System and method for remote tracking of vital signs with millimeter wave radar

Abstract

Systems and methods for monitoring vital signs of a subject. A radar unit includes generating raw data, and a processor unit receives the raw data and identifies an oscillatory signal in a series of complex value arrays representing radiation reflected from each voxel of a target region during a given time segment. A phase value is determined for each voxel in each frame, and a waveform representing the change in phase over time is generated for each voxel. Voxel(s) indicative of the oscillatory signal are selected and vital sign parameters are extracted.

Description

[0001] Cross-reference to related applications

[0002] This application claims the entire contents of U.S. Provisional Patent Application No. 62 / 977,902, filed February 18, 2020; U.S. Provisional Patent Application No. 63 / 030,943, filed May 28, 2020; U.S. Provisional Patent Application No. 63 / 021,373, filed May 7, 2020; U.S. Provisional Patent Application No. 63 / 069,050, filed August 23, 2020; and U.S. Provisional Patent Application No. 63 / 135,782, filed January 11, 2021. Technical Field

[0003] This disclosure relates to systems and methods for monitoring an individual's vital signs using millimeter-wave radar. Specifically, it relates to systems and methods for tracking displacement of body parts caused by the respiratory cycle and for monitoring respiratory characteristics, heart rate, or pulse patterns. Background Technology

[0004] A well-known method for detecting small displacements in narrowband radar systems is to utilize the phase change of the received signal over time. This method is typically best performed by first selecting a voxel most suitable for frequency determination.

[0005] However, selecting the optimal voxel for extracting the breathing pattern is challenging, while also being robust to other motion and noise. This disclosure relates to selecting voxels, extracting waveforms of phase changes from selected voxels, and combining breathing waveforms from multiple voxels.

[0006] The need for efficient and reliable remote monitoring of vital signs remains. The invention described herein addresses this need. Summary of the Invention

[0007] Based on one aspect of the currently disclosed subject matter, a method for monitoring the vital signs of an object within a target area is taught herein. The method may include providing a radar unit comprising: at least one transmitter antenna connected to an oscillator and configured to transmit electromagnetic waves into the target area; and at least one receiver antenna configured to receive electromagnetic waves reflected by an object within the target area and operable to generate raw data. The method may further include providing a processor unit configured to receive the raw data from the radar unit and operable to identify oscillating signals in the raw data.

[0008] The radar can transmit and receive scanning radiation over a target region; the radar can produce a series of frames, each frame comprising an array of complex values representing radiation reflected from each voxel of the target region during a given time segment. Thus, the processor can collate, for each voxel, a series of complex values representing reflected radiation of the associated voxel in a plurality of frames. The method can further comprise: determining, for each voxel, a center point in the complex plane; determining, in each frame, a phase value for each voxel; generating, for each voxel, a smoothed waveform representing the phase as a function of time; selecting a subset of voxels indicative of an oscillatory signal; and extracting a vital sign parameter from the oscillatory signal.

[0009] Optionally, the step of selecting the subset of voxels indicative of an oscillatory signal comprises selecting the subset of voxels indicative of a breathing pattern.

[0010] Additionally or alternatively, the step of selecting the subset of voxels indicative of an oscillatory signal comprises: calculating, for the phase value associated with each voxel, an arc fit metric; and selecting voxels having an arc fit metric above a threshold value.

[0011] Still additionally or alternatively, the step of selecting the subset of voxels indicative of an oscillatory signal comprises: calculating, for the phase value associated with each voxel, a time-dependent function; and selecting voxels having a periodic characteristic indicative of respiration. For example, the periodic characteristic indicative of respiration is selected from the group consisting of: frequency, inhalation duration, exhalation duration, respiratory rate, and combinations thereof.

[0012] In yet another example of the method, the step of selecting the subset of voxels indicative of an oscillatory signal comprises: generating a line segment model for the oscillatory signal; extracting characteristic features from the line segment model; and performing a validation function to determine whether the line segment model is indicative of a true respiration signal.

[0013] Optionally, the step of generating the line segment model comprises: selecting, for each significant tilt segment in the oscillatory signal, a boundary point, wherein the boundary point for each significant tilt segment comprises: a slope start point indicative of a significant change from a previous extreme; and a slope end point sharing a Y coordinate with a subsequent slope start point of a next significant tilt segment.

[0014] Thus, the step of generating the line segment model can further comprise constructing a trapezoidal line segment model by: constructing a line segment between the boundary points of each significant tilt segment, and constructing a line segment between the end point of each significant tilt segment and the start point of a next significant tilt segment.

[0015] Additionally or alternatively, the step of selecting the boundary points can comprise: defining a detection threshold; sampling the signal values in the oscillatory signal until an extreme signal is detected; detecting the extreme signal point; sampling the signal values in the oscillatory signal after the extreme signal is detected; calculating the difference between each sampled signal and the extreme signal; and, if the difference is greater than the detection threshold: setting a slope start point for the line segment after the extreme signal; and setting a slope end point for the line segment before the extreme signal. Thus, the step of setting a slope start point for the line segment after the extreme signal can comprise selecting a first signal after the extreme signal. Optionally, the step of setting a slope end point for the line segment before the extreme signal comprises selecting a signal sample before the extreme, the signal value of which is equal to the signal value of the slope start value.

[0016] In a particular example of the method, the step of selecting the subset of voxels indicative of the breathing pattern comprises: providing a breathing parameter determination module comprising a pre-filter, a point generation module, a feature extraction module, a metric generation module, and a respiratory signal identification module; the pre-filter providing candidate signals to the point generation module; the point generation module generating a line segment model corresponding to the candidate signals; the feature extraction module receiving the line segment model and extracting characteristic features from the line segment model; the metric generation module processing the characteristic features to calculate a characteristic metric indicative of a probability that each feature is consistent with the breathing pattern; the respiratory signal identification module receiving the characteristic metric and determining a probability that the candidate signal represents a respiratory signal.

[0017] Thus, the step of the metric generation module processing the characteristic features to calculate a characteristic metric can comprise: generating a log-likelihood ratio (LLR) for each feature; assigning a positive or negative sign to each LLR for each feature; and summing the LLR values.

[0018] Optionally, the step of extracting vital sign parameters from the oscillatory signal comprises determining a heart rate parameter. Where required, this can comprise removing breathing-related oscillations from the oscillatory signal, which step can comprise: obtaining a set of reference signal curves with known heart rates; applying a line segment approximation to a plot of the phase values over time; applying a high-pass filter to identify only oscillations above a threshold frequency value; applying a correlator to compare the filtered phase signal to each of the set of reference signal curves; obtaining characteristic reference correlation curves; selecting a peak correlation heart rate value; and, optionally, also applying a validation rule such that only possible heart rates are selected.

[0019] In appropriate cases, the step of providing a radar monitor comprises providing a first radar transceiver device directed at the common target region from a first angle and providing a second radar transceiver device directed at the common target region from a second angle, and the step of removing from the data the oscillations related to respiration comprises summing the phase signals received from the first radar transceiver and the phase signals received from the second radar transceiver such that radial movements are removed.

[0020] Thus, the step of removing from the oscillation signal the oscillations related to respiration can comprise: obtaining a set of phase values associated with energy reflected from a selected voxel of the target region in a set of consecutive time periods; creating a delta file by subtracting the phase value of a previous time period from the phase value of each time period; and subtracting a moving average of the phase values.

[0021] Additionally or alternatively, the step of determining the heart rate parameter can comprise: obtaining a heart rate signal; obtaining a characteristic reference correlation curve indicative of a degree of periodicity of the heart rate signal as a function of a candidate heart rate; and selecting a heart rate value corresponding to a peak periodicity correlation. Optionally, the step of obtaining a characteristic reference correlation curve indicative of a degree of periodicity of the heart rate signal as a function of a candidate heart rate comprises: sampling a section of the heart rate signal; for each of a set of candidate heart rates: segmenting the section of the heart rate signal into a plurality of segment windows of equal duration corresponding to a period of the particular candidate heart rate; and correlating adjacent segment windows to obtain a correlation index indicative of a degree of periodicity corresponding to the candidate heart rate.

[0022] In various examples of the method, the step of determining a phase value for each voxel in each frame comprises: the processor collating for each voxel a series of complex values representing reflected radiation of the associated voxel in a plurality of frames; determining a center point in a complex plane for each voxel; and calculating an inverse tangent of a ratio of an imaginary component and a real component of a difference between the frame value and the center point. Furthermore, the step of generating for each voxel a smoothed waveform representing a change in phase over time can comprise rounding each phase value.

[0023] According to other aspects of the present disclosure, systems for monitoring vital signs of objects within a target area are introduced. Such a system can include a radar unit including at least one transmitter antenna connected to an oscillator and configured to transmit electromagnetic waves into a target area, and at least one receiver antenna configured to receive electromagnetic waves reflected by objects within the target area and operable to generate raw data, and a processor unit configured to receive the raw data from the radar unit and operable to identify oscillatory signals in the raw data. The processor unit can include a frame assembler including a memory unit configured to store a series of complex frame values for each voxel within the target area, a voxel segmentation module including a processor configured to generate, for each voxel, a smoothed waveform representing a change in phase over time, a respiration parameter determination module, and a heart rate parameter determination module. Optionally, the heart rate parameter determination module includes a signal filter module and a correlation module.

[0024] Optionally, the system includes an overhead radar unit directed to the target area, and a seat positioned such that an upper body of an object seated on the seat is within the target area. The seat can be positioned such that a back of the object faces the radar-based remote heart rate monitor. Optionally, the seat is positioned such that a chest of the object faces the radar remote heart rate monitor.

[0025] In appropriate cases, the heart rate monitor includes a radar unit directed to the target area, and a foot rest positioned such that an upper side of a foot of an object placed on the foot rest is within the target area.

[0026] Additionally or alternatively, the remote heart rate monitor includes a first radar unit directed to the target area, and a second radar unit directed to the target area.

[0027] In some examples, the signal filter unit includes a line segment approximation module and a high pass filter module. Optionally, the signal filter unit includes a delta file generation unit configured and operable to: obtain a set of phase values associated with energy reflected from a selected voxel of the target area in a set of consecutive time periods; create a delta file by subtracting a phase value of a previous time period from a phase value of each time period; and subtract a moving average of the phase values. BRIEF DESCRIPTION OF DRAWINGS

[0028] For a better understanding of the present embodiments, and to show how they can be put into effect, reference will now be made, by way of example, to the accompanying drawings in which:

[0029] Referring now to the details of the accompanying drawings, it should be emphasized that the details shown are by way of example and are intended only to illustrate selected embodiments, and are meant to provide a description that is considered most useful and readily understood in terms of principles and concepts. In this regard, no attempt is made to show more detailed structural details than are necessary for a basic understanding; the description accompanying the drawings clearly shows those skilled in the art how the various selected embodiments are put into practice. In the drawings:

[0030] FIG. 1A This is a schematic diagram of a possible vital signs monitoring and communication system;

[0031] FIG. 1B to FIG. 1D Various configurations of a device for directing radar waves to an object are shown;

[0032] FIG. 2 This is a flowchart illustrating the actions involved in vital sign monitoring methods;

[0033] FIG. 3A to FIG. 3F An example of plotting a series of complex values ​​representing the reflected radiation of a single voxel within a target area across multiple frames;

[0034] FIG. 4A and FIG. 4B A system configured to direct radar waves from two cells toward an object from different directions is shown.

[0035] FIG. 4C A single chest guide unit is shown for a standing object;

[0036] FIG. 4D A radar device integrated into a headrest and directed towards the neck is shown;

[0037] FIG. 5A The diagram illustrates the use of in, for example FIG. 4D The method involves processing radar signals acquired in a dual-radar system, such as the system shown, to generate a single displacement signal and to separate a heart rate signal from the single displacement signal.

[0038] FIG. 5B It is a graph illustrating an example of a spiked signal generated by a moving object;

[0039] FIG. 5C It is illustrated with FIG. 5B A graph showing the standard deviation of the average value of each frame corresponding to the signal;

[0040] FIG. 5D The illustration shows another method for separating the heart rate signal from the displacement signal;

[0041] FIG. 6A A possible method for generating a correlation index for a heart rate signal corresponding to a particular heart rate is shown;

[0042] FIG. 6B A possible method for generating a correlation curve for a heart rate signal and selecting a possible heart rate is shown; and

[0043] FIG. 6C An alternative method for generating a correlation curve and selecting a possible heart rate is shown;

[0044] FIG. 7 An example of a possible Morse wavelet for estimating RPM is shown;

[0045] FIG. 8 is a flow chart representing a possible method for determining whether a candidate periodic waveform is consistent with characteristics of a respiratory signal;

[0046] FIG. 9 is a plot indicating how a line segment model comprising a series of trapezoidal segments can represent a periodic signal;

[0047] FIG. 10 Elements of a respiratory parameter determination module are schematically represented, the respiratory parameter determination module being configured to process a periodic waveform and determine a probability that a candidate waveform is indicative of a respiratory signal;

[0048] FIG. 11A to FIG. 11C are combined to represent a schematic block diagram indicating how components of a system can be combined to process a candidate signal;

[0049] FIG. 12A and FIG. 12B are plots representing examples of superimposing line segment models on a periodic signal from which the line segment models were generated;

[0050] FIG. 12C and FIG. 12D indicate characteristic features extracted from line segment models of plots of FIG. 12A and FIG. 12B respectively;

[0051] FIG. 13A is a flow chart illustrating main steps of a method for determining key points defining a start and an end of each line segment;

[0052] FIG. 13B is a flow chart illustrating a possible method for setting values of a slope start point and a slope end point;

[0053] FIG. 14A andFIG. 14B is a plot of a section of a periodic signal indicating extreme points and threshold values that can be used to calculate a slope start point and a slope end point;

[0054] FIG. 15 is a plot illustrating how a swing index varies over time for a fluctuating signal; and

[0055] FIG. 16A is a flowchart, and FIG. 16B is a schematic diagram illustrating a method for generating a feature metric. DETAILED DESCRIPTION

[0056] Various aspects of the present disclosure relate to systems and methods for monitoring vital signs of an individual by using millimeter wave radar. The systems and methods described herein are used to track displacement of a body part during a respiratory cycle. Displacement patterns can be analyzed to extract respiratory characteristics and identify heart rate and pulse patterns.

[0057] It is important to note that tracking such displacement during a respiratory cycle can be used to monitor physiological parameters during sleep time and / or wake-up time. Furthermore, respiratory tracking can be used to identify humans and other targets and their posture for additional radar applications, such as fall detection.

[0058] The detailed implementations of the present invention are disclosed herein as needed; however, it is to be understood that the disclosed implementations are merely exemplary of the present invention, which can be embodied in various forms and alternative forms. The drawings are not necessarily to scale; some features can be exaggerated or minimised for clarity of illustrative purposes. Therefore, specific structural and functional details disclosed herein are not to be interpreted as limiting, but merely as a representative basis for teaching one skilled in the art to employ the present invention in a variety of ways.

[0059] In various implementations of the present disclosure, one or more tasks described herein can be performed by a data processor, such as by a computing platform or distributed computing system for executing a plurality of instructions. Optionally, the data processor includes or has access to a volatile memory for storing instructions, data, etc. Additionally or alternatively, the data processor can access a non-volatile memory device for storing instructions and / or data, e.g., a magnetic hard disk, a flash drive, removable media, etc.

[0060] It is important to note that the systems and methods disclosed herein can not be limited to the details of the structures and arrangements of components or methods set forth in the descriptions of the application or illustrated in the drawings and examples. The systems and methods of the present disclosure can have other implementations or be practiced with various ways and techniques.

[0061] Alternative methods and materials similar or equivalent to those described herein can be used in the practice or testing of embodiments of the present disclosure. However, particular methods and materials are described herein for illustrative purposes. These materials, methods, and examples are not intended to be limiting.

[0062] Reference is now made to FIG. 1A , which FIG. 1A is a schematic representation of a vital signs monitor 100 according to an embodiment of the present application. The vital signs monitor can include a radar unit 110, a processor unit 120, and a communication module 130. It is noted that the processor unit 120 can include a respiration characteristic determination module 126 operable to extract respiration data, and a pulse determination module 128 operable to extract heart rate data.

[0063] The radar unit includes a transmitter array 112 and a receiver array 114. An oscillator 116 can be connected to at least one transmitter antenna or transmitter antenna array. Thus, the transmitter 112 can be configured to produce a beam of electromagnetic radiation, such as microwave radiation, directed towards a monitored area 105, such as an enclosed room or the like. The receiver 114 can be operable to receive electromagnetic radiation reflected from an object 102 within the monitored area 105.

[0064] It is noted that the object 102 can stand, sit, recline or walk in various ways within the monitored area. If desired, the object 102 can sit with his or her back to the device 110B, such that the radiation is directed towards the object's back, as shown in FIG. 1B Alternatively, the object can sit with his or her chest towards the device 110C, such that the radiation is directed towards the object's chest, as shown in FIG. 1C As another alternative, the device 110D can be configured such that the radiation is directed towards the top of the object's foot, as shown in FIG. 1D for example, where it can receive strong reflections from the dorsal artery 103. Yet other systems can use multiple transceivers, as described below with respect to FIG. 4A and FIG. 4B .

[0065] The processor unit 120 can include modules such as a frame assembler 122, a voxel segmentation module 124, a respiration parameter determination module 126, and a heart rate parameter determination module 128. Thus, the processor unit 120 can be configured to receive raw data from the radar unit 110, and operable to generate respiration parameters and pulse parameters based on the received data.

[0066] The communication module 130 is configured and operable to communicate the pulse parameters to a third party 138. Optionally, the communication module 130 can communicate with a computer network 132, such as the Internet, via which it can communicate alerts to a third party (e.g., via a phone, computer, wearable device, etc.).

[0067] Reference is now made to FIG. 2 the flowchart of FIG. 1, presenting selected steps in a possible method for determining respiratory parameters within a target region.

[0068] The method includes providing a radar unit 202, providing a processor unit 204 to identify at least one heart rate pattern within a target region, the radar transmitting and receiving scanning radiation 206 over the target region, the processor unit: collating a series of complex values 208 for each voxel, the complex values representing reflected radiation of the associated voxel in a plurality of frames; determining a center point 210 in the complex plane for each voxel; determining a phase value 212 for each voxel in each frame; generating a smoothed waveform 214 representing the change in phase over time for each voxel; selecting a subset of voxels 216 indicative of a breathing pattern; and determining a breathing function 218 for the detected breathing pattern; selecting a subset of voxels 220 indicative of a pulse pattern; and determining a pulse function 222 for the detected pulse pattern.

[0069] The radar unit can include at least one transmitter antenna connected to an oscillator and configured to transmit electromagnetic waves into a monitored region, and at least one receiver antenna configured to receive electromagnetic waves reflected by objects within the monitored region and operable to generate raw data. The processor unit is generally configured to receive the raw data from the radar unit and operable to derive respiratory and pulse parameters from the raw data.

[0070] Optionally, the processor can generate a series of frames, wherein each frame includes an array of complex values representing radiation reflected from each voxel of the target region during a given time period.

[0071] In one embodiment, the method of the present application monitors a plurality of voxels in parallel over a period of time. The signal received by the receiver can be given by:

[0072]

[0073] where v is the index of the voxel, n is the time index; A v is the DC part of the voxel, due to leakage and static objects; R v is the amplitude (or radius) of the phase variation part of the voxel v; φ v is the nuisance phase offset of the voxel v; λ is the wavelength; Bv is the effective displacement amplitude of voxel v; v v [n] is the additional noise; and w[n] is the waveform at time n.

[0074] For each monitored voxel, a reference center point By way of illustration, the center can be determined from a linear mean square error estimator of the circle center. For example, the estimation can be based on the moments of the real and imaginary parts of the received signal. The moments can be averaged using an infinite impulse response (IIR) filter. The forgetting factor of the IIR filter has an adaptive control that balances the need to rapidly converge to new values when the environment changes (e.g. movement of an object) with the need to maintain consistency of the estimate.

[0075] The phase value for each voxel in each frame can be determined by a processor that: collates, for each voxel, a series of complex values representing reflected radiation of the associated voxel in a plurality of frames; and determines, for each voxel, a center point in a complex plane; and computes an inverse tangent of a ratio of an imaginary component to a real component of a difference between the frame value and the center point.

[0076] Thus, given the reference center point, the phase of voxel v at a given time n can be computed as:

[0077]

[0078] To create a smooth waveform representing the phase as a function of time, the phase values can be rounded. The phase can be unwrapped to generate a smooth waveform without discontinuities greater than π according to the following equation:

[0079]

[0080] In another implementation, the phase unwrapping can be based on predicting the next phase from the previous few phases. Such a prediction can be used to reduce the frame rate and / or increase resilience to noise while avoiding periodic drift. For example, the following predictor tends to take into account the momentum of the phase (linear progression of the phase) and the stability (zeroth order hold):

[0081]

[0082] where 0 < a < 1 is a parameter that controls the weight of the momentum (linear progression of the phase) and the stability (zeroth order hold).

[0083] FIG. 3AFigures 3A-3D show examples of plots representing a series of complex values of reflected radiation for a single voxel within a target region over multiple frames. Note that the complex values form an approximate circle centered at a single point in the complex plane. Periodicity can be seen over multiple frames, giving rise to an oscillatory function of the characteristic, such as FIG. 3B and FIG. 3E The plots shown in Figures 3A-3D represent the change in phase over time (using frame number as a proxy for time) for FIG. 3A and FIG. 3B The plots shown in Figures 3A-3D represent the change in phase over time (using frame number as a proxy for time) for FIG. 3C and FIG. 3F show the corresponding change in frequency space. Such an oscillatory function can be indicative of a breathing or pulse rate, for example. The amplitude is referenced to the plot.

[0084] For example, the voxels indicative of breathing and pulse characteristics can be found by selecting a subset of voxels that meet a selection rule, such as using a metric that evaluates the suitability of the voxels. Such a metric can include a fit to a circular arc model, a fit to a predetermined pattern pulse waveform with strong periodicity, and the spatial location of the voxel.

[0085] In many cases, the voxels that are best suited for breathing tracking are located near the chest and abdomen of the respiration. In other cases, the best choices are other voxels, such as reflections from walls or ceilings, or other objects that move due to respiration.

[0086] An arc fit metric can be computed for the phase values associated with each voxel; and the selected voxels will be those with an arc fit metric above a predetermined threshold. For example, the metric can evaluate the accuracy of fitting the data to the model described herein. The relative stability can be measured from the distance between the received signal and the estimated reference center point.

[0087]

[0088] For example, the metric can be computed as:

[0089]

[0090] Additionally or alternatively, a time dependence function can be computed for the phase values associated with each voxel; and voxels with periodic characteristics indicative of a pulse, such as a systolic phase, diastolic phase, pulse rate, and the like, and combinations thereof, can be selected.

[0091] Such a metric can spread out the phase The suitability of the extracted signal from a voxel can be assessed by the cleanness of the pulse waveform. The Fourier transform of this signal can be computed and it can be checked whether the peak is achieved at a frequency in a reasonable range of frequencies expected for a respiration or a normal pulse and it can be checked that the energy of this peak is divided by the average energy in other frequencies.

[0092] By way of example, the periodicity characteristics indicative of respiration can include a ratio of inhalation to exhalation between 1 : 1 and 1 : 6, a respiration frequency between 1 and 10 seconds. By way of further example, the periodicity characteristics indicative of the pulse of the subject at rest can include a pulse or heart rate, for example, between 45 and 150 times per minute, and a ratio of diastolic to systolic phase of about 2: 1.

[0093] The two above metrics can be smoothed and then combined into a single metric representing the suitability of the extraction of the pulse per voxel.

[0094] In one possible embodiment of the application, a single voxel is selected for pulse determination based on the above metrics, with a hysteresis to avoid frequent jumping between voxels.

[0095] In another embodiment, a plurality of voxels with high magnitude are selected and the waveforms of the plurality of voxels are averaged using SVD (PCA) after weighting by the suitability metric.

[0096] The low frequency oscillations of the phase signal indicative of respiration can be filtered from the phase curve signal to leave the high phase oscillations indicative of the heart rate.

[0097] In other embodiments, the voxel selection can use additional metrics such as signal quality (SNR) to verify that the signal extracted from the voxel has a good enough signal for respiration detection to verify that the observed signal is consistent with a true respiration signal. These two metrics can be combined to determine that the voxel is suitable for selection.

[0098] Reference is now made to FIG. 4A and FIG. 4B , which illustrate a system configured to direct radar waves from two units from different directions to the subject. The system can include dual or multiple transceivers, where radiation is directed to the back and chest of the subject. It is noted that respiration creates more exaggerated movement in the chest than in the back. Thus, in appropriate cases, the transceiver directed to the chest can be used to detect the oscillations indicative of respiration, and the transceiver directed to the back can be used to separate the heart rate signal. Furthermore, it has been found that methods using multiple transceivers can be performed to separate the heart rate signal from the respiration signal described herein.

[0099] In particular, FIG. 4AA system is shown in which the back sensor 412 is directed at the back 402 of a seated subject and the chest sensor 414 is directed at the chest 404 of the subject. Depending on the implementation, the sensors 412, 414 can be located about 90 cm above the ground, with the back sensor 412 being about 50 cm or so from the back 402 of the subject and the chest sensor 414 being about 80 cm or so from the chest 402 of the subject. Other heights can be used to direct the radar sensors at, for example, the lower or upper back, lower and upper chest, as desired.

[0100] As shown in yet another system, a single sensor 419 can be directed at a subject in a standing position, for example, at a distance of about 40 cm. FIG. 4C

[0101] In particular, it has been found that a good heart rate signal has been observed when at least one sensor is directed at the neck and upper back 405 of a subject. This can be due to the strong pulse through the carotid artery 406. Thus, as shown, it is noted that a sensor arrangement 417 located in a car headrest can be well positioned to monitor the vital signs of an occupant in a car seat. Such a sensor can also monitor the health and alertness of a driver. FIG. 4D

[0102] A similar system can arrange sensors 416, 418 at a height of, for example, 150 cm, so that when a subject is in a standing position as shown, the sensors are directed at the chest and back. FIG. 4B

[0103] FIG. 5A A method for processing radar signals acquired in a dual radar system, such as the system shown in FIG. 4A and 4B is shown. In one implementation, in an optional first step, the two acquired phase signals can be combined 510, for example, by adding the two phase signals. When the phase signals 512, 514 from the two transceivers are added, the radial variation of the phase tends to be cancelled out. Thus, the oscillations in the remaining signal 516 indicate the pulse due to the radial movement of the body generated by the respiration. Alternatively, a single sensor can be used to produce a single phase signal, which can be processed to isolate the pulse.

[0104] ​​​A delta process can be applied to the combined signal 520, or alternatively to the individual phase signals, where the phase difference between each pair of consecutive phase values in the combined signal is calculated. This will result in a derivative displacement curve 522, which can then be smoothed. For example, smoothing can be achieved by generating a moving time window, and the average of the displacement curve in the time window can be subtracted from the phase difference at the center of the window 530. Thus, the heart rate signal 532 can be isolated.

[0105] It is noted that, if necessary, e.g. when the subject's movement causes the direct current fluctuation to create spikes 534 in the signal, a spike removal mechanism can be applied to the resulting heart rate signal 532 to filter out the spikes. One example of a spike removal mechanism can be to filter out any data that exceeds a threshold value of the median in the environment, or data that is a certain number of standard deviations from the mean, as shown in FIG. 5B FIG. 5C FIG. 5B where clear spikes 536 correspond to jumps 534 at frame 1000 in

[0106] The jump removal algorithm can be based on a relatively small number of sliding windows, e.g. a sliding window of around 9 samples, such that for each sample i in the signal (the window is centered on the point)

[0107] d i = X i - X i-1

[0108] where X is the input signal, the following measures are calculated:

[0109] σ W - the standard deviation of the values in the window W.

[0110] M w - the median of the values in the window W.

[0111] If the sample d i satisfies the following condition, it will be replaced by the previous sample to remove the jump:

[0112]

[0113] where SF is a relative threshold, and A is an absolute threshold for the displacement difference value.

[0114] In FIG. 5D ​​Another method for processing displacement signals is shown. Displacement signal 542 is acquired at 540, and a line segment is approximated to this displacement signal at 550. The resulting signal 552 can be subtracted from displacement signal 542 to remove drift that may be caused by respiratory movement. This signal is then subjected to a high-pass filter at 560 to remove the respiratory component, thereby obtaining a separated signal 562.

[0115] Now for reference FIG. 6A ,Should FIG. 6A The illustration shows possible methods for generating an association index that indicates the probability that a heart rate signal corresponds to a specific heart rate. The heart rate signal 612 is sampled (610), for example, a 12-second sample 614 can be selected for analysis. The sample 614 is segmented into several segmented windows 622A, 622B, 622C, etc. (collectively referred to as 622), the duration of which corresponds to the specific heart rate. For example, for a heart rate of 40 beats per minute, the expected periodicity is 1.5 seconds; therefore, the 12-second segment can be divided into eight segmented windows, each lasting 1.5 seconds. Adjacent segmented windows can be associated (630), for example, each of the eight 1.5-second segmented windows 622 can be compared with its adjacent segmented windows, such as comparing 622A and 622B, 622B and 622C, etc., to assess the degree of periodicity (630). The overall correlation can be recorded as correlation index 640, which, for example, corresponds to a heart rate of 40 bpm with a specific window size of 1.5 seconds.

[0116] like FIG. 6B As shown, association indexes for various heart rates can be generated by segmenting the samples into window segments 650 of different lengths. Typically, an association index is generated for a heart rate of 45 beats per minute by dividing the samples into window segments 652 of size 1.33 seconds, and for other heart rates by segmenting the samples into progressively shorter windows, such as: up to a window segment of length 0.6 seconds, corresponding to a heart rate of 100 beats per minute, or a window segment of length 0.43 seconds, corresponding to a heart rate of 140 beats per minute 654. It is understood that other ranges of heart rates can be associated as needed.

[0117] Typically, for each integer heart rate between the maximum value (max_BPM) and the minimum value (min_BPM), the correlation window w can be limited. BPM Therefore, T DET The second-segmented window can be divided into N BPM Segmented window w BPM，i Therefore, for each hypothesized candidate heart rate, the number of associated windows is...

[0118] The correlation between each pair of consecutive windows w BPM，i and w BPM，i-1 may be calculated to obtain a confidence measure, such as:

[0119]

[0120] After the calculations have been performed for all candidate heart rate values, the plurality of correlation indices can be plotted on a graph 662 to produce a correlation curve 660 characteristic of the signal 664. The peak of this curve can be used to indicate the most likely heart rate of the individual being monitored. Thus, the most likely heart rate 670 corresponding to the signal can be selected. Thus, the heart rate can be selected to be the value:

[0121]

[0122] where the confidence given by the autocorrelation:

[0123] Conf = max(C BPM )

[0124] It is also noted that the resulting characteristic correlation curve itself can be used as a novel diagnostic measure.

[0125] FIG. 6C Another method for determining the heart rate from a displacement signal generated by subtracting a moving average is shown, as described above. FIG. 6C The method shown makes use of two or more reference curves 710. In generating each reference curve, the subject is exposed to radiation from a dual transceiver system as described above. At the same time as being exposed to the radiation, the heart rate of the subject is determined by any known method. For example, the heart rate can be determined by taking an electrocardiogram of the individual, or by taking an acoustic map by placing a microphone on the chest over the heart of the subject. The reference signal is processed to produce a reference curve as explained above with reference to FIG. 5A The reference signal is processed to produce a reference curve as explained above with reference to

[0126] Still referring to FIG. 6C The displacement signal taken on the subject can be filtered 720 using a high pass filter. Thus, a heart rate signal 730 of the subject can be taken. The heart rate signal can be analysed by generating a cross-correlation 740 of the filtered signal with each reference curve. The reference curve having the greatest correlation with the displacement signal of the subject is selected 742, and the heart rate associated with the reference curve having the greatest correlation is determined to be the heart rate of the subject 750. Validation rules can be applied so that only possible heart rates are selected.

[0127] The RPM can be determined by finding the maximum of the unwrapped phase Fast Fourier Transform (FFM).

[0128]

[0129] A Hamming window can be used to reduce sidelobes.

[0130] It is also noted that a high-pass filter can be used to counteract patient movement. Thus, to support various breathing rates, a set of high-pass filters based on moving averages (MA subtraction) can be introduced, e.g., a 10-second average for RPM of 6, a 5-second average for RPM of 12, a 3-second average for RPM of 20, a 1-second average for RPM of 60, etc.

[0131] RPM estimation can be performed using a continuous wavelet transform. In the example of FIG. 7 a Morlet wavelet is used, where the time-BW product is 60, and the symmetry parameter γ = 3.

[0132] The continuous wavelet transform yields a quantity map S, and each scale corresponds to a pseudo-frequency F a which can be given by:

[0133]

[0134] where a is the scale, and F c is the wavelet center frequency.

[0135] The maximum pseudo-frequency in the center of the measurement interval T corresponds to the estimated RPM

[0136]

[0137] and the confidence can be defined by:

[0138]

[0139] It is noted that since the scale is chosen according to the length of the signal, the wavelet cannot be scaled to be longer than the signal. Thus, lower frequencies can require longer signals. Thus, to measure slow RPMs, longer recordings are required.

[0140] Best voxel selection can be performed by picking a combination of voxel candidates coupled with a high-pass filter length that yields the best confidence.

[0141] Since the filter length affects the RPM accuracy and also the quality of the respiratory waveform, the system can be configured to penalize [P L1 P L2 ..P LMadded to the confidence matrix row, thus giving preference to certain filter lengths over others. This can help improve accuracy in certain ranges.

[0142] Referring back to FIG. 3B and FIG. 3E respiratory signals, it is noted that identifying respiratory signals by detecting peaks in the signal is surprisingly difficult. Peak detection is surprisingly unsuccessful due to factors such as false negatives associated with a lack of periodicity in the true respiratory signal, and false positives associated with phase jumps, scans, random-walk data caused by phase wrapping of high energy in the low frequency signal. Thus, a more robust approach to respiratory signal determination is desired.

[0143] Thus, the extracted periodic waveform signal can be processed by a line segment mechanism to determine whether the periodic waveform signal is consistent with characteristics of a respiratory signal.

[0144] Referring now to the flowchart of FIG. 8 , the line segment mechanism is operable to perform a method by estimating a line segment model (LSM) for a candidate waveform 802, extracting characteristic features from the line segment model 804, executing a respiratory recognition function 806, and determining a probability that the candidate waveform is a respiratory signal 808.

[0145] Referring now to FIG. 9 , the line segment model of a candidate signal is characterized by four line segments per period. Various features can be determined for each trapezoidal period of the LSM, such as depth, amplitude, and period. Thus, characteristic features can be determined for a candidate signal, such as average depth, amplitude error, average period, period variation, etc. These features can be converted to a metric suitable for determining a probability that the signal represents a true respiratory signal.

[0146] It is noted that advantages of the LSM include its modularity and physical meaning associated with the parameters, resulting in ease of extraction of certain respiratory parameters, such as RPM. In addition, it is relatively easy to remove trends explicitly, the time axis is independent of RPM, because the window that is updated each time is the respiratory period itself, and the number of variables per period is relatively small, which simplifies implementation and reduces memory requirements for storing the signal.

[0147] Thus, as schematically represented in FIG. 10 , the respiratory parameter determination module 1000 can include a pre-filter 1002, a point generation module 1004, a feature extraction module 1006, a metric generation module 1008, and a respiratory signal recognition module 1010.

[0148] The pre-filter 1002 can be configured to provide a candidate signal 1003 to a point generation module 1004. The point generation module 1004 receives the candidate signal 1003 and generates a line segment model (LSM) 1005, typically by identifying a first point and a last point on each significant slope in the candidate signal 1004.

[0149] The feature extraction module 1006 is configured and operable to extract characteristic features 1007 from the LSM 1005, such as the rate per minute, rate change, depth change, slope change, rest state amplitude change, rest state length change, and so on, as described herein.

[0150] The metric generation module 1008 is configured and operable to process the features and compute a characteristic metric 1009 for each feature, which indicates a probability that the feature is consistent with a respiratory signal.

[0151] The respiratory signal identification module 1010 receives the characteristic metrics 1009 for the features and applies a respiratory identification function to integrate the metrics and determine a probability 1011 that the candidate signal represents a respiratory signal.

[0152] By way of example, FIG. 11A to FIG. 11C The block diagram of FIG. 1 indicates how the components of the system can be combined to process a candidate signal.

[0153] The pre-filter receives a periodic phase change signal, for example, in order to determine whether the voxel generating the phase change signal is suitable for monitoring a respiratory signal. The pre-filter can be used to reduce noise levels or concentrate the extracted signal to the frequency domain of the expected respiratory signal, and extract a candidate periodic signal, which is transmitted to the point generation module.

[0154] The point generation module can apply a line segment sequential approximation by identifying a slope start point and a slope end point for each significant slope in the candidate signal.

[0155] Reference is now made to FIG. 12A and FIG. 12B The graphs of FIGS. 12 and 13 represent two examples showing line segment models superimposed on the periodic signals from which the line segment models were generated. The input periodic signals were processed to generate line segment models having four line segments per cycle. The four line segments include an ascending segment 1202 having a positive gradient, a hold segment 1204 following the ascending segment and having a zero gradient, a descending segment 1206 having a negative gradient, and a rest segment 1208 following the descending segment and having a zero gradient.

[0156] As FIG. 13AThe flowchart of Figure 1 1 illustrates a method performed by the point generation module to determine the key points that define the start and end of each line segment. The method comprises: identifying significant rising and falling sections within the periodic signal 1310; detecting a slope start point for each significant slope section 1320; and detecting a slope end point for each significant slope section 1330.

[0157] Various definitions can be used to set the value of the slope start and slope end points, FIG. 13B The flowchart of Figure 1 1 illustrates a method. In this method, a detection threshold is defined THR 131 1, signal values are sampled during the pre-extreme phase 1312, extreme signal points are detected 1313 and recorded, and the post-extreme phase begins during which signal values are again sampled 1314. During the post-extreme phase, the difference DIFF between each signal value and the previous extreme signal value is calculated 1315. The difference DIFF is compared to the threshold THR. If the difference DIFF is greater than the threshold THR, a new slope start point is set for the line segment after the extreme signal point by selecting the slope_start_value point as the first signal sample after the previous extreme 1317, and a new slope end point is set for the line segment before the extreme signal point by selecting the previous slope_end_value point as the signal sampled during the previous pre-extreme phase 1319, which has the same value as the new slope_start_value.

[0158] Reference can also be made to FIG. 14A and FIG. 14B to understand the steps of the method. In the method of this example, the detection threshold can be selected such that only significant slopes are detected and smaller fluctuations are discarded. In particular reference is made to FIG. 14A The plot of Figure 14 shows an example of sections of a periodic signal, indicating the detection thresholds rise_fall_detect_thresh 1402, 1404. The detection thresholds are measured from the extreme points 1406, 1408 of the signal towards the centre, and thus define an upper and lower margin within which small peaks are not considered to be significant. By way of example, FIG. 14A The plot of Figure 14 indicates the limits of the margins 1402, 1404 with circular markers.

[0159] Reference is made back to FIG. 13BThe signal is sampled 1312 to detect a local extremum 1314 in the signal value, which can be a local minimum signal value 1408 or a local maximum signal value 1406. Upon detection of an extremum signal value, each post-extremum sample is analyzed by calculating the difference 1315 between the current sampled signal value and the previous extremum signal. If the absolute value of the difference (DIFF) is less than a threshold (THR), the next post-extremum signal value is sampled.

[0160] If the absolute value of the difference (DIFF) is greater than the threshold (THR), a slope start point 1317 is set, which can optionally be chosen as the first signal sample after the previous extremum, although other samples can be chosen as desired. For example, FIG. 14B A plot of the signal is indicated by diamond markers 1410, 1412 at the slope start points.

[0161] Upon setting a new slope start point, the previous slope end point can be determined retroactively as the point 1318 before the previous extremum that has the same signal value as the new slope start point. The signal is then again sampled to detect the next local extremum. For example, FIG. 14B A plot of the signal is indicated by octagon markers at the slope end points 1414, 1416.

[0162] It is noted that the signal can experience various amplitude and frequency swings that are not dependent on the effects of respiration. These swings include large amplitude excursions that can be due to body movement, phase jumps due to phase unwrapping errors, other noise, etc.

[0163] The algorithm can also normalize the swings of the signal, where appropriate, for example, an upswing can be identified when the following expression is true

[0164] (Signal-min(Signal)) > (risefall_detect_thresh * Swing),

[0165] Similarly, a downswing can be identified when the following expression is true

[0166] (Signal-max(Signal)) < (risefall_detect_thresh * Swing).

[0167] where "min" and "max" in the above expressions are measured over a period of time from the previous rise / fall detection.

[0168] Swing can be evaluated to identify the overall shape of the signal and adapt to it. A simple way to estimate the swing can be by taking the RMS of the signal, e.g. using an IIR to measure the RMS of the signal over a time window, and multiplying by a factor; or estimating the swing as the difference between the last high extreme and the low extreme, e.g. Swing = RunMax(s n )- RunMin(s n ), where RunMax / RunMin are the maximum / minimum values over a time window or an approximation of these values. Possible implementations of RunMax / RunMin can be defined by the following recursive equations. The relation is defined as:

[0169] y n = (1 - a) max(y n-1 , x n ) + a x n

[0170] for some constant a, and min and max are the minimum and maximum values, respectively.

[0171] One way to achieve this and be flexible to unidirectional jumps is to evaluate a swing index S, where there is a positive peak with two signal values on one or both sides of S that are smaller than S as needed. For example, one can take the largest time window and the smallest time window and the swing as the minimum between the "up" and "down" min(p + [k] - p - [k-1], p + [k-1] - p - [k]).

[0172] The swing index can be computed by measuring the distance between the signal and the upper and lower limits and taking the side of the smallest distance. Thus, the swing index can be given by:

[0173] Swing = min(RunMax(U n - s n ), RunMax(s n - L n ))

[0174] where U n = RunMax(s n ) and L n = RunMin(s n ), and RunMax / Min are defined as above, so that if a sudden rise in the bias of the signal occurs, U n will track this rise, but L nwill lag. The graph of Figure 11 illustrates how the values of the swing index and U n and L n vary over time for a fluctuating signal.

[0175] Now referring back to the block diagram of FIG. 11A to FIG. 11C , the point generation module transmits the line segment model to the feature extraction module, which can execute a line segment breathing feature sequence function on the line segment, thereby generating breathing signal features for the overall signal. Among others, some examples of such features are introduced below for illustrative purposes.

[0176] avg_rpm - this value can approximate the RPM of the true breathing signal, which is expected to have a value of about 3-60, with a typical value of 15. This value can be calculated using the formula

[0177] avg rpm = 60 / AvgPeriod

[0178] rate_variation_short_term - this value can indicate the consistency of the breathing rate. This value can be calculated using the formula rate_variation_short_term = PeriodVariation / AvgPeriod

[0179] breath_depth_variation_short_term - this value can indicate the periodicity of the breathing amplitude. This value can be calculated using the formula breath_depth_variation_short_term = DepthVariation / AvgDepth

[0180] breath_slope_variation - this value can indicate the periodicity of the breathing signal slope. This value can be calculated using the formula breath_slope_variation = SlopeVariation / AvgSlope Mean[rise / fall]

[0181] rest_state_amp_variation - this value can indicate the variation of the rest state amplitude, which tends to have less variation and is generally independent of the depth. This value can be calculated using the formula rest_state_amp_variation = AmpVariation[rest] / AvgDepth

[0182] rest_state_length_variation - This value can indicate the variation in the rest state duration. This value can be calculated using the formula rest_state_length_variation = LengthVariation[rest] / AvgDepth.

[0183] rest_length_minus_hold_length_norm - This value can indicate how much longer the rest state is than the hold state, which should have a positive value for a typical breath. This value can be calculated using the formula rest_length_minus_hold_length_norm = (AvgLength[Rest] - AvgLength[Hold]) / avgPeriod.

[0184] hold_state_minus_rest_state_amp_variation - This value can indicate how much more variation there is in the hold state than in the rest state. This is typically a positive value for a typical breath and can be used to compensate for periodicity. This value can be calculated using the formula hold_state_minus_rest_state_amp_variation = (AmpVariation[hold] - AmpVariation[rest]) / (AmpVariation[hold] + AmpVariation[rest]).

[0185] exhale_minus_inhale_norm - This value can indicate how much longer the exhalation is than the inhalation, which is typically about twice as long for a typical breath. This value can be calculated using the formula exhale_minus_inhale_norm = (AvgLength[Fall] - AvgLength[Rise]) / avgPeriod.

[0186] For the above purposes, the following terms are used as given by:

[0187] where represents the running exponentially averaged value.

[0188]

[0189]

[0190] Depth = s 2i - s 2i-1

[0191]

[0192]

[0193]

[0194]

[0195]

[0196]

[0197]

[0198] As an example, reference is made to FIG. 12C and FIG. 12D indicate characteristic features extracted from the signals shown in FIG. 12C indicate characteristic features extracted from the signals shown in FIG. 12A FIG. 12D indicate characteristic features extracted from the signals shown in FIG. 12B

[0199] It is to be noted that the selected characteristic features can be independent of the time scale and of each other.

[0200] Referring again to the block diagram of FIG. 11A to FIG. 11C the feature extraction module transmits the extracted characteristic features to a metric generation module which can perform a line segment feature to metric function to aggregate the characteristic features into a metric.

[0201] It is to be understood that the characteristic features can be divided into two broad categories, namely, symbol dependent features and symbol independent features.

[0202] ​​The sign-dependent features include rest_length_minus_hold_length_norm, hold_state_minus_rest_state_amp_variation, and exhale_minus_inhale_norm. These features can be expected to be positive for any normal breathing signal, and the inversion of the breathing signal will also invert these features. It can be expected that the higher the value of these features, the more likely the signal is a breathing signal. These features can be used to determine the sign (inversion / non-inversion) of the received signal; after detecting the sign (e.g., by integrating these features over one or more breaths), these features can be multiplied by the detected sign to align them.

[0203] The sign-independent features include avg_rpm, rate_variation_short_term, breath_depth_variation_short_term, breath_slope_variation, rest_state_amp_variation, and rest_state_length_variation. These features are generally independent of the inverted signal.

[0204] Thus, in converting the features to a characteristic metric, a log likelihood ratio (LLR) can be generated to show the probability that the signal is breathing. For example, the probability that the signal is breathing can be given by where LLR = log{Pr(breathing) / Pr(~breathing)}.

[0205] By way of example, one possible approach is set forth below, in which the overall probability is determined by passing each feature through a piecewise linear function to convert to an LLR, and then summing the LLRs and adding a global offset where appropriate.

[0206] Referring to the flowchart of FIG. 16A and the schematic diagram of FIG. 16B the method for generating a metric can include calculating an LLR value for each feature 1602, correcting the sign of the LLR value for the feature 1604, and summing the corrected LLR values 1606.

[0207] The step of computing the LLR for each feature can use a piecewise linear function, such that for each feature, a piecewise linear description of the LLR can be provided by a function of the feature, given in the form of a set of points (each point is a collection of a feature value and an LLR value), typically the LLR is constant below the first point and above the last point.

[0208] LLR perFeature [n] = interp1(points[n], feature[n])

[0209] Because, for sign dependent features, the inversion of the signal implies the inversion of the feature, the LLR can be computed for both hypotheses: first, sign = +1, and second, sign = -1. Thus, the function can be given by:

[0210] LLR perFeature [n; sign] = interp1(points[n], sign * feature[n])

[0211] Both the feature generation (LineSegmentBreathingFeaturesSequential) and the breathing LLR computation (LineSegmentFeaturesToMetrics) need to know the sign, and because this is not available, it can be made resilient to the sign by taking the "min / max / abs" of the relevant parameters in the best case (i.e. choosing the option that gives the better LLR). This "un-encoded" decision increases the LLR and reduces the difference to the null hypothesis - for example, in the presence of noise, the LLR is found to increase by almost +1.

[0212] Thus, the step of correcting the sign feature can depend on whether the sign decision feedback has already determined the sign.

[0213] In the case where the sign is known, then the input to LLR_per_feature[n, sign] is selected according to the known sign.

[0214] In the case where no such sign is determined, the sign LLR-s for each sign hypothesis can be summed, and the maximum of the two cases is selected.

[0215] The step of summing the corrected LLR values can generate the breathing LLR by summing the LLR_per_feature for all features after the sign feature with the correct polarity is selected, and additionally adding a possible offset:

[0216]

[0217] The sign LLR = log(Pr(Sign = 1) / Pr(Sign = -1)) can be given by the difference between the LLRs taken separately for each hypothesis:

[0218]

[0219] It is noted that smoothing of the sign LLR can be done outside the "LineSegmentGenerateMetrics" module, however incorporating the sign decision is useful to improve the respiration detection LLR by feeding it with the correct sign. The equation is as follows:

[0220] SignLLRAccum[n] = limiter(SignLLRAccum[n-1] + SignLLR[n], [-K, K])

[0221] It is further noted that although a specific model is presented above, one skilled in the art can prefer alternative models that can generate LSMs. Indeed, the metrics of other features can be determined as needed. Furthermore, in addition to or in parallel with respiration signal analysis, LSMs can be generated for other signals such as heart rate signals as needed.

[0222] In appropriate cases, the line segment model can be used to separate a signal that includes both heartbeats and respiration into two components. For example, if a signal s[n] includes two components s[n] = b[n] + h[n], where b is respiration, and h is heartbeat, then b[n] can be estimated by the LSM described herein, and h[n] can be determined by the difference of s[n] and b[n].

[0223] TECHNICAL DESCRIPTION

[0224] Technical terms and scientific terms used herein should have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure belongs. It is expected, however, that many related systems and methods will be developed during the life of the patent for this application. Accordingly, the scope of the terms, such as computing unit, network, display, memory, server, etc., is intended to include all such new technologies a priori.

[0225] The term "about" as used herein means at least ± 10%.

[0226] The terms "comprises", "comprising", "includes", "including", "has", "having", "contains", "containing", and variations thereof, are meant to encompass the terms "including", "comprising", and "consisting of" and are intended to be used in their broadest sense.

[0227] The phrase "consisting essentially of means that the composition or method can include additional ingredients and / or steps, but only if the additional ingredients and / or steps do not materially alter the basic and novel characteristics of the claimed composition or method.

[0228] As used herein, the singular forms "a", "an" and "the" can include plural referents unless the context clearly dictates otherwise. For example, the terms "a compound" or "at least one compound" can include more than one compound, including mixtures thereof.

[0229] The word "exemplary" is used herein to mean "serving as an example, instance, or illustration." Any implementation described as "exemplary" is not necessarily to be construed as preferred or advantageous over other implementations.

[0230] The word "optionally" is used herein to mean "in some embodiments, but not in others." Any particular implementation of the present disclosure can include a plurality of "optional" features, unless such features conflict.

[0231] Whenever a numerical range is indicated herein, it is meant to include any cited number (fractional or integral) within the specified range. The phrases "ranging / range between" a first indicate number and a second indicate number and "ranging / range from" a first indicate number "to" a second indicate number are used herein interchangeably and are meant to include the first and second indicate number and all the fractional and integral values therebetween. Accordingly, it is understood that the description of a range format is merely a shorthand for specifying a range starting with a first numerical value and ending with a second numerical value. Therefore, it is understood that all ranges format descriptions are "shorthand" notation that describes and claims almost every possible combination of endpoints in between the first numerical value and the second numerical value. For example, a range of "1 to 6" is understood to include the following: the range of "1 to 3," the range of "1 to 4," the range of "1 to 5," the range of "2 to 4," the range of "2 to 6," the range of "3 to 6," and the like; and, the individual numbers 1, 2, 3, 4, 5, and 6, and fractional values therein such as, 1.1, 2.3, 3.75, 4.522, 5.675, 6.100 and the like. This applies regardless of the breadth of the range.

[0232] It is to be understood that certain features that are, for clarity, described above in the context of separate embodiments can also be provided in combination in a single embodiment. Conversely, various features that are, for brevity, described above in the context of a single embodiment can also be provided separately or in any appropriate subcombination. Certain features described in the context of various embodiments are not to be interpreted as essential features of those embodiments, unless explicitly stated otherwise.

[0233] Although the present disclosure has been described in connection with specific embodiments thereof, it will be evident that many alternatives, modifications, and variations will be apparent to those of ordinary skill in the art. Accordingly, it is intended to embrace all such alternatives, modifications, and variations that fall within the spirit and broad scope of the appended claims.

[0234] All publications, patents, and patent applications mentioned in this specification are herein incorporated by reference in their entirety for the purpose of describing and disclosing, for example, the methodologies described in such publications, which might be used in connection with the methods described in this specification. The reference herein to any reference document is not intended to imply that the reference document is prior art to this disclosure. To the extent that section headings are used, they should not be construed as necessarily limiting.

[0235] The scope of the disclosed subject matter is limited only by the claims that are appended hereto, and any equivalents thereof, and includes combinations and subcombinations of the various features described above and variations and modifications thereof that would occur to persons of ordinary skill in the art upon reading the foregoing description.

Claims

1. A method for monitoring the vital signs of objects within a target area, the method comprising: Provide a radar unit, the radar unit comprising: At least one transmitter antenna, connected to an oscillator, and configured to transmit electromagnetic waves into the target region, and At least one receiver antenna, the at least one receiver antenna being configured to receive electromagnetic waves reflected by an object within the target area, and the at least one receiver antenna being operable to generate raw data; A processor unit is provided, the processor unit being configured to receive raw data from the radar unit, and the processor unit being operable to identify oscillation signals in the raw data; The radar transmits and receives scanning radiation over the target area; The radar generates a series of frames, each frame including a complex-valued array representing the radiation reflected from each voxel of the target region during a given time segment; The processor organizes a series of complex values ​​for each voxel, the series of complex values ​​representing the reflected radiation of the voxel associated in multiple frames; Determine the center point in the complex plane for each voxel; Determine the phase value for each voxel in each frame; For each voxel, a smooth waveform representing the phase change over time is generated; Select a subset of voxels indicating the oscillation signal; and Extract vital sign parameters from the oscillation signal, and The step of selecting the voxel subset indicating the oscillation signal includes: Generate a line segment model for the oscillation signal; Extract characteristic features from the line segment model; and Execute the verification function to determine whether the line segment model indicates a real breathing signal.

2. A method for monitoring the vital signs of objects within a target area, the method comprising: Provide a radar unit, the radar unit comprising: At least one transmitter antenna, connected to an oscillator, and configured to transmit electromagnetic waves into the target region, and At least one receiver antenna, the at least one receiver antenna being configured to receive electromagnetic waves reflected by an object within the target area, and the at least one receiver antenna being operable to generate raw data; A processor unit is provided, the processor unit being configured to receive raw data from the radar unit, and the processor unit being operable to identify oscillation signals in the raw data; The radar transmits and receives scanning radiation over the target area; The radar generates a series of frames, each frame including a complex-valued array representing the radiation reflected from each voxel of the target region during a given time segment; The processor organizes a series of complex values ​​for each voxel, the series of complex values ​​representing the reflected radiation of the voxel associated in multiple frames; Determine the center point in the complex plane for each voxel; Determine the phase value for each voxel in each frame; For each voxel, a smooth waveform representing the phase change over time is generated; Select a subset of voxels indicating the oscillation signal; and Extract vital sign parameters from the oscillation signal, and The step of selecting the voxel subset indicating the oscillation signal includes: Generate a line segment model for the oscillation signal; Extract characteristic features from the line segment model; and Execute the verification function to determine whether the line segment model indicates a real breathing signal; The steps for generating the line segment model include: For each significantly tilted segment in the oscillation signal, select a boundary point; The boundary points of each significantly inclined segment include: The slope initiation point, which indicates a significant change from the previous extreme; and The slope termination point shares the Y-coordinate with the subsequent slope start point of the next significant inclined segment.

3. The method according to claim 2, wherein, The steps for generating a line segment model also include: Construct the trapezoidal line segment model in the following way: Construct line segments between the boundary points of each significantly inclined segment; Construct a line segment between the end point of each significantly inclined segment and the beginning point of the next significantly inclined segment.

4. The method according to claim 3, wherein, The steps for selecting boundary points include: Limit the detection threshold; The signal values ​​in the oscillation signal are sampled until an extreme signal is detected; Detect extreme signal points; After the extreme signal is detected, the signal value in the oscillation signal is sampled; Calculate the difference between each sampled signal and the extreme signal; If the difference is greater than the detection threshold, then: Set the slope starting point for the line segment following the extreme signal; and Set a slope termination point for the line segment preceding the extreme signal.

5. The method according to claim 4, wherein, The step of setting the slope starting point for the line segment following the extreme signal includes: selecting the first signal following the extreme signal.

6. The method according to claim 4, wherein, The step of setting a slope termination point for the line segment before the extreme signal includes: selecting a signal sample before the extreme, wherein the signal value of the signal sample is equal to the signal value of the slope initiation value.

7. A method for monitoring the vital signs of objects within a target area, the method comprising: Provide a radar unit, the radar unit comprising: At least one transmitter antenna, connected to an oscillator, and configured to transmit electromagnetic waves into the target region, and At least one receiver antenna, the at least one receiver antenna being configured to receive electromagnetic waves reflected by an object within the target area, and the at least one receiver antenna being operable to generate raw data; A processor unit is provided, the processor unit being configured to receive raw data from the radar unit, and the processor unit being operable to identify oscillation signals in the raw data; The radar transmits and receives scanning radiation over the target area; The radar generates a series of frames, each frame including a complex-valued array representing the radiation reflected from each voxel of the target region during a given time segment; The processor organizes a series of complex values ​​for each voxel, the series of complex values ​​representing the reflected radiation of the voxel associated in multiple frames; Determine the center point in the complex plane for each voxel; Determine the phase value for each voxel in each frame; For each voxel, a smooth waveform representing the phase change over time is generated; Select a subset of voxels indicating the oscillation signal; and Extract vital sign parameters from the oscillation signal, and The step of generating a smooth waveform representing the phase change over time for each voxel includes rounding each phase value.

8. A system for monitoring the vital signs of objects within a target area, the system comprising: Radar unit, the radar unit comprising: At least one transmitter antenna, connected to an oscillator, and configured to transmit electromagnetic waves into the target region, and At least one receiver antenna, configured to receive electromagnetic waves reflected by an object within the target area, and operable to generate raw data; and A processor unit configured to receive raw data from the radar unit, and the processor unit operable to identify oscillation signals in the raw data; The processor unit includes: A frame organizer, the frame organizer including a memory unit configured to store a series of multiframe values ​​for each voxel in the target region; A voxel segmentation module, the voxel segmentation module including a processor, the processor being configured to generate a smooth waveform representing the phase change over time for each voxel; Respiratory parameter determination module; and Heart rate parameter determination module The system further includes a heart rate monitor, which includes a footrest positioned such that the upper part of the object's foot placed on the footrest is located within the target area, and the radar unit is directed to the target area.

Images