectrum amplitude information Po is greater than the fourth body movement threshold Thod4, and the distance gate diffusion information Diff is greater than the third body movement threshold Thod3, it is judged as the first amplitude body movement, and the target is in the first amplitude body movement state. At this time, the target's sleep stage may change, and the first amplitude body movement information will be input into the sleep monitoring module as a sleep feature, and the data of the body movement occurrence time needs to be separated from the data entering the vital signs monitoring.

[0140] Step 4: When it is detected that the target has no body movement event, the phase signal P(t) is extracted from the X(t) signal. Perform Fourier transform on the phase signal P(t). The number of Fourier transform points is NumFFT. In this example, the number of Fourier transform points is 512. After the transformation, the periodic information PF(n) in the phase signal is obtained, where n = 1, 2...NumFFT. The adaptive noise extraction algorithm is used again to smooth out the background noise n2, and the periodic information in the Fourier transformed phase periodic information that is higher than the set signal-to-noise ratio SNR2 is confirmed as the target's breathing signal period f. RR According to the normal distribution of human respiratory rate, which is 0.1 Hz to 0.6 Hz, a Butterworth bandpass filter is designed to extract the respiratory signal RR(t) from P(t). In this example, let the current extracted respiratory signal period f RR is 0.3Hz. The phase signal P(t) is IIR filtered to extract the heartbeat signal HR(t). According to the normal heart rate distribution of the human body of 0.8Hz~1.2Hz, a 0.8Hz~1.2Hz Butterfly bandpass filter is designed to amplify the heartbeat signal contained in the original signal, which is more conducive to the extraction of the heartbeat signal. After passing P(t) through the designed filter, the periodic information HF(n) of the heartbeat signal can be obtained. The adaptive noise extraction algorithm n3 is used again to smooth the filter to filter out the background noise, and the periodic information in the Fourier transformed phase period information that is higher than the set signal-to-noise ratio SNR3 is confirmed as the target heartbeat signal period f HR In this example, let the currently extracted respiratory signal period f HR The real-time heart rate and respiration rate are calculated using the formula:

[0141] Respiration rate = 60*f RR=60*0.3=18(bpm), heart rate=60*f HR =60*1.2=72(bpm)

[0142] Slow breathing, fast breathing, heart rate variability, etc. are reported as abnormal conditions in the sleep report.

[0143] Step 5: When apnea occurs, in addition to the frequency peak weakening of the corresponding respiratory frequency band, the intensity, standard score, and range of the respiratory signal are used as features to determine apnea. The intensity P, standard score Z, and range R of the original phase signal P(t) processed in step 4 are used for apnea detection.

[0144] First, calculate the intensity P. Intensity refers to the absolute amplitude of the signal. In this example, the object of intensity is P(t), so it can be calculated by the formula:

[0145] P=max(P(t)), t=1,2...T

[0146] Where T is the observation window length.

[0147] Next, calculate the standard score Z. The standard score indicates the degree of deviation between a raw score and the mean of the group of data. In this example, the object of the standard score is P(t), so it can be calculated by the formula:

[0148]

[0149] Where μ is the mean of the observations within the observation window, i.e., the mean of the signal P(t), and σ is the standard deviation within the observation window, i.e., the standard deviation of the signal P(t).

[0150] Next, calculate the range R. The range is a basic statistic that describes the dispersion of data. It represents the difference between the maximum and minimum values ​​in a set of data. In this example, the object of the standard score is P(t), so it can be calculated by the formula:

[0151] R=max(P(t))-min(P(t)),t=1,2...T

[0152] For the three features, three thresholds are set accordingly. When the intensity P of the respiratory signal is less than the respiratory feature threshold Thod5, the intensity feature of the respiratory signal is valid. When the standard score Z of the respiratory signal is less than the respiratory feature threshold Thod6, the standard score feature of the respiratory signal is valid. When the range R of the respiratory signal is less than the respiratory feature threshold Thod7, the range feature of the respiratory signal is valid. When the three features are valid at the same time, it is confirmed that a apnea event has occurred. Multi-feature fusion signal detection greatly improves the detection accuracy and real-time performance.

[0153] Step 6: Sleeping posture judgment mainly judges the three sleeping postures of supine, side and prone. The radar echo signals of the human body in different sleeping postures will be different, because the sleeping posture affects the body's reflection characteristics of radar waves.

[0154] When lying on your back, your chest and abdomen face the radar, your body is relatively flat, and because the main organs of the body (such as the lungs and heart) face the radar directly, a stronger echo signal will be generated. The rise and fall of the chest caused by breathing and heartbeat will form a more obvious periodic change in the signal.

[0155] When lying on your side, the rise and fall of one side of the body (especially the side facing the radar) will produce a more significant signal change compared to lying on your back. The signal strength and pattern will also vary depending on the part of the body facing the radar (such as the left chest facing the radar when lying on the left side).

[0156] When lying prone, the radar waves mainly come into contact with the back. Although the ups and downs of the back caused by breathing can still be detected, the signal is not as strong as when in frontal contact.

[0157] Therefore, after a large number of experimental tests, the one-dimensional range image spectrum RA, the two-dimensional range velocity spectrum RD, the heart rate change HR, and the respiratory rate change RR are selected as the sleep posture feature information. The four feature information needs to be obtained from the results of the previous steps.

[0158] First, the one-dimensional range image spectrum RA is obtained by performing a one-dimensional range image Fourier transform on the intermediate frequency signal obtained in step 1. A statistical analysis is performed on the one-dimensional range image spectrum RA. When the characteristic condition of one of the three sleeping positions is met, the characteristic information of the one-dimensional range image spectrum RA is valid.

[0159] Secondly, the two-dimensional distance velocity spectrum RD. The one-dimensional distance image spectrum RA is obtained by performing a one-dimensional distance image Fourier transform on the intermediate frequency signal obtained in step 1, and then a Doppler dimension Fourier transform is performed to obtain the two-dimensional distance velocity spectrum RD. A statistical analysis is performed on the two-dimensional distance velocity spectrum RD. When the characteristic conditions of one of the three sleeping positions are met, the characteristic information of the two-dimensional distance velocity spectrum RD is valid.

[0160] Finally, the heart rate change HR and respiratory rate change RR refer to the speed of the change of heart rate and respiratory rate. The heart rate and respiratory rate results obtained in step 4 will be stored in a repository. Statistical analysis will be performed on this data repository at regular intervals to obtain the heart rate and respiratory rate change information within the observation window. When the characteristic conditions of one of the three sleeping positions are met, the heart rate change HR and respiratory rate change RR characteristic information are valid.

[0161] When the three types of characteristic information simultaneously meet the characteristic conditions of a certain sleeping posture, it can be confirmed that the current sleeping posture of the person being tested is detected, such as Fig.11 The data shown in (a) and (b) in the figure are used as an example to illustrate. Fig.11 This is a schematic diagram of the sleeping posture judgment effect.

[0162] When lying on the back, the number of peak points in the one-dimensional distance image spectrum RA is small and the diffusion distance is small; the speed of the distance speed spectrum information is stable and balanced; and the chest cavity is in the most comfortable state without any compression. At this time, the heart rate change HR and the respiratory rate change RR are relatively slow, and the change statistics show that the change trend is slow and the degree of change is low; the above conditions are jointly judged by a threshold structure (which contains the number of peak points, diffusion distance, speed stability, and change rate mentioned above). When the threshold structure Thod5 is met as a whole, it is lying on the back.

[0163] When lying on one's side, there are more peak points in the one-dimensional distance image spectrum RA and the diffusion distance is larger; the speed of the distance speed spectrum information is unstable and drastic; at this time, the heart rate change HR and the respiratory rate change RR are relatively rapid, and the change statistics show that the change trend is unstable and the degree of change is large; the above conditions are jointly judged by a threshold structure (which contains the number of peak points, diffusion distance, speed stability, and change rate mentioned above). When the threshold structure Thod6 is met as a whole, it means lying on one's side.

[0164] When lying prone, the number of peak points in the one-dimensional distance image spectrum RA is large and the diffusion distance is large; the distance velocity spectrum information speed is stable and balanced but the overall signal strength is weak; the chest cavity is in a compressed state, at this time the heart rate HR and respiratory rate RR changes are relatively rapid, and the change statistics show that the change trend is rapid and the degree of change is large; when the overall threshold structure Thod7 is met, it is lying prone.

[0165] The threshold structures Thod5, Thod6, and Thod7 of this embodiment are determined as follows:

[0166] Data was collected from 10 test subjects (7 males, 3 females, 5 obese subjects, and 5 normal subjects). The specific threshold was determined as follows: Fig.11 As shown, based on the 3-minute single posture data collection of 10 test persons, only the data after stabilization in the middle minute was taken for statistical analysis to obtain the value of the threshold structure used in sleeping posture judgment, and the accurate judgment is applicable to 90% of the population.

[0167] Step 7: The bed status, heart rate, breathing, body movement, breathing status (pause, bradyslexia, tachycardia), and sleeping posture judgment data obtained through the above steps will be used for target sleep monitoring and analysis in this step. The sleep monitoring part analyzes the target person's wakefulness time, light sleep time, and deep sleep time through the above characteristic data information.

[0168] A sleep quality analysis report of the target person is formed based on the monitored heartbeat, breathing, body movement, respiratory status and sleeping posture.

[0169] In this method, there are four main sleep states: wakefulness, light sleep, deep sleep, and REM sleep. And the sleep state is mainly assessed by combining body movement recording with cardiopulmonary coupling method.

[0170] Actigraphy is a non-invasive monitoring method that is used to assess an individual's activity level and sleep and wake patterns. Actigraphy records an individual's movements, which can be used to infer their sleep status. Long periods of low activity levels usually mean that the individual is asleep. Frequent movements or higher activity levels indicate that the individual may be awake. Occasional minor movements during sleep (such as turning over) can also be captured, and this information helps assess the quality of sleep.

[0171] In this example, the body movement results obtained in step 3 are input to the sleep monitoring module, and the recording window length is T3. Thresholds Thod8, Thod9, and Thod10 are used as thresholds for sleep staging to define the number of strong body movements in the recording window T3. The state machine is switched for the three sleep states: wakefulness, light sleep, and deep sleep. The deep sleep stage includes deep sleep and deep REM stage. Figure 7 As shown:

[0172] When the number of body movements within the recording window T3 is greater than the eighth threshold value Thod8, the number of body movements is large, indicating that the target person has more physical activities and is in a sober period;

[0173] When the number of body movements within the recording window T3 is less than the eighth threshold value Thod8, but greater than the ninth threshold value Thod9, the number of body movements has slightly decreased, indicating that the target person is basically in the stage of entering sleep, in a light sleep period;

[0174] When the number of body movements within the recording window T3 is less than the ninth threshold value Thod9 but greater than the tenth threshold value Thod10, it means that the target person is basically in resting sleep, with basically no movements, and has entered the deep sleep stage and deep sleep REM stage.

[0175] The eighth threshold value Thod8, the ninth threshold value Thod9, and the tenth threshold value Thod10 in this embodiment are determined as follows:

[0176] According to some existing professional literature and research reports, the eighth threshold Thod8, the ninth threshold Thod9, and the tenth threshold Thod10 can be basically determined. This device defines the number of body movements per minute less than the tenth threshold Thod10 as "low activity". For example, some literatures define that if the number of body movements per minute is less than 3 times, it is defined as "low activity". If it is in low activity for a long time, it can be judged as a deep sleep period; the number of body movements per minute exceeding the eighth threshold Thod8 times is defined as "high activity". For example, it can be defined that if the number of body movements per minute is more than 5 times, it is defined as "high activity". If it is in high activity for a long time, it can be judged as a wakeful period; if it is between the two, it can be judged as a light sleep period. Data collection was performed on 10 testers (7 males, 3 females, 5 obese people, and 5 normal people) to verify this method, which can achieve basically accurate light sleep and deep sleep analysis.

[0177] The cardiopulmonary coupling method is based on the natural interaction between the heart and the respiratory system. The interaction between the heart and the respiratory system changes during different stages of sleep. These changes can be reflected in the coupling strength of the ECG signal and the respiratory signal. Studies have shown that in deep sleep, breathing becomes more regular, heart rate and blood pressure decrease, and the cardiopulmonary coupling strength is generally higher. In light sleep and REM sleep, breathing and heart rate are more irregular, and the cardiopulmonary coupling strength is lower. In this example, although the millimeter wave radar cannot directly obtain the ECG signal, the heart rate signal is the RR interval in the ECG signal, which can be obtained by the detected heart rate signal. Therefore, a similar analysis can be performed by coupling the heart rate signal obtained from the echo signal with the respiratory signal to obtain the coupling strength. The respiratory signal RR(t) obtained in step 4 is coupled with the heart rate signal HR(t) to obtain the cardiopulmonary coupling strength γ(f).

[0178] First, calculate the autocorrelation function of RR(t) and HR(t). The autocorrelation function of RR(t) and HR(t) can be obtained according to the formula:

[0179]

[0180]

[0181] where RR * (t) and HR * (t) are the conjugate functions of the respiratory signal RR(t) and the heart rate signal HR(t), respectively.

[0182] Then, by performing Fourier transform on the autocorrelation function, we can get the power spectrum density function of the signal. The formula is as follows:

[0183]

[0184]

[0185] As above, the cross spectral density function of RR(t) and HR(t) is calculated. First, the cross correlation function of RR(t) and HR(t) is calculated:

[0186]

[0187] By performing Fourier transform on the cross-correlation function, we can get the cross spectral density function of the two signals. The formula is as follows:

[0188]

[0189] Therefore, the coupling strength γ(f) between RR(t) and HR(t) is calculated by the following formula:

[0190]

[0191] The combination of body movement recording and cardiopulmonary coupling method can successfully distinguish the four sleep states: wakefulness, light sleep, deep sleep, and deep sleep REM.

[0192] Finally, the monitoring information in the previous steps is combined to generate the target's sleep status and health analysis suggestions as a sleep report, which users can view by clicking through the mini program.

[0193] The above-described embodiments are only preferred embodiments for fully illustrating the present invention, and the protection scope of the present invention is not limited thereto. Equivalent substitutions or changes made by those skilled in the art based on the present invention are within the protection scope of the present invention. The protection scope of the present invention shall be subject to the claims.

Claims

1. A method for monitoring vital signs and sleep based on millimeter wave radar, characterized in that: The following steps are involved: Use millimeter wave radar to collect radar echo signals obtained by chest wall vibration during sleep; Determine the location of the monitoring target through the bed area positioning algorithm; The adaptive noise extraction algorithm can effectively eliminate interference and extract the target's vital signs. The first amplitude body movement and the second amplitude body movement of the target are accurately detected through a multi-feature body movement detection algorithm; Extract breathing and heartbeat signals, detect whether the target has apnea events in real time through a multi-feature apnea judgment algorithm, and prompt an alarm message when the event occurs for a time exceeding the threshold time; Effectively detect different sleeping positions through a multi-feature sleeping position judgment algorithm; Different sleeping positions are monitored and analyzed through the sleep monitoring algorithm, and the monitoring information is combined to form a user sleep quality analysis report.

2. The method for monitoring vital signs and sleep based on millimeter wave radar as claimed in claim 1, characterized in that: The monitoring target position is performed according to the following steps: Determine the installation location of the millimeter-wave radar and the bed parameters, and determine the bed monitoring area; By detecting whether there is a target exceeding the threshold in the window monitoring area, if a target appears, continue to detect whether there is a signal containing vital signs. If a vital sign signal is detected, it is determined that there is a monitoring target in the bed monitoring area.

3. The method for monitoring vital signs and sleep based on millimeter wave radar as claimed in claim 1, characterized in that: The multi-feature body motion detection algorithm is performed according to the following steps: Calculate the range time second-order derivative information, range image spectrum amplitude information, and range gate diffusion information based on the range gate data of the target; When it is detected that: the second-order derivative information of distance time is less than the first body motion threshold, the amplitude information of the range image spectrum is less than the second body motion threshold, and the range gate diffusion information is less than the third body motion threshold, it is judged that there is no body motion and the target is in a resting state; When it is detected that: the distance time second-order derivative information is less than the first body motion threshold, the distance image spectrum amplitude information is greater than the second body motion threshold and less than the fourth body motion threshold, and the range gate diffusion information is greater than the third body motion threshold, it is judged as the second amplitude body motion and the target is in the second amplitude body motion state; When it is detected that: the second-order derivative information of distance time is greater than the first body motion threshold, the distance image spectrum amplitude information is greater than the fourth body motion threshold, and the distance gate diffusion information is greater than the third body motion threshold, it is judged as first-amplitude body motion, and the first-amplitude body motion information is input into the sleep monitoring module as a sleep feature, and the data of the body movement occurrence time is separated from the data entering the vital signs monitoring.

4. The method for monitoring vital signs and sleep based on millimeter wave radar as claimed in claim 1, characterized in that: The extraction of breathing and heartbeat signals is performed according to the following steps: Phase information is extracted from resting data without body movement, and the original phase information is obtained by phase unwrapping; Performing Fourier transform on the original phase information to obtain phase change period information, and confirming the period information of the Fourier transformed phase period information that is higher than the set signal-to-noise ratio as the target breathing signal period; Perform IIR filtering on the original phase information to extract the heartbeat signal, perform spectrum detection on the heartbeat signal, calculate the Fourier transform of the heartbeat signal, and find out the period information of the detected heartbeat signal; Calculate real-time heart rate and breathing.

5. The method for monitoring vital signs and sleep based on millimeter wave radar as claimed in claim 1, characterized in that: The multi-feature apnea determination algorithm is performed according to the following steps: When apnea occurs, the intensity, standard score, and range of the respiratory signal are used as features to determine apnea: When the intensity of the breathing signal is less than the first breathing feature threshold, the intensity feature of the breathing signal is valid; When the standard score of the respiratory signal is less than the second respiratory feature threshold, the standard score feature of the respiratory signal is valid; When the range of the breathing signal is less than the third breathing feature threshold, the range feature of the breathing signal is valid; When the three characteristics are simultaneously effective for a period exceeding the set apnea judgment time, an apnea event is confirmed to have occurred.

6. The method for monitoring vital signs and sleep based on millimeter wave radar as claimed in claim 1, characterized in that: The multi-feature sleeping posture judgment algorithm is performed according to the following steps: The sleeping position is determined according to the one-dimensional range image spectrum, the two-dimensional range speed spectrum, the heart rate change, and the breathing rate change; the sleeping position includes three sleeping positions: supine, side-lying, and prone.

7. The method for monitoring vital signs and sleep based on millimeter wave radar as claimed in claim 1, characterized in that: The sleep monitoring algorithm is performed according to the following steps: A sleep status analysis report is obtained by monitoring and analyzing different sleeping positions, wherein the sleep status includes wakefulness, light sleep, deep sleep, and deep sleep REM period.

8. The method for monitoring vital signs and sleep based on millimeter wave radar as claimed in claim 1, characterized in that: The awake period, light sleep period, deep sleep period, and deep sleep REM period are determined according to the following steps: When the number of body movements within the recording window T3 is greater than the threshold Thod8, it is in the awake period; When the number of body movements within the recording window T3 is less than the threshold Thod8 but greater than Thod9, it is in the light sleep stage; When the number of body movements within the recording window T3 is less than the ninth threshold value Thod9 but greater than the threshold value Thod10, the body enters the deep sleep period and the deep sleep REM period.

9. A millimeter wave radar-based vital sign and sleep monitoring system, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the method according to any one of claims 1 to 8 is implemented.

Images