Respiratory status monitoring method and device

Abstract

The present invention provides a respiratory state monitoring method and device, comprising: extracting a respiratory signal s(t) from an echo signal of a millimeter-wave radar within a single monitoring period; extracting multiple window respiratory signals from the respiratory signal s(t) by means of a sliding window, and determining the respiratory rate BPM of each window respiratory signal respectively; calculating the probability density information P of the respiratory rate BPM of the multiple window respiratory signals. i ; Using the probability density information P of the same monitoring object in multiple monitoring periods i Generate probability density baseline information B; by the probability density information P of the monitoring object in the target monitoring period j Compare with the probability density baseline information B to obtain the deviation feature b; determine whether the breathing state of the monitored object in the target monitoring period is normal according to the deviation feature b.

Description

Technical Field

[0001] The present invention relates to the field of medical equipment, and in particular to a respiratory status monitoring method and equipment. Background Art

[0002] Respiratory movements can be controlled at will but can also be performed autonomously without consciousness. The normal respiratory rate of an adult is 16 to 24 times per minute, and the respiratory rate during sleep is 12 to 20 times per minute. In-depth research on changes in respiratory rate, rhythm, and waveform can uncover a lot of health information. Slowed breathing is common in cases of reduced metabolic rate, overdose of anesthesia, shock, and significant increase in intracranial pressure. Faster breathing is mainly seen in fever, pain, anemia, hyperthyroidism, heart failure, pneumonia pulmonary embolism, pleurisy, bronchial asthma, and neurological and mental disorders. Understanding changes in respiratory rate can help detect and record vital signs early, especially for the detection of respiratory failure, which is crucial to the patient's prognosis.

[0003] Millimeter wave radar equipment has the characteristics of continuous signal collection and non-contact. It can collect the breathing signal of the human body in the scene for a long time, and the breathing rate can be calculated through the signal. The existing breathing abnormality detection method determines whether the breathing state is normal by comparing the detected breathing frequency with the threshold, or combining other vital signs such as pulse and blood oxygen. In actual application, the actual situation of the monitored object is diverse, and other vital signs such as blood oxygen are not convenient for long-term continuous collection. Therefore, the accuracy and convenience of the existing breathing abnormality detection method need to be improved. Summary of the invention

[0004] In view of this, the present application provides a respiratory status monitoring method, comprising:

[0005] Extract the breathing signal s (t) from the echo signal of the millimeter-wave radar within a single monitoring period;

[0006] Extracting multiple window breathing signals from the breathing signal s(t) by sliding window, and determining the breathing rate BPM of each window breathing signal respectively;

[0007] Calculate the probability density information P of the respiratory rate BPM of the multiple window respiratory signals i ;

[0008] Using the probability density information P in multiple monitoring periods of the same monitoring object i Generate probability density baseline information B;

[0009] By using the probability density information P of the monitored object in the target monitoring period j Compare with the probability density baseline information B to obtain the deviation feature b;

[0010] Determine whether the breathing state of the monitored object during the target monitoring period is normal based on the deviation feature b.

[0011] Optionally, determining the breathing rate BPM of each window breathing signal separately further comprises:

[0012] Performing short-time Fourier transform on each of the windowed respiratory signals to obtain corresponding spectrum information, and calculating the confidence C of the spectrum information using the peak point and the data points near the peak point in the spectrum information, wherein the energy value of the data points near the peak point is negatively correlated with the confidence C;

[0013] Extracting the frequency of the peak point from the spectrum information corresponding to each of the windowed respiratory signals;

[0014] Calculate the breathing rate BPM corresponding to each of the window breathing signals according to the frequency of the peak point;

[0015] The confidence level C is used to screen the respiratory rates BPM of the respiratory signals in each window to obtain the respiratory rates BPM of the plurality of respiratory signals in the window that are credible.

[0016] Optionally, calculating the confidence C of the spectrum information using the peak point in the spectrum information and data points in its vicinity further includes:

[0017] Determine the maximum energy value and its corresponding frequency value in the spectrum information to obtain the highest peak point;

[0018] Extracting energy values ​​of data points within a first range centered on the highest peak point and energy values ​​of data points within a second range centered on the highest peak point;

[0019] The confidence C is calculated using the extracted energy value.

[0020] Optionally, extracting the frequency of the peak point for each spectrum information corresponding to the windowed respiratory signal further includes:

[0021] Extracting frequency values ​​between a preset upper frequency limit and a preset lower frequency limit from the frequency spectrum information;

[0022] The frequency of the peak point is obtained by determining the maximum frequency value among the extracted frequency values.

[0023] Optionally, the probability density information P of the respiratory rate BPM of the respiratory signals in the multiple windows is calculated. i , further comprising:

[0024] Using the kernel function K, a one-dimensional probability density function of the respiratory rate BPM of the respiratory signals in the multiple windows is obtained;

[0025] Normalize the one-dimensional probability density function to obtain the probability density information P i , which includes the values ​​of each respiratory rate and its corresponding probability value.

[0026] Optionally, the probability density information P in multiple monitoring periods of the same monitoring object is used. i Generating probability density baseline information B further includes:

[0027] Determine the probability density information P in multiple monitoring time periods respectively i The eigenvalues ​​of X i , and calculate the average eigenvalue X mean ;

[0028] Calculate the characteristic value X in each monitoring period respectively i With the average eigenvalue X mean The difference between the two, and all the differences are used to obtain the stability information bias of the monitoring period;

[0029] Determining whether the stability information bias exceeds a full-time stability threshold;

[0030] If the stability information bias does not exceed the full-time stability threshold, the probability density baseline information B is generated using the probability density information of the monitoring period in which the difference does not exceed the single-time period stability threshold.

[0031] Optionally, the eigenvalue X i is the probability density information P i The probability of at least one peak in corresponds to the breathing rate.

[0032] Optionally, generating probability density baseline information B further includes:

[0033] The coefficient w is calculated based on the number of the reliable multiple window breathing signals and the number of all window breathing signals in a single monitoring period. i , where the number of credible multiple window breathing signals and the coefficient w i There is a positive correlation;

[0034] Using the probability density information P of each monitoring period i and the corresponding coefficient w i Get the probability density baseline information B.

[0035] Optionally, the probability density baseline information B is calculated as follows:

[0036] ;

[0037] Wherein, B is the probability density baseline information, P i represents the probability density information of the i-th monitoring period, w i represents the coefficient of the i-th monitoring period, and N represents the number of the monitoring periods.

[0038] Optionally, obtaining the deviation feature b further includes:

[0039] Determine the respiratory rate B corresponding to the maximum probability in the probability density baseline information B max , and the probability density information P during the target monitoring period j Determine the respiratory rate P corresponding to the maximum probability max , using the respiratory rate B max and respiratory rate P max Obtaining a deviation characteristic value b1; and / or

[0040] The probability density baseline information B and the probability density information P of the target monitoring period are respectively j Extract the probability value B corresponding to the same respiratory rate value x x and P x , and calculate the difference between each pair of probability values ​​extracted;

[0041] The differences between each pair of probability values ​​are accumulated to obtain the deviation characteristic value b2.

[0042] Optionally, determining whether the breathing state of the monitored object in the target monitoring period is normal according to the deviation feature b further includes:

[0043] Determine whether the deviation characteristic value b1 and / or the deviation characteristic value b2 exceeds a set threshold;

[0044] When the deviation characteristic value b1 and / or the deviation characteristic value b2 exceeds a set threshold, it is determined that the breathing state of the monitored object during the target monitoring period is abnormal.

[0045] Optionally, the target monitoring period is a period before, during or after multiple monitoring periods.

[0046] Optionally, the monitoring period is a period when the monitored object is in a sleeping state.

[0047] Optionally, the monitoring period further includes various sleep staging periods.

[0048] Accordingly, the present application provides a respiratory status monitoring device, characterized in that it includes: a processor and a memory connected to the processor; wherein the memory stores instructions that can be executed by the processor, and the instructions are executed by the processor so that the processor performs the above-mentioned respiratory status monitoring method.

[0049] According to the respiratory status monitoring method and device provided by the present application, first, the respiratory status monitoring period can be set, and a sliding window can be performed for the respiratory signal in each monitoring period to determine the respiratory rate of each window respectively, and the probability density information of the respiratory rate in a single monitoring period can be calculated, thereby obtaining the probability distribution of the respiratory rate of the monitored object in each monitoring period; then, for this monitored object, its long-term breathing rate probability distribution is used to generate its personalized probability density baseline information; finally, for a certain period that needs to be determined, by comparing the probability density information of the respiratory rate in this period with its personalized probability density baseline information, it is determined whether the respiratory state of the monitored object deviates from its long-term breathing state, thereby improving the accuracy of the respiratory state determination result; and this scheme does not need to rely on other vital sign signals, and performs non-contact, long-term respiratory status monitoring based on millimeter-wave radar equipment, which has high convenience and practicality. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0051] Figure 1 is a flow chart of a respiratory status monitoring method in an embodiment of the present invention;

[0052] Figure 2 is a waveform diagram of a breathing signal in an embodiment of the present invention;

[0053] Figure 3 It is a schematic diagram of performing sliding window on the respiratory signal and obtaining the respiratory rate in an embodiment of the present invention. DETAILED DESCRIPTION

[0054] The technical solution of the present invention will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0055] In addition, the technical features involved in the different embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0056] The embodiment of the present invention provides a method for monitoring respiratory status, which can be executed by an electronic device such as a computer or a server. Figure 1The following operations are shown:

[0057] S1, extract the breathing signal s (t) from the echo signal of the millimeter wave radar in a single monitoring period. The millimeter wave radar device sends a radar signal to the monitored object and collects the radar echo signal. This solution is to extract the vital sign signal from the echo signal and filter out the following Figure 2 A continuous breathing signal is shown.

[0058] In one embodiment, the millimeter wave radar device is installed on the wall, and the height of the radar from the monitoring position is about 1 meter, so that the radar beam points to the chest position of the monitored object. The radar transmits a frequency modulated continuous wave (FMCW) signal, and the echo signal received by the radar is mixed with the transmission signal to obtain a difference frequency signal, which is then high-pass filtered, low-noise amplified, and ADC sampled to obtain a digitized echo signal.

[0059] The collected human body radar echo data are processed and divided into a body motion segment and a stable segment through body motion detection. For the stable segment, the range gate where the chest cavity is located is first determined, and then the original phase signal is extracted along the range gate. The extracted original phase signal is unwrapped to obtain a phase signal after unwrapping. The unwrapped phase signal is the vital sign signal, which contains information related to respiratory activity, heartbeat activity, body movement, etc.

[0060] In order to avoid interference of other high-frequency vital signs such as heartbeat on the respiratory signal, a moving mean filtering method with a preset data volume length can be used to filter the vital sign signal, and the filtered result is used as the respiratory signal s (t).

[0061] The monitoring period can be set according to the application scenario, for example, a fixed time period, or it can be combined with a sleep state recognition scheme to determine the sleep period by identifying whether the monitored object is in a sleep state.

[0062] S2, extract multiple window breathing signals from the breathing signal s(t) by sliding window, and determine the breathing rate BPM of each window breathing signal respectively. Figure 3 As shown, the respiratory signal s(t) is continuous data at time t. In this step, a fixed window length is used to sample / divide the continuous signal. Adjacent windows can be partially overlapped or connected end to end without overlapping.

[0063] Through the processing of this step, the respiration rate BPM corresponding to each window respiration signal can be obtained, which represents the number of respirations per unit time. For example, the respiration signal s(t) is divided into window respiration signal #1, window respiration signal #2, ... window respiration signal #n, and then the corresponding respiration rate BPM#1, respiration rate BPM#2, ... respiration rate BPM#n are obtained.

[0064] There are many ways to determine the respiratory rate from the respiratory signal collected by the millimeter wave radar. Usually, the signal is processed based on Fourier transform to obtain spectrum data, and then the frequency information is obtained based on the spectrum data, and finally the respiratory rate is obtained based on the frequency information.

[0065] S3, calculate the probability density information P of the respiratory rate BPM of multiple window respiratory signals i Probability density information is a kind of statistical information. The probability density function of breathing rate BPM#1, breathing rate BPM#2…breathing rate BPM#n is expressed by the kernel density estimation method to construct a continuous probability density curve. When the data is shown by the curve, the value of the breathing rate is usually used as the horizontal axis and the probability value is used as the vertical axis, that is, the probability density information P i The probability values ​​corresponding to different respiratory rate values ​​are included. In the multiple window respiratory signals, the more the same respiratory rate BPM is, the higher the probability value corresponding to the respiratory rate value is.

[0066] S4, using the probability density information P of the same monitoring object in multiple monitoring periods i Generate probability density baseline information B. Figure 3 The same method as shown in the figure is used to calculate the probability density information P for N monitoring periods. i (i=1, 2...N), and then combine these probability density information P i , calculate the probability density baseline information B. The data type of the probability density baseline information B is the same as the probability density information P i The same is the probability distribution of the respiratory rate, but the probability value is a combination of the probability density information P of multiple monitoring periods. i The results obtained.

[0067] There are many ways to generate probability density baseline information B. As a simple example, for example, N probability density information P i Accumulate and normalize, or for N probability density information P i Find the average, etc.

[0068] The probability density baseline information B is used to characterize the comprehensive respiratory state of the monitored object over a long period of time. For example, if the monitoring period is during the daily sleep period, and the multiple monitoring periods are one month, the probability density baseline information B obtained through the above steps S1 to S4 represents the respiratory state of the monitored object in the sleep state during this one month.

[0069] S5, by calculating the probability density information P of the monitored object in the target monitoring period jCompare with the probability density baseline information B to obtain the deviation feature b. The so-called target monitoring period in this step depends on the purpose of executing this method. The target monitoring period can be a historical period or a current period, and the length of the target monitoring period can be set arbitrarily.

[0070] The respiratory signal of the target monitoring period is processed in the same manner as steps S1 to S3 to obtain probability density information. i To distinguish the literal expression, the probability density information of the target monitoring period is recorded as P j .

[0071] There are many ways to determine the difference between two probability density distribution information. For example, the probability density information P j The indicators of probability density baseline information B are then used to obtain the deviation characteristic value b through the values ​​of one or more indicators. The indicators for measuring probability density distribution include but are not limited to expected value, variance and standard deviation of each point data and expected value, skewness and kurtosis, etc.

[0072] The magnitude of the deviation feature b represents the difference between the breathing state of the monitoring subject during the so-called target monitoring period and the breathing state over a long period of time.

[0073] S6, determining whether the breathing state of the monitored object in the target monitoring period is normal according to the deviation feature b. The deviation feature b can be one or more numerical values, or multi-dimensional data, depending on the calculation method used in step S5. For example, if the deviation feature b is a numerical value and is positively correlated with the difference in breathing state, then the larger the value of the deviation feature b, the more different the breathing state of the monitored object in the (any) monitoring period is from the long-term breathing state, that is, the probability density information P j It deviates from the probability density baseline information B, thereby determining that the breathing state is abnormal; if the deviation feature b is not a numerical value, it can be judged according to the specific data type whether its characteristics are obvious or meet certain preset conditions, thereby determining whether the breathing state of the monitored object during the target monitoring period is normal.

[0074] The target monitoring period may be a period before, during or after the above-mentioned multiple monitoring periods. For example, in the above steps, the daily respiratory signals of a monitored object in a certain month of a year are used to generate the probability density baseline information B. Then, any period may be a monitoring period of a certain day in the month, used to determine whether the respiratory state of the monitored object on that day is normal; any period may be a monitoring period of a certain day before the month, used to determine whether the respiratory state of the monitored object is normal before generating the baseline; any period may be a monitoring period of a certain day after the month, used to determine whether the respiratory state of the monitored object is normal after generating the baseline.

[0075] Taking the application scenario as an example, for example, when the present solution is applied to sleep breathing monitoring of a human body, the above steps S1 to S4 can be used to generate the above probability density baseline information B based on the breathing signal of a user during a long period of sleep (such as a week or a month, etc.), and then the above steps S5 to S6 can be used to generate the probability density information P based on the breathing signal of the user in each subsequent sleep period. j , so as to monitor whether the sleep state of the user is normal. And the probability density baseline information B can be updated regularly or irregularly over time, that is, the probability density baseline information B is recalculated using the breathing signal of a period of time close to the current time or a specified arbitrary period of time.

[0076] Furthermore, the monitoring period further includes various sleep staging periods. Taking five stages as an example, the sleep period is divided into REM stage (rapid eye movement stage), N1 stage, N2 stage, N3 stage, and W stage (wakefulness stage). According to the above method, the respiratory signals in each sleep stage are processed respectively to generate probability density baseline information corresponding to each sleep stage, and then the probability density information of the monitored object in each sleep stage is compared with the corresponding probability density baseline information to determine whether the respiratory state of the monitored object in each sleep stage in a certain monitoring period is normal.

[0077] In one embodiment, determining the breathing rate BPM of each window breathing signal in the above step S2 further includes the following operations:

[0078] S21, respectively perform short-time Fourier transform on each window breathing signal to obtain corresponding spectrum information, and use the peak point in the spectrum information and the data points near it to calculate the confidence C of the spectrum information, wherein the energy value of the data points near the peak point is negatively correlated with the confidence C. Specifically, the window breathing signal is a signal of a certain time length (such as several seconds or minutes), and the specific implementation process may be affected by the behavior of the monitored object or the noise in the environment, for example, the monitored object may have micro-body movements such as turning over, and other moving objects may appear in the environment, all of which may affect the echo signal of the millimeter wave radar.

[0079] Spectrum information is the data of the corresponding relationship between frequency and energy. In an ideal state (without body movement and environmental influence), the spectrum information should have an obvious energy peak. When affected, the spectrum information will fluctuate to varying degrees. There may be multiple adjacent energy peaks in the spectrum information. For example, there are other slightly lower peaks near the highest energy peak point. The confidence C calculated in this step is used to reflect whether the data of each window is affected by the above influence.

[0080] S22, extracting the frequency of the peak point for the spectrum information corresponding to the respiratory signal in each window, that is, the frequency corresponding to the highest energy peak point in the spectrum.

[0081] S23, calculate the respiratory rate BPM corresponding to each window respiratory signal according to the frequency of the peak point. Specifically, first determine the sampling frequency Fs and the number of sampling points N. The sampling frequency refers to the number of samples per second, the unit is Hz, and the number of sampling points refers to the number of sample points used in an FFT calculation. Then calculate the frequency resolution. The frequency resolution Δf is the interval between each point on the frequency axis, Δf=Fs / N. For example, if the sampling frequency is 10Hz and the number of sampling points is 128 points, the frequency resolution is 0.078Hz. Then calculate BPM, determine the peak point index, assuming it is x, calculate the respiratory frequency f=Δf * x, assuming it is a Hz, BPM = a * 60.

[0082] S24, using the confidence C to screen the respiratory rate BPM of each window respiratory signal to obtain the respiratory rate BPM of multiple window respiratory signals that are credible. For example, an absolute threshold of the confidence can be set, and the respiratory rate BPM of the window respiratory signal with a confidence C higher than the absolute threshold is determined as credible data; or a relative threshold method can be used, such as determining that a certain proportion of the respiratory rate BPM of the window respiratory signal with a relatively higher confidence is credible data.

[0083] Based on the screening method of this embodiment, the disturbed window data can be discarded in the subsequent steps, and the respiratory rate BPM of multiple window respiratory signals with high credibility can be used to generate the probability density information P i , making the probability density information of a single monitoring period more accurate, and thus making the probability density baseline information B more accurate.

[0084] Further, step S21 may include:

[0085] S211, determine the maximum energy value and its corresponding frequency value in the spectrum information to obtain the highest peak point. Regardless of whether the window breathing signal is disturbed (whether there are multiple obvious peaks), the highest energy value can be found therein.

[0086] S212, extracting energy values ​​of data points within a first range centered on the highest peak point, and energy values ​​of data points within a second range centered on the highest peak point. For example, the first range is a frequency band close to the highest peak point, and the second range can be a relatively larger frequency band (the second range includes the first range), or a relatively larger frequency band minus the frequency band of the first range, thereby obtaining two segments of energy data.

[0087] S213, using the extracted energy values ​​to calculate confidence C. If the energy values ​​in the first range are significantly higher than the energy values ​​in the second range, it means that the highest peak point is the only obvious peak point, otherwise it means that there are other slightly lower peak points near the highest peak point.

[0088] As an example, the calculation can be performed as follows:

[0089] ;

[0090] Where C is the confidence level, P i is the energy value corresponding to the i-th frequency point in the spectrum, max is the frequency point corresponding to the energy peak point in the spectrum, L is the half-width of the main lobe interval (preset parameter, corresponding to the above-mentioned first range), A is the half-width of the side lobe interval (preset parameter, corresponding to the above-mentioned second range), and the larger the C value is, the more stable the waveform breathing characteristics in the window are, and the higher the confidence level is.

[0091] Further, step S22 may include:

[0092] S221, extracting frequency values ​​between a preset upper frequency limit and a preset lower frequency limit in the spectrum information. The maximum respiratory cycle and the minimum respiratory cycle that meet the monitoring object (such as the upper and lower limits of a human respiratory cycle) can be preset, and the upper and lower frequency limits are obtained according to the cycle.

[0093] S222, determine the frequency of the peak point with the maximum frequency value in the extracted frequency values. In the spectrum information obtained in step S21, only the energy peak point is extracted within this frequency range.

[0094] This embodiment extracts a reasonable energy peak in the spectrum information by setting the upper and lower frequency limits, thereby avoiding interference to the respiratory signal caused by the body movement of the monitored object or environmental factors, and improving the reliability of the respiratory rate of the respiratory signal in each window.

[0095] Based on the above embodiment of calculating the confidence level, in one embodiment, step S4 may include the following operations:

[0096] S41A, calculating the coefficient w according to the number of credible multiple window breathing signals and the number of all window breathing signals in a single monitoring period i , where the number of credible multiple window breathing signals and the coefficient w i According to the above embodiment, the total number of windows in each monitoring period is the same, but the number of credible multiple window respiratory signals may be different, then the corresponding coefficient w i There are many specific calculation methods, as long as they can reflect the above quantity and coefficient w i For example, the coefficient wi It can be the result of dividing the number of credible window breathing signals on a certain day by the number of all window breathing signals on that day (the proportion of credible windows).

[0097] S42A, using the probability density information P of each monitoring period i and the corresponding coefficient w i Get the probability density baseline information B. Coefficient w i Used to measure the probability density information P of each monitoring period i In the weight used to generate the probability density baseline information B, the higher the coefficient of a certain monitoring period, the greater the impact / contribution to the generation of the probability density baseline information B. This allows the generated result to more accurately reflect the long-term respiratory state of the monitored subject, while reducing the impact of the unstable respiratory state on a certain day on the baseline.

[0098] As an example, the probability density baseline information B can be calculated as follows:

[0099] ;

[0100] Among them, B is the probability density baseline information, P i represents the probability density information of the ith monitoring period, w i represents the coefficient of the i-th monitoring period, and N represents the number of monitoring periods.

[0101] In one embodiment, in step S3, the probability density information P of the breathing rate BPM of the breathing signals in the multiple windows is calculated. i , further comprising the following operations:

[0102] S31, using the kernel function K to obtain the one-dimensional probability density function of the respiratory rate BPM of the respiratory signals in multiple windows. This embodiment adopts the kernel density estimation method, as an example, the one-dimensional probability density function can be

[0103] ;

[0104] Where N is the number of data points (the number of windows / the number of breathing rate BPM), h is the bandwidth (i.e. the preset width of the kernel function), and x is the i is the value of the data point (the respiratory rate of the respiratory signal in the i-th window), and x is the one-dimensional probability density function P (x) The independent variable, K(u) is the kernel function. There are many optional kernel functions, and the Gaussian kernel function can usually be used:

[0105] ;

[0106] S32, normalize the one-dimensional probability density function to obtain probability density information P i, which includes the values ​​of each respiratory rate and its corresponding probability value. As mentioned above, the probability density information P i It can be understood as a curve data, in which the higher the probability value corresponding to a certain respiratory rate value, the more window data with this respiratory rate there are in all the window data in a monitoring period.

[0107] This embodiment uses the kernel density estimation method to calculate the probability density curve, which has the advantages of non-parametric, smoothness, visualization, and density estimation.

[0108] In one embodiment, the above step S4 further comprises:

[0109] S41B, respectively determine the probability density information P in multiple monitoring periods i The eigenvalues ​​of X i , and calculate the average eigenvalue X mean .

[0110] Eigenvalue X i It can be the probability density information P i A certain breathing rate in the probability density information P i There may be one or more obvious probability peaks in the eigenvalue X i It can be the breathing rate corresponding to the peak probability, where the peak probability can be the maximum probability value (unique) or multiple relatively large probability values ​​(multiple peaks), that is, the eigenvalue X i It can be one or more. Eigenvalue X i It can also be used to measure the probability density information P i One or more indicators of the data, such as skewness and kurtosis, etc.

[0111] Average eigenvalue X mean It can be the characteristic value X of multiple monitoring periods i The arithmetic mean, weighted mean, etc. of the average eigenvalue X mean It reflects the long-term respiratory characteristics of the monitored object.

[0112] S42B, calculate the characteristic value X in each monitoring period respectively i With the average eigenvalue X mean The difference between the two is calculated, and the stability information bias of multiple monitoring periods is obtained based on all the differences. The characteristic value X in a certain monitoring period i With the average eigenvalue X mean The greater the difference, the more abnormal the breathing state of the monitored object during this period. For example, if the monitoring is performed every day for one month, the daily characteristic value X i ; Daily characteristic value X i With the average eigenvalue X meanThe difference of a single day is obtained by subtracting the difference, and the difference of a month is accumulated to obtain the stability information bias, which reflects the stability of the breathing state of the monitored object over a long period of time.

[0113] The data type of the stability information bias depends on the eigenvalue X i , which can be one or more values, specifically the difference value of respiratory rate, the difference value of probability density curve index, etc.

[0114] It should be noted that the difference described in this embodiment may be subtraction, and in addition to subtraction, it may at least be similarity information, and the higher the similarity, the smaller the difference.

[0115] S43B, determine whether the stability information bias exceeds the full-time stability threshold. In this embodiment, the so-called "exceed" can be greater than or less than. If the stability information bias is based on the result obtained by difference, the larger the stability information bias is, the more unstable the breathing state is; if the stability information bias is based on the result obtained by similarity, the smaller the stability information bias is, the more unstable the breathing state is.

[0116] S44B, if the stability information bias does not exceed the full-time stability threshold, the probability density baseline information B is generated using the probability density information of the monitoring period whose difference does not exceed the single-time period stability threshold.

[0117] If the stability information bias exceeds the full-time stability threshold, it means that the long-term breathing state of the monitored object is inherently unstable, and the data during these periods are not suitable for generating a baseline. The data from other monitoring times can be used for re-judgment, or the reasons for the instability can be analyzed.

[0118] If the stability information bias does not exceed the full-time stability threshold, it means that the long-term breathing state of the monitored object is relatively stable, but the data can be further filtered. As mentioned above, the characteristic value X can be calculated for each monitoring period. i With the average eigenvalue X mean In this embodiment, a single-day threshold (single-period stability threshold) is further set. If the difference on a certain day exceeds the single-day threshold, the probability density information of the day is discarded.

[0119] As an example, suppose there are 30 days (monitoring period) of respiratory signals, corresponding to 30 sets of eigenvalues ​​X 1~ X 30 Its average eigenvalue is X mean30 , further calculate the stability information bias; if the bias exceeds the full-time stability threshold Y 30Continue, otherwise it means that the data of these 30 days or the patient is not suitable for respiratory monitoring using this scheme. Then judge based on X1~X 30 and X mean30 The stability information obtained is bias1~bias 30 Whether it exceeds the single period stability threshold Y i , take the data that does not exceed the probability density baseline information B, assuming that there are X3, X 19 , X 20 Corresponding bias3, bias 19 、bias 20 Exceeds Y i , then the data of these 3 days are enabled, and the probability density baseline information B is generated based on the probability density information of the remaining 27 days of respiratory signals 27 .

[0120] It should be noted that the operation of step S44B of this embodiment can be combined with the operations of steps S41A to S42A in the above embodiment, and the coefficient w is further calculated for the probability density information of the monitoring period whose difference does not exceed the single period stability threshold. i , and combined with the probability density information P i and the corresponding coefficient w i Generate probability density baseline information B.

[0121] In one embodiment, in step S5, the deviation feature b (deviation feature value b1) is obtained in the following manner:

[0122] Determine the respiratory rate B corresponding to the maximum probability in the probability density baseline information B max , and the probability density information P during the target monitoring period j Determine the respiratory rate P corresponding to the maximum probability max , using the respiratory rate B max and respiratory rate P max Get the deviation characteristic value b1, which can be b1=P max -B max . Respiration rate B max It can be understood as the most common respiratory rate of the monitored object over a long period of time, the respiratory rate P max It can be understood as the respiratory rate that the monitored subject most often presents on a single day, so b1 can be regarded as the deviation or fluctuation of the respiratory rate of the monitored subject on that day relative to his long-term respiratory rate. b1 reflects the amplitude of the respiratory rate change on that day, indicating the difference between the respiratory rate on that day and his baseline respiratory rate.

[0123] In one embodiment, in step S5, the deviation feature b (deviation feature value b2) is obtained in the following manner:

[0124] The probability density baseline information B and the probability density information P of the target monitoring period are respectively j Extract the probability value B corresponding to the same respiratory rate value x x and P x , and calculate the difference between each pair of probability values ​​extracted; the difference between each pair of probability values ​​is accumulated to obtain the deviation characteristic value b2. As a reference, the calculation formula is used:

[0125] ;

[0126] bpm min Represents probability density information P j and the minimum value of the respiratory rate in the probability density baseline information B, bpm max Represents probability density information P j And the maximum value of the respiratory rate in the probability density baseline information B.

[0127] b2 is a comprehensive characteristic value that describes the difference between the respiratory rate distribution during the target monitoring period and the baseline distribution, and is used to evaluate the stability and change trend of the respiratory rate.

[0128] b1 and b2 in the above two embodiments can be used either selectively or simultaneously.

[0129] Further, step S6 includes the following operations:

[0130] Determine whether the deviation characteristic value b1 and / or the deviation characteristic value b2 exceeds the set threshold; when the deviation characteristic value b1 and / or the deviation characteristic value b2 exceeds the set threshold, determine that the breathing state of the monitored object in the target monitoring period is abnormal.

[0131] If b1 is large, it means that the respiratory rate of the monitored subject during the monitoring period has increased significantly compared with the baseline respiratory rate, which may indicate a change in some physiological or health status.

[0132] If b2 is large, it means that the respiratory rate distribution of the monitored subject during the monitoring period is significantly different from the baseline, which may indicate that the breathing pattern of the monitored subject has changed, which may be due to the influence of health status, activity level, environment and other factors.

[0133] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0134] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0135] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0136] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0137] Obviously, the above embodiments are merely examples for the purpose of clear explanation, and are not intended to limit the implementation methods. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the implementation methods here. The obvious changes or modifications derived therefrom are still within the scope of protection of the invention.

Claims

1. A respiratory status monitoring method, characterized in that: include: Extract the breathing signal s (t) from the echo signal of the millimeter-wave radar within a single monitoring period; Extracting multiple window breathing signals from the breathing signal s(t) by sliding window, and determining the breathing rate BPM of each window breathing signal respectively; Calculate the probability density information P of the respiratory rate BPM of the multiple window respiratory signals i ; Using the probability density information P in multiple monitoring periods of the same monitoring object i Generate probability density baseline information B, including respectively determining the characteristic values ​​Xi of the probability density information Pi in a plurality of the monitoring periods, and calculating the average characteristic value Xmean; respectively calculating the difference between the characteristic value Xi in each of the monitoring periods and the average characteristic value Xmean, and obtaining the stability information bias of the monitoring period from all the differences; and judging whether the stability information bias exceeds the stability threshold of the entire period; If the stability information bias does not exceed the full-time stability threshold, the probability density baseline information B is generated using the probability density information of the monitoring period in which the difference does not exceed the single-time stability threshold; By using the probability density information P of the monitored object in the target monitoring period j Compare with the probability density baseline information B to obtain the deviation feature b; Determine whether the breathing state of the monitored object during the target monitoring period is normal based on the deviation feature b.

2. The method according to claim 1, characterized in that Determining the breathing rate BPM of each window breathing signal separately further includes: Performing short-time Fourier transform on each of the windowed respiratory signals to obtain corresponding spectrum information, and calculating the confidence C of the spectrum information using the peak point and the data points near the peak point in the spectrum information, wherein the energy value of the data points near the peak point is negatively correlated with the confidence C; Extracting the frequency of the peak point from the spectrum information corresponding to each of the windowed respiratory signals; Calculate the breathing rate BPM corresponding to each of the window breathing signals according to the frequency of the peak point; The confidence level C is used to screen the respiratory rates BPM of the respiratory signals in each window to obtain the respiratory rates BPM of the plurality of respiratory signals in the window that are credible.

3. The method according to claim 2, characterized in that Calculating the confidence C of the spectrum information using the peak point in the spectrum information and the data points near the peak point, further comprising: Determine the maximum energy value and its corresponding frequency value in the spectrum information to obtain the highest peak point; Extracting energy values ​​of data points within a first range centered on the highest peak point and energy values ​​of data points within a second range centered on the highest peak point; The confidence C is calculated using the extracted energy value.

4. The method according to claim 2, characterized in that: Extracting the frequency of the peak point from the spectrum information corresponding to each of the windowed respiratory signals further includes: Extracting frequency values ​​between a preset upper frequency limit and a preset lower frequency limit from the frequency spectrum information; The frequency of the peak point is obtained by determining the maximum frequency value among the extracted frequency values.

5. The method according to claim 1, characterized in that Calculate the probability density information P of the respiratory rate BPM of the multiple window respiratory signals i , further comprising: Using the kernel function K, a one-dimensional probability density function of the respiratory rate BPM of the respiratory signals in the multiple windows is obtained; Normalize the one-dimensional probability density function to obtain the probability density information P i , which includes the values ​​of each respiratory rate and its corresponding probability value.

6. The method according to claim 1, characterized in that The eigenvalue X i is the probability density information P i The probability of at least one peak in corresponds to the breathing rate.

7. The method according to claim 2, characterized in that Generating probability density baseline information B further includes: The coefficient w is calculated based on the number of the reliable multiple window breathing signals and the number of all window breathing signals in a single monitoring period. i , where the number of credible multiple window breathing signals and the coefficient w i There is a positive correlation; Using the probability density information P of each monitoring period i and the corresponding coefficient w i Obtain probability density baseline information B.

8. The method according to claim 7, characterized in that The probability density baseline information B is calculated as follows: ; Wherein, B is the probability density baseline information, P i represents the probability density information of the i-th monitoring period, w i represents the coefficient of the i-th monitoring period, and N represents the number of the monitoring periods.

9. The method according to claim 1, characterized in that: Obtaining the deviation feature b further includes: Determine the respiratory rate B corresponding to the maximum probability in the probability density baseline information B max , and the probability density information P during the target monitoring period j Determine the respiratory rate P corresponding to the maximum probability max , using the respiratory rate B max and respiratory rate P max Obtaining a deviation characteristic value b1; and / or The probability density baseline information B and the probability density information P of the target monitoring period are respectively j Extract the probability value B corresponding to the same respiratory rate value x x and P x , and calculate the difference between each pair of probability values ​​extracted; The differences between each pair of probability values ​​are accumulated to obtain the deviation characteristic value b2.

10. The method according to claim 9, characterized in that Determining whether the breathing state of the monitored object in the target monitoring period is normal according to the deviation feature b further includes: Determine whether the deviation characteristic value b1 and / or the deviation characteristic value b2 exceeds a set threshold; When the deviation characteristic value b1 and / or the deviation characteristic value b2 exceeds a set threshold, it is determined that the breathing state of the monitored object during the target monitoring period is abnormal.

11. The method according to claim 1, characterized in that: The target monitoring period is a period before, during or after the plurality of monitoring periods.

12. The method according to claim 1, characterized in that The monitoring period is a period when the monitored object is in a sleeping state.

13. The method according to claim 1, characterized in that The monitoring period further includes various sleep staging periods.

14. A respiratory status monitoring device, characterized in that: include: A processor and a memory connected to the processor; wherein the memory stores instructions executable by the processor, and the instructions are executed by the processor so that the processor executes the respiratory status monitoring method as described in any one of claims 1-13.

Images