Vital sign detection device and method

Abstract

To provide a vital sign detection device and method.SOLUTION: A method comprises: acquiring a radar multi-antenna zero speed distance FFT signal; selecting a first radar signal from the radar multi-antenna zero speed distance FFT signal and generating an amplitude feature and a phase feature of the first radar signal; using a detection model based on a deep neural network to generate candidate physiological signals for multiple segments on the basis of the amplitude feature and the phase feature; selecting a target physiological signal from the candidate physiological signals of the multiple segments, the candidate physiological signal and the target physiological signal being respiratory signals or heartbeat signals; and detecting a vital sign parameter on the basis of the target physiological signal.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to the technical field of radar detection. [Background technology]

[0002] Physiological indicators such as respiration and heart rate are basic vital sign information of living organisms and can directly reflect the health status of living organisms. Abnormalities in vital sign parameters can often lead to medical emergencies, so real-time detection of vital sign information has great practical value.

[0003] Currently, vital signs detection methods can be divided into wearable sensor-based detection methods and wireless signal-based detection methods. Wearable sensor-based detection methods mainly use contact sensors / patch electrodes for measurements. Wireless signal-based detection methods mainly use radar, such as Doppler radar, ultra-wide band (UWB) radar, and frequency modulated continuous wave (FMCW) radar.

[0004] The detection methods based on wearable sensors require the sensor to directly or indirectly contact the human body, which not only limits the scope of application but also has the problem of user dependency: when the user does not want to wear or forgets to wear the sensor device, these systems cannot perform detection.

[0005] In the wireless signal-based detection method, the radar detects the change in vibration amplitude of the cardiac and / or thoracic scattering centers, extracts physiological signals based on the detection results, and determines vital sign parameters by the physiological signals. Summary of the Invention [Problem to be solved by the invention]

[0006] However, the inventor has found that in the detection method based on wireless signals, due to the influence of various disturbing factors, the radar detection result may have relatively large noise, and the physiological signal waveform of one segment extracted based on the detection result cannot realistically reflect the physiological parameters, so the accuracy of the physiological parameters determined from such physiological signals cannot be guaranteed.

[0007] In view of at least one of the above technical problems, the present invention provides a vital signs detection apparatus and method. [Means for solving the problem]

[0008] According to one aspect of an embodiment of the present invention, there is provided a vital signs detection device, comprising: A data acquisition unit that acquires the radar multi-antenna zero speed distance FFT signal output by the radar; a data pre-processing unit for selecting a first radar signal from the radar multi-antenna zero speed range FFT signal and generating an amplitude characteristic and a phase characteristic of the first radar signal; a generating unit for generating candidate physiological signals of a plurality of segments based on the amplitude features and the phase features using a detection model based on a deep neural network; a selection unit for selecting a target physiological signal from the candidate physiological signals of the plurality of segments, the candidate physiological signals and the target physiological signal being a respiratory signal or a cardiac signal; and A detection unit is included for detecting a vital sign parameter based on the target physiological signal.

[0009] According to another aspect of an embodiment of the present invention, there is provided a method for vital signs detection, the method comprising: Acquire radar multi-antenna zero speed distance FFT signal; Selecting a first radar signal from the radar multi-antenna zero speed range FFT signal, and generating an amplitude characteristic and a phase characteristic of the first radar signal; generating candidate physiological signals of a plurality of segments based on the amplitude features and the phase features using a deep neural network based detection model; selecting a target physiological signal from the plurality of segments of candidate physiological signals, the candidate physiological signal and the target physiological signal being a respiratory signal or a cardiac signal; and Detecting a vital sign parameter based on the target physiological signal. Effect of the Invention

[0010] The advantageous effects of the embodiment of the present invention are at least as follows: based on the amplitude and phase features of a first radar signal, a detection model based on a deep neural network is used to generate candidate physiological signals of multiple segments, a target physiological signal is selected from the candidate physiological signals of the multiple segments, and a vital sign parameter is detected based on the target physiological signal, thereby improving the accuracy (precision) of the detection result of the vital sign parameter. [Brief description of the drawings]

[0011] [Figure 1] 1 is a schematic diagram of a vital signs detection method according to an embodiment of the present invention; [Diagram 2] 4A to 4C are schematic diagrams of complex signals at distance frequency points and velocity frequency points according to an embodiment of the present invention. [Diagram 3] FIG. 2 is a schematic diagram of step 102 in an embodiment of the present invention. [Figure 4] FIG. 2 is a schematic diagram of distance frequency point vs. time vs. amplitude / phase of a radar multi-antenna zero-speed distance FFT signal according to an embodiment of the present invention. [Diagram 5] FIG. 2 illustrates determining a region of interest in an embodiment of the present invention. [Figure 6] FIG. 2 is a schematic diagram of a detection model according to an embodiment of the present invention. [Figure 7] FIG. 6 is a schematic diagram of a single-time point feature extraction module 601 according to an embodiment of the present invention. [Figure 8]FIG. 6 is a schematic diagram of a local time feature extraction module 602 according to an embodiment of the present invention. [Figure 9] FIG. 6 is a schematic diagram of a global feature extraction module 603 according to an embodiment of the present invention. [Figure 10] FIG. 4 is another schematic diagram of a detection model according to an embodiment of the present invention. [Figure 11] 2A-2C are time and frequency domain diagrams of real and predicted physiological signals according to an embodiment of the present invention; [Figure 12] 4 is a schematic diagram showing the amount of change in the waveform of a real physiological signal according to an embodiment of the present invention. FIG. [Figure 13] 2 is a time domain diagram of a real physiological signal and a predicted physiological signal according to an embodiment of the present invention. [Figure 14] FIG. 13 illustrates a detection model outputting candidate physiological signals of multiple segments. [Figure 15] FIG. 2 is a schematic diagram of step 104 in an embodiment of the present invention. [Figure 16] FIG. 13 is a schematic diagram of step 1041 in an embodiment of the present invention. [Figure 17] FIG. 11 is a schematic diagram of step 1042 in an embodiment of the present invention. [Figure 18] 1 is a schematic diagram of a vital sign detection device according to an embodiment of the present invention. [Figure 19] 1 is a schematic diagram of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0012] Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. Note that the embodiments described below are merely examples and do not limit the present invention.

[0013] In the embodiment of the present invention, the radar may be, but is not limited to, a millimeter wave (mmWave) radar. The radar emits electromagnetic waves through a transmitting antenna, and receives corresponding reflected waves (which may be called radar echo information or radar echo signals) after the electromagnetic waves are reflected by different objects. By processing and analyzing the radar echo information, information such as physiological signals of the detection target can be extracted, and this information can be applied to many application scenarios.

[0014] In the embodiment of the present invention, the detection target may be, for example, an elderly person, a child, an elderly person and / or a caregiver, a child and / or a guardian. The present invention is not limited thereto, and the detection target may further be an animal having a vital sign. The following description will be given taking the human body as an example.

[0015] <Example of the first aspect> An embodiment of the present invention provides a vital signs detection method. Figure 1 is a schematic diagram of the vital signs detection method according to an embodiment of the present invention. As shown in Figure 1, the method includes the following steps:

[0016] 101: Radar multi-antenna zero-speed range FFT (Fast Fourier Transform) signal acquisition; 102: Select a first radar signal from the radar multi-antenna zero-speed range FFT signal, and generate an amplitude characteristic and a phase characteristic of the first radar signal; 103: generating candidate physiological signals of a plurality of segments based on the amplitude features and the phase features using a detection model based on a deep neural network; 104: Selecting a target physiological signal from the candidate physiological signals of a plurality of segments, the candidate physiological signal and the target physiological signal being a respiratory signal or a cardiac signal; and 105: Detect a vital sign parameter based on the target physiological signal.

[0017] It should be noted that, although the above-mentioned FIG. 1 is a schematic diagram of an embodiment of the present invention, the present invention is not limited thereto. For example, the execution order between the above-mentioned steps may be appropriately adjusted, and some other steps may be added or some steps may be removed. Without being limited to the description of the above-mentioned FIG. 1, a person skilled in the art may appropriately modify the above-mentioned contents.

[0018] According to the above embodiment, a detection model based on a deep neural network is used to generate candidate physiological signals of a plurality of segments based on the amplitude characteristics and phase characteristics of the first radar signal, a target physiological signal is selected from the candidate physiological signals of the plurality of segments, and a vital sign parameter is detected based on the target physiological signal, thereby enabling a physiological signal that can more accurately reflect the vital sign parameter to be selected from the candidate physiological signals of the plurality of segments as the target physiological signal, thereby improving the accuracy of the detection result of the vital sign parameter.

[0019] According to the above-mentioned embodiment, vital sign parameters are detected based on the radar multi-antenna zero-speed distance FFT signal, which on the one hand can reduce the data transmission amount and improve the data transmission efficiency, which is advantageous for realizing real-time detection after commercialization, and on the other hand can enable the vital sign detection method to be applied to more scenes.

[0020] For example, the beating process of the human heart is similar to the expansion and contraction of vibration, and the amplitude of the expansion and contraction is about 0.01 to 0.2 mm. In addition, the speed of the human heartbeat changes periodically within a stable range, so the heartbeat can be approximated by a sinusoidal vibration model. In addition, breathing is performed by the expansion and contraction of the thoracic cavity, which is also similar to a sinusoidal vibration, so it can also be approximated by a sinusoidal vibration model, and the amplitude of the expansion and contraction is about 0.1 to 0.5 mm.

[0021] Therefore, as can be seen, the change in the vibration magnitude (amplitude) of the heart and thoracic scattering center is very weak against the magnitude of body movement. Therefore, some human vital sign detection schemes based on wireless signals do not consider body shaking. For example, these schemes require that the detection of vital signs is performed when the subject is stationary in front of the signal source or the signal source is attached behind the seat or under the bed, that is, when the subject is in a stationary or quasi-stationary state. Therefore, the mounting method and monitoring range of the device are limited. Alternatively, some schemes only consider limited shaking. For example, these schemes only allow occasional shaking after the body is fixed for a certain period of time, and noise removal is performed on the collected radar signal using conventional signal processing methods (such as filtering and signal decomposition), but the noise removal effect is not ideal.

[0022] Furthermore, these methods mainly use the radar's original I / Q signal or Doppler-range spectrum to detect vital signs, and the data capacity of the original I / Q signal or Doppler-range spectrum is large, the real-time transmission speed is slow, and it is limited by the SPI port (buffer size), making it disadvantageous for real-time monitoring after commercialization.

[0023] In the embodiment of the present invention, the radar multi-antenna zero speed range FFT signal is used to detect vital sign parameters, that is, by performing a Fourier transform of the speed dimension and the distance dimension, signals at static or quasi-static distance points with respect to the radar are extracted for subsequent detection processing, so that even if there is a subject with large fluctuations with respect to the radar in the detection range, appropriate signals can be accurately extracted from the radar echo signal and detection processing can be performed, and there is no need to limit the installation form of the device or the monitoring range, and the vital sign detection method can be applied to more situations. In addition, by performing detection using the radar multi-antenna zero speed range FFT signal, the amount of data to be transmitted or processed can be reduced, which is advantageous for improving data transmission and processing speed.

[0024] In some embodiments, 101 can obtain radar multi-antenna zero-speed range FFT signals in various ways. For example, taking FMCW radar signals as an example, 1D-FFT (one-dimensional fast Fourier transform) and 2D-FFT (two-dimensional fast Fourier transform) can be performed on the received radar echo signals to obtain complex signals at each distance frequency point and each speed frequency point, i.e., Doppler-range frequency spectrum.

[0025] 2 is a schematic diagram of complex signals at each distance frequency point and each velocity frequency point according to an embodiment of the present invention. It shows that, for a radar echo signal of one frame in a preset (predetermined) period (for example, T frame), after performing distance TTF and velocity FFT on the radar echo signal of one antenna among A (A>1) antennas, D*R complex signals are obtained. In which, D is the number of velocity frequency points (also called Doppler frequency points), for example, D is 64 or 128, and R is the number of distance frequency points, for example, R is 128.

[0026] As shown in FIG. 2, from the complex signal, a signal at each range frequency point where the Doppler velocity is zero or is less than a preset (predetermined) threshold is selected as a zero-velocity range FFT signal.

[0027] The signal at each distance frequency point where the Doppler velocity is 0 (i.e., (rd in FIG. 2) 00 ,rd 01 ,…,rd 0(R-1) ) as the zero-speed distance FFT signal. Compared to selecting all complex signals, the amount of data waiting to be processed can be reduced, thereby reducing the amount of calculation and the demand for calculation resources.

[0028] In some embodiments, in 102, a first radar signal can be selected from the acquired multi-antenna zero speed range FFT signal, and an amplitude characteristic and a phase characteristic of the first radar signal can be generated, so that a radar signal that can be optimally mapped to a potential change characteristic of the physiological movement of the human body can be selected from the multi-antenna zero speed range FFT signal, thereby further improving the accuracy of the vital sign detection result.

[0029] In some embodiments, the first radar signal can be selected in various ways to generate amplitude and phase characteristics of the first radar signal. Figure 3 is a schematic diagram of 102 in an embodiment of the present invention. As shown in Figure 3, 102 can include the following steps.

[0030] 301: determining a target position based on time domain information and / or frequency domain information of a radar multi-antenna zero-speed range FFT signal; 302: A target region is defined as a target position and the positions of N1 distance frequency points adjacent to the target position, where N1≧1; 303: Selecting a first radar signal within a target area; and 304: Generate amplitude and phase signatures of the first radar signal.

[0031] 4 is a schematic diagram of the distance frequency point vs. time vs. amplitude / phase of the radar multi-antenna zero speed range FFT signal according to an embodiment of the present invention. In some embodiments, as shown in FIG. 4, in 301, based on the time domain information and / or frequency domain information of the radar multi-antenna zero speed range FFT signal on the multiple distance frequency points, a distance frequency point corresponding to the detection target can be determined as the target position of the detection target from the multiple distance frequency points.

[0032] In some embodiments, the time domain information may include information on the time-varying amplitude of the radar multi-antenna zero-speed range FFT signal at each range frequency point within a preset period. When determining the target position based on the time domain information, the amplitude fluctuation (wave) amount of each range frequency point within a preset period is obtained (calculated), and the range frequency point corresponding to the maximum amplitude fluctuation amount is set as the target position. Among them, the signal amplitude fluctuation amount may include the variance of the amplitude in a preset period, the standard deviation of the amplitude in a preset period, the average value of the amplitude change amount in a preset period (as shown in Equation 9 and Equation 10 described later), etc.

[0033] In some embodiments, the frequency domain information may include a result of performing FFT on the phase of the radar multi-antenna zero-speed range FFT signal at each range frequency point. When determining the target position based on the frequency domain information, the target position may be determined to be the range frequency point corresponding to the maximum value of the sum of the energy of the radar multi-antenna zero-speed range FFT signal at each frequency point within a physiological signal band (e.g., a respiration band (e.g., 0.083 to 0.5 Hz) and / or a heart rate band (e.g., 0.83 to 2 Hz)).

[0034] However, the present invention is not limited to this, and when determining a target position based on frequency domain information, the target position may be the distance frequency point corresponding to the maximum value of the ratio between the sum of energy at each frequency point in the physiological signal band of the radar multi-antenna zero speed distance FFT signal and the sum of energy at each frequency point in the non-physiological signal band.

[0035] For example, based on frequency domain information at each distance frequency point of the radar multi-antenna zero speed range FFT signal, the sum of the energy at each frequency point within the respiratory band (e.g., 0.083 to 0.5 Hz) of the radar multi-antenna zero speed range FFT signal and the sum of the energy at each frequency point outside the respiratory band can be calculated, and the distance frequency point corresponding to the maximum ratio of the two can be set as the target position; alternatively, the sum of the energy at each frequency point within the heart rate band (e.g., 0.83 to 2 Hz) of the radar multi-antenna zero speed range FFT signal and the sum of the energy at each frequency point outside the heart rate band can be calculated, and the distance frequency point corresponding to the maximum ratio of the two can be set as the target position.

[0036] In some embodiments, after determining the target position where the detection target exists, a target region can be determined based on the target position. Fig. 5 is a schematic diagram showing the determination of the target region in an embodiment of the present invention. In some embodiments, as shown in Fig. 5, when the r-th distance frequency point is selected as the target position, the target region is determined by centering the r-th distance frequency point and determining the r-th distance frequency point before the distance frequency point. limit distance frequency points, r limit The r-th distance frequency point and the r-th distance frequency point may be set as the target region. limit = (N1-1) / 2. Note that the present invention is not limited to this, and in 302, the target position may be determined without being centered on the target position, and for example, the target position may be positioned further forward or further backward within the target area.

[0037] In some embodiments, at 303, the first radar signal can be selected from the target area in various ways.

[0038] For example, among the amplitude fluctuation amounts of the radar multi-antenna zero speed range FFT signal at each distance frequency point (N1 distance frequency points) within the target area, the radar multi-antenna zero speed range FFT signal corresponding to the largest N2 (N1≧N2≧1) amplitude fluctuation amounts can be used as the first radar signal.

[0039] Also, for example, the radar multi-antenna zero speed range FFT signal corresponding to the maximum N2 energy values ​​among the energy values ​​within the physiological signal band of the radar multi-antenna zero speed range FFT signal at each distance frequency point (N1 distance frequency points) within the target area can be the first radar signal.

[0040] For example, for each of the N1 distance frequency points, the sum of the energy of the radar multi-antenna zero speed range FFT signal at each frequency point within the heartbeat band is calculated, and the radar multi-antenna zero speed range FFT signal at the N2 distance frequency points where the sum of the energy is maximum is defined as the first radar signal; alternatively, the sum of the energy of the radar multi-antenna zero speed range FFT signal at each frequency point within the breathing band is calculated, and the radar multi-antenna zero speed range FFT signal at the N2 distance frequency points where the sum of the energy is maximum is defined as the first radar signal.

[0041] Also, for example, the radar multi-antenna zero speed range FFT signal corresponding to the largest N2 ratios among the ratios of energy values ​​in the physiological signal band to energy values ​​in the non-physiological signal band of the radar multi-antenna zero speed range FFT signal at each range frequency point within the target area can be the first radar signal.

[0042] For example, for each of the N1 distance frequency points, the sum of the energy at each frequency point within the heartbeat band and the sum of the energy at each frequency point outside the heartbeat band of the radar multi-antenna zero speed range FFT signal are calculated, and the radar multi-antenna zero speed range FFT signal at the N2 distance frequency points where the ratio of the two is maximum is defined as the first radar signal; alternatively, the sum of the energy at each frequency point within the breathing band of the radar multi-antenna zero speed range FFT signal and the sum of the energy at each frequency point outside the breathing band are calculated, and the radar multi-antenna zero speed range FFT signal at the N2 distance frequency points where the ratio of the two is maximum is defined as the first radar signal.

[0043] However, the present invention is not limited to this, and the first radar signal may be a multi-antenna zero-speed range FFT signal at N1 range frequency points within the target area.

[0044] In some embodiments, at 304, after selecting the first radar signal, the first radar signal may be pre-processed to generate amplitude and phase characteristics of the first radar signal.

[0045] In some embodiments, the amplitude characteristic of the first radar signal may be various characteristics related to the amplitude of the first radar signal.

[0046] For example, the amplitude characteristic of the first radar signal may be a difference between an amplitude of the first radar signal at each time point within a preset time period and an average value of the amplitude of the first radar signal within the preset time period, thereby reducing low-frequency interference in the first radar signal.

[0047] For example, the amplitude characteristic of the first radar signal may be generated by subtracting the average value of the amplitude of the first radar signal from the amplitude of the first radar signal at each time point according to Equation 1.

[0048]

number

[0049] However, the present invention is not limited thereto, and the amplitude characteristic of the first radar signal may be other quantities, such as the amplitude of the first radar signal at each point in time.

[0050] In some embodiments, the phase characteristic of the first radar signal may be various characteristics related to the phase of the first radar signal.

[0051] For example, the phase feature of the first radar signal may be a difference (i.e., a phase difference) between a phase value of a next frame of the first radar signal within a preset period and a phase value of a previous frame of the first radar signal. This can eliminate interference caused by slow fluctuations of the detection target and improve signal quality. For example, the phase feature of the first radar signal can be generated by subtracting the phase value of the previous frame from the phase value of the next frame of the first radar signal according to Equation 2.

[0052]

number

[0053] However, the present invention is not limited thereto, and the phase characteristic of the first radar signal may be other quantities, such as the phase of the first radar signal at each point in time.

[0054] In some examples, after amplitude and phase features of the first radar signal are generated, the amplitude and phase features can be input to a deep neural network based detection model, which can then be used to generate candidate physiological signals of multiple segments.

[0055] For example, the first radar signal includes radar multi-antenna zero-speed range FFT signals at N1 range-frequency points within a target area, and the target area includes a target position (the position of the rth range-frequency point), r limit distance frequency points and r limit distance frequency points, the amplitude and phase features can be concatenated as input to the detection model according to Equation 3.

[0056]

number

[0057] In some embodiments, at 103, a plurality of segments of candidate physiological signals can be generated based on the amplitude and phase features of the first radar signal using a deep neural network-based detection model, where the detection model can be various types of deep neural network-based models, such as a convolutional neural network, a recurrent neural network, an adversarial neural network, etc.

[0058] In some embodiments, the detection model may include a multi-scale feature extraction module, which may improve the accuracy of the candidate physiological signal by considering information in different time ranges when generating the candidate physiological signal, although the present invention is not limited in this respect and the detection model 600 may also include a single-scale feature extraction module.

[0059] 6 is a schematic diagram of a detection model in an embodiment of the present invention. In some embodiments, the detection model 600 may include a single-time feature extraction module 601, a local time feature extraction module 602, and a global feature extraction module 603, as shown in FIG. 6. Among them, the single-time feature extraction module 601 can obtain first spatial feature information on each distance frequency point in a single frame time based on the amplitude feature and phase feature of the first radar signal. The local time feature extraction module 602 can obtain second spatial feature information on each distance frequency point in multiple frames time based on the first spatial feature information. The global feature extraction module 603 can obtain global time feature information on each distance frequency point based on the second spatial feature information.

[0060] The first spatial feature information can correspond to a time of one frame, for example, 0.1s. The second spatial feature information can correspond to a time of multiple frames, for example, 0.3s or 0.7s. The global temporal feature information can correspond to a time longer than the time of multiple frames, for example, 10s or 20s.

[0061] 7 is a schematic diagram of a single-time feature extraction module 601 according to an embodiment of the present invention. As shown in FIG. 7, in some embodiments, the single-time feature extraction module 601 may include a module 6011, a module 6012, and a module 6013, and the outputs of the modules 6011 and 6012 may be cascaded as inputs to the module 6013. The modules 6011, 6012, and 6013 may include a one-dimensional batch normalization layer BN1d, a one-dimensional convolution layer Conv1d_single, and a rectified linear unit ReLU (Rectified Linear Unit), in which the one-dimensional convolution layer Conv1d_single may have a kernel size of 1, a step length of 1, and a number of channels of N. However, the present invention is not limited thereto, and the single-time feature extraction module 601 may have other configurations.

[0062] FIG. 8 is a schematic diagram of a local time feature extraction module 602 according to an embodiment of the present invention. As shown in FIG. 8, in some embodiments, the local time feature extraction module 602 may include a module 6021, a module 6022, and a module 6023, and the outputs of the modules 6021 and 6022 may be cascaded as inputs to the module 6023. The modules 6021, 6022, and 6023 may include a one-dimensional batch normalization layer BN1d, a one-dimensional convolution layer Conv1d_short, and a rectified linear unit ReLU. Here, the one-dimensional convolution layer Conv1d_single may have a kernel size of K1 (K1>1), a step length of 1, and a number of channels of N. However, the present invention is not limited thereto, and the local time feature extraction module 602 may have other configurations. Additionally, in the detection model 600, multiple local time feature extraction modules 602 may be cascaded, and the processing within each local time feature extraction module 602 may be the same or different, thereby forming more complex local time feature extraction modules.

[0063] 9 is a schematic diagram of the global feature extraction module 603 according to an embodiment of the present invention. As shown in FIG. 9, in some embodiments, the global feature extraction module 603 may include a module 6031 and a module 6032. Here, the module 6031 may perform a transpose process, and the module 6032 may include a one-dimensional batch normalization layer BN1d, a one-dimensional convolution layer Conv1d_single, and a rectified linear unit ReLU. However, the present invention is not limited thereto, and the global feature extraction module 603 may have other configurations.

[0064] In some embodiments, the detection model 600 may further include a feature fusion module 604, as shown in Figure 6. The feature fusion module 604 may fuse the first spatial feature information, the second spatial feature information, and the global temporal feature information to generate a candidate physiological signal.

[0065] In some embodiments, the feature fusion module 604 may fuse the first spatial feature information, the second spatial feature information, and the global temporal feature information in various manners, for example, the feature fusion module 604 may cascade the first spatial feature information, the second spatial feature information, and the global temporal feature information, or the feature fusion module 604 may add the first spatial feature information, the second spatial feature information, and the global temporal feature information.

[0066] 10 is another schematic diagram of a detection model according to an embodiment of the present invention. As shown in FIG. 10, the detection model can include a single-time feature extraction module 601, a global feature extraction module 603, and multiple local time feature extraction modules 602. Note that, the specific method of the detection model generating an output Y based on an input X can be referred to the related art, and the description thereof will be omitted here.

[0067] In some embodiments, a loss function may be used to train the detection module before candidate physiological signals are generated by the detection module. The loss function used to train the detection model is a waveform reconstruction loss function Loss re , waveform change loss function Loss changes , waveform fluctuation loss function Loss v The loss function Loss may be a function generated from at least one of the following:

[0068] Loss=a*Loss re +b*Loss changes +c*Loss v (Formula 4) where a, b, and c are real numbers.

[0069] In some embodiments, the waveform reconstruction loss function Loss re may be the mean square error between the predicted physiological signal waveform values ​​output by the detection model within a certain period of time and the actual physiological signal waveform values, as shown in Equation 5.

[0070]

number

[0071] Note that the present invention is not limited to this, and the waveform reconstruction loss function Loss re may take other forms. For example, the waveform reconstruction loss function Loss re may be the mean square error between the predicted physiological signal waveform value output by the detection model within a certain period of time and the inverse of the actual physiological signal waveform value, as shown in Equation 6.

[0072]

number

[0073] Alternatively, the waveform reconstruction loss function Loss re is the mean square error (Loss) between the predicted physiological signal waveform value output by the detection model within a certain period and the actual physiological signal waveform value. mse , and the mean square error Loss between the waveform value of the physiological signal output by the detection model within a certain period and the inverse value of the waveform value of the actual physiological signal inverse The value may be the smaller of the two.

[0074] The waveform reconstruction loss function Loss re By including in the loss function, it is possible to ignore the representation of state transitions of the physiological signal. For example, for a respiratory signal, one can ignore tracking the inhalation or exhalation states, and Loss mse and Loss inverse By using the smaller value of the two, an optimal physiological parameter value can be obtained. FIG. 11 shows a time domain diagram and a frequency domain diagram of a real physiological signal and a predicted physiological signal according to an embodiment of the present invention. As shown in FIG. 11, the peaks and valleys of a predicted waveform 1001 and an actual waveform 1002 are reversed, but the frequency values ​​of both are the same (see frequency spectrum plot 1003 of predicted waveform 1001 and frequency spectrum plot 1004 of actual waveform 1002). In this case, it may be considered as follows, that is, since the predicted waveform 1001 output by the detection model is a correct result, its loss function value is small.

[0075] In some embodiments, the waveform change loss function Loss changes may be the mean square error between the waveform change amount of the real physiological signal and the waveform change amount of the predicted physiological signal output by the detection model, as shown in Equation 7.

[0076]

number

[0077] 12 is a diagram showing the amount of change in the waveform of a real physiological signal according to an embodiment of the present invention. As shown in FIG. 12, the amount of change in the time-domain waveform of the physiological signal can be obtained by subtracting the waveform value at a second time tk from the time-domain waveform value at a first time t. Here, the magnitude of the k value may be correlated with the period or frequency of the physiological signal. For example, the k value becomes smaller as the frequency of the physiological signal becomes higher.

[0078] The above waveform change loss function Loss changes By including in the loss function, it is possible to pay more attention to the state transition of the physiological signal. For example, in the case of a respiratory signal, the waveform change loss function can pay more attention to the state transition process from inspiration to expiration or from expiration to inspiration, which is advantageous to more accurately track the features of the physiological signal.

[0079] In some embodiments, the waveform variation loss function Loss v may be the mean square error between the waveform fluctuation amount of the actual physiological signal and the waveform fluctuation amount of the physiological signal output by the detection model, as shown in Equation 8.

[0080]

number

[0081] The amount of waveform variation of the actual physiological signal can be set to the average value of the amount of waveform variation of the time domain waveform of the actual physiological signal in a preset period, as shown in Equation 9.

[0082]

number

[0083]

number

[0084] In some embodiments, the target physiological signal may be selected from the candidate physiological signals of the multiple segments generated by the detection model at 104. Here, the candidate physiological signal output by the detection model each time may be the candidate physiological signal of one segment. The candidate physiological signal may be a respiratory signal or a heartbeat signal, and the target physiological signal may also be a respiratory signal or a heartbeat signal. Here, the respiratory signal may be various signals related to the respiratory rate, and the heartbeat signal may be various signals related to the heartbeat.

[0085] 14 is a diagram showing that the detection model outputs candidate physiological signals of multiple segments. As shown in FIG. 14, the detection model outputs a candidate physiological signal Y m and output the candidate physiological signal Y mmay include multiple frames. For example, a candidate physiological signal of one segment includes 100 frames, and the time length of each frame is 0.1 s. The detection model can output a candidate physiological signal of one segment at a preset time interval. For example, the detection model outputs a candidate physiological signal of one segment at an interval of one frame. The predetermined number of candidate physiological signals (Y1, ...Y m ,…Y M ) are stored, one or more segments from these candidate physiological signals can be selected as a target physiological signal.

[0086] 15 is a schematic diagram of step 104 in an embodiment of the present invention. As shown in FIG. 15, 104 may include the following steps:

[0087] 1041: determining validity of candidate physiological signals of a plurality of segments; and 1042: When the candidate physiological signals of the multiple segments are determined to be valid, a target physiological signal is determined from the candidate physiological signals of the multiple segments, and when the candidate physiological signals of the multiple segments are determined to be invalid, a target physiological signal is not determined.

[0088] In some embodiments, whether the candidate physiological signals of the plurality of segments are valid can be determined based on the similarity between the candidate physiological signals of the plurality of segments. For example, if the similarity between the candidate physiological signals of the plurality of segments is high, the candidate physiological signals of the plurality of segments can be regarded as valid, and if the similarity between the candidate physiological signals of the plurality of segments is low, the candidate physiological signals of the plurality of segments can be regarded as invalid. Only if the candidate physiological signals of the plurality of segments are valid, a target physiological signal is further determined based on the candidate physiological signals of the plurality of segments, thereby improving the reliability of the determined target.

[0089] In some embodiments, the similarity between the candidate physiological signals of the plurality of segments may be expressed in various ways, for example, the similarity between the candidate physiological signals of the plurality of segments may be determined based on a similarity of a combination of the candidate physiological signals of two segments among the candidate physiological signals of the plurality of segments.

[0090] 16 is a schematic diagram of step 1041 in an embodiment of the present invention. As shown in FIG. 16, 1041 may include the following steps:

[0091] 10411: Determine a temporal overlap length (temporal overlap length) of candidate physiological signals of any two segments among the candidate physiological signals of a plurality of segments, and obtain a candidate physiological signal combination having a temporal overlap length longer than a time of one breath or heartbeat and a first combination number L of the candidate physiological signal combination; 10412: Using time domain correlation and / or frequency domain coherence to determine a similarity of the candidate physiological signals of the two segments in each candidate physiological signal combination, and determining a second combination number K of the candidate physiological signal combinations whose similarity is greater than a first threshold; 10413: If K / L (ratio) is greater than a second threshold, the candidate physiological signals of the multiple segments are determined to be valid, and if K / L is less than or equal to the second threshold, the candidate physiological signals of the multiple segments are determined to be invalid.

[0092] Hereinafter, the candidate physiological signals are respiratory signals, and the number M of the candidate physiological signals is 10 (10 segments), and the validity of the candidate physiological signals of multiple segments will be described as an example. m ,…Y 10 ), then, for the candidate physiological signals of these 10 segments, pairwise combinations are performed, and C 10 2 = 45 combinations (Y i ,Y j), where 1≦i≦10, 1≦j≦10, and i≠j. For each of the 45 combinations, the two candidate physiological signals Y i and Y j The overlap length in the time domain of the two candidate physiological signals Y i and Y j A similarity between the candidate physiological signals is calculated. The calculated similarity between the candidate physiological signal combinations is compared with a first threshold to determine a number K of combinations whose similarity is greater than the first threshold. If a ratio of the number K to the number L is greater than a second threshold, it can be determined that the candidate physiological signals of the 10 segments are valid, and if the ratio of the number K to the number L is equal to or less than the second threshold, it can be determined that the candidate physiological signals of the 10 segments are invalid.

[0093] In some embodiments, the two candidate physiological signals Y in the candidate physiological signal combination i and Y j The similarity of two time series can be determined in various ways. For example, the similarity can be determined using time-domain correlation. The time-domain correlation can be expressed by a time-domain correlation coefficient ρ(T), where ρ(T) has a numerical range of [-1, 1], as shown in the following Equation 11. The closer ρ(T) is to 1, the higher the time-domain similarity of the two time series.

[0094]

number

[0095] Also, for example, the similarity is determined using frequency domain coherence, which can be expressed by a frequency spectrum coherence coefficient C(f), where C(f) has a numerical range of [0,1], as shown in the following Equation 12, where the closer C(f) is to 1, the higher the frequency domain similarity of two candidate physiological signals.

[0096]

number

[0097] Also, for example, the similarity may be determined using both the time-domain correlation and the frequency-domain coherence. For example, the first threshold may include a first value and a second value, and if the time-domain correlation coefficient ρ(T) is greater than the first value and the frequency-domain coherence coefficient C(f) is greater than the second value, the frequency-domain similarity of the two candidate physiological signals may be considered to be greater than the first threshold.

[0098] In some examples, at 1042, if the candidate physiological signals of the plurality of segments are valid, a target physiological signal of one or more segments may be selected from the candidate physiological signals of the plurality of segments based on the time domain evaluation index and / or the frequency domain evaluation index.

[0099] 17 is a schematic diagram of step 1042 in an embodiment of the present invention. As shown in FIG. 17, 1042 may include the following steps:

[0100] 10421: Based on the valid candidate physiological signals of the plurality of segments, evaluating the candidate physiological signals of each segment among the candidate physiological signals of the plurality of segments using a time domain evaluation index and / or a frequency domain evaluation index; and 10422: A target physiological signal is determined based on the time domain and / or frequency domain evaluation index values ​​of the candidate physiological signals for each segment.

[0101] In some embodiments, the time-domain evaluation index may include a maximum time-domain amplitude of the candidate physiological signal. For example, the candidate physiological signals of the multiple segments are sorted according to the maximum time-domain amplitude from largest to smallest, and the candidate physiological signal of the segment with the largest maximum time-domain amplitude is selected as the target physiological signal, or the candidate physiological signal of the multiple segments with the highest maximum time-domain amplitude is selected as the target physiological signal.

[0102] In some embodiments, the time-domain evaluation index may include an average value of correlation coefficients between the time-domain waveform of the candidate physiological signal and the time-domain waveforms of other candidate physiological signals of the plurality of segments, so that the candidate physiological signal of one or more segments that best represents the candidate physiological signal of the plurality of segments can be determined as the target physiological signal.

[0103] For example, for each of the candidate physiological signals of a plurality of segments, a time-domain correlation coefficient between the candidate physiological signal of the one segment and the other candidate physiological signals is calculated, and an average value of these time-domain correlation coefficients is calculated as a time-domain evaluation index value of the candidate physiological signal of the one segment. The calculated time-domain evaluation index values ​​of the candidate physiological signals of each segment are sorted in order from large to small, and the candidate physiological signal of the segment with the largest time-domain evaluation index value is selected as the target physiological signal, or the candidate physiological signal of the plurality of segments with the highest time-domain evaluation index value is selected as the target physiological signal.

[0104] In some embodiments, the frequency domain evaluation index may include a maximum value of frequency spectrum energy in a physiological signal band of the candidate physiological signal. For example, perform FFT on the candidate physiological signals of multiple segments, calculate the frequency spectrum energy of the candidate physiological signals of each segment at each frequency point in the respiratory band or the cardiac band, determine the maximum value of the frequency spectrum energy, and sort the maximum values ​​of the frequency spectrum energy according to the order of "largest to smallest", and select the candidate physiological signal of the segment with the largest maximum value of the frequency spectrum energy as the target physiological signal, or select the candidate physiological signal of the multiple segments with the highest maximum value of the frequency spectrum energy as the target physiological signal.

[0105] In some embodiments, the frequency domain evaluation index may include an energy value in a physiological signal band of the candidate physiological signal. For example, perform FFT on the candidate physiological signals of multiple segments, calculate the sum of the energy of the candidate physiological signals of each segment at each frequency point in the respiratory band or the cardiac band, and use it as the energy value of the candidate physiological signals of each segment in the respiratory band or the cardiac band, and sort the energy values ​​according to the order of "largest to smallest", and select the candidate physiological signal of the segment with the largest energy value as the target physiological signal, or select the candidate physiological signals of the multiple segments with the highest energy value as the target physiological signal.

[0106] In some embodiments, the frequency domain evaluation index may include a ratio of an energy value in a physiological band to an energy value in a non-physiological band of the candidate physiological signal. For example, perform FFT on the candidate physiological signals of multiple segments, calculate the ratio of an energy value in a respiratory band to an energy value in a non-respiratory band, or the ratio of an energy value in a cardiac band to an energy value in a non-cardiac band of the candidate physiological signal of each segment, and sort the candidate physiological signals according to the ratio of the energy values ​​from "large to small", and select the candidate physiological signal of the segment with the largest ratio of the energy values ​​as the target physiological signal, or select the candidate physiological signals of the multiple segments with the highest ratio of the energy values ​​as the target physiological signal.

[0107] In some embodiments, the frequency domain evaluation index may include an average value of a frequency spectrum coherence coefficient between the candidate physiological signal and other candidate physiological signals of the candidate physiological signals of the plurality of segments in a physiological frequency band, so that the candidate physiological signal of one or more segments that can best represent the candidate physiological signal of the plurality of segments can be determined as the target physiological signal.

[0108] For example, an FFT is performed on each candidate physiological signal of a plurality of segments to obtain frequency spectrum information of the candidate physiological signal of each segment, a frequency spectrum coherence coefficient between the candidate physiological signal of the one segment and other candidate physiological signals is calculated based on the frequency spectrum information, and an average value of these frequency spectrum coherence coefficients is calculated as a frequency domain evaluation index value of the candidate physiological signal of the one segment. The calculated frequency domain evaluation index values ​​of the candidate physiological signals of each segment are sorted in order of "largest to smallest", and the candidate physiological signal of the segment with the largest frequency domain evaluation index value is selected as the target physiological signal, or a candidate physiological signal of the plurality of segments with the highest frequency domain evaluation index value is selected as the target physiological signal.

[0109] In some embodiments, when selecting a target physiological signal from candidate physiological signals of multiple segments at 104, the candidate physiological signal of one segment may be selected as the target physiological signal, i.e., the target physiological signal includes the candidate physiological signal of one segment, or the candidate physiological signals of two or more segments may be selected as the target physiological signal, i.e., the target physiological signal includes signals of multiple segments.

[0110] In some embodiments, the target physiological signal may be selected based on a time domain metric or a frequency domain metric, as described above.

[0111] For example, a candidate physiological signal of a segment having a maximum time domain evaluation index value or a maximum frequency domain evaluation index value may be selected as a target physiological signal. For example, a candidate physiological signal of a segment having a maximum time domain amplitude value is selected as a target physiological signal.

[0112] Alternatively, for example, the candidate physiological signals of two or more segments that are ranked higher after being sorted according to the order of one time domain evaluation index value or one frequency domain evaluation index value from "large to small" may be selected as the target physiological signal. For example, the candidate physiological signals may be sorted according to the order of the maximum time domain amplitude from "large to small", and the candidate physiological signals of the two segments that are ranked higher may be selected as the target physiological signal.

[0113] In some embodiments, the target physiological signal may be selected based on a plurality of evaluation indexes, which may include a plurality of time domain evaluation indexes, a plurality of frequency domain evaluation indexes, or one or more time domain evaluation indexes and one or more frequency domain evaluation indexes.

[0114] In some embodiments, when selecting a target physiological signal based on a plurality of evaluation indexes, candidate physiological signals of one or more segments may be selected based on each of the plurality of evaluation indexes as a set of candidate physiological signals corresponding to the evaluation index, and a target physiological signal may be determined based on the set of candidate physiological signals corresponding to each evaluation index.

[0115] In some embodiments, when determining a target physiological signal based on a set of candidate physiological signals corresponding to each evaluation index, the target physiological signal may be a union of each set, or the target physiological signal may be a union of each set, or the target physiological signal may be determined based on the number of times the candidate physiological signal appears in the set, and / or the priority of each evaluation index, and / or the priority of the candidate physiological signals in the same set (e.g., the higher the evaluation index value of a candidate physiological signal, the higher the priority of the candidate physiological signal).

[0116] The above-mentioned selection method is hereinafter illustrated by taking as an example the selection of a target physiological signal of one segment (i.e., the target physiological signal includes a candidate physiological signal of one segment) from candidate physiological signals of ten segments based on index 1, index 2, and index 3. Among them, index 1, index 2, and index 3 may be any one of the time domain or frequency domain evaluation indexes as mentioned above.

[0117] From the candidate physiological signals of the 10 segments, a candidate physiological signal with the maximum value of index 1 is selected based on index 1, for example, the candidate physiological signal of the seventh segment; a candidate physiological signal with the maximum value of index 2 is selected based on index 2, for example, the candidate physiological signal of the first segment; and a candidate physiological signal with the maximum value of index 3 is selected based on index 3, for example, the candidate physiological signal of the first segment. Since the candidate physiological signal of the first segment has been selected twice, the candidate physiological signal of the first segment may be used as the target physiological signal.

[0118] Alternatively, when the candidate physiological signals selected by indexes 1, 2, and 3 are all different, for example, when the physiological signals of the 7th, 1st, and 10th segments are selected based on indexes 1, 2, and 3, respectively, the target physiological signal may be determined based on the index with the highest priority, for example, index 1, i.e., the candidate physiological signal of the 7th segment may be selected as the target physiological signal.

[0119] In some embodiments, 105 detects vital sign parameters based on the target physiological signal. When the target physiological signal selected from the candidate physiological signals of a plurality of segments is a signal of one segment, the heartbeat frequency or respiration frequency may be determined based on the frequency of the target physiological signal. For example, the target physiological signal may be subjected to FFT to obtain frequency spectrum information of the target physiological signal, and the frequency of a frequency point with the maximum frequency spectrum energy may be determined as the heartbeat frequency or respiration frequency based on the frequency spectrum information, or the frequency of each frequency point may be weighted using the frequency spectrum energy of each frequency point within the range of the heartbeat or respiration frequency based on the frequency spectrum information, and the weighted sum result may be determined as the heartbeat frequency or respiration frequency. Note that the present invention is not limited to this, and the heartbeat frequency or respiration frequency may be determined by other methods.

[0120] In some embodiments, when the target physiological signal selected from the candidate physiological signals of a plurality of segments includes signals of a plurality of segments, the heartbeat frequency or the respiratory frequency may be determined based on the frequency of the target physiological signal of each section of the target physiological signals of the plurality of segments. The frequency of the target physiological signal of each segment may be determined by the above-mentioned method, and the detailed description thereof will be omitted here. After obtaining the frequency of the target physiological signal of each segment, the average value of the frequency of the target physiological signal of each segment may be determined as the heartbeat frequency or the respiratory frequency. Alternatively, the frequency of the target physiological signal of each segment may be weighted, and the weighted sum result may be determined as the heartbeat frequency or the respiratory frequency. The weighting coefficient of the frequency of the target physiological signal of each segment may be, for example, the energy value in the physiological frequency band of the target physiological signal of the section.

[0121] It should be noted that the above drawings are for the purpose of briefly explaining the embodiments of the present invention, but the present invention is not limited thereto. The execution order between each operation (step) can be appropriately adjusted, or some operations can be added or removed. In addition, the present invention is not limited to the description of the above drawings, and those skilled in the art can appropriately modify the above contents.

[0122] The above describes only the steps or processes related to the present invention, but the present invention is not limited thereto. The above method may also include other steps or processes, and the details of these steps or processes may be found in the prior art. In addition, the above describes the embodiments of the present invention using only some configurations of the vital signs detection model as examples, but the present invention is not limited to these configurations, and these configurations may be modified as appropriate, and all of these modified embodiments are included in the scope of the embodiments of the present invention.

[0123] Furthermore, although the present invention has been illustrated above by taking various examples as examples, the present invention is not limited to these, and these various examples can be modified as appropriate. Furthermore, these various examples can be used, for example, alone or in combination with one or more of them.

[0124] According to the above-described embodiment, candidate physiological signals of a plurality of segments are generated using a detection model based on a deep neural network based on the amplitude and phase features of the first radar signal, a target physiological signal is selected from the candidate physiological signals of the plurality of segments, and vital sign parameters are detected based on the target physiological signal. This makes it possible to select a physiological signal that can more accurately reflect the vital sign parameters as the target physiological signal from the candidate physiological signals of the plurality of segments, thereby improving the accuracy of the detection results of the vital sign parameters.

[0125] In addition, according to the above-mentioned embodiment, vital sign parameters are detected based on the radar multi-antenna zero-speed distance FFT signal, which can reduce the amount of data transmission and improve the data transmission efficiency, which is advantageous for realizing real-time detection after commercialization, and on the other hand, the vital sign detection method can be applied to more situations.

[0126] <Example of the second aspect> In the embodiment of the present invention, a vital sign detection device is provided, and the same contents as those in the embodiment of the first aspect are omitted.

[0127] 18 is a schematic diagram of a vital sign detection device 1800 according to an embodiment of the present invention. As shown in FIG. 18, the vital sign detection device 1800 includes:

[0128] Data acquisition unit 1801: acquires radar multi-antenna zero speed range FFT signal; A data pre-processing unit 1802: selects a first radar signal from the radar multi-antenna zero-speed range FFT signal, and generates amplitude and phase characteristics of the first radar signal; A generating unit 1803: generating candidate physiological signals of a plurality of segments based on the amplitude features and the phase features using a detection model based on a deep neural network; A selection unit 1804: selecting a target physiological signal from the candidate physiological signals of the plurality of segments, the candidate physiological signal and the target physiological signal being a respiratory signal or a cardiac signal; and A detection unit 1805: detects a vital sign parameter based on the target physiological signal.

[0129] In some embodiments, the selection unit 1804 includes:

[0130] A determining unit for determining the validity of candidate physiological signals of the plurality of segments; and A determination unit: when the judgment unit determines that the candidate physiological signals of the plurality of segments are valid, the target physiological signal is determined from the candidate physiological signals of the plurality of segments, and when the judgment unit determines that the candidate physiological signals of the plurality of segments are invalid, the target physiological signal is not determined.

[0131] In some embodiments, the determination unit determines a temporal overlap length of candidate physiological signals of any two segments among the candidate physiological signals of the plurality of segments, and obtains candidate physiological signal combinations whose temporal overlap length is longer than a time of one breath or heartbeat and a first combination number L of the candidate physiological signal combinations; determines a similarity of the candidate physiological signals of the two segments in each candidate physiological signal combination using time-domain correlation and / or frequency-domain coherence, and determines a second combination number K of the candidate physiological signal combinations whose similarity is greater than a first threshold; and determines that the candidate physiological signals of the plurality of segments are valid if the ratio K / L is greater than a second threshold, and determines that the candidate physiological signals of the plurality of segments are invalid if the ratio K / L is less than or equal to the second threshold.

[0132] In some embodiments, the determining unit evaluates a candidate physiological signal of each segment among the candidate physiological signals of the plurality of segments using time domain evaluation indexes and / or frequency domain evaluation indexes based on the valid candidate physiological signals of the plurality of segments; and determines the target physiological signal based on the time domain evaluation index values ​​and / or frequency domain evaluation index values ​​of the candidate physiological signals of each segment.

[0133] In some embodiments, the time domain evaluation metric is: a maximum time domain amplitude of the candidate physiological signal; and / or an average correlation coefficient between the time-domain waveform of the candidate physiological signal and the time-domain waveforms of other candidate physiological signals of the plurality of segments; and / or The frequency domain evaluation index is a maximum frequency spectral energy in a physiological signal band of the candidate physiological signal; and / or and / or an energy value in a physiological signal band of the candidate physiological signal; and / or a ratio of energy values ​​in physiological signal bands of the candidate physiological signal to energy values ​​in non-physiological signal bands of the candidate physiological signal; and / or an average value of frequency spectral coherence coefficients in a physiological signal band between the candidate physiological signal and other candidate physiological signals among the candidate physiological signals of the plurality of segments; Includes.

[0134] In some embodiments, the data pre-processing unit 1802 includes:

[0135] A target location determination unit: determining a target location according to time domain information and / or frequency domain information of the multi-antenna zero-speed range FFT signal; A target area determination unit: the target position and the positions of N1 distance frequency points adjacent to the target position are set as a target area; and A radar signal selection unit: selects the first radar signal in the target area and generates an amplitude signature and a phase signature of the first radar signal.

[0136] In some embodiments, the time domain information includes information on the change in amplitude of the multi-antenna zero speed range FFT signal at each range frequency point over time within a predetermined period, and the target position determination unit calculates the amplitude fluctuation amount at each range frequency point during the predetermined period and determines the range frequency point corresponding to the maximum value of the amplitude fluctuation amount as the target position; and / or the frequency domain information includes a result after performing a fast Fourier transform on the phase of the multi-antenna zero speed range FFT signal at each range frequency point, and the target position determination unit determines the range frequency point corresponding to the maximum value of the sum of energy at each frequency point in the physiological signal band as the target position, and / or the target position determination unit determines the range frequency point corresponding to the maximum value of the ratio between the sum of energy at each frequency point in the physiological signal band and the sum of energy at each frequency point in the non-physiological signal band as the target position.

[0137] In some embodiments, the radar signal selection unit determines as the first radar signal the multi-antenna zero speed range FFT signal corresponding to the maximum N2 amplitude fluctuation amounts among the amplitude fluctuation amounts of the multi-antenna zero speed range FFT signal at each distance frequency point in the target area; and / or the radar signal selection unit determines as the first radar signal the multi-antenna zero speed range FFT signal corresponding to the maximum N2 energy values ​​among the energy values ​​within a physiological signal band of the multi-antenna zero speed range FFT signal at each distance frequency point in the target area; and / or the radar signal selection unit determines as the first radar signal the multi-antenna zero speed range FFT signal corresponding to the maximum N2 ratios among the ratios between the energy values ​​within a physiological signal band of the multi-antenna zero speed range FFT signal at each distance frequency point in the target area and the energy values ​​within a non-physiological signal band.

[0138] In some embodiments, the radar signal selection unit subtracts an average value of the amplitude of the first radar signal from the amplitude of the first radar signal at each time point to generate an amplitude characteristic of the first radar signal; and / or the radar signal selection unit calculates a difference between a phase value of a next frame of the first radar signal and a phase value of a frame immediately before that to generate a phase characteristic of the first radar signal.

[0139] In some embodiments, the detection model includes the following:

[0140] A single time point feature extraction module: obtains first spatial feature information at each distance frequency point within a single frame time according to the amplitude feature and phase feature of the first radar signal; A local time feature extraction module: obtains second spatial feature information at each distance frequency point within a plurality of frames of time according to the first spatial feature information; A global feature extraction module: obtains global time feature information at each distance frequency point according to the second spatial feature information; and A feature fusion module: fuses the first spatial feature information, the second spatial feature information and the global temporal feature information to generate the candidate physiological signal.

[0141] In some embodiments, the feature fusion module cascades the first spatial feature information, the second spatial feature information, and the global temporal feature information, or adds the first spatial feature information, the second spatial feature information, and the global temporal feature information.

[0142] In some embodiments, the loss function for training the detection model is a waveform reconstruction loss function Loss re , waveform change loss function Loss changes , or waveform fluctuation loss function Loss v is a function generated according to at least one of

[0143] In some embodiments, the loss function Loss is:

[0144] Loss=a*Loss re +b*Loss changes +c*Loss v where a, b, and c are real numbers.

[0145] In some embodiments, the waveform reconstruction loss function Loss re is the mean square error between the waveform value of the actual physiological signal and the waveform value of the physiological signal output by the detection model; and / or the waveform change loss function Loss changes is the mean square error between the waveform change amount of the actual physiological signal and the waveform change amount of the physiological signal output by the detection model; and / or the waveform change amount loss function Loss v is the mean square error between the waveform fluctuation amount of the actual physiological signal and the waveform fluctuation amount of the physiological signal output by the detection model.

[0146] Although only the components or modules related to the present invention have been described above, the present invention is not limited thereto. Furthermore, the vital sign detection device 1800 may include other components or modules, and the details of these components or modules may be found in related art.

[0147] For convenience, Fig. 18 illustrates only the connection relationships or signal directions between each component or module, but a person skilled in the art would recognize that various related technologies such as bus connection can be applied. The above-described various components or modules can be realized by hardware such as a processor (processing device) and a memory (storage device). The above-described various embodiments are not limited thereto.

[0148] The above-mentioned various embodiments are merely illustrative of the embodiments of the present invention, but the present invention is not limited thereto, and appropriate modifications may be made based on the above-mentioned various embodiments. For example, the above-mentioned various embodiments may be used alone, or one or more of the above-mentioned various embodiments may be used in combination.

[0149] According to the above-described embodiment, candidate physiological signals of a plurality of segments are generated using a detection model based on a deep neural network based on the amplitude and phase features of the first radar signal, a target physiological signal is selected from the candidate physiological signals of the plurality of segments, and vital sign parameters are detected based on the target physiological signal. This makes it possible to select a physiological signal that can more accurately reflect the vital sign parameters as the target physiological signal from the candidate physiological signals of the plurality of segments, thereby improving the accuracy of the detection results of the vital sign parameters.

[0150] In addition, according to the above-mentioned embodiments, vital sign parameters are detected based on radar multi-antenna zero-speed distance FFT signals, which can reduce the amount of data transmission and improve data transmission efficiency, which is advantageous for realizing real-time detection after commercialization, and on the other hand, the vital sign detection method can be applied to more scenes.

[0151] <Example of the third aspect> In an embodiment of the present invention, an electronic device is provided, which includes the vital signs detection device 1800 in the embodiment of the second aspect, the contents of which are incorporated herein. The electronic device may be, for example, a computer, a server, a workstation, a laptop computer, a smartphone, etc. However, the present invention is not limited thereto.

[0152] Fig. 19 is a schematic diagram of an electronic device according to an embodiment of the present invention. As shown in Fig. 19, the electronic device 1900 may include a processor (e.g., CPU) 1910 and a memory (storage unit) 1920. The memory 1920 is connected to the central processor 1910. The memory 1920 may store various data, and may also store a program for information processing executed under the control of the processor 1910.

[0153] In some embodiments, the functionality of the vital signs detection device 1800 is integrated and implemented in the processor 1910. In some embodiments, the processor 1910 is configured to implement the vital signs detection method described in the embodiments of the first aspect.

[0154] In some embodiments, the vital sign detection device 1800 may be configured separately from the processor 1910, for example, the vital sign detection device 1800 may be configured as a chip connected to the processor 1910 that implements the functions of the vital sign detection device 1800 under the control of the processor 1910.

[0155] In some embodiments, the processor 1910 is configured to perform the following controls, for example: acquire a radar multi-antenna zero speed range FFT signal; select a first radar signal from the multi-antenna zero speed range FFT signal and generate amplitude and phase features of the first radar signal; generate candidate physiological signals of a plurality of segments based on the amplitude and phase features using a detection model based on a deep neural network; select a target physiological signal from the candidate physiological signals of the plurality of segments; and detect vital sign parameters based on the target physiological signal.

[0156] As shown in Fig. 19, the electronic device 1900 may further include an input / output (I / O) device 1930 and a display (display) 1940. The functions of these components are similar to those of the conventional technology, and detailed descriptions thereof will be omitted here. The electronic device 1900 does not necessarily have to include all the components shown in Fig. 19. The electronic device 1900 may further include components not shown in Fig. 19 with reference to the related art.

[0157] In some embodiments, a computer readable program is further provided, which, when executed in the electronic device, causes the computer to perform the vital signs detection method described in any embodiment of the first aspect.

[0158] In an embodiment of the present invention, a storage medium storing a computer-readable program is further provided, which, when executed in an electronic device, causes the computer to execute the vital signs detection method described in the embodiment of the first aspect.

[0159] The above-mentioned devices and methods may be realized by software or hardware, or may be realized by a combination of hardware and software. The present invention further relates to a computer-readable program, which, when executed by a logic component, causes the logic component to realize the above-mentioned devices or components, or causes the logic component to realize each of the above-mentioned methods or steps. The logic component may be, for example, a Field Programmable Gate Array (FPGA), a microprocessor, a processor for use in a computer, etc. The present invention also relates to a storage medium, such as a hard disk, a magnetic disk, an optical hard disk, a DVD, a flash memory, etc., storing the above-mentioned program.

[0160] Additionally, one or more combinations of the functional blocks illustrated in the figures and / or one or more combinations of the functional blocks may be implemented as a general purpose processor, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic component, a discrete gate or transistor logic component, a discrete hardware assembly, or any other suitable combination for performing the functions described herein. Also, one or more combinations of the functional blocks illustrated in the figures and / or one or more combinations of the functional blocks may be further configured as a combination of computing devices, such as a combination of a DSP and a microprocessor, a combination of multiple microprocessors, one or more microprocessors in communication with a DSP, or any other configuration.

[0161] Furthermore, with respect to the above-mentioned embodiments, the following supplementary notes are disclosed.

[0162] (Appendix 1) A vital sign detection device, comprising: Data acquisition unit to acquire radar multi-antenna zero speed range FFT signal; a data pre-processing unit for selecting a first radar signal from the radar multi-antenna zero speed range FFT signal, and generating an amplitude characteristic and a phase characteristic of the first radar signal; a generating unit for generating candidate physiological signals of a plurality of segments based on the amplitude features and the phase features and utilizing a detection model based on a deep neural network; a selection unit for selecting a target physiological signal from the candidate physiological signals of the plurality of segments, the candidate physiological signals and the target physiological signal being a respiratory signal or a cardiac signal; and The apparatus includes a detection unit for detecting a vital sign parameter based on the target physiological signal.

[0163] (Appendix 2) 2. The apparatus of claim 1, The detection model is a single-time feature extraction module for obtaining first spatial feature information at each distance frequency point within a single frame time according to the amplitude feature and the phase feature of the first radar signal; a local time feature extraction module for obtaining second spatial feature information at each distance frequency point within a plurality of frames of time according to the first spatial feature information; a global feature extraction module for obtaining global time feature information at each distance frequency point according to the second spatial feature information; and the first spatial feature information, the second spatial feature information, and the global temporal feature information to generate the candidate physiological signal.

[0164] (Appendix 3) 3. The apparatus of claim 2, The feature fusion module cascades the first spatial feature information, the second spatial feature information and the global temporal feature information, or adds the first spatial feature information, the second spatial feature information and the global temporal feature information.

[0165] (Appendix 4) 2. The apparatus of claim 1, The loss function for training the detection model is the waveform reconstruction loss function Loss re , waveform change loss function Loss changes , or waveform fluctuation loss function Loss v The apparatus is a function generated based on at least one of the above.

[0166] (Appendix 5) 5. The apparatus of claim 4, The loss function Loss is Loss=a*Loss re +b*Loss changes +c*Loss v where a, b, and c are real numbers.

[0167] (Appendix 6) 5. The apparatus of claim 4, The waveform reconstruction loss function Loss re is the mean square error between the waveform values ​​of the actual physiological signal and the waveform values ​​of the physiological signal output by the detection model; and / or The waveform change loss function Loss changes is the mean square error between the waveform change of the actual physiological signal and the waveform change of the physiological signal output by the detection model; and / or The waveform fluctuation loss function Loss v is the mean square error between the waveform fluctuation of an actual physiological signal and the waveform fluctuation of a physiological signal output by the detection model.

[0168] (Appendix 7) 1. A method for detecting vital signs, comprising: Acquire radar multi-antenna zero speed distance FFT signal; Selecting a first radar signal from the radar multi-antenna zero speed range FFT signal, and generating an amplitude characteristic and a phase characteristic of the first radar signal; generating candidate physiological signals of a plurality of segments based on the amplitude features and the phase features using a deep neural network based detection model; selecting a target physiological signal from the plurality of segments of candidate physiological signals, the candidate physiological signal and the target physiological signal being a respiratory signal or a cardiac signal; and detecting a vital sign parameter based on the target physiological signal.

[0169] (Appendix 8) 8. The method according to claim 7, The step of selecting a target physiological signal from the plurality of segments of candidate physiological signals comprises: determining validity of candidate physiological signals of the plurality of segments; The method includes determining the target physiological signal from the candidate physiological signals of the plurality of segments when the candidate physiological signals of the plurality of segments are determined to be valid, and not determining the target physiological signal when the candidate physiological signals of the plurality of segments are determined to be invalid.

[0170] (Appendix 9) 9. The method according to claim 8, comprising: The step of determining validity of the candidate physiological signals of the plurality of segments includes: determining a temporal overlap length of candidate physiological signals of any two segments among the candidate physiological signals of the plurality of segments, and obtaining a candidate physiological signal combination having a temporal overlap length longer than a time of one breath or heartbeat and a first combination number L of the candidate physiological signal combinations; Using time-domain correlation and / or frequency-domain coherence to determine a similarity of the candidate physiological signals of the two segments in each of the candidate physiological signal combinations, and determining a second number K of the candidate physiological signal combinations in which the similarity is greater than a first threshold; and determining that the candidate physiological signals of the plurality of segments are valid if K / L is greater than a second threshold, and determining that the candidate physiological signals of the plurality of segments are invalid if K / L is less than or equal to a second threshold.

[0171] (Appendix 10) 9. The method according to claim 8, comprising: The step of determining a target physiological signal from the candidate physiological signals of the plurality of segments comprises: Evaluating the candidate physiological signals of each segment among the candidate physiological signals of the plurality of segments using time domain and / or frequency domain evaluation indices based on the valid candidate physiological signals of the plurality of segments; and determining the target physiological signal based on time domain and / or frequency domain performance index values ​​of candidate physiological signals for each segment.

[0172] (Appendix 11) 11. The method according to claim 10, The time domain evaluation index is a maximum time domain amplitude of the candidate physiological signal; and / or an average correlation coefficient between the time-domain waveform of the candidate physiological signal and the time-domain waveforms of other candidate physiological signals of the plurality of segments; Includes and / or The frequency domain evaluation index is a maximum frequency spectral energy in a physiological signal band of the candidate physiological signal; and / or a ratio of energy values ​​in physiological signal bands to energy values ​​in non-physiological signal bands of the candidate physiological signal; and / or comprising an average value of frequency spectral coherence coefficients in physiological signal bands of the candidate physiological signal and other candidate physiological signals of the plurality of segments of candidate physiological signals.

[0173] (Appendix 12) 8. The method according to claim 7, The step of selecting a first radar signal from the multi-antenna zero speed range FFT signal and generating an amplitude signature and a phase signature of the first radar signal includes: determining a target location based on time domain information and / or frequency domain information of the multi-antenna zero speed range FFT signal; The target position and the positions of N1 distance frequency points adjacent to the target position are defined as a target region; and selecting the first radar signal within the target region and generating an amplitude signature and a phase signature of the first radar signal.

[0174] (Appendix 13) 13. The method of claim 12, The time domain information includes information on the time-varying amplitude of the multi-antenna zero-speed range FFT signal at each range frequency point within a preset period, and determining the target region includes calculating an amplitude variation amount at each range frequency point within the preset period, and determining the range frequency point corresponding to the maximum value of the amplitude variation amount as the target position; and / or The method, wherein the frequency domain information includes a result of performing a fast Fourier transform on the phase of the multi-antenna zero-speed range FFT signal at each range frequency point, and determining the target position includes determining as the target position a range frequency point corresponding to a maximum value of a sum of energy at each frequency point of a physiological signal band, and / or determining as the target position a range frequency point corresponding to a maximum value of a ratio between the sum of energy at each frequency point of a physiological signal band and the sum of energy at each frequency point of a non-physiological signal band.

[0175] (Appendix 14) 13. The method of claim 12, Selecting the first radar signal within the target region includes: Among the amplitude fluctuation amounts of the multi-antenna zero speed range FFT signal at each range frequency point within the target area, the multi-antenna zero speed range FFT signal corresponding to the maximum N2 amplitude fluctuation amounts is set as the first radar signal; and / or Among the energy values ​​within the physiological signal band of the multi-antenna zero speed range FFT signal at each range frequency point within the target area, the multi-antenna zero speed range FFT signal corresponding to the maximum N2 energy values ​​is the first radar signal; and / or The method includes determining as the first radar signal the multi-antenna zero speed range FFT signal corresponding to the largest N2 ratios between the energy values ​​in the physiological signal band of the multi-antenna zero speed range FFT signal and the energy values ​​in the non-physiological signal band at each range frequency point in the target area.

[0176] (Appendix 15) 13. The method of claim 12, The step of generating amplitude and phase characteristics of the first radar signal comprises: subtracting an average value of the amplitude of the first radar signal from the amplitude at each time point of the first radar signal to generate an amplitude signature of the first radar signal; and / or subtracting a phase value of a previous frame from a phase value of a next frame of the first radar signal to generate a phase signature of the first radar signal.

[0177] (Appendix 16) 8. The method according to claim 7, The detection model is a single-time feature extraction module for obtaining first spatial feature information at each distance frequency point within a single frame time according to the amplitude feature and the phase feature of the first radar signal; a local time feature extraction module for obtaining second spatial feature information at each distance frequency point within a plurality of frames of time according to the first spatial feature information; a global feature extraction module for obtaining global time feature information at each distance frequency point according to the second spatial feature information; and a feature fusion module that fuses the first spatial feature information, the second spatial feature information, and the global temporal feature information to generate the candidate physiological signal.

[0178] (Appendix 17) 17. The method of claim 16, comprising: The method, wherein the feature fusion module cascades the first spatial feature information, the second spatial feature information and the global temporal feature information, or adds the first spatial feature information, the second spatial feature information and the global temporal feature information.

[0179] (Appendix 18) 8. The method according to claim 7, The loss function for training the detection model is the waveform reconstruction loss function Loss re , waveform change loss function Loss changes , or waveform fluctuation loss function Loss v The method of claim 1, wherein the function is generated from at least one of

[0180] (Appendix 19) 19. The method according to claim 18, comprising: The loss function Loss is Loss=Loss re +a*Loss changes or Loss=Loss re +a*Loss v That is, the method.

[0181] (Appendix 20) 19. The method according to claim 18, comprising: The waveform reconstruction loss function Loss re is the mean square error between the waveform values ​​of the actual physiological signal and the waveform values ​​of the physiological signal output by the detection model; and / or The waveform change loss function Loss changes is the mean square error between the waveform change of the actual physiological signal and the waveform change of the physiological signal output by the detection model; and / or The waveform fluctuation loss function Loss v is the mean square error between the waveform variation of an actual physiological signal and the waveform variation of a physiological signal output by the detection model.

[0182] (Appendix 21) An electronic device including a memory storing a computer program and a processor, The electronic device is configured to realize the vital sign detection method described in any one of Supplementary Notes 7 to 20 by executing the computer program.

[0183] (Appendix 22) A storage medium storing a computer program, A storage medium, the computer program being for causing a computer to execute the vital sign detection method according to any one of appendices 7 to 20.

[0184] Although the preferred embodiment of the present invention has been described above, the present invention is not limited to this embodiment, and any modification to the present invention falls within the technical scope of the present invention, provided that the modification does not depart from the spirit of the present invention.

Claims

1. 1. An apparatus for detecting vital signs, comprising: A data acquisition unit for acquiring radar multi-antenna zero-speed range FFT signals; a data pre-processing unit for selecting a first radar signal from the radar multi-antenna zero speed range FFT signal and generating amplitude and phase characteristics of the first radar signal; a generating unit for generating candidate physiological signals of a plurality of segments based on the amplitude features and the phase features and utilizing a detection model based on a deep neural network; a selection unit for selecting a target physiological signal from the candidate physiological signals of the plurality of segments, the candidate physiological signals and the target physiological signal being a respiratory signal or a cardiac signal; and The apparatus includes a detection unit for detecting a vital sign parameter based on the target physiological signal.

2. 2. The apparatus of claim 1, The selection unit is a determining unit for determining validity of the candidate physiological signals of the plurality of segments; and a determination unit that determines the target physiological signal from the candidate physiological signals of the plurality of segments when the determination unit determines that the candidate physiological signals of the plurality of segments are valid, and that does not determine the target physiological signal when the determination unit determines that the candidate physiological signals of the plurality of segments are invalid.

3. 3. The apparatus of claim 2, The determination unit is Obtain a temporal overlap length of candidate physiological signals of any two segments among the candidate physiological signals of the plurality of segments, and obtain a candidate physiological signal combination having a temporal overlap length greater than a time period for one breath or heartbeat, and a first combination number L of the candidate physiological signal combinations; Obtaining a similarity of the candidate physiological signals of the two segments in each of the candidate physiological signal combinations using time domain correlation and / or frequency domain coherence, and determining a second number K of the candidate physiological signal combinations in which the similarity is greater than a first threshold; and The apparatus determines that the candidate physiological signals of the plurality of segments are valid if K / L is greater than a second threshold, and determines that the candidate physiological signals of the plurality of segments are invalid if K / L is less than or equal to the second threshold.

4. 3. The apparatus of claim 2, The determination unit is Evaluating the candidate physiological signals of each segment among the candidate physiological signals of the plurality of segments using time domain and / or frequency domain evaluation indices based on the valid candidate physiological signals of the plurality of segments; and The apparatus determines the target physiological signal based on time domain and / or frequency domain performance index values ​​of candidate physiological signals for each segment.

5. 5. The apparatus of claim 4, The time domain evaluation index is a maximum time domain amplitude of the candidate physiological signal; and / or an average correlation coefficient between the time-domain waveform of the candidate physiological signal and the time-domain waveforms of other candidate physiological signals of the plurality of segments; and / or The frequency domain evaluation index is a maximum frequency spectral energy in a physiological signal band of the candidate physiological signal; and / or and / or the energy values ​​in physiological signal bands of the candidate physiological signals; a ratio of energy values ​​in physiological signal bands to energy values ​​in non-physiological signal bands of the candidate physiological signal; and / or an average value of wavenumber spectral coherence coefficients in a physiological signal band of the candidate physiological signal and other candidate physiological signals among the candidate physiological signals of the plurality of segments; 13. An apparatus comprising:

6. 2. The apparatus of claim 1, The data pre-processing unit includes: a target location determination unit for determining a target location based on time domain information and / or frequency domain information of the radar multi-antenna zero speed range FFT signal; The target position and N adjacent to the target position １ (N １ A target area determination unit that determines the target area as the position of the distance frequency points (≧1); and a radar signal selection unit for selecting the first radar signal at the target area and generating an amplitude signature and a phase signature of the first radar signal.

7. 7. The apparatus of claim 6, The time domain information includes information on the amplitude of the radar multi-antenna zero speed range FFT signal at each range frequency point changing over time within a predetermined period, and the target position determination unit calculates the amplitude fluctuation amount of the radar multi-antenna zero speed range FFT signal at each range frequency point during the predetermined period, and determines the range frequency point corresponding to the maximum amplitude fluctuation amount as the target position; and / or the frequency domain information includes a result of performing an FFT on the phase of the radar multi-antenna zero speed range FFT signal at each range frequency point, and the target position determination unit determines the target position to be a range frequency point corresponding to a maximum value of a sum of energy at each frequency point in a physiological signal band, and / or the target position determination unit determines the target position to be a range frequency point corresponding to a maximum value of a ratio between the sum of energy at each frequency point in a physiological signal band and the sum of energy at each frequency point in a non-physiological signal band.

8. 7. The apparatus of claim 6, The radar signal selection unit selects the maximum N of the amplitude fluctuation amounts of the radar multi-antenna zero-speed range FFT signal at each range frequency point in the target area. ２ the radar multi-antenna zero speed range FFT signal corresponding to the amplitude fluctuation amount is the first radar signal; and / or The radar signal selection unit selects the maximum N of energy values ​​in a physiological signal band of the radar multi-antenna zero-speed range FFT signal at each range frequency point in the target area. ２ the radar multi-antenna zero speed range FFT signal corresponding to the energy values ​​is the first radar signal; and / or The radar signal selection unit selects a maximum N of ratios of energy values ​​in a physiological signal band to energy values ​​in a non-physiological signal band of the radar multi-antenna zero-speed range FFT signal at each range frequency point in the target area. ２ The radar multi-antenna zero speed range FFT signal corresponding to the ratio is the first radar signal.

9. 7. The apparatus of claim 6, the radar signal selection unit subtracting an average value of the amplitude values ​​of the first radar signal from the amplitude values ​​of the first radar signal at each time point, respectively, to generate an amplitude value characteristic of the first radar signal; and / or The radar signal selection unit generates a phase signature of the first radar signal by determining a difference between a phase value of a next frame of the first radar signal and a phase value of a previous frame of the next frame of the first radar signal.

10. 1. A method for detecting vital signs, comprising: Obtain radar multi-antenna zero speed range FFT signal; Selecting a first radar signal from the radar multi-antenna zero speed range FFT signal and generating an amplitude signature and a phase signature of the first radar signal; generating candidate physiological signals of a plurality of segments based on the amplitude features and the phase features and utilizing a deep neural network based detection model; selecting a target physiological signal from the candidate physiological signals of the plurality of segments, the candidate physiological signal and the target physiological signal being a respiratory signal or a cardiac signal; and detecting a vital sign parameter based on the target physiological signal.

Images