CN109965858B

CN109965858B - Ultra-wideband radar-based human body vital sign detection method and device

Info

Publication number: CN109965858B
Application number: CN201910241638.7A
Authority: CN
Inventors: 李秀萍; 李剑菡; 李昱冰
Original assignee: Beijing University of Posts and Telecommunications
Current assignee: Beijing University of Posts and Telecommunications
Priority date: 2019-03-28
Filing date: 2019-03-28
Publication date: 2020-10-30
Anticipated expiration: 2039-03-28
Also published as: CN109965858A

Abstract

The invention discloses a human body vital sign detection method and a device based on an ultra-wideband radar, wherein the method comprises the following steps: acquiring a plurality of first echo signals of normal breath and/or heartbeat of a human body and a plurality of second echo signals of abnormal breath and/or heartbeat of the human body through an ultra-wideband radar, and preprocessing and extracting to respectively obtain a normal detection sequence and an abnormal detection sequence; setting a convolutional neural network; training the convolutional neural network based on the normal detection sequence and the abnormal detection sequence; and detecting the respiration and/or heartbeat of the current detection object through the trained convolutional neural network, and outputting a judgment result of whether the current detection object belongs to a normal heartbeat. The device comprises a signal processing module, a convolutional neural network setting module, a training module and a detection module. The method and the device improve the detection accuracy of the heartbeat frequency through deep learning, improve the discrimination capability of normal heartbeat and abnormal heartbeat, and realize more accurate discrimination of real-time detection.

Description

Ultra-wideband radar-based human body vital sign detection method and device

Technical Field

The invention relates to the technical field of people number detection, in particular to a human body vital sign detection method and device based on an ultra-wideband radar.

Background

Human vital sign detection is widely used in rescue, intelligent house, mainly with breathing and/or heartbeat detection. Contact detection methods are commonly used in modern society, but contact of the probe with the measurer may cause physical discomfort or secondary injury to the measurer. Contact measurement cannot be performed on a person to be measured in the ruin at a long distance. The radar has the advantages of high resolution, low power consumption, strong anti-interference capability, penetrability, capability of detecting in dark complex environment and the like, can be used for clinical monitoring of infectious disease patients and severe patients, can also monitor vital signs of infants in real time, finds illness state and hidden danger in time, and makes up for the defect of measuring human breath and/or heartbeat in contact type measurement. In recent years, radar has been widely used, and there are many types of radar, including X-band radar, CW (continuous wave) radar, UWB (ultra wide band) radar, FMCW (frequency modulated continuous wave) radar, MIMO (multiple input multiple output) radar, and the like. At present, CW radar, FMCW radar and MIMO radar are used for checking the respiration and/or heartbeat of a human body, and have high detection precision, while UWB radar is less involved and has low detection precision.

In the prior art, chinese patent application 201810288607.2 discloses a method for measuring human respiration rate and heart rate based on ultra-wideband radar, and the key points of the technical scheme thereof include: initializing an ultra-wideband radar; acquiring an echo signal; calculating the position information of the person according to the echo signals; performing Butterworth band-pass filtering on the one-dimensional time sequence signal; respectively enhancing the signals by using a Min-Max normalization method; smoothing the enhanced signals respectively by using a Hanning window; performing fast discrete Fourier transform on the smoothed signal; the peak values of the signals in the respiration interval and the heart rate interval are searched as parameters of respiration and/or heartbeat. Although the patent application of the invention proposes that the non-contact type respiration and heart rate measurement is realized based on the ultra-wideband radar, the respiration and/or heart rate is extracted by using the band-pass filter and the fast Fourier method, because the heart rate of the human body is weak and is easily influenced by factors such as environment and the like, for example, a large low-frequency interference component is generated by detecting the slight movement of the human body, a measured target frequency signal deviates from the actual frequency, so that an accurate respiration and/or heart rate value cannot be obtained, and the resolution capability of judging whether the heart rate of a measurement object is abnormal or not is reduced.

Disclosure of Invention

The invention aims to provide a human body vital sign detection method and device based on an ultra-wideband radar so as to solve the technical problem.

In order to achieve the purpose, the invention provides the following scheme:

in a first aspect of the embodiments of the present invention, a method for detecting human vital signs based on an ultra-wideband radar is provided, which includes the steps of:

a, acquiring a plurality of first echo signals of normal breath and/or heartbeat of a human body and a plurality of second echo signals of abnormal breath and/or heartbeat of the human body through an ultra-wideband radar, and preprocessing and extracting to respectively obtain a normal detection sequence and an abnormal detection sequence;

step B, setting a convolution neural network;

step C, training the convolutional neural network based on a normal detection sequence and the abnormal detection sequence;

and D, detecting the breath and/or heartbeat of the current detection object through the trained convolutional neural network, and outputting a judgment result of whether the current detection object belongs to normal heartbeat.

Optionally, step a includes:

a1, acquiring a plurality of first echo signals of normal breath and/or heartbeat of a human body through an ultra-wideband radar, and converting the first echo signals into first complex signals;

step A2, performing FFT transformation on the first complex signal in a fast time dimension, taking a zero frequency component to obtain phase information of the first complex signal, and eliminating a distance dimension component of a UWB (ultra wideband) radar;

step A3, eliminating the phase aliasing of the first complex signal by using an extended difference and cross multiplication method, namely, using DACM to obtain an angular velocity function of the inching signal relative to the velocity function;

step A5, eliminating low-frequency components caused by human body displacement in the first complex signal by using a median filter, namely eliminating large-amplitude jitter caused by the human body displacement;

step A6, performing z-transformation on the first complex signal after the low-frequency components are eliminated, and suppressing zero-mean noise through an accumulator;

a step a7 of extracting a normal respiration and/or heartbeat frequency signal (high signal-to-noise ratio) from the first complex signal using a State Space Method (SSM);

step A8, extracting normal respiration and/or heartbeat frequency by FFT;

step A9, converting the extracted respiration and/or heartbeat frequency signals in the windows with preset number into normal detection sequences with preset length;

and step A10, referring to steps A1-A9, correspondingly extracting abnormal breathing and/or heartbeat signals of the human body, and converting the abnormal breathing and/or heartbeat signals into an abnormal detection sequence with a preset length.

Optionally, step B includes setting the convolutional neural network to the following structure:

the hierarchical data structure sequentially comprises an input layer, a5 × 5 first convolution layer of 8 cores, a first RELU layer, a first BN layer, a first maximum pooling layer, a5 × 5 second convolution layer of 16 cores, a second RELU layer, a second BN layer, a second maximum pooling layer, a5 × 5 third convolution layer of 32 cores, a third RELU layer, a third BN layer, an average pooling layer, a softmax layer and a classification output layer.

Optionally, step C includes:

respectively and correspondingly converting the normal detection sequence and the abnormal detection sequence into a normal detection RGB picture and an abnormal detection RGB picture;

and randomly disordering the normal detection RGB pictures and the abnormal detection RGB pictures, dividing the normal detection RGB pictures and the abnormal detection RGB pictures into training group data and testing group data, and training the convolutional neural network.

Optionally, step A8 extracts the normal respiration and/or heartbeat frequency signal from the first complex signal using an FFT transformation, including extracting the normal respiration and/or heartbeat frequency signal based on phase using an FFT transformation.

Optionally, step D further includes:

and outputting and displaying the heartbeat frequency, the heartbeat waveform, the respiratory frequency and the respiratory waveform in real time.

In a second aspect of the embodiments of the present invention, an ultra-wideband radar-based human body vital sign detection apparatus is provided, including a signal processing module, a convolutional neural network setting module, a training module, and a detection module;

the signal processing module is used for acquiring a plurality of first echo signals of normal breathing and/or heartbeat of a human body and a plurality of second echo signals of abnormal breathing and/or heartbeat of the human body through an ultra-wideband radar, preprocessing and extracting the first echo signals and the second echo signals to respectively obtain a normal detection sequence and an abnormal detection sequence;

the convolutional neural network setting module is used for setting a convolutional neural network;

the training module is used for training the convolutional neural network based on the normal detection sequence and the abnormal detection sequence;

and the detection module is used for detecting the breath and/or heartbeat of the current detection object through the trained convolutional neural network and outputting a judgment result of whether the current detection object belongs to normal heartbeat.

Optionally, the signal processing module includes an acquisition unit, a preprocessing unit and an extraction unit;

the system comprises an acquisition unit, a processing unit and a processing unit, wherein the acquisition unit is used for acquiring a plurality of first or second echo signals of normal or abnormal breath and/or heartbeat of a human body through an ultra-wideband radar and respectively and correspondingly converting the first or second echo signals into first or second complex signals;

the preprocessing unit is used for carrying out FFT (fast Fourier transform) on the first or second complex signal in a fast time dimension, taking a zero-frequency component to obtain phase information of the first or second complex signal and eliminating a distance dimension component of the ultra-wideband radar; eliminating phase aliasing of the first or second complex signal using an extended difference and cross-product method; eliminating low-frequency components caused by human body displacement in the first or second complex signals by using a median filter; performing z-transformation on the first or second complex signal after the low-frequency component is removed, and suppressing zero-mean noise through an accumulator; increasing the signal-to-noise ratio of the first or second complex signal using a state space method;

the extraction unit is used for extracting a normal respiration and/or heartbeat frequency signal from the first or second complex signal processed by the preprocessing unit by utilizing FFT (fast Fourier transform); and converting the extracted respiration and/or heartbeat frequency signals in the windows with the preset number into normal detection sequences or abnormal detection sequences with preset lengths.

Optionally, the convolutional neural network setting module is configured to set a convolutional neural network with a structure as follows:

Optionally, the extracting unit is configured to extract the normal respiratory and/or heartbeat frequency signal based on the phase by using FFT.

According to the specific embodiment provided by the invention, the invention discloses the following technical effects:

according to the ultra-wideband radar-based human body vital sign detection method and device, normal and abnormal breathing and/or heartbeat of a human body are respectively obtained through the ultra-wideband radar, a normal detection sequence and an abnormal detection sequence are obtained after preprocessing and extraction, a convolutional neural network with a specific structure is trained by taking two kinds of sequence data as training or test data, the heartbeat frequency is extracted more accurately by the convolutional neural network through deep learning, particularly, the discrimination capability of the convolutional neural network on the normal heartbeat and the abnormal heartbeat is improved, the discrimination result for recognizing the normal heartbeat and the abnormal heartbeat is automatically output, and accurate discrimination of real-time detection is achieved;

furthermore, the method uses a DACM-SSMF method to extract normal respiration and/or heartbeat signals for echo signals, compared with band-pass filtering and fast Fourier transform, the problem of phase boundary crossing aliasing based on a phase extraction method is solved, a median filter is applied to eliminate the influence of slight movement of human body displacement, the influence that the heartbeat signals cannot be normally extracted due to the position movement of the human body is effectively improved, and meanwhile, the accuracy of heartbeat frequency extraction is improved by using an SSM method.

Drawings

In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings needed to be used in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings without inventive exercise.

FIG. 1 is a schematic flow chart of an embodiment of a method for detecting human vital signs based on ultra-wideband radar according to the present invention;

fig. 2 is a schematic diagram (part of the flow is not shown) of a main flow of another embodiment of the ultra-wideband radar-based human body vital sign detection method according to the present invention;

FIG. 3 is a schematic diagram of a main process in yet another embodiment of the ultra-wideband radar-based human vital sign detection method according to the present invention;

FIG. 4 is a FFT spectrogram result in the ultra-wideband radar-based human body vital sign detection method of the present invention;

FIG. 5 is a DACM spectrogram result in the ultra-wideband radar-based human body vital sign detection method of the present invention;

FIG. 6 is a DACM-SSM spectrogram result in the ultra-wideband radar-based human body vital sign detection method.

Detailed Description

In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the drawings of the embodiments of the present invention. It is to be understood that the embodiments described are only a few embodiments of the present invention, and not all embodiments. All other embodiments, which can be derived by a person skilled in the art from the described embodiments of the invention without any inventive step, are within the scope of protection of the invention.

Example 1

The embodiment 1 of the invention provides a human body vital sign detection method based on an ultra-wideband radar, which is shown in figure 1 and mainly comprises the following steps:

step S100, acquiring a plurality of first echo signals of normal respiration and/or heartbeat of a human body and a plurality of second echo signals of abnormal respiration and/or heartbeat of the human body through an ultra-wideband radar, preprocessing and extracting to respectively obtain a normal detection sequence and an abnormal detection sequence.

The normality and the abnormality of the heartbeat and the respiration in the invention are distinguished according to the medical general knowledge as a judgment standard, for example, the normal heart rate of a common adult (non-athlete) is normal when 60-100 times/minute, and the other is abnormal, so that the specific indications of the normality and the abnormality in the invention can be clarified by a person skilled in the art according to the medical general knowledge, and are not repeated.

Step S101, a convolutional neural network is set.

Step S102, training the convolutional neural network based on the normal detection sequence and the abnormal detection sequence;

and S103, detecting the respiration and/or heartbeat of the current detection object through the trained convolutional neural network, and outputting a judgment result of whether the current detection object belongs to normal heartbeat.

The main application of the non-contact detection of vital signs such as human heartbeat, respiration and the like can comprise daily human health examination and rescue of wounded persons in the scene of accidents or natural disasters, and the rapid judgment of whether the heartbeat of a measured object is abnormal has important significance for timely finding the state of an illness and timely rescuing the wounded persons. The ultra-wideband radar-based human body vital sign detection method provided by the embodiment of the invention can quickly judge whether the heartbeat of a measurement object is abnormal, undoubtedly provides reference data for rescue of patients or wounded persons, and strives for time.

Example 2

The embodiment 2 of the invention provides another embodiment of a human body vital sign detection method based on an ultra-wideband radar, and in the embodiment, the detection method comprises the following steps:

step S200, initializing the radar and configuring radar parameters.

Step S201, acquiring a normal respiration and/or heartbeat I/Q channel signal of a human body, and converting a first echo signal of the I/Q channel into a first complex signal.

Step S202, the phase of the first complex signal is obtained by applying fast time dimension Fourier transform, and the distance dimension component is eliminated.

In step S203, DACM (extended differential and cross multiplication Method) is used to solve the phase aliasing problem.

Step S204, the DACM processed signal is z-transformed and an accumulator is used for filtering zero mean noise.

In step S205, low frequency components caused by human body displacement are filtered out using median filtering.

Step S206, increasing signal to noise ratio by using SSM (State space model) algorithm, extracting normal respiration and/or heartbeat signals with high signal to noise ratio, and improving the accuracy of extracted respiration and/or heartbeat frequency.

Step S207, converting the heartbeat frequencies extracted from the windows with the preset number into a normal detection sequence with a preset length, so as to ensure real-time effectiveness. Preferably, the number of windows can be preset to 10-20, and the preset length of the normal detection sequence is also 10-20.

Preferably, this step also converts the respiratory rate signal into a normal detection sequence of a preset length. Namely, it is

And S208, acquiring abnormal respiration and/or heartbeat I/Q channel signals of the human body as second echo signals, converting the second echo signals into second complex signals, referring to the steps S202-S207, correspondingly processing the second complex signals, extracting the abnormal respiration and/or heartbeat signals of the human body, and converting the abnormal respiration and/or heartbeat signals into an abnormal detection sequence with a preset length.

And step S209, converting the normal and abnormal heartbeat detection sequences into RGB pictures for convolutional neural network training.

And step S210, randomly disordering the normal heartbeat detection picture and the abnormal heartbeat detection picture, dividing the normal heartbeat detection picture and the abnormal heartbeat detection picture into training data and testing data, and training the training data and the testing data by using a convolutional neural network to obtain a normal result or an abnormal result.

Step S211 displays the result of the determination as to whether or not there is an abnormality, and displays the detected heart rate.

As one of the possible implementation manners, the main flow of the detection method according to the embodiment of the present invention is shown in fig. 2 (a part of the flow is not shown).

In the prior art, after the heartbeat frequency is displayed, whether the heartbeat is abnormal or not is judged manually, so that certain hysteresis is often provided, the real-time performance is poor, the judgment accuracy completely depends on the extraction accuracy of the heartbeat frequency, and when the extraction of the heartbeat frequency deviates from the actual condition, the judgment result based on the heartbeat frequency is also meaningless; the invention is based on the deep learning of the convolution neural network on the abnormal heartbeat signal and the normal heartbeat signal, so that the method has certain discrimination capability on whether the heartbeat is abnormal or not and can automatically output, thereby simplifying the link of manual discrimination, saving the time and improving the discrimination efficiency.

Example 3

Embodiment 3 of the present invention provides a preferred embodiment of a method for detecting human vital signs based on ultra wideband radar.

Specifically, as an implementable manner, in this embodiment, the detection method includes the steps of:

step S300, initializing the radar and configuring radar parameters.

Step S301, acquiring respiratory and/or heartbeat radar echo signals of normal or abnormal human bodies, and converting the echo signals of the I/Q channel in each frame into complex signals, wherein the conversion formula is that y (t, tau) is y_I(t,τ)+jy_Q(t, τ), j is the unit of imaginary number, t is the interval of each frame, τ is the echo position in each frame, y (t, τ) is the complex signal, y_I(t, τ) is the echo signal of the I channel, y_Q(t, τ) is an echo signal of the Q channel; further, the complex signal matrix may be represented as

And N is the total number of echo signals.

Step S302, performing FFT (Fast Fourier transform) on the complex signal obtained in step S301 in a Fast time dimension, namely a distance dimension, removing distance dimension components, and applying a formula by taking a real part R (t) of a zero-frequency component and an imaginary part M (t) of the zero-frequency component

A corresponding phase function ψ (t) is obtained, where F is the multi-valued aliasing component of the phase.

Step S303, using DACM (extended difference and cross power method) eliminates the aliasing problem of the phase, and the formula is

Obtaining an angular velocity function ω (t) related to the velocity function of the inching signal, wherein

And

the derivatives of Q (t) and I (t), respectively. Q (t) is a Q channel signal of the first complex signal, I (t) is an I channel signal of the first complex signal, and the obtained omega (t) becomes a single-value function, so that the multi-value problem of the co-domain limitation is solved.

Step S304, ω (t) is sensitive to noise, z transformation is carried out on signals in order to inhibit noise, and finally φ (n) is obtained through a discrete accumulator, wherein the formula is

Zero-mean noise is effectively suppressed. Phi (n) includes signals of breathing and/or heartbeat of the human body and clutter introduced by random movement of the human body.

Step S305, the low-frequency clutter component caused by human body micro-motion displacement in the respiration and/or heartbeat measuring process exists in the phi (n) signal, and a median filter h is used_m(n) extracting a low-frequency clutter component phi from phi (n)_f(n)，φ_f(n)＝φ(n)*h_m(n), where denotes the convolution operation and subtracting φ from φ (n) by a subtractor_f(n) obtaining a signal phi eliminating low-frequency clutter components introduced by human body micromotion displacement_C(n), the signal may be expressed as φ_C(n)＝φ(n)-φ_f(n)。

And S306, extracting the respiration and/or heartbeat of the human body by using an SSM algorithm on the signal. The extraction process is described as follows, and the SSM algorithm is composed of a state transition matrix A, a state independent matrix B and a state modulation matrix C. Firstly phi is measured_c(n) arranged in a hankel matrix form,

the method is convenient for extracting the features of the breath and the heartbeat, and after Singular Value Decomposition (SVD), the principal component components and the noise components of the breath and the heartbeat are respectively extracted according to the formula

U_snLeft unitary matrix of respiration and heartbeat, Σ_snA matrix of characteristics of the respiration and the heartbeat,

right unitary matrix of respiration and heartbeat, U_nLeft unitary matrix of noise, sigma_nA feature matrix of the noise is generated,

right unitary matrix of noise. Selecting principal component components of respiration and heartbeat to synthesize a hand matrix only containing the respiration and heartbeat components again

Through coordinate transformation

Splitting the obtained object into an observation matrix omega and a state transition matrix with the formula of

Wherein

The state transition matrix A is solved by a formula

Find out, wherein the first temporary variable

The first row of the matrix is subtracted from the Ω matrix, the second temporary variable Ω_-rlSubtracting moments from omega matrixThe last row of the array is obtained as the conjugate transpose operator. And the state modulation matrix C is the first column of the omega matrix. C ═ Ω (1,: state independent matrix B,

wherein phi_C ^TFor the original input signal phi_CTranspose of (t), transfer equation, third temporary variable Ω_NIs given by

Derived, and thus finally fitted

Can be represented by formula

And (4) pushing out, namely outputting the signal to noise ratio of the respiration and the heartbeat.

Preferably, in the present embodiment, the SSM extraction algorithm is a phase-based extraction algorithm, and the comparison of the breathing and/or heartbeat results obtained by the SSM algorithm with the FFT-transformed extraction results is shown in fig. 4 to 6. As can be seen from fig. 4, the FFT spectrum results have a large number of noise waves, which cover the heartbeat signal, as indicated by the arrows in the figure, where fb is the breathing frequency, fh is the heartbeat frequency, and 2fb, 2fh, and 4fb respectively represent several harmonics of the corresponding multiple (e.g., 2fb represents the second harmonic of the breathing frequency). Most of the clutter signals are suppressed after the DACM algorithm is extracted, and the heartbeat signals can be clearly shown, as shown in FIG. 5. After the SSM algorithm, as shown in fig. 6, only the pure heartbeat signal is left, so that the SSM algorithm improves the signal-to-noise ratio of the heartbeat signal.

In step S307, the breathing and/or heartbeat frequency is extracted using FFT.

Therefore, in practice, the embodiment of the present invention provides a DACM-SSMF algorithm (i.e., an extended and cross pulse Method-state space model Fourier) for preprocessing and extracting echo signals of heartbeat and respiration of a human body acquired by a radar, which can effectively suppress clutter, improve a signal-to-noise ratio of the signals, and improve accuracy of extracting heartbeat frequency signals.

Step S308, converting the heartbeat frequency extracted from the windows with preset number into a normal detection sequence with preset length;

and repeating the steps S301-S306 to extract the breathing and/or heartbeat signals of the abnormal human body and converting the signals into an abnormal detection sequence with a preset length.

Step S309, converting the normal and abnormal heartbeat detection sequences into RGB pictures for convolutional neural network training.

And S310, randomly disordering the normal heartbeat detection picture and the abnormal heartbeat detection picture, dividing the normal heartbeat detection picture and the abnormal heartbeat detection picture into training data and testing data, and training by using a convolutional neural network to obtain a normal result or an abnormal result.

Preferably, in this embodiment, the convolutional neural network includes an input layer, a5 × 5 first convolutional layer of 8 cores, a first RELU layer, a first BN layer, a first maximum pooling layer, a5 × 5 second convolutional layer of 16 cores, a second RELU layer, a second BN layer, a second maximum pooling layer, a5 × 5 third convolutional layer of 32 cores, a third RELU layer, a third BN layer, an average pooling layer, a softmax layer, and a classification output layer in this order.

Step S3011, outputting and displaying the heartbeat frequency, heartbeat waveform, respiratory frequency and respiratory waveform in real time.

As one of the possible implementation manners, the main flow of the preferred embodiment is shown in fig. 3.

Based on the method provided by the embodiment of the invention, the test precision of the convolutional neural network is 90%, which shows that the convolutional neural network can be used for classifying normal and abnormal heartbeats of a human body, overcomes the defect that the normal and abnormal heartbeats cannot be classified in the common method, and has strong resolving power.

Example 4

The embodiment 4 of the invention provides a human body vital sign detection device based on an ultra-wideband radar, which comprises a signal processing module, a convolutional neural network setting module, a training module and a detection module, wherein the signal processing module is used for processing a signal;

The method for detecting the heartbeat and the breath of the human body by the ultra-wideband radar in the prior art also has the following defects: 1. cannot be used for real-time processing; 2. the method is only suitable for ideal data and later static processing; 3. since the algorithm determines the breathing and/or heartbeat of the target according to the target position, when a plurality of targets exist in the environment, the breathing and heartbeat of other targets are easily misjudged, and the test is wrong.

The invention adopts UWB radar to detect the breath and/or heartbeat of the human body, utilizes DACM to solve the problem of phase out-of-range aliasing based on a phase extraction method, and applies a median filter to eliminate the influence of slight movement of the human body displacement, thereby effectively improving the influence that heartbeat signals cannot be normally extracted due to the movement of the human body, improving the detection precision of the UWB radar to detect the breath and/or heartbeat of the human body, outputting a normal or abnormal judgment result in real time, alarming in real time if the abnormal condition exists, undoubtedly increasing the probability that serious patients and valuable wounded persons are discovered, and striving for the time for rescuing the wounded persons or patients.

In one or more exemplary designs, the functions may be implemented in hardware, software, firmware, or any combination thereof. If implemented in software, the functions may be stored on or transmitted over as one or more instructions or code on a computer-readable medium. Computer-readable media includes both computer storage media and communication media including any medium that facilitates transfer of a computer program from one place to another. A storage media may be any available media that can be accessed by a general purpose or special purpose computer. By way of example, and not limitation, such computer-readable media can comprise RAM, ROM, EEPROM, CD-ROM or other optical disk storage, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to carry or store desired program code in the form of instructions or data structures and that can be accessed by a general-purpose or special-purpose computer, or a general-purpose or special-purpose processor. Also, any connection is properly termed a computer-readable medium. For example, if the software is transmitted from a website, server, or other remote source using a coaxial cable, fiber optic cable, twisted pair, Digital Subscriber Line (DSL), or wireless technologies such as infrared, radio, and microwave, then the coaxial cable, fiber optic cable, twisted pair, DSL, or wireless technologies such as infrared, radio, and microwave are included in the definition of medium. Disk and disc, as used herein, includes Compact Disc (CD), laser disc, optical disc, Digital Versatile Disc (DVD), floppy disk, blu-ray disc where disks usually reproduce data magnetically, while discs reproduce data optically with lasers. Combinations of the above should also be included within the scope of computer-readable media.

The embodiments in the present description are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments are referred to each other. For the system disclosed by the embodiment, the description is relatively simple because the system corresponds to the method disclosed by the embodiment, and the relevant points can be referred to the method part for description.

The principle and the implementation manner of the present invention are explained by applying specific examples, the above description of the embodiments is only used to help understanding the method of the present invention and the core idea thereof, the described embodiments are only a part of the embodiments of the present invention, not all embodiments, and all other embodiments obtained by a person of ordinary skill in the art based on the embodiments of the present invention without creative efforts belong to the protection scope of the present invention.

Claims

1. A human body vital sign detection method based on an ultra-wideband radar is characterized by comprising the following steps:

step A, acquiring a plurality of first echo signals of normal breath and/or heartbeat of a human body and a plurality of second echo signals of abnormal breath and/or heartbeat of the human body through an ultra-wideband radar, preprocessing and extracting to respectively obtain a normal detection sequence and an abnormal detection sequence, wherein the step A specifically comprises the following steps:

step A2, performing FFT transformation on the first complex signal in a fast time dimension, taking a zero frequency component to obtain phase information of the first complex signal, and eliminating a distance dimension component of the ultra-wideband radar;

a step a3 of eliminating phase aliasing of the first complex signal using an extended difference and cross-product method;

step A4, eliminating low-frequency components caused by human body displacement in the first complex signal by using a median filter;

step A5, performing z-transform on the first complex signal after the low-frequency components are removed, and suppressing zero-mean noise through an accumulator;

step A6, increasing the signal-to-noise ratio of the first complex signal by using a state space method;

step A7, extracting normal respiration and/or heartbeat frequency signals from the first complex signals by using FFT;

step A8, converting the extracted respiration and/or heartbeat frequency signals in the windows with preset number into normal detection sequences with preset length;

step A9, referring to the step A1-A8, acquiring a second echo signal, converting the second echo signal into a second complex signal, and correspondingly extracting abnormal breathing and/or heartbeat frequency signals of the human body to obtain an abnormal detection sequence with a preset length;

step B, setting a convolution neural network;

step C, training the convolutional neural network based on the normal detection sequence and the abnormal detection sequence;

2. The ultra-wideband radar-based human vital sign detection method according to claim 1, wherein the step B comprises setting the convolutional neural network to the following structure:

3. The ultra-wideband radar-based human vital sign detection method according to claim 1, wherein the step C comprises:

4. The ultra-wideband radar based human vital signs detection method according to claim 1, wherein the step a7 extracts normal breathing and/or heartbeat frequency signals from the first complex signal using FFT transformation, including phase-based extraction of the normal breathing and/or heartbeat frequency signals using FFT transformation.

5. The ultra-wideband radar based human vital signs detection method according to any one of claims 1 to 4, wherein the step D further comprises:

6. The human body vital sign detection device based on the ultra-wideband radar is characterized by comprising a signal processing module, a convolutional neural network setting module, a training module and a detection module;

the signal processing module is used for acquiring a plurality of first echo signals of normal breathing and/or heartbeat of a human body and a plurality of second echo signals of abnormal breathing and/or heartbeat of the human body through an ultra-wideband radar, preprocessing and extracting the first echo signals to respectively obtain a normal detection sequence and an abnormal detection sequence, acquiring a plurality of first echo signals of normal breathing and/or heartbeat of the human body through the ultra-wideband radar, and converting the first echo signals into first complex signals; performing FFT (fast Fourier transform) on the first complex signal in a fast time dimension, taking a zero-frequency component to obtain phase information of the first complex signal, and eliminating a distance dimension component of the ultra-wideband radar; eliminating phase aliasing of the first complex signal using an extended difference and cross-product method; eliminating low-frequency components caused by human body displacement in the first complex signal by using a median filter; z-transforming the first complex signal after the low-frequency components are eliminated, and suppressing zero-mean noise through an accumulator; increasing the signal-to-noise ratio of the first complex signal by using a state space method; extracting a normal breathing and/or heartbeat frequency signal from the first complex signal using an FFT transformation; converting the extracted respiration and/or heartbeat frequency signals in the windows with the preset number into normal detection sequences with preset lengths; acquiring a second echo signal, converting the second echo signal into a second complex signal, and correspondingly extracting abnormal breathing and/or heartbeat frequency signals of the human body to obtain an abnormal detection sequence with a preset length;

7. The ultra-wideband radar based human vital signs detection device according to claim 6, wherein the convolutional neural network setting module is configured to set a convolutional neural network of the following structure:

8. The ultra wideband radar-based human vital signs detection device according to claim 6, wherein the device is configured to extract the normal respiration and/or heartbeat frequency signals based on phase using FFT transformation.