Vital sign detection device and method

Abstract

The embodiment of the invention provides a vital sign detection device and method. The method comprises the following steps: acquiring a radar multi-antenna zero-speed distance FFT signal; selecting a first radar signal from the radar multi-antenna zero-speed distance FFT signals, and generating an amplitude feature and a phase feature of the first radar signal; according to the amplitude features and the phase features, multiple segments of candidate physiological signals are generated by using a detection model based on a deep neural network; selecting a target physiological signal from the multiple sections of candidate physiological signals, and detecting vital sign parameters according to the target physiological signal. Therefore, the accuracy of the vital sign detection result can be improved.

Description

Technical Field

[0001] The embodiments of the present application relate to the field of radar detection technology. Background Art

[0002] Physiological indicators such as breathing and heart rate are basic vital signs of living organisms, which can directly reflect the health status of living organisms. Abnormal vital sign parameters can often lead to sudden medical emergencies. Therefore, real-time detection of vital sign information has very important practical value.

[0003] Current vital sign 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 measurement. Wireless signal-based detection methods mainly use radars, such as Doppler radar, ultra-wideband (UWB) radar, frequency modulated continuous wave (FMCW) radar, etc.

[0004] Wearable sensor-based detection methods require the sensors to directly or indirectly contact the human body, which not only limits their application scope, but also has user compliance issues. When users are unwilling to wear or forget to wear the sensor device, these systems cannot detect.

[0005] In the detection method based on wireless signals, the radar detects changes in the vibration amplitude of the heart and / or chest scattering center, extracts physiological signals based on the detection results, and determines vital sign parameters based on the physiological signals.

[0006] It should be noted that the above introduction to the technical background is only for the convenience of providing a clear and complete description of the technical solutions of the present application and for the convenience of understanding by those skilled in the art. It cannot be considered that the above technical solutions are well known to those skilled in the art simply because these solutions are described in the background technology part of the present application. Summary of the invention

[0007] However, the inventors found that in the detection method based on wireless signals, due to the influence of various interference factors, the detection result of the radar may have a large noise, and the physiological signal waveform extracted from the detection result cannot truly reflect the physiological parameters. Therefore, the accuracy of the physiological parameters determined from the physiological signal cannot be guaranteed.

[0008] In response to at least one of the above technical problems, an embodiment of the present application provides a vital sign detection device and method.

[0009] According to one aspect of an embodiment of the present application, a vital sign detection device is provided, including: a data acquisition unit, which acquires a radar multi-antenna zero-speed distance FFT signal output by a radar; a data preprocessing unit, which selects a first radar signal from the radar multi-antenna zero-speed distance FFT signal, and generates an amplitude feature and a phase feature of the first radar signal; a generation unit, which generates a plurality of segments of candidate physiological signals according to the amplitude feature and the phase feature using a detection model based on a deep neural network; a selection unit, which selects a target physiological signal from the plurality of segments of candidate physiological signals, the candidate physiological signals and the target physiological signal being a breathing signal or a heartbeat signal; and a detection unit, which detects vital sign parameters according to the target physiological signal.

[0010] According to another aspect of an embodiment of the present application, a vital sign detection method is provided, including: obtaining 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; generating a plurality of segments of candidate physiological signals according to the amplitude feature and the phase feature using a detection model based on a deep neural network; selecting a target physiological signal from the plurality of segments of candidate physiological signals, the candidate physiological signals and the target physiological signal being a breathing signal or a heartbeat signal, and detecting vital sign parameters according to the target physiological signal.

[0011] One of the beneficial effects of the embodiment of the present application is that: according to the amplitude characteristics and phase characteristics of the first radar signal, a detection model based on a deep neural network is used to generate multiple candidate physiological signals, a target physiological signal is selected from the multiple candidate physiological signals, and a vital sign parameter is detected according to the target physiological signal. Thus, the accuracy of the detection result of the vital sign parameter can be improved.

[0012] With reference to the following description and accompanying drawings, the specific implementation of the embodiment of the present application is disclosed in detail, indicating the way in which the principle of the embodiment of the present application can be adopted. It should be understood that the implementation of the present application is not limited in scope. Within the scope of the spirit and clauses of the attached claims, the implementation of the present application includes many changes, modifications and equivalents. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] The included drawings are used to provide a further understanding of the embodiments of the present application, which constitute a part of the specification, are used to illustrate the implementation methods of the present application, and together with the text description, explain the principles of the present application. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other implementation methods can be obtained based on these drawings without creative work. In the drawings:

[0014] Figure 1 is a schematic diagram of a vital sign detection method according to an embodiment of the present application;

[0015] Figure 2 is a schematic diagram of complex signals at various distance frequency points and various speed frequency points in an embodiment of the present application;

[0016] Figure 3 is a schematic diagram of 102 according to an embodiment of the present application;

[0017] Figure 4 It is a schematic diagram of the range frequency point-time-amplitude / phase of the radar multi-antenna zero-speed range FFT signal in an embodiment of the present application;

[0018] Figure 5 is a schematic diagram of determining a target area in an embodiment of the present application;

[0019] Figure 6 is a schematic diagram of a detection model according to an embodiment of the present application;

[0020] Figure 7 is a schematic diagram of a single time point feature extraction module 601 according to an embodiment of the present application;

[0021] Figure 8 is a schematic diagram of the local time feature extraction module 602 of an embodiment of the present application;

[0022] Fig. 9 is a schematic diagram of the global feature extraction module 603 of an embodiment of the present application;

[0023] Fig.10 is another schematic diagram of the detection model of an embodiment of the present application;

[0024] Fig.11 It is a time domain diagram and a frequency domain diagram of the real physiological signal and the predicted physiological signal of the embodiment of the present application;

[0025] Fig.12 is a schematic diagram of the waveform variation of a real physiological signal according to an embodiment of the present application;

[0026] Fig.13 is a time domain diagram of a real physiological signal and a predicted physiological signal of an embodiment of the present application;

[0027] Fig.14 It is a schematic diagram of the detection model outputting multiple candidate physiological signals;

[0028] Fig.15 is a schematic diagram of 104 according to an embodiment of the present application;

[0029] Fig.16 is a schematic diagram of 1041 of an embodiment of the present application;

[0030] Fig.17 is a schematic diagram of 1042 of an embodiment of the present application;

[0031] Fig.18 is a schematic diagram of a vital sign detection device according to an embodiment of the present application;

[0032] Fig.19 It is a schematic diagram of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0033] With reference to the accompanying drawings, the foregoing and other features of the embodiments of the present application will become apparent through the following description. In the description and the accompanying drawings, specific embodiments of the present application are specifically disclosed, which show some embodiments in which the principles of the embodiments of the present application can be adopted. It should be understood that the present application is not limited to the described embodiments. On the contrary, the embodiments of the present application include all modifications, variations and equivalents that fall within the scope of the attached claims.

[0034] In the embodiments of the present application, the terms "first", "second", etc. are used to distinguish different elements in terms of title, but do not indicate the spatial arrangement or temporal order of these elements, etc., and these elements should not be limited by these terms. The term "and / or" includes any one and all combinations of one or more of the associated listed terms. The terms "comprising", "including", "having", etc. refer to the presence of the stated features, elements, components or components, but do not exclude the presence or addition of one or more other features, elements, components or components.

[0035] In the embodiments of the present application, the singular forms "a", "the", etc. include plural forms and should be broadly understood as "a kind" or "a type" rather than being limited to the meaning of "one"; in addition, the term "said" should be understood to include both singular and plural forms, unless the context clearly indicates otherwise. In addition, the term "according to" should be understood as "at least in part according to...", and the term "based on" should be understood as "at least in part based on...", unless the context clearly indicates otherwise.

[0036] Features described and / or shown for one embodiment may be used in one or more other embodiments in the same or similar manner, combined with features in other embodiments, or replace features in other embodiments. The term "includes / comprising" when used herein refers to the presence of a feature, an entirety, a step, or a component, but does not exclude the presence or addition of one or more other features, entireties, steps, or components.

[0037] In the embodiment of the present application, the radar may be a millimeter wave (mmWave) radar, but is not limited thereto. The radar transmits electromagnetic waves through a transmitting antenna, and receives corresponding reflected waves (which may be referred to as radar echo information or radar echo signals) after being reflected by different objects. By processing and analyzing the radar echo information, information such as physiological signals of the detected target can be extracted, and this information can meet the needs of many application scenarios.

[0038] In the embodiment of the present application, the detection target can be people of various ages, for example, it can be an elderly person, it can also be a child, it can also be an elderly person and / or a caregiver, a child and / or a guardian. The present application is not limited thereto, and the detection target can also be an animal with life characteristics, etc. The following is described using the human body as an example.

[0039] Embodiments of the first aspect

[0040] An embodiment of the present application provides a method for detecting vital signs. Figure 1 is a schematic diagram of a vital sign detection method according to an embodiment of the present application. Figure 1 As shown, the method includes:

[0041] 101, obtaining a radar multi-antenna zero-speed range FFT (Fast Fourier Transform) signal;

[0042] 102, selecting a first radar signal from the radar multi-antenna zero-speed range FFT signal, and generating an amplitude feature and a phase feature of the first radar signal;

[0043] 103, generating multiple candidate physiological signals using a detection model based on a deep neural network according to the amplitude characteristics and the phase characteristics;

[0044] 104, selecting a target physiological signal from a plurality of candidate physiological signals, wherein the candidate physiological signal and the target physiological signal are breathing signals or heartbeat signals, and

[0045] 105, detecting vital sign parameters according to the target physiological signal.

[0046] It is worth noting that the above Figure 1 The embodiments of the present application are only schematically described, but the present application is not limited thereto. For example, the execution order of the various operations can be appropriately adjusted, and other operations can be added or some operations can be reduced. Those skilled in the art can make appropriate modifications based on the above content, and are not limited to the above appended Figure 1 Records of.

[0047] According to the above embodiment, based on the amplitude characteristics and phase characteristics of the first radar signal, a detection model based on a deep neural network is used to generate multiple candidate physiological signals, a target physiological signal is selected from the multiple candidate physiological signals, and a vital sign parameter is detected based on the target physiological signal. Thus, a physiological signal that can more accurately reflect the vital sign parameter can be selected from the multiple candidate physiological signals as the target physiological signal, thereby improving the accuracy of the detection result of the vital sign parameter.

[0048] According to the above embodiment, the detection of vital sign parameters is performed based on the radar multi-antenna zero-speed distance FFT signal. On the one hand, it can reduce the amount of data transmission, improve data transmission efficiency, and facilitate the realization of productized real-time detection; on the other hand, it can make the vital sign detection method applicable to more scenarios.

[0049] For example, the beating process of the human heart is similar to the expansion and contraction of vibration, and its expansion and contraction amplitude is about 0.01 to 0.2 mm. And the rate of the human heartbeat changes periodically within a stable range, so the heartbeat can be approximated to a sinusoidal vibration model. Breathing is completed through the expansion and contraction of the chest cavity, which is also similar to sinusoidal vibration and can also be approximated to a sinusoidal vibration model, and its fluctuation amplitude is about 0.1 to 0.5 mm.

[0050] It can be seen that the change in the vibration amplitude of the heart and chest scattering center is very weak compared to the amplitude of body movement. Therefore, the scenario of body shaking is not considered in some human vital signs detection schemes based on wireless signals. For example, in these schemes, the person being detected is required to be stationary in front of the signal source, or the signal source is installed behind the seat or under the bed, that is, the vital signs are detected when the person being detected is in a static or quasi-static state. Therefore, the installation method and monitoring range of the equipment are limited. Alternatively, only limited shaking scenarios are considered in some schemes. For example, in these schemes, only occasional shaking of the body after being fixed for a period of time is allowed, and the collected radar signals are denoised using traditional signal processing methods (such as filtering or signal decomposition, etc.), and the denoising effect is not ideal.

[0051] In addition, among these methods, the original I / Q signal or Doppler-range spectrum of the radar is mainly used for vital sign detection. The original I / Q signal or Doppler-range spectrum has a large data capacity, a slow real-time transmission speed, and is limited by the SPI port (buffer size), which is not conducive to the realization of real-time monitoring of products.

[0052] In the embodiment of the present application, the radar multi-antenna zero-speed distance FFT signal is used to detect vital sign parameters, that is, by performing Fourier transform in the speed dimension and the distance dimension, the signal at the distance point that is static or quasi-static relative to the radar is extracted for subsequent detection and processing. Therefore, even if there are detected persons who shake significantly relative to the radar within the detection range, the appropriate signal can be accurately extracted from the radar echo signal for detection and processing, without limiting the installation method and monitoring range of the equipment, so that the vital sign detection method can be applied to more scenarios. In addition, by using the radar multi-antenna zero-speed distance FFT signal for detection, the amount of data to be transmitted or processed can be reduced, which helps to improve the data transmission and processing speed.

[0053] In some embodiments, in 101, the radar multi-antenna zero-speed range FFT signal can be obtained in various ways. For example, taking the FMCW radar signal as an example, the received radar echo signal can be subjected to 1D-FFT (one-dimensional fast Fourier transform) and 2D-FFT (two-dimensional fast Fourier transform) to obtain the complex signal at each distance frequency point and each speed frequency point, that is, the Doppler-range spectrum.

[0054] Figure 2 It is a schematic diagram of the complex signals at each distance frequency point and each speed frequency point of an embodiment of the present application, which shows that for a frame of radar echo signal within a preset time period (for example, T frame), after performing distance TTF and speed FFT on the radar echo signal of one antenna among A (A>1) antennas, D*R complex signals are obtained, where D is the number of speed frequencies (also called Doppler frequencies), for example, D is 64 or 128, and R is the number of distance frequencies, for example, D is 128.

[0055] like Figure 2 As shown, the signals at each range frequency point where the Doppler velocity is 0 or the Doppler velocity is less than a preset threshold are selected from the complex signal as zero-speed range FFT signals.

[0056] To select the signal at each distance frequency point where the Doppler velocity is 0 (i.e., Figure 2 (rd 00 ,rd 01 ,…,rd 0(R-1) )) Taking the zero-speed distance FFT signal as an example, compared with selecting all complex signals, the amount of data to be processed can be reduced, thereby reducing the amount of calculation and lowering the requirements for computing resources.

[0057] In some embodiments, in 102, a first radar signal can be selected from the acquired multi-antenna zero-speed distance FFT signal, and the amplitude characteristics and phase characteristics of the first radar signal can be generated, so that a radar signal that can best map the potential change characteristics of human physiological movements can be selected from the multi-antenna zero-speed distance FFT signal, thereby further improving the accuracy of vital sign detection results.

[0058] In some embodiments, the first radar signal may be selected in various ways, and an amplitude feature and a phase feature of the first radar signal may be generated. Figure 3 102 is a schematic diagram of an embodiment of the present application. Figure 3 As shown, 102 may include:

[0059] 301, determining the target position according to the time domain information and / or frequency domain information of the radar multi-antenna zero-speed range FFT signal;

[0060] 302, taking the target position and the positions of N1 frequency points adjacent to the target position as the target area, where N1 ≥ 1; and

[0061] 303, selecting a first radar signal in the target area; and

[0062] 304, generate an amplitude feature and a phase feature of the first radar signal.

[0063] Figure 4 Schematic diagram of the range frequency-time-amplitude / phase of the radar multi-antenna zero-speed range FFT signal of the embodiment of the present application. In some embodiments, Figure 4 As shown, in 301, the range frequency point corresponding to the detection target can be determined from the multiple range frequency points according to the time domain information and / or frequency domain information of the radar multi-antenna zero-speed range FFT signal at the multiple range frequency points as the target position of the detection target.

[0064] In some embodiments, the time domain information may include information on how the amplitude of the radar multi-antenna zero-speed range FFT signal at each range frequency point changes over time within a preset time period. When determining the target position based on the time domain information, the amplitude fluctuation of each range frequency point within the preset time period may be calculated, and the range frequency point corresponding to the maximum amplitude fluctuation may be used as the target position. Among them, the signal amplitude fluctuation may include the variance of the amplitude within the preset time, the standard deviation of the amplitude within the preset time, or the mean of the amplitude change within the preset time (as shown in Formulas 9 and 10 described later), etc.

[0065] In some embodiments, the frequency domain information may include the result of performing FFT on the phase of the radar multi-antenna zero-speed distance FFT signal at each distance frequency point. When determining the target position based on the frequency domain information, the distance frequency point corresponding to the maximum value of the sum of the energy of each frequency point of the radar multi-antenna zero-speed distance FFT signal in the physiological signal frequency band (e.g., the breathing frequency band (e.g., 0.083-0.5 Hz) and / or the heartbeat frequency band (e.g., 0.83-2 Hz)) can be used as the target position.

[0066] The present application is not limited to this. When determining the target position based on frequency domain information, the distance frequency point corresponding to the maximum value of the ratio of the sum of the energy of each frequency point in the physiological signal frequency band of the radar multi-antenna zero-speed distance FFT signal to the sum of the energy of each frequency point in the non-physiological signal frequency band can also be used as the target position.

[0067] For example, the sum of the energies of each frequency point in the breathing frequency band (for example, 0.083 to 0.5 Hz) and the sum of the energies of each frequency point outside the breathing frequency band of the radar multi-antenna zero-speed distance FFT signal can be calculated based on the frequency domain information of each frequency point in the breathing frequency band, and the sum of the energies of each frequency point outside the breathing frequency band, and the distance frequency point corresponding to the maximum value of the ratio between the two is taken as the target position; or, the sum of the energies of each frequency point in the heartbeat frequency band (for example, 0.83 to 2 Hz) and the sum of the energies of each frequency point outside the heartbeat frequency band of the radar multi-antenna zero-speed distance FFT signal are calculated, and the distance frequency point corresponding to the maximum value of the ratio between the two is taken as the target position.

[0068] In some embodiments, after the target position where the detection target is located is determined, the target area can be determined according to the target position. Figure 5 is a schematic diagram of determining a target area in an embodiment of the present application. Figure 5 As shown in the figure, when the rth distance frequency point is selected as the target position, when determining the target area, the rth distance frequency point can be taken as the center and the rth distance frequency point before the distance frequency point can be taken as the center. limit The distance frequency point, the r after the distance frequency point limit The rth distance frequency point and the rth distance frequency point are taken as the target area, wherein, The present application is not limited thereto. In 302 , the target area may not be determined with the target position as the center. For example, the target position may be located further forward or further back in the target area.

[0069] In some embodiments, in 303 , the first radar signal may be selected in the target area in various ways.

[0070] For example, the radar multi-antenna zero-speed range FFT signals corresponding to the largest N2 (N1≥N2≥1) amplitude fluctuations of the amplitude fluctuations of the radar multi-antenna zero-speed range FFT signals at each range frequency point (N1 range frequency points) in the target area can be used as the first radar signal.

[0071] For another example, the radar multi-antenna zero-speed range FFT signals corresponding to the maximum N2 energy values ​​of the radar multi-antenna zero-speed range FFT signals at each range frequency point (N1 range frequency points) in the target area within the physiological signal frequency band can be used as the first radar signal.

[0072] For example, for each of the N1 distance frequency points, the total energy of the radar multi-antenna zero-speed distance FFT signal at each frequency point in the heartbeat frequency band is calculated respectively, and the radar multi-antenna zero-speed distance FFT signal at the N2 distance frequency points with the largest total energy is used as the first radar signal; or, the total energy of the radar multi-antenna zero-speed distance FFT signal at each frequency point in the breathing frequency band is calculated respectively, and the radar multi-antenna zero-speed distance FFT signal at the N2 distance frequency points with the largest total energy is used as the first radar signal.

[0073] For another example, the radar multi-antenna zero-speed range FFT signals corresponding to the largest N2 ratios of the energy values ​​of the radar multi-antenna zero-speed range FFT signals at each range frequency point in the target area in the physiological signal frequency band to the energy values ​​in the non-physiological signal frequency band can be used as the first radar signal.

[0074] For example, for each of the N1 distance frequency points, the total energy of the radar multi-antenna zero-speed distance FFT signal at each frequency point within the heartbeat frequency band and the total energy of each frequency point outside the heartbeat frequency band are calculated respectively, and the radar multi-antenna zero-speed distance FFT signal at the N2 distance frequency points with the largest ratio between the two is used as the first radar signal; or, the total energy of the radar multi-antenna zero-speed distance FFT signal at each frequency point within the breathing frequency band and the total energy of each frequency point outside the breathing frequency band are calculated respectively, and the radar multi-antenna zero-speed distance FFT signal at the N2 distance frequency points with the largest ratio between the two is used as the first radar signal.

[0075] The present application is not limited thereto, and a multi-antenna zero-speed range FFT signal of N1 range frequency points within the target area may also be used as the first radar signal.

[0076] In some embodiments, in 304, after the first radar signal is selected, the first radar signal may be preprocessed to generate an amplitude feature and a phase feature of the first radar signal.

[0077] In some embodiments, the amplitude feature of the first radar signal may be various features related to the amplitude of the first radar signal.

[0078] For example, the amplitude feature of the first radar signal may be the difference between the amplitude of the first radar signal at each time point within a preset time period and the average value of the amplitude of the first radar signal within the time period. Thus, low-frequency interference in the first radar signal can be reduced.

[0079] For example, the amplitude feature of the first radar signal may be generated by subtracting the mean value of the amplitude of the first radar signal from the amplitude of the first radar signal at each time point according to Formula 1:

[0080]

[0081] Among them, A r is the amplitude characteristic of the rth range frequency point within T time, T is the number of frames of the first radar signal, and a is the amplitude of the rth range frequency point within T time.

[0082] The present application is not limited thereto, and the amplitude feature of the first radar signal may also be other quantities. For example, the amplitude feature of the first radar signal may also be the amplitude of the first radar signal at each time point.

[0083] In some embodiments, the phase feature of the first radar signal may be various features related to the phase of the first radar signal.

[0084] For example, the phase characteristic of the first radar signal may be the difference between the phase value of the next frame and the phase value of the previous frame of the first radar signal within a preset time period (i.e., the phase difference). Thus, the interference caused by the slow shaking of the detection target can be eliminated and the signal quality can be improved. For example, the phase value of the next frame of the first radar signal may be subtracted from the phase value of the previous frame according to Formula 2 to generate the phase characteristic of the first radar signal:

[0085] P r =[p t0+1 -p t0 ,…,p j+1 -p j ,…,p t0+T-1 -p t0+T-2 ] (Formula 2)

[0086] Among them, P r is the phase difference of the rth distance frequency point in T time, T is the number of frames of the first radar signal, p j is the phase of the jth frame within the time T of the rth distance frequency point.

[0087] The present application is not limited thereto, and the phase characteristic of the first radar signal may also be other quantities. For example, the phase characteristic of the first radar signal may also be the phase of the first radar signal at each time point.

[0088] In some embodiments, after the amplitude characteristics and phase characteristics of the first radar signal are generated, the amplitude characteristics and phase characteristics can be used as inputs of a detection model based on a deep neural network, thereby using the detection model to generate multiple segments of candidate physiological signals.

[0089] For example, the first radar signal includes radar multi-antenna zero-speed range FFT signals of N1 range frequency points in the target area, and the target area includes the target position (the position of the rth range frequency point), the rth range frequency point before the target position ... target area before the target position, and the target area includes the target position (the position of the rth range frequency point), the target area before the target position, and the target area limit The distance frequency point and the r after the target position limit In the case of a distance frequency point, you can follow

[0090] Formula 3 connects the amplitude feature and the phase feature as the input of the detection model:

[0091]

[0092] For another example, when the first radar signal includes radar multi-antenna zero-speed range FFT signals of N2 range frequency points selected from N1 range frequency points in the target area based on the amplitude fluctuation amount and / or the energy value of the physiological signal frequency band and / or the ratio of the energy values ​​of the physiological signal frequency band to the non-physiological signal frequency band, similarly, the amplitude characteristics and phase characteristics of the first radar signal can also be connected as input to the detection model.

[0093] In some embodiments, in 103, a plurality of candidate physiological signals may be generated according to the amplitude characteristics and phase characteristics of the first radar signal using a detection model based on a deep neural network. The detection model may be a model based on various types of deep neural networks, such as a convolutional neural network, a recurrent neural network, an adversarial neural network, etc.

[0094] In some embodiments, the detection model may include a multi-scale feature extraction module, so that when generating candidate physiological signals, information within different time ranges can be considered, thereby improving the accuracy of the candidate physiological signals. The present application is not limited thereto, and the detection model 600 may also include a single-scale feature extraction module.

[0095] Figure 6 is a schematic diagram of a detection model of an embodiment of the present application. In some embodiments, Figure 6As shown, the detection model 600 may include a single time point feature extraction module 601, a local time feature extraction module 602, and a global feature extraction module 603. The single time point feature extraction module 601 may obtain first spatial feature information at each distance frequency point within a single frame time based on the amplitude feature and phase feature of the first radar signal; the local time feature extraction module 602 may obtain second spatial feature information at each distance frequency point within multiple frames time based on the first spatial feature information; and the global feature extraction module 603 may obtain global time feature information at each distance frequency point based on the second spatial feature information.

[0096] The first spatial feature information may correspond to one frame time, such as 0.1s; the second spatial feature information may correspond to multiple frame times, such as 0.3s or 0.7s; the global time feature information may correspond to a time longer than multiple frame times, such as 10s or 20s.

[0097] Figure 7 601 is a schematic diagram of a single time point feature extraction module 601 in an embodiment of the present application. Figure 7 As shown, the single time point 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 the input of 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 (RectifiedLinear Unit), wherein the kernel size of the one-dimensional convolution layer Conv1d_single may be 1, the step size may be 1, and the number of channels may be N. The present application is not limited thereto, and the single time point feature extraction module 601 may also be other structures.

[0098] Figure 8 602 is a schematic diagram of the local time feature extraction module 602 of the embodiment of the present application. Figure 8As shown, 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 the input of 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. Among them, the kernel size of the one-dimensional convolution layer Conv1d_single may be K1 (K1>1), the step size is 1, and the number of channels is N. The present application is not limited to this, and the local time feature extraction module 602 may also be other structures. In addition, in the detection model 600, multiple local time feature extraction modules 602 may be cascaded, and the processing in each local time feature extraction module 602 may be the same or different to form a more complex local time feature extraction module.

[0099] Fig. 9 603 is a schematic diagram of the global feature extraction module 603 of the embodiment of the present application. Fig. 9 As shown, the global feature extraction module 603 may include a module 6031 and a module 6032. Among them, the module 6031 performs 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. The present application is not limited to this, and the global feature extraction module 603 may also be other structures.

[0100] In some embodiments, Figure 6 As shown, the detection model 600 may further include a feature fusion module 604. 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.

[0101] 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 ways. 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, and so on.

[0102] Fig.10 is another schematic diagram of the detection model of the embodiment of the present application. Fig.10 As shown, the detection model may include a single time point feature extraction module 601, a global feature extraction module 603 and multiple local time feature extraction modules 602. For the specific method of generating output Y according to input X by the detection model, reference may be made to related technologies, which will not be described in detail here.

[0103] In some embodiments, before using the detection module to generate candidate physiological signals, the detection module may be trained using a loss function. The loss function used to train the detection model may be a function generated according to at least one of the following: waveform reconstruction loss function Loss re , waveform change loss function Loss changes Or waveform fluctuation loss function Loss v The loss function Loss can be expressed as follows:

[0104] Loss = a*Loss re +b*Loss changes +c*Loss v (Formula 4)

[0105] Where a, b, and c are real numbers.

[0106] In some embodiments, the waveform reconstruction loss function Loss re It can be the mean square error between the waveform value of the predicted physiological signal output by the detection model and the waveform value of the real physiological signal over a period of time:

[0107]

[0108] Among them, y t is the waveform value of the real physiological signal, y t ′ is the waveform value of the predicted physiological signal output by the detection model.

[0109] The present application is not limited thereto. The waveform reconstruction loss function Loss re It can also be in other forms. For example, the waveform reconstruction loss function Loss re It can be the mean square error between the waveform value of the predicted physiological signal output by the detection model over a period of time and the inverse value of the waveform value of the real physiological signal:

[0110]

[0111] Among them, 1-y t It is the inverse value of the waveform value of the real physiological signal.

[0112] Alternatively, the waveform reconstruction loss function Loss re It can also be the mean square error Loss between the waveform value of the predicted physiological signal output by the detection model and the waveform value of the real physiological signal over a period of time. mse , the mean square error Loss between the waveform value of the physiological signal output by the detection model and the inverse value of the waveform value of the real physiological signal over a period of time inverse The smaller of the two.

[0113] By including the above waveform reconstruction loss function Loss in the loss function re , can ignore the representation of physiological signal state changes. For example, for breathing signals, it is possible to ignore the tracking of inhalation or exhalation states and use Loss mse and Loss inverse The smaller value in obtains the optimal physiological parameter value. Fig.11 ] are time domain diagrams and frequency domain diagrams of the real physiological signal and the predicted physiological signal of the embodiment of the present application. Fig.11 As shown, the peaks and troughs of the predicted waveform 1001 and the true waveform 1002 are opposite, but their frequency values ​​are consistent (see the spectrum diagram 1003 of the predicted waveform 1001 and the spectrum diagram 1004 of the true waveform 1002). In this case, it can be considered that the predicted waveform 1001 output by the detection model is the correct result, so its loss function value is small.

[0114] In some embodiments, the waveform change loss function Loss changes It can be the mean square error between the waveform change of the real physiological signal and the waveform change of the predicted physiological signal output by the detection model:

[0115]

[0116] Among them, y t is the waveform value of the real physiological signal, y t ′ is the waveform value of the predicted physiological signal, y t -y t-k is the waveform change of the real physiological signal, y t ′-y t ' -k To predict the waveform change of physiological signals.

[0117] Fig.12 FIG. 1 is a schematic diagram of the waveform variation of a real physiological signal according to an embodiment of the present application. Fig.12 As shown, the time domain waveform value at the first moment t minus the waveform value at the second moment tk is used to obtain the change in the time domain waveform of the physiological signal. The value of k may be related to the period or frequency of the physiological signal. For example, the larger the frequency of the physiological signal, the smaller the value of k is set.

[0118] By including the above waveform change loss function Loss in the loss function changes , can pay more attention to the transition of physiological signal states. For example, for respiratory signals, the waveform change loss function pays more attention to the state transition process from inhalation to exhalation or from exhalation to inhalation, which is conducive to more accurate tracking of the characteristics of physiological signals.

[0119] In some embodiments, the waveform fluctuation momentum loss function Loss v It can be the mean square error between the waveform fluctuation of the real physiological signal and the waveform fluctuation of the physiological signal output by the detection model:

[0120]

[0121] Among them, y t is the waveform value of the real physiological signal, y t ′ is the waveform value of the predicted physiological signal.

[0122] The waveform fluctuation amount of the real physiological signal may be the average value of the waveform variation amount of the time domain waveform of the real physiological signal within a preset time period:

[0123]

[0124] The waveform fluctuation amount of the predicted physiological signal may be the average value of the waveform variation amount of the time domain waveform of the predicted physiological signal within a preset time period:

[0125]

[0126] By including the above waveform fluctuation momentum loss function Loss in the loss function v , we can pay more attention to the accuracy of the frequency of physiological signals and ignore the translation of the waveform. Fig.13 is a time domain diagram of the real physiological signal and the predicted physiological signal of the embodiment of the present application. Fig.13 As shown in (a), the predicted physiological waveform 1201 fails to completely track the real physiological waveform 1202, but the frequency value of the predicted physiological waveform 1201 is consistent with the frequency value of the real physiological waveform 1202. In this case, it can be considered that the predicted physiological waveform 1201 output by the detection model is the correct result, so its loss function value is small. Fig.13 As shown in (b), the frequency value of the predicted physiological waveform 1203 is significantly different from the frequency value of the real physiological waveform 1204, and the loss function value is relatively large.

[0127] In some embodiments, in 104, a target physiological signal may be selected from multiple segments of candidate physiological signals generated by the detection model. The candidate physiological signal output by the detection model each time may be used as a segment of candidate physiological signal. The candidate physiological signal may be a breathing signal or a heartbeat signal, and the target physiological signal may also be a breathing signal or a heartbeat signal. The breathing signal may be various signals related to breathing, and the heartbeat signal may be various signals related to heartbeat.

[0128] Fig.14 It is a schematic diagram of the detection model outputting multiple candidate physiological signals. Fig.14As shown, the detection model can output a candidate physiological signal Y each time. m , this section of candidate physiological signal Y m It may include multiple data frames. For example, a segment of candidate physiological signals includes 100 frames, each frame lasting 0.1s. The detection model may output a segment of candidate physiological signals at a preset interval. For example, a segment of candidate physiological signals is output at a 1-frame interval. After storing a preset number of candidate physiological signals (Y1…Y m …Y M )After that, one or more segments can be selected from these candidate physiological signals as target physiological signals.

[0129] Fig.15 104 is a schematic diagram of an embodiment of the present application. Fig.15 As shown, 104 may include:

[0130] 1041, determining the validity of multiple candidate physiological signals;

[0131] 1042, when the multiple segments of candidate physiological signals are determined to be valid, a target physiological signal is determined from the multiple segments of candidate physiological signals; when the multiple segments of candidate physiological signals are determined to be invalid, the target physiological signal is not determined.

[0132] In some embodiments, whether the multiple segments of candidate physiological signals are valid can be determined based on the similarity between the multiple segments of candidate physiological signals. For example, when the similarity between the multiple segments of candidate physiological signals is relatively high, the multiple segments of candidate physiological signals are considered valid, and when the similarity between the multiple segments of candidate physiological signals is relatively low, the multiple segments of candidate physiological signals are considered invalid. Only when the multiple segments of candidate physiological signals are valid, the target physiological signal is further determined based on the multiple segments of candidate physiological signals, thereby improving the reliability of the determined target.

[0133] In some embodiments, the similarity between the multiple candidate physiological signals can be represented in various ways. For example, the similarity between the multiple candidate physiological signals can be determined based on the similarity of a combination of two candidate physiological signals in the multiple candidate physiological signals.

[0134] Fig.16 1041 is a schematic diagram of an embodiment of the present application. Fig.16 As shown, 1041 may include:

[0135] 10411, determining the time overlap length of any two candidate physiological signals among the multiple candidate physiological signals, obtaining a candidate physiological signal combination whose time overlap length is greater than one breathing or heartbeat time and a first combination number L of candidate physiological signal combinations;

[0136] 10412, using time domain correlation and / or frequency domain coherence to determine the similarity between two segments of candidate physiological signals in each candidate physiological signal combination, and determining a second number K of candidate physiological signal combinations whose similarity is greater than a first threshold;

[0137] 10413, when the ratio K / L is greater than the second threshold, it is determined that the multiple segments of candidate physiological signals are valid; when the ratio K / L is less than or equal to the second threshold, it is determined that the multiple segments of candidate physiological signals are invalid.

[0138] Below, taking the candidate physiological signal as a breathing signal and the number of candidate physiological signals M as 10 as an example, the validity of determining multiple segments of candidate physiological signals is exemplified. m …Y 10 ) After that, these 10 candidate physiological signals are combined in pairs to obtain Combination (Y i ,Y j ), where 1≤i≤10, 1≤j≤10, i≠j. For each of the 45 combinations, determine the two candidate physiological signals Y in the combination. i With Y j The overlap length in the time domain determines the number of combinations L (L≤45) whose overlap length is greater than one breathing time. For combinations whose overlap length is greater than one breathing time, calculate the two candidate physiological signals Y in the combination. i With Y j The calculated similarity of the candidate physiological signal combination is compared with the first threshold, and the number K of the combinations whose similarity is greater than the first threshold is determined. When the ratio of the number K to the number L is greater than the second threshold, the 10 candidate physiological signals can be determined to be valid; when the ratio of the number K to the number L is less than or equal to the second threshold, the 10 candidate physiological signals can be determined to be invalid.

[0139] In some embodiments, the two candidate physiological signals Y in the candidate physiological signal combination can be determined in various ways. i With Y j For example, the similarity is determined by using time domain correlation. The time domain correlation can be represented by the time domain correlation coefficient ρ(T), and the value range of ρ(T) is [-1,1]. The closer ρ(T) is to 1, the higher the time domain similarity of the two time series is:

[0140]

[0141] Among them, u i Y i The mean value of u j Y jThe mean value, σ i Y i The variance, σ j Y j The variance of .

[0142] For another example, the similarity is determined by using frequency domain coherence. The frequency domain coherence can be represented by the spectral coherence coefficient C(f), the value range of C(f) is [0,1], and the closer C(f) is to 1, the higher the frequency domain similarity of the two candidate physiological signals is:

[0143]

[0144] Among them, Y i (f) is Y i Spectrum information, Y j (f) is Y j spectral information.

[0145] For another example, the similarity can also be determined using both time domain correlation and frequency domain coherence. For example, the first threshold may include a first value and a second value, and when the time domain correlation coefficient ρ(T) is greater than the first value, and the spectral coherence coefficient C(f) is greater than the second value, it is considered that the frequency domain similarity of the two candidate physiological signals is greater than the first threshold.

[0146] In some embodiments, in 1042 , when multiple segments of candidate physiological signals are valid, one or more segments of target physiological signals may be selected from the multiple segments of candidate physiological signals according to a time domain evaluation index and / or a frequency domain evaluation index.

[0147] Fig.17 1042 is a schematic diagram of an embodiment of the present application. Fig.17 As shown, 1042 may include:

[0148] 10421, based on the valid multiple candidate physiological signals, evaluate each candidate physiological signal in the multiple candidate physiological signals using a time domain evaluation index and / or a frequency domain evaluation index; and

[0149] 10422. Determine a target physiological signal according to the time domain evaluation index value and / or the frequency domain evaluation index value of each segment of the candidate physiological signal.

[0150] In some embodiments, the time domain evaluation index may include: the maximum value of the time domain amplitude of the candidate physiological signal. For example, multiple candidate physiological signals are sorted from large to small according to the maximum value of the time domain amplitude, and the candidate physiological signal with the largest maximum value of the time domain amplitude is selected as the target physiological signal, or multiple candidate physiological signals with the highest maximum value of the time domain amplitude are selected as the target physiological signal.

[0151] 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 in the multiple candidate physiological signals. Thus, one or more candidate physiological signals that best represent the multiple candidate physiological signals can be used as target physiological signals.

[0152] For example, for each of the multiple candidate physiological signals, the time domain correlation coefficients between the candidate physiological signal and other candidate physiological signals are calculated, and the average value of the time domain correlation coefficients is calculated as the time domain evaluation index value of the candidate physiological signal. The calculated time domain evaluation index values ​​of each candidate physiological signal are sorted from large to small, and the candidate physiological signal with the largest time domain evaluation index value is selected as the target physiological signal, or multiple candidate physiological signals with top ranked time domain evaluation index values ​​are selected as the target physiological signal.

[0153] In some embodiments, the frequency domain evaluation index may include: the maximum value of the spectrum energy of the candidate physiological signal in the physiological signal frequency band. For example, perform FFT on multiple candidate physiological signals, calculate the spectrum energy of each candidate physiological signal at each frequency point in the respiratory frequency band or the heartbeat frequency band, determine the maximum value of the spectrum energy, sort the spectrum energy from large to small, and select the candidate physiological signal with the largest maximum value of the spectrum energy as the target physiological signal, or select multiple candidate physiological signals with the highest maximum value of the spectrum energy as the target physiological signal.

[0154] In some embodiments, the frequency domain evaluation index may include: the energy value of the candidate physiological signal in the physiological signal frequency band. For example, perform FFT on multiple candidate physiological signals, calculate the sum of the energy of each frequency point in the respiratory frequency band or the heartbeat frequency band of each candidate physiological signal, and use it as the energy value of each candidate physiological signal in the respiratory frequency band or the heartbeat frequency band, sort them in descending order according to the energy value, and select the candidate physiological signal with the largest energy value as the target physiological signal, or select multiple candidate physiological signals with the highest energy value as the target physiological signal.

[0155] In some embodiments, the frequency domain evaluation index may include: the ratio of the energy value of the candidate physiological signal in the physiological frequency band to the energy value in the non-physiological frequency band. For example, perform FFT on multiple segments of candidate physiological signals, calculate the ratio of the energy value of each segment of candidate physiological signals in the respiratory frequency band to the energy value in the non-respiratory frequency band, or the ratio of the energy value in the heartbeat frequency band to the energy value in the non-heartbeat frequency band, sort them in descending order according to the energy value ratio, select the segment of candidate physiological signal with the largest energy value ratio as the target physiological signal, or select multiple segments of candidate physiological signals with the highest energy value ratio as the target physiological signal.

[0156] In some embodiments, the frequency domain evaluation index may include: an average value of the spectral coherence coefficients of the candidate physiological signal and other candidate physiological signals in the multiple candidate physiological signals in the physiological frequency band. Thus, one or more candidate physiological signals that best represent the multiple candidate physiological signals can be used as target physiological signals.

[0157] For example, perform FFT on each of the multiple candidate physiological signals to obtain the spectrum information of each candidate physiological signal; calculate the spectrum coherence coefficients between the candidate physiological signal and other candidate physiological signals based on the spectrum information, and calculate the average value of these spectrum coherence coefficients as the frequency domain evaluation index value of the candidate physiological signal. For the calculated frequency domain evaluation index value of each candidate physiological signal, sort them from large to small, and select the candidate physiological signal with the largest frequency domain evaluation index value as the target physiological signal, or select multiple candidate physiological signals with the highest frequency domain evaluation index value as the target physiological signal.

[0158] In some embodiments, in 104, when selecting a target physiological signal from multiple candidate physiological signals, one segment of the candidate physiological signal can be selected as the target physiological signal, that is, the target physiological signal includes one segment of the signal; or more than two segments of the candidate physiological signal can be selected as the target physiological signal, that is, the target physiological signal includes multiple segments of the signal.

[0159] In some embodiments, the target physiological signal may be selected based on a time domain evaluation index or a frequency domain evaluation index as described above.

[0160] For example, a candidate physiological signal segment with the largest time domain evaluation index value or frequency domain evaluation index value may be selected as the target physiological signal. For example, a candidate physiological signal segment with the largest maximum time domain amplitude value may be selected as the target physiological signal.

[0161] For another example, more than two candidate physiological signals with the time domain evaluation index value or the frequency domain evaluation index value ranked top in descending order may be selected as target physiological signals. For example, the candidate physiological signals are ranked from large to small according to the maximum value of the time domain amplitude, and the top two candidate physiological signals are selected as target physiological signals.

[0162] In some embodiments, the target physiological signal may be selected according to a plurality of evaluation indicators, wherein the plurality of evaluation indicators may include a plurality of time domain evaluation indicators, or a plurality of frequency domain evaluation indicators, or one or more time domain evaluation indicators and one or more frequency domain evaluation indicators.

[0163] In some embodiments, when selecting a target physiological signal based on multiple evaluation indicators, one or more segments of candidate physiological signals can be selected according to each of the multiple evaluation indicators as a set of candidate physiological signals corresponding to the evaluation indicator, and the target physiological signal can be determined based on the set of candidate physiological signals corresponding to each evaluation indicator.

[0164] In some embodiments, when determining the target physiological signal based on the sets of candidate physiological signals corresponding to various evaluation indicators, the sum of the sets can be used as the target physiological signal; or, the union of the sets can be used as the target physiological signal; or, the target physiological signal can be determined based on the number of times the candidate physiological signal appears in each set, and / or the priority of each evaluation indicator, and / or the priority of the candidate physiological signals in the same set (for example, the larger the evaluation indicator value of the candidate physiological signal, the higher the priority of the candidate physiological signal), etc.

[0165] The following is an example of selecting one target physiological signal (i.e., the target physiological signal includes one segment of signal) from ten candidate physiological signals according to indicator 1, indicator 2, and indicator 3, to illustrate the above selection method. Indicator 1, indicator 2, and indicator 3 may be any time domain or frequency domain evaluation indicator as described above.

[0166] Among the 10 candidate physiological signals, according to index 1, the candidate physiological signal with the largest value of index 1 is selected, for example, the 7th candidate physiological signal; according to index 2, the candidate physiological signal with the largest value of index 2 is selected, for example, the 1st candidate physiological signal; according to index 3, the candidate physiological signal with the largest value of index 3 is selected, for example, the 1st candidate physiological signal. Since the 1st candidate physiological signal is selected twice, the 1st candidate physiological signal can be used as the target physiological signal.

[0167] Alternatively, when the candidate physiological signals selected according to indicators 1, 2, and 3 are all different, for example, the 7th, 1st, and 10th segment physiological signals are selected according to indicators 1, 2, and 3, respectively, the target physiological signal can be determined according to the indicator with the highest priority, for example, indicator 1, that is, the 7th segment candidate physiological signal can be used as the target physiological signal.

[0168] In some embodiments, in 105, vital sign parameters are detected according to the target physiological signal. When the target physiological signal selected from multiple candidate physiological signals is a segment signal, the heart rate or respiratory rate can be determined based on the frequency of the target physiological signal. For example, the target physiological signal is subjected to FFT to obtain the spectrum information of the target physiological signal, and based on the spectrum information, the frequency of the frequency point with the maximum spectrum energy is used as the heart rate or respiratory rate; or, based on the spectrum information, within the range of the heart rate or respiratory rate, the frequency of each frequency point is weighted using the spectrum energy of each frequency point, and the result of the weighted sum is used as the heart rate or respiratory rate. The present application is not limited to this, and the heart rate or respiratory rate can also be determined in other ways.

[0169] In some embodiments, when the target physiological signal selected from multiple candidate physiological signals includes multiple signal segments, the heart rate or respiratory rate can be determined based on the frequency of each target physiological signal in the multiple target physiological signals. The frequency of each target physiological signal can be determined in the aforementioned manner, which will not be described in detail here. After obtaining the frequency of each target physiological signal, the average value of the frequency of each target physiological signal can be used as the heart rate or respiratory rate. Alternatively, the frequency of each target physiological signal can be weighted, and the result of the weighted summation can be used as the heart rate or respiratory rate. The weighting coefficient of the frequency of each target physiological signal can be the energy value of the target physiological signal in the physiological frequency band, etc.

[0170] It is worth noting that the above drawings only schematically illustrate the embodiments of the present application, but the present application is not limited thereto. For example, the execution order between the various operations can be appropriately adjusted, and other operations can be added or some operations can be reduced. Those skilled in the art can make appropriate modifications based on the above content, and are not limited to the records of the above drawings.

[0171] The above only describes the steps or processes related to the present application, but the present application is not limited thereto. The motion detection method may also include other steps or processes, and the specific contents of these steps or processes may refer to the prior art. In addition, the above only uses some structures of the vital sign detection model as an example to illustrate the embodiments of the present application, but the present application is not limited to these structures, and these structures may be appropriately modified, and the implementation methods of these modifications should all be included in the scope of the embodiments of the present application.

[0172] The above embodiments are merely exemplary of the embodiments of the present application, but the present application is not limited thereto, and appropriate modifications may be made based on the above embodiments. For example, the above embodiments may be used alone, or one or more of the above embodiments may be combined.

[0173] According to the above embodiment, based on the amplitude characteristics and phase characteristics of the first radar signal, a detection model based on a deep neural network is used to generate multiple candidate physiological signals, a target physiological signal is selected from the multiple candidate physiological signals, and a vital sign parameter is detected based on the target physiological signal. Thus, a physiological signal that can more accurately reflect the vital sign parameter can be selected from the multiple candidate physiological signals as the target physiological signal, thereby improving the accuracy of the detection result of the vital sign parameter.

[0174] According to the above embodiment, the detection of vital sign parameters is performed based on the radar multi-antenna zero-speed distance FFT signal. On the one hand, it can reduce the amount of data transmission, improve data transmission efficiency, and facilitate the realization of productized real-time detection; on the other hand, it can make the vital sign detection method applicable to more scenarios.

[0175] Embodiments of the second aspect

[0176] The embodiment of the present application provides a vital sign detection device, and the contents that are the same as those in the embodiment of the first aspect are not repeated here.

[0177] Fig.18 is a schematic diagram of a vital sign detection device according to an embodiment of the present application. Fig.18 As shown, the vital sign detection device 1800 includes:

[0178] A data acquisition unit 1801, which acquires a radar multi-antenna zero-speed range FFT signal;

[0179] A data preprocessing unit 1802, which selects a first radar signal from the multi-antenna zero-speed range FFT signal, and generates an amplitude feature and a phase feature of the first radar signal;

[0180] A generating unit 1803, which generates multiple candidate physiological signals according to the amplitude feature and the phase feature by using a detection model based on a deep neural network;

[0181] A selection unit 1804 is configured to select a target physiological signal from the plurality of candidate physiological signals, wherein the candidate physiological signals and the target physiological signal are respiratory signals or heartbeat signals, and

[0182] The detection unit 1805 detects vital sign parameters according to the target physiological signal.

[0183] In some embodiments, the selection unit 1804 includes:

[0184] a determination unit, configured to determine the validity of the plurality of candidate physiological signals;

[0185] A determination unit determines the target physiological signal from the multiple segments of candidate physiological signals when the determination unit determines that the multiple segments of candidate physiological signals are valid, and does not determine the target physiological signal when the determination unit determines that the multiple segments of candidate physiological signals are invalid.

[0186] In some embodiments, the determination unit determines the time overlap length of any two candidate physiological signals among the multiple candidate physiological signals, and obtains a candidate physiological signal combination whose time overlap length is greater than one breath or heartbeat time and a first combination number L of the candidate physiological signal combination;

[0187] Determine the similarity between two segments of candidate physiological signals in each candidate physiological signal combination by using time domain correlation and / or frequency domain coherence, and determine a second number K of combinations of candidate physiological signal combinations whose similarity is greater than a first threshold;

[0188] When the ratio K / L is greater than the second threshold, it is determined that the multiple segments of candidate physiological signals are valid; when the ratio K / L is less than or equal to the second threshold, it is determined that the multiple segments of candidate physiological signals are invalid.

[0189] In some embodiments, the determining unit evaluates each of the multiple candidate physiological signals using a time domain evaluation index and / or a frequency domain evaluation index according to the multiple valid candidate physiological signals;

[0190] The target physiological signal is determined according to the time domain evaluation index value and / or the frequency domain evaluation index value of each segment of the candidate physiological signal.

[0191] In some embodiments, the time domain evaluation index includes:

[0192] the maximum value of the time domain amplitude of the candidate physiological signal; and / or

[0193] 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 in the plurality of segments of candidate physiological signals;

[0194] and / or

[0195] The frequency domain evaluation indicators include:

[0196] the maximum value of the spectrum energy of the candidate physiological signal in the physiological signal frequency band; and / or

[0197] The energy value of the candidate physiological signal in the physiological signal frequency band; and / or

[0198] The ratio of the energy value of the candidate physiological signal in the physiological signal frequency band to the energy value of the candidate physiological signal in the non-physiological signal frequency band; and / or

[0199] an average value of spectral coherence coefficients of the candidate physiological signal and other candidate physiological signals in the multiple segments of candidate physiological signals in a physiological signal frequency band.

[0200] In some embodiments, the data preprocessing unit 1802 includes:

[0201] a target position determination unit, which determines the target position according to the time domain information and / or frequency domain information of the multi-antenna zero-speed range FFT signal;

[0202] a target area determination unit, which takes the target position and positions of N1 frequency points adjacent to the target position as the target area; and

[0203] A radar signal selection unit selects the first radar signal in the target area and generates an amplitude feature and a phase feature of the first radar signal.

[0204] In some embodiments, the time domain information includes information on how the amplitude of the multi-antenna zero-speed range FFT signal at each distance frequency point changes over time within a preset time period, and the target position determination unit calculates the amplitude fluctuation of each distance frequency point within the preset time period, and takes the distance frequency point corresponding to the maximum amplitude fluctuation as the target position; and / or

[0205] The frequency domain information includes the result of fast Fourier transform on the phase of the multi-antenna zero-speed distance FFT signal at each distance frequency point, and the target position determination unit takes the distance frequency point corresponding to the maximum value of the sum of the energies of each frequency point in the physiological signal frequency band as the target position; and / or, the target position determination unit takes the distance frequency point corresponding to the maximum value of the ratio of the sum of the energies of each frequency point in the physiological signal frequency band to the sum of the energies of each frequency point in the non-physiological signal frequency band as the target position.

[0206] In some embodiments, the radar signal selection unit uses the multi-antenna zero-speed range FFT signal corresponding to the largest N2 amplitude fluctuations among the amplitude fluctuations of the multi-antenna zero-speed range FFT signal at each range frequency point in the target area as the first radar signal; and / or

[0207] The radar signal selection unit uses the multi-antenna zero-speed range FFT signals corresponding to the maximum N2 energy values ​​of the multi-antenna zero-speed range FFT signals at each range frequency point in the target area within the physiological signal frequency band as the first radar signal; and / or

[0208] The radar signal selection unit uses the multi-antenna zero-speed range FFT signals corresponding to the largest N2 ratios of the energy values ​​of the multi-antenna zero-speed range FFT signals at each range frequency point in the target area in the physiological signal frequency band to the energy values ​​in the non-physiological signal frequency band as the first radar signal.

[0209] In some embodiments, the radar signal selection unit subtracts the average of the amplitudes of the first radar signal from the amplitudes of the first radar signal at each time point to generate an amplitude feature of the first radar signal; and / or

[0210] The radar signal selection unit subtracts a phase value of a subsequent frame from a phase value of a previous frame of the first radar signal to generate a phase feature of the first radar signal.

[0211] In some embodiments, the detection model includes:

[0212] A single time point feature extraction module, which obtains first spatial feature information at each range frequency point within a single frame time based on the amplitude feature and phase feature of the first radar signal;

[0213] A local time feature extraction module, which obtains second spatial feature information at each distance frequency point within a multi-frame time according to the first spatial feature information;

[0214] a global feature extraction module, which obtains global time feature information at each distance frequency point according to the second spatial feature information; and

[0215] A feature fusion module is configured to fuse the first spatial feature information, the second spatial feature information and the global temporal feature information to generate the candidate physiological signal.

[0216] 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.

[0217] In some embodiments, the loss function used to train the detection model is a function generated according to at least one of the following: waveform reconstruction loss function Loss re , waveform change loss function Loss changes Or waveform fluctuation loss function Loss v .

[0218] In some embodiments, the loss function Loss is in the following form:

[0219] Loss = a*Loss re+b*Loss changes +c*Loss v , where a, b, and c are real numbers.

[0220] In some embodiments, the waveform reconstruction loss function Loss re is the mean square error between the waveform value of the real physiological signal and the waveform value of the physiological signal output by the detection model; and / or

[0221] The waveform change loss function Loss changes is the mean square error between the waveform variation of the real physiological signal and the waveform variation of the physiological signal output by the detection model; and / or

[0222] The waveform fluctuation momentum loss function Loss v It is the mean square error between the waveform fluctuation amount of the real physiological signal and the waveform fluctuation amount of the physiological signal output by the detection model.

[0223] It is worth noting that the above only describes the components or modules related to the present application, but the present application is not limited thereto. The vital sign detection device 1800 may also include other components or modules, and the specific contents of these components or modules may refer to the relevant technology.

[0224] To keep it simple, Fig.18 The connection relationship or signal direction between various components or modules is only exemplified, but it should be clear to those skilled in the art that various related technologies such as bus connection can be used. The above-mentioned components or modules can be implemented by hardware facilities such as processors and memories; the embodiments of the present application are not limited to this.

[0225] The above embodiments are merely exemplary of the embodiments of the present application, but the present application is not limited thereto, and appropriate modifications may be made based on the above embodiments. For example, the above embodiments may be used alone, or one or more of the above embodiments may be combined.

[0226] According to the above embodiment, based on the amplitude characteristics and phase characteristics of the first radar signal, a detection model based on a deep neural network is used to generate multiple candidate physiological signals, a target physiological signal is selected from the multiple candidate physiological signals, and a vital sign parameter is detected based on the target physiological signal. Thus, a physiological signal that can more accurately reflect the vital sign parameter can be selected from the multiple candidate physiological signals as the target physiological signal, thereby improving the accuracy of the detection result of the vital sign parameter.

[0227] According to the above embodiment, the detection of vital sign parameters is performed based on the radar multi-antenna zero-speed distance FFT signal. On the one hand, it can reduce the amount of data transmission, improve data transmission efficiency, and facilitate the realization of productized real-time detection; on the other hand, it can make the vital sign detection method applicable to more scenarios.

[0228] Embodiments of the third aspect

[0229] The embodiment of the present application provides an electronic device, including the vital sign detection device 1800 as described in the embodiment of the second aspect, the contents of which are incorporated herein. The electronic device can be, for example, a computer, a server, a workstation, a laptop computer, a smart phone, etc., but the embodiment of the present application is not limited thereto.

[0230] Fig.19 Schematic diagram of an electronic device according to an embodiment of the present application. Fig.19 As shown, the electronic device 1900 may include: a processor (e.g., a central processing unit CPU) 1910 and a memory 1920; the memory 1920 is coupled to the central processing unit 1910. The memory 1920 may store various data; in addition, it may store information processing programs and execute the programs under the control of the processor 1910.

[0231] In some embodiments, the functions of the vital sign detection device 1800 are integrated into the processor 1910. The processor 1910 is configured to implement the vital sign detection method as described in the embodiment of the first aspect.

[0232] In some embodiments, the vital sign detection device 1800 is configured separately from the processor 1910 . For example, the vital sign detection device 1800 can be configured as a chip connected to the processor 1910 , and the functions of the vital sign detection device 1800 are realized through the control of the processor 1910 .

[0233] For example, processor 1910 is configured to perform the following control: obtain a radar multi-antenna zero-speed distance FFT signal; select a first radar signal from the multi-antenna zero-speed distance FFT signal, and generate an amplitude feature and a phase feature of the first radar signal; generate multiple segments of candidate physiological signals based on the amplitude feature and the phase feature using a detection model based on a deep neural network; select a target physiological signal from the multiple segments of candidate physiological signals, and detect vital sign parameters based on the target physiological signal.

[0234] In addition, if Fig.19 As shown, the electronic device 1900 may also include: an input / output (I / O) device 1930 and a display 1940, etc.; wherein the functions of the above components are similar to those of the prior art and are not described in detail here. It is worth noting that the electronic device 1900 does not necessarily have to include Fig.19 In addition, the electronic device 1900 may also include Fig.19 For components not shown in the figure, reference may be made to the related art.

[0235] An embodiment of the present application also provides a computer-readable program, wherein when the program is executed in an electronic device, the program enables a computer to execute the vital sign detection method as described in the embodiment of the first aspect in the electronic device.

[0236] An embodiment of the present application also provides a storage medium storing a computer-readable program, wherein the computer-readable program enables a computer to execute the vital sign detection method as described in the embodiment of the first aspect in an electronic device.

[0237] The above devices and methods of the present application can be implemented by hardware, or by hardware combined with software. The present application relates to such a computer-readable program, which, when executed by a logic component, enables the logic component to implement the above-mentioned devices or components, or enables the logic component to implement the various methods or steps described above. The present application also relates to a storage medium for storing the above program, such as a hard disk, a magnetic disk, an optical disk, a DVD, a flash memory, etc.

[0238] The method / device described in conjunction with the embodiments of the present application may be directly embodied as hardware, a software module executed by a processor, or a combination of the two. For example, one or more of the functional block diagrams shown in the figure and / or one or more combinations of the functional block diagrams may correspond to various software modules of the computer program flow or to various hardware modules. These software modules may correspond to the various steps shown in the figure, respectively. These hardware modules may be implemented by solidifying these software modules, for example, using a field programmable gate array (FPGA).

[0239] The software module may be located in a RAM memory, a flash memory, a ROM memory, an EPROM memory, an EEPROM memory, a register, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art. A storage medium may be coupled to a processor so that the processor can read information from the storage medium and write information to the storage medium; or the storage medium may be an integral part of the processor. The processor and the storage medium may be located in an ASIC. The software module may be stored in a memory of a mobile terminal or in a memory card that can be inserted into the mobile terminal. For example, if a device (such as a mobile terminal) uses a large-capacity MEGA-SIM card or a large-capacity flash memory device, the software module may be stored in the MEGA-SIM card or the large-capacity flash memory device.

[0240] For one or more of the functional blocks described in the drawings and / or one or more combinations of functional blocks, it can be implemented as a general-purpose processor, digital signal processor (DSP), application-specific integrated circuit (ASIC), field programmable gate array (FPGA) or other programmable logic device, discrete gate or transistor logic device, discrete hardware component or any appropriate combination thereof for performing the functions described in the present application. For one or more of the functional blocks described in the drawings and / or one or more combinations of functional blocks, it can also be implemented as a combination of computing devices, such as a combination of a DSP and a microprocessor, multiple microprocessors, one or more microprocessors in communication with a DSP, or any other such configuration.

[0241] The present application is described above in conjunction with specific implementation methods, but it should be clear to those skilled in the art that these descriptions are exemplary and are not intended to limit the scope of protection of the present application. Those skilled in the art can make various modifications and variations to the present application based on the principles of the present application, and these modifications and variations are also within the scope of the present application.

[0242] Regarding the implementation methods including the above embodiments, the following additional notes are also disclosed:

[0243] 1. A vital sign detection device, characterized in that the device comprises:

[0244] A data acquisition unit, which acquires a radar multi-antenna zero-speed range FFT signal;

[0245] A data preprocessing unit, which selects a first radar signal from the multi-antenna zero-speed range FFT signal and generates an amplitude feature and a phase feature of the first radar signal;

[0246] A generating unit, which generates multiple segments of candidate physiological signals by using a detection model based on a deep neural network according to the amplitude characteristics and the phase characteristics;

[0247] a selection unit, which selects a target physiological signal from the plurality of candidate physiological signals, wherein the candidate physiological signals and the target physiological signal are respiratory signals or heartbeat signals, and

[0248] A detection unit detects vital sign parameters according to the target physiological signal.

[0249] 2. The device according to Note 1, characterized in that the detection model comprises:

[0250] A single time point feature extraction module, which obtains first spatial feature information at each range frequency point within a single frame time based on the amplitude feature and phase feature of the first radar signal;

[0251] A local time feature extraction module, which obtains second spatial feature information at each distance frequency point within a multi-frame time according to the first spatial feature information;

[0252] a global feature extraction module, which obtains global time feature information at each distance frequency point according to the second spatial feature information; and

[0253] A feature fusion module is configured to fuse the first spatial feature information, the second spatial feature information and the global temporal feature information to generate the candidate physiological signal.

[0254] 3. The device according to Note 2, characterized in that:

[0255] The feature fusion module cascades the first spatial feature information, the second spatial feature information and the global time feature information, or adds the first spatial feature information, the second spatial feature information and the global time feature information.

[0256] 4. The device according to Note 1, characterized in that:

[0257] The loss function used to train the detection model is a function generated according to at least one of the following: waveform reconstruction loss function Loss re , waveform change loss function Loss changes Or waveform fluctuation loss function Loss v .

[0258] 5. The device according to Note 4 is characterized in that the loss function Loss is in the following form:

[0259] Loss = a*Loss re +b*Loss changes +c*Loss v , where a, b, and c are real numbers.

[0260] 6. The device according to Note 4, characterized in that:

[0261] The waveform reconstruction loss function Loss re is the mean square error between the waveform value of the real physiological signal and the waveform value of the physiological signal output by the detection model; and / or

[0262] The waveform change loss function Loss changes is the mean square error between the waveform variation of the real physiological signal and the waveform variation of the physiological signal output by the detection model; and / or

[0263] The waveform fluctuation momentum loss function Loss vIt is the mean square error between the waveform fluctuation amount of the real physiological signal and the waveform fluctuation amount of the physiological signal output by the detection model.

[0264] 7. A method for detecting vital signs, characterized in that the method comprises:

[0265] Get the radar multi-antenna zero-speed range FFT signal;

[0266] Selecting a first radar signal from the multi-antenna zero-speed range FFT signal, and generating an amplitude feature and a phase feature of the first radar signal;

[0267] Generate multiple candidate physiological signals using a detection model based on a deep neural network according to the amplitude characteristics and the phase characteristics;

[0268] selecting a target physiological signal from the plurality of candidate physiological signals, wherein the candidate physiological signals and the target physiological signal are respiratory signals or heartbeat signals, and

[0269] Vital sign parameters are detected according to the target physiological signal.

[0270] 8. The method according to Supplement 7, characterized in that selecting a target physiological signal from the plurality of candidate physiological signals comprises:

[0271] Determining the validity of the plurality of candidate physiological signals;

[0272] When the multiple segments of candidate physiological signals are determined to be valid, the target physiological signal is determined from the multiple segments of candidate physiological signals; when the multiple segments of candidate physiological signals are determined to be invalid, the target physiological signal is not determined.

[0273] 9. The method according to Supplementary Note 8, characterized in that determining the validity of the plurality of candidate physiological signals comprises:

[0274] Determine the time overlap length of any two candidate physiological signals among the multiple candidate physiological signals, and obtain a candidate physiological signal combination whose time overlap length is greater than one breathing or heartbeat time and a first combination number L of the candidate physiological signal combinations;

[0275] Determine the similarity between two segments of candidate physiological signals in each candidate physiological signal combination by using time domain correlation and / or frequency domain coherence, and determine a second number K of combinations of candidate physiological signal combinations whose similarity is greater than a first threshold;

[0276] When the ratio K / L is greater than the second threshold, it is determined that the multiple segments of candidate physiological signals are valid; when the ratio K / L is less than or equal to the second threshold, it is determined that the multiple segments of candidate physiological signals are invalid.

[0277] 10. The method according to Supplement 8, characterized in that determining the target physiological signal from the multiple candidate physiological signals comprises:

[0278] According to the valid multiple candidate physiological signals, each candidate physiological signal in the multiple candidate physiological signals is evaluated by using a time domain evaluation index and / or a frequency domain evaluation index;

[0279] The target physiological signal is determined according to the time domain evaluation index value and / or the frequency domain evaluation index value of each segment of the candidate physiological signal.

[0280] 11. The method according to Supplement 10, characterized in that:

[0281] The time domain evaluation indicators include:

[0282] the maximum value of the time domain amplitude of the candidate physiological signal; and / or

[0283] 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 in the plurality of segments of candidate physiological signals;

[0284] and / or

[0285] The frequency domain evaluation indicators include:

[0286] the maximum value of the spectrum energy of the candidate physiological signal in the physiological signal frequency band; and / or

[0287] The ratio of the energy value of the candidate physiological signal in the physiological signal frequency band to the energy value of the candidate physiological signal in the non-physiological signal frequency band; and / or

[0288] an average value of spectral coherence coefficients of the candidate physiological signal and other candidate physiological signals in the multiple segments of candidate physiological signals in a physiological signal frequency band.

[0289] 12. The method according to Note 7, characterized in that the first radar signal is selected from the multi-antenna zero-speed range FFT signal, and the amplitude feature and phase feature of the first radar signal are generated, including:

[0290] Determine the target position according to the time domain information and / or frequency domain information of the multi-antenna zero-speed range FFT signal;

[0291] The target position and the positions of N1 frequency points adjacent to the target position are taken as the target area; and

[0292] In the target area, the first radar signal is selected, and an amplitude feature and a phase feature of the first radar signal are generated.

[0293] 13. The method according to Supplement 12, characterized in that:

[0294] The time domain information includes information on how the amplitude of the multi-antenna zero-speed range FFT signal at each range frequency point changes over time within a preset time period, and determining the target position includes: calculating the amplitude fluctuation of each range frequency point within the preset time period, and taking the range frequency point corresponding to the maximum amplitude fluctuation as the target position;

[0295] and / or

[0296] The frequency domain information includes the result of fast Fourier transform on the phase of the multi-antenna zero-speed distance FFT signal at each distance frequency point, and determines the target position, including: taking the distance frequency point corresponding to the maximum value of the sum of the energies of each frequency point in the physiological signal frequency band as the target position; and / or, taking the distance frequency point corresponding to the maximum value of the ratio of the sum of the energies of each frequency point in the physiological signal frequency band to the sum of the energies of each frequency point in the non-physiological signal frequency band as the target position.

[0297] 14. The method according to supplementary note 12, characterized in that, in the target area, selecting the first radar signal comprises:

[0298] Using the multi-antenna zero-speed range FFT signals corresponding to the largest N2 amplitude fluctuations among the amplitude fluctuations of the multi-antenna zero-speed range FFT signals at each range frequency point in the target area as the first radar signal; and / or

[0299] Using the multi-antenna zero-speed range FFT signals corresponding to the maximum N2 energy values ​​of the multi-antenna zero-speed range FFT signals at each range frequency point in the target area within the physiological signal frequency band as the first radar signal; and / or

[0300] The multi-antenna zero-speed range FFT signals corresponding to the largest N2 ratios of the energy values ​​of the multi-antenna zero-speed range FFT signals at each range frequency point in the target area in the physiological signal frequency band to the energy values ​​in the non-physiological signal frequency band are used as the first radar signal.

[0301] 15. The method according to supplementary note 12, characterized in that generating the amplitude characteristic and the phase characteristic of the first radar signal comprises:

[0302] subtracting the mean of the amplitude of the first radar signal from the amplitude of the first radar signal at each time point to generate an amplitude feature of the first radar signal; and / or

[0303] The phase value of a subsequent frame of the first radar signal is subtracted from the phase value of a previous frame to generate a phase feature of the first radar signal.

[0304] 16. The method according to Supplement 7, characterized in that the detection model includes:

[0305] A single time point feature extraction module, which obtains first spatial feature information at each range frequency point within a single frame time based on the amplitude feature and phase feature of the first radar signal;

[0306] A local time feature extraction module, which obtains second spatial feature information at each distance frequency point within a multi-frame time according to the first spatial feature information;

[0307] a global feature extraction module, which obtains global time feature information at each distance frequency point according to the second spatial feature information; and

[0308] A feature fusion module is configured to fuse the first spatial feature information, the second spatial feature information and the global temporal feature information to generate the candidate physiological signal.

[0309] 17. The method according to Supplement 16, characterized in that:

[0310] The feature fusion module cascades the first spatial feature information, the second spatial feature information and the global time feature information, or adds the first spatial feature information, the second spatial feature information and the global time feature information.

[0311] 18. The method according to Supplement 7, characterized in that:

[0312] The loss function used to train the detection model is a function generated according to at least one of the following: waveform reconstruction loss function Loss re , waveform change loss function Loss changes Or waveform fluctuation loss function Loss v .

[0313] 19. The method according to Note 18 is characterized in that the loss function Loss is in the following form:

[0314] Loss=Loss re +a*Loss changes or

[0315] Loss=Loss re +a*Loss v .

[0316] 20. The method according to Supplement 18, characterized in that:

[0317] The waveform reconstruction loss function Loss re is the mean square error between the waveform value of the real physiological signal and the waveform value of the physiological signal output by the detection model; and / or

[0318] The waveform change loss function Loss changes is the mean square error between the waveform variation of the real physiological signal and the waveform variation of the physiological signal output by the detection model; and / or

[0319] The waveform fluctuation momentum loss function Loss v It is the mean square error between the waveform fluctuation amount of the real physiological signal and the waveform fluctuation amount of the physiological signal output by the detection model.

[0320] 21. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, and wherein the processor is configured to execute the computer program to implement the vital sign detection method as described in any one of Notes 7 to 20.

[0321] 22. A storage medium storing a computer-readable program, wherein the computer-readable program enables a computer to execute the vital sign detection method as described in any one of Notes 7 to 20 in an electronic device.

Claims

1. A vital sign detection device, characterized in that: The device comprises: A data acquisition unit, which acquires a radar multi-antenna zero-speed range FFT signal; A data preprocessing unit, which selects a first radar signal from the radar multi-antenna zero-speed range FFT signal and generates an amplitude feature and a phase feature of the first radar signal; A generating unit, which generates multiple segments of candidate physiological signals by using a detection model based on a deep neural network according to the amplitude characteristics and the phase characteristics; a selection unit, which selects a target physiological signal from the plurality of candidate physiological signals, wherein the candidate physiological signals and the target physiological signal are respiratory signals or heartbeat signals, and A detection unit detects vital sign parameters according to the target physiological signal.

2. The device according to claim 1, characterized in that The selection unit comprises: a determination unit, configured to determine the validity of the plurality of candidate physiological signals; A determination unit determines the target physiological signal from the multiple segments of candidate physiological signals when the determination unit determines that the multiple segments of candidate physiological signals are valid, and does not determine the target physiological signal when the determination unit determines that the multiple segments of candidate physiological signals are invalid.

3. The device according to claim 2, characterized in that The determination unit determines the time overlap length of any two candidate physiological signals among the multiple candidate physiological signals, and obtains a candidate physiological signal combination whose time overlap length is greater than one breath or heartbeat time and a first combination number L of the candidate physiological signal combination; Determine the similarity between two segments of candidate physiological signals in each candidate physiological signal combination by using time domain correlation and / or frequency domain coherence, and determine a second number K of combinations of candidate physiological signal combinations whose similarity is greater than a first threshold; When the ratio K / L is greater than the second threshold, the multiple segments of candidate physiological signals are determined to be valid; when the ratio K / L is less than or equal to the second threshold, the multiple segments of candidate physiological signals are determined to be invalid.

4. The device according to claim 2, characterized in that The determining unit evaluates each of the multiple candidate physiological signals using a time domain evaluation index and / or a frequency domain evaluation index according to the multiple valid candidate physiological signals; The target physiological signal is determined according to the time domain evaluation index value and / or the frequency domain evaluation index value of each segment of the candidate physiological signal.

5. The device according to claim 4, characterized in that The time domain evaluation indicators include: the maximum value of the time domain amplitude of the candidate physiological signal; and / or 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 in the plurality of segments of candidate physiological signals; and / or The frequency domain evaluation indicators include: the maximum value of the spectrum energy of the candidate physiological signal within the physiological signal frequency band; and / or The energy value of the candidate physiological signal within the physiological signal frequency band; and / or The ratio of the energy value of the candidate physiological signal in the physiological signal frequency band to the energy value in the non-physiological signal frequency band; and / or an average value of spectral coherence coefficients of the candidate physiological signal and other candidate physiological signals in the multiple segments of candidate physiological signals in a physiological signal frequency band.

6. The device according to claim 1, characterized in that The data preprocessing unit comprises: A target position determination unit, which determines the target position according to the time domain information and / or frequency domain information of the radar multi-antenna zero-speed range FFT signal; a target area determination unit, which takes the target position and positions of N1 frequency points adjacent to the target position as the target area, where N1≥1; and A radar signal selection unit selects the first radar signal in the target area and generates an amplitude feature and a phase feature of the first radar signal.

7. The device according to claim 6, characterized in that The time domain information includes information on how the amplitude of the radar multi-antenna zero-speed range FFT signal at each range frequency point changes over time within a preset time period, and the target position determination unit calculates the amplitude fluctuation of the radar multi-antenna zero-speed range FFT signal at each range frequency point within the preset time period, and uses the range frequency point corresponding to the maximum amplitude fluctuation as the target position; and / or The frequency domain information includes the result of FFT on the phase of the radar multi-antenna zero-speed distance FFT signal at each distance frequency point, and the target position determination unit takes the distance frequency point corresponding to the maximum value of the sum of the energies of each frequency point in the physiological signal frequency band as the target position; and / or, the target position determination unit takes the distance frequency point corresponding to the maximum value of the ratio of the sum of the energies of each frequency point in the physiological signal frequency band to the sum of the energies of each frequency point in the non-physiological signal frequency band as the target position.

8. The device according to claim 6, characterized in that The radar signal selection unit uses the radar multi-antenna zero-speed range FFT signals corresponding to the largest N2 amplitude fluctuations among the amplitude fluctuations of the radar multi-antenna zero-speed range FFT signals at each range frequency point in the target area as the first radar signal; and / or The radar signal selection unit uses the radar multi-antenna zero-speed range FFT signals corresponding to the maximum N2 energy values ​​of the radar multi-antenna zero-speed range FFT signals at each range frequency point in the target area within the physiological signal frequency band as the first radar signal; and / or The radar signal selection unit takes the radar multi-antenna zero-speed range FFT signals corresponding to the largest N2 ratios of the energy values ​​of the radar multi-antenna zero-speed range FFT signals in the physiological signal frequency band to the energy values ​​in the non-physiological signal frequency band at each range frequency point in the target area as the first radar signal, N1≥N2≥1.

9. The device according to claim 6, characterized in that The radar signal selection unit subtracts the mean of the amplitude of the first radar signal from the amplitude of the first radar signal at each time point to generate an amplitude feature of the first radar signal; and / or The radar signal selection unit subtracts a phase value of a subsequent frame from a phase value of a previous frame of the first radar signal to generate a phase feature of the first radar signal.

10. A method for detecting vital signs, characterized in that: The method comprises: Get the 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 feature and a phase feature of the first radar signal; Generate multiple candidate physiological signals using a detection model based on a deep neural network according to the amplitude characteristics and the phase characteristics; selecting a target physiological signal from the plurality of candidate physiological signals, wherein the candidate physiological signals and the target physiological signal are respiratory signals or heartbeat signals, and Vital sign parameters are detected according to the target physiological signal.

Images