CN112998668A

CN112998668A - Millimeter wave-based non-contact far-field multi-human-body respiration heart rate monitoring method

Info

Publication number: CN112998668A
Application number: CN202110173869.6A
Authority: CN
Inventors: 关山
Original assignee: Lusheng Youbai Chongqing Technology Co ltd
Current assignee: Lusheng Youbai Chongqing Technology Co ltd
Priority date: 2021-02-06
Filing date: 2021-02-06
Publication date: 2021-06-22
Anticipated expiration: 2041-02-06
Also published as: CN112998668B

Abstract

The invention provides a millimeter wave-based non-contact far-field multi-human-body respiration heart rate monitoring method, which comprises the following steps: performing micro Doppler operation on the monitored millimeter wave signals, and demodulating operation data into micro-distance spectrum data and micro-angle spectrum data; separating the two kinds of frequency spectrum data by adopting a band-pass filter; constructing a distance set and an angle set based on the micro-distance spectrum data and the micro-angle spectrum data, wherein the distance set is used for storing multi-target human body signal data at corresponding distances, and the angle set is used for storing multi-target human body signal data at corresponding angles at the same distance; calculating the breathing frequency of each multi-target human body in each distance set and each angle set based on wavelet transformation, and calculating the heartbeat frequency of each multi-target human body in each distance set and each angle set based on a clustering density algorithm. The method has the advantages of remote monitoring, simultaneous monitoring of multiple target human bodies, high accuracy of human heart rate and respiration monitoring results, great referential significance of the monitoring results and the like.

Description

Millimeter wave-based non-contact far-field multi-human-body respiration heart rate monitoring method

Technical Field

The invention relates to the field of communication of the Internet of things and medical monitoring, in particular to a millimeter wave-based non-contact far-field multi-human-body respiration and heart rate monitoring method.

Background

The millimeter wave refers to an electromagnetic wave with a wavelength of 1-10 mm, and is located in a wavelength range where microwave and far-infrared wave are overlapped, so that the millimeter wave has the characteristics of two wave spectrums. Compared with light waves, the millimeter waves are less influenced by natural light and a thermal radiation source; the device has extremely wide bandwidth, the frequency range of millimeter waves is 30GHz-300GHz, and the device has the characteristics of high precision and high resolution; the beam of the millimeter wave is narrow, and the beam of the millimeter wave is much narrower than that of the microwave under the same antenna size, so that small targets which are closer to each other can be distinguished or the details of the targets can be observed more clearly. The propagation of millimeter waves is much less affected by weather than laser light and can be considered to be all-weather. Compared with microwaves, millimeter wave components are much smaller in size and easier to miniaturize.

In application scenes such as intelligent home old people monitoring, sleep monitoring of patients with intermittent sleep breathing, patient state monitoring, early warning of fatigue driving, vital sign tracking after disasters occur, detection of hidden personnel in security scenes and the like, real-time monitoring of human heart rate and breathing is the most direct and important. By utilizing the characteristics of millimeter wave high precision and high resolution, the radial motion of the target human body generates phase modulation on millimeter wave carrier waves, so that frequency difference, namely Doppler frequency, is generated between a reflection signal and a receiving signal of the millimeter wave. The heart beat caused by human heartbeat and the chest fluctuation caused by human respiration can generate sideband frequency, namely micro Doppler frequency, near the Doppler frequency of a target human body; the capture of the heart rate and the breath of the human body by adopting the millimeter waves is realized based on the physical phenomenon.

At present, some methods for detecting human respiration and heart rate by millimeter waves exist in the prior art, for example, chinese patent CN111481184A discloses a multi-target respiration heart rate monitoring method and system based on millimeter wave radar technology, non-contact heart rate and respiration rate detection is performed by a millimeter wave radar module, tedious contact monitoring programs are avoided, privacy of a detected person is not violated, respiration heart rate data of a monitored object can be monitored and analyzed in real time, unexpected diseases of the monitored object can be prevented, or timely alarm is achieved when an emergency occurs. For example, chinese patent CN201811596373.4 discloses a non-contact vital sign monitoring method for a bedridden patient, which utilizes the advantage that a millimeter wave radar does not need to be in direct contact with a measured object, and utilizes the bopler effect to obtain vibration information of the chest surface of the patient through a series of processing on the received millimeter wave radar, thereby calculating the heart rate and the respiratory rate. However, the prior art still has a plurality of defects:

1. the existing method for detecting heart rate and breath of a human body by using a millimeter wave technology can only be applied to a close-range scene, and the detection accuracy can be ensured only if the distance between millimeter wave detection equipment and the detected human body is within one meter. The actual landing of the technology is greatly limited, and the actual user experience is greatly reduced.

2. Most of the existing methods for detecting heart rate and breath of human bodies by using the millimeter wave technology are one-to-one detection, and if a plurality of human body targets exist in a detection scene, the human body targets are difficult to distinguish and identify and can not be accurately monitored.

3. In an actual monitoring scene, a plurality of interference factors may exist, such as interference of heart rate respiration of household pets in home life, interference of human walking and limb movement, interference of water mist and shower water flow in a bathroom, interference of household electrical equipment (electric fans, sweeping robots and the like), and the like.

Disclosure of Invention

In view of the above-mentioned shortcomings of the prior art, the present invention aims to: the millimeter wave-based non-contact far-field multi-human-body respiration and heart rate monitoring method is provided, on the premise of ensuring the detection accuracy, the respiration and heart rate of a human body can be monitored in real time at a long distance of more than five meters, and a plurality of targets can be effectively distinguished, identified and monitored simultaneously; the interference of a plurality of complex signals existing in an actual monitoring scene can be solved. The method has the advantages of remote monitoring, simultaneous monitoring of multiple target human bodies, high accuracy of human heart rate and respiration monitoring results, great referential significance of the monitoring results and the like.

A millimeter wave-based non-contact far-field multi-human-body respiration heart rate monitoring method comprises the following steps:

emitting linear frequency modulation continuous millimeter waves into a monitoring space through at least one millimeter wave sensing device, receiving millimeter wave sensing signals reflected in the monitoring space in real time, and preprocessing the received millimeter wave sensing signals;

performing micro Doppler operation on the preprocessed millimeter wave sensing signals, and demodulating micro Doppler data obtained through operation into micro distance spectrum data and micro angle spectrum data; based on the difference between the respiratory frequency and the heartbeat frequency, a band-pass filter is adopted to separate the two kinds of frequency spectrum data;

constructing a distance set and an angle set based on the separated micro-distance spectrum data and micro-angle spectrum data, wherein the distance set is used for storing multi-target human body signal data of corresponding distances, and the angle set is used for storing multi-target human body signal data of the same distance and corresponding angles;

analyzing and determining the spatial position of each target human body through a neural network model, and classifying and identifying the signal data of each target human body by adopting a pattern recognition classifier;

calculating the breathing frequency of each multi-target human body in each distance set and each angle set based on wavelet transformation, and calculating the heartbeat frequency of each multi-target human body in each distance set and each angle set based on a clustering density algorithm;

and transmitting the respiratory frequency, the heartbeat frequency and the spatial position information of each target human body to a monitoring server or a user side in real time.

Further, the preprocessing the received millimeter wave sensing signal specifically includes:

carrying out autocorrelation function and cross-correlation function analysis on received millimeter wave sensing signals x (t), and screening out real millimeter wave sensing signals s (t), wherein x (t) s (t) n (t), and n (t) represents noise signals;

converting the real millimeter wave sensing signal from an analog signal into a digital signal, performing inverse Fourier transform, and converting a frequency domain digital signal into a time domain digital signal;

and carrying out digital filtering processing, spatial noise processing and spatial multi-path interference crosstalk elimination on the time domain digital signals.

Further, the performing micro doppler operation on the preprocessed millimeter wave sensing signal, and demodulating the micro doppler data obtained by the operation into micro distance spectrum data and micro angle spectrum data specifically includes:

the millimeter wave sensing signals comprise initial millimeter wave sensing signals and millimeter wave sensing signals to be recognized, and a space environment static model is constructed according to the preprocessed initial millimeter wave sensing signals;

performing Doppler operation and micro-Doppler operation on the preprocessed millimeter wave sensing signals to be recognized, and shielding the sensing signals of the static objects in the space and the sensing signals without Doppler effect based on the static object model in the space environment to obtain micro-Doppler data of the millimeter wave sensing signals to be recognized;

performing I-Q demodulation on the micro Doppler data obtained by operation to obtain an RSS spectrum of an analytic amplitude and a phi spectrum of an analytic phase; and respectively converting the RSS frequency spectrum and the phi frequency spectrum of the analytic amplitude into micro-distance frequency spectrum data and micro-angle frequency spectrum data, and determining the polar coordinates of the multi-human body signal data.

Further, the removing of animal-related signals in the millimeter wave sensing signals specifically comprises:

setting an amplitude RSS threshold and an angular frequency omega threshold aiming at the amplitude RSS and the angular frequency omega of the demodulated data, and keeping all periodic signal data in the amplitude RSS threshold and the angular frequency omega threshold;

and performing energy spectrum integral summation of amplitude RSS on each demodulated periodic signal data, comparing the energy spectrum integral summation with a preset human body reflection amplitude RSS energy spectrum integral summation value, and if the demodulated energy spectrum integral sum is smaller than the preset energy spectrum integral sum, judging as an animal related signal and rejecting related data.

Further, the procedure of I-Q demodulating the micro-doppler data is as follows:

ADC sampling is carried out on the micro Doppler data, the nth data output after sampling is f (n), and sin (omega) is generated by a digital oscillation algorithm NCO based on the sampling rate f(s)_an) and cos (. omega.) of_an)；

Data f (n) and sin (ω) are multiplied by a first multiplier_an) multiplication, filtering by using a low-pass filter algorithm LPF after multiplicationAfter filtering, extracting data by adopting a data extraction algorithm in integral multiple to obtain data I (n);

the data f (n) and cos (omega) are processed by a second multiplier_an), multiplying, filtering by adopting a low-pass filtering algorithm LPF after multiplying, and extracting data by adopting a data extraction algorithm in integral multiple after filtering to obtain data Q (n).

Further, the constructing a distance set and an angle set based on the separated micro-distance spectrum data and micro-angle spectrum data specifically includes:

performing inverse Fourier transform (FFT) on the demodulated micro-distance spectrum data and micro-angle spectrum data to obtain a distance pulse function D and an angle pulse function phi;

performing moving target discrimination filtering MTI on all data of function points in the distance pulse function D and the angle pulse function phi; based on the difference between the respiratory frequency and the heartbeat frequency, a band-pass filter is adopted to separate the respiratory signal and the heartbeat signal;

performing inverse distance Fourier transform (FFT) on data in the distance pulse function D by adopting Hamming function Hamming weighting, and constructing a distance set based on the data of the inverse distance Fourier transform (FFT);

carrying out inverse angle Fourier transform (FFT) on data in the angle pulse function phi by adopting Hamming function Hamming weighting; and constructing an angle set based on the data of the angle inverse fourier transform FFT.

Further, a digital band-pass filter is adopted to separate the respiratory signal and the heartbeat signal, the digital band-pass filter comprises a first adder and a second adder, the input signal x (n) is input into the input end of the first adder, the output end of the first adder outputs a filtering signal Y '(n), the filtering signal Y' (n) is input into the input end of the second adder, the output end of the second adder outputs a signal Y (n), and the algorithm of the digital band-pass filter is as follows:

input signal X (n) Z-transformed^-iFor the signal X (n-1), the signal X (n-1) is Z-transformed into Z^-iIs signal X (n-2); the filtered signal Y' (n) is Z-transformed Z^-iThe signal Y' (n-1) is Z-transformed to Z^-iSignal Y' (n-2); the output signal Y (n) being Z-transformed^-iFor the signal Y (n-1), the signal Y (n-1) is Z-transformed into Z^-iIs signal Y (n-2);

the input signal X (n) is multiplied by a positive coefficient b0 and then input to the input end of the first adder, the signal X (n-1) is multiplied by a positive coefficient b1 and then input to the input end of the first adder, and the signal X (n-2) is multiplied by a positive coefficient b2 and then input to the input end of the first adder; the signal Y '(n-1) is multiplied by a negative coefficient-a 1 and then input to the first adder input, and the signal Y' (n-2) is multiplied by a negative coefficient-a 2 and then input to the first adder input;

the filtered signal Y ' (n) is multiplied by a positive coefficient b0 and then input to the second adder input, the signal Y ' (n-1) is multiplied by a positive coefficient b1 and then input to the second adder input, and the signal Y ' (n-2) is multiplied by a positive coefficient b2 and then input to the second adder input; the signal Y (n-1) is multiplied by the negative coefficient-a 1 and input to the second adder input, and the signal Y (n-2) is multiplied by the negative coefficient-a 2 and input to the second adder input.

Further, the calculating the respiratory frequency of each of the multiple persons in each distance set and each angle set based on the wavelet transform, and the calculating the heartbeat frequency of each of the multiple persons in each distance set and each angle set based on the clustering density algorithm specifically include:

in a time window T, calculating a periodic peak value in a respiratory spectrum based on a wavelet transform algorithm WT, and removing aperiodic interference signals to obtain respective respiratory frequencies of multiple human bodies in a corresponding distance set and an angle set;

in a time window T, based on a cluster density algorithm DBSCAN, classifying different peak value data sets of inverse Fourier transform FFT after band-pass digital filtering, removing a maximum peak value density set and a minimum peak value density set, and calculating a peak value period in the sets by taking a middle peak value density set as a reference to obtain respective heartbeat frequencies of a plurality of human bodies in each distance set and each angle set.

Further, the calculating of the respective heartbeat frequencies of the multiple human bodies in each distance set and each angle set based on the clustering density algorithm specifically includes:

inputting all distance sets D' and angle sets A, setting the numerical value interval radius epsilon of each set, and setting a density threshold rho, wherein all input data are unread data;

randomly selecting one data k as read data, if the data meeting the density threshold value rho exists in the designated radius interval epsilon of the read data k, creating a new set C, and adding the data k into the new set C;

setting N as a set of all data in a radius interval epsilon of the data k, marking each data k in the set N as read data if the data k is unread, and adding the data into the set N if the data k conforms to a density threshold rho in the radius interval epsilon of a numerical value interval;

the marking data set C is noise to be discarded, and the marking data set N is access data; and repeating the above process for all unread data until no unread data exists, wherein the obtained data set N is the respective heartbeat frequency of a plurality of human bodies.

Further, the method for constructing the static model of the space environment comprises the following steps:

scanning a static monitoring space for N times through millimeter wave sensing equipment; acquiring distance, angle and amplitude data corresponding to each scanning, wherein the amplitude is a two-dimensional correlation function of the distance and the angle;

comparing whether the amplitude data at the moment k and the amplitude data at the moment k +1 are within the tolerance range of the tolerance delta;

if the comparison data of the k +1 moment at the same distance and angle is smaller than the tolerance delta, the reflection point cloud of the static object of the real space environment is obtained;

and repeating N times of data comparison on each real reflection point cloud, determining all real reflection point cloud sets of the static objects of the space environment, and storing data and coordinates of all environment static object point cloud sets.

Compared with the prior art, the invention has the following advantages:

the invention provides a millimeter wave-based non-contact far-field multi-human-body respiration and heart rate monitoring method, which can monitor the respiration and heart rate of a human body in real time at a long distance of more than five meters on the premise of ensuring the detection accuracy, and can effectively distinguish, identify and monitor a plurality of targets simultaneously; the interference of a plurality of complex signals existing in an actual monitoring scene can be solved. The method has the advantages of remote monitoring, simultaneous monitoring of multiple target human bodies, high accuracy of human heart rate and respiration monitoring results, great referential significance of the monitoring results and the like.

Drawings

FIG. 1 is a control flow chart of a millimeter wave-based non-contact far-field multi-human respiration heart rate monitoring method in an embodiment of the present invention;

FIG. 2 is a flow chart illustrating the preprocessing of millimeter wave sensor signals according to an embodiment of the present invention;

FIG. 3 is a flowchart illustrating micro-Doppler operations performed on the preprocessed millimeter wave sensor signals according to an embodiment of the present invention;

fig. 4 is a flowchart of removing animal-related signals from millimeter wave sensor signals according to an embodiment of the present invention;

FIG. 5 is a flow chart of I-Q demodulation of micro-Doppler data in an embodiment of the invention;

FIG. 6 is a flow chart of the construction of a distance set and an angle set according to an embodiment of the present invention;

FIG. 7 is a flow chart of an algorithm of a digital band pass filter according to an embodiment of the present invention;

FIG. 8 is a flowchart of an algorithm for a cluster density algorithm in an embodiment of the present invention;

FIG. 9 is a flowchart illustrating a static model of a spatial environment according to an embodiment of the present invention;

FIG. 10 is a flowchart of MTI moving object discrimination digital filtering according to an embodiment of the present invention;

FIG. 11 is a flowchart illustrating the learning of the long-short term memory recurrent neural network LSTM according to an embodiment of the present invention;

FIG. 12 is a flow chart of the KNN classifier identification according to an embodiment of the present invention;

FIG. 13 is a flow chart of signal digital filtering of a time domain digital signal according to an embodiment of the present invention;

FIG. 14 is a flow chart of spatial multi-path interference cancellation for time domain digital signals according to an embodiment of the present invention;

FIG. 15 is a flowchart illustrating the control of spatial noise processing on a time-domain digital signal according to an embodiment of the present invention;

fig. 16 is a control block diagram of the millimeter wave sensing device in the embodiment of the present invention;

fig. 17 is a flowchart of an overall principle of a millimeter wave-based non-contact far-field multi-human respiration and heart rate monitoring method in the embodiment of the present invention.

Detailed Description

Embodiments of the present invention will be described in detail below with reference to the accompanying drawings. The following examples are only for illustrating the technical solutions of the present invention more clearly, and therefore are only examples, and the protection scope of the present invention is not limited thereby.

Example (b):

referring to fig. 1 to 17, a millimeter wave based non-contact far field multi-human body respiration and heart rate monitoring method includes the following steps:

emitting linear frequency modulation continuous millimeter waves into a monitoring space through at least one millimeter wave sensing device, receiving millimeter wave sensing signals reflected in the monitoring space in real time, and preprocessing the received millimeter wave sensing signals. Specifically, the monitoring space is freely selected by a user, and can be a living room, a bedroom and the like of an intelligent home, so that the health of the old is conveniently monitored, and the sleep monitoring of a sleep breathing intermittent patient can also be realized; the monitoring space can be a ward, so that the state of the patient can be conveniently monitored; the monitoring space can be the internal space of the automobile, so that the early warning of fatigue driving of a driver is facilitated; the monitoring space can be a wide disaster site, and the vital signs after the disaster happens can be conveniently tracked; the monitoring space can be a public space, and detection of hidden personnel in a security scene is facilitated. The number of the millimeter wave sensing devices arranged may be specifically determined according to the size of the monitoring space, whether there is a barrier wall, or the like. The millimeter waves act on objects or human bodies in the space to form reflection point clouds, the objects or the human bodies can be effectively identified and distinguished according to millimeter wave signals of the reflection point clouds, and then the respiration heart rate, the space position and the like of the human bodies are analyzed.

Performing micro Doppler operation on the preprocessed millimeter wave sensing signals, and demodulating micro Doppler data obtained through operation into micro distance spectrum data and micro angle spectrum data; and based on the difference between the respiratory frequency and the heartbeat frequency, a band-pass filter is adopted to separate the two kinds of frequency spectrum data. Specifically, by utilizing the characteristics of high precision and high resolution of millimeter wave band (30GHz-300GHz), the radial motion of the target human body generates phase modulation on millimeter wave carrier, so that frequency difference, i.e. doppler frequency, is generated between the reflected signal and the received signal of millimeter wave. When the heart beats caused by the heartbeat of a human body and the chest fluctuates caused by the respiration of the human body, the side band frequency, namely the micro Doppler frequency, is generated at the Doppler frequency accessory of the target human body. The capture of the heart rate and the breath of the human body by the millimeter waves is realized based on the physical phenomenon.

And constructing a distance set and an angle set based on the separated micro-distance spectrum data and micro-angle spectrum data, wherein the distance set is used for storing multi-target human body signal data of corresponding distances, and the angle set is used for storing multi-target human body signal data of corresponding angles at the same distance.

Analyzing and determining the spatial position of each target human body through a neural network model, and classifying and identifying the signal data of each target human body by adopting a pattern recognition classifier; specifically, the RNN/LSTM neural network model is used for determining the position of each human body in the FOV of the monitoring space, and a KNN pattern recognition classifier algorithm is used for classifying and recognizing the signal data. In the space after the static elimination, all the amplitudes (RSS) and corresponding polar coordinate sets form an array matrix, and a long-short term memory recurrent neural network (LSTM) under the Recurrent Neural Network (RNN) is used for machine learning to obtain one or more space positions of the human body.

Calculating the breathing frequency of each multi-target human body in each distance set and each angle set based on wavelet transformation, and calculating the heartbeat frequency of each multi-target human body in each distance set and each angle set based on a clustering density algorithm; and transmitting the respiratory frequency, the heartbeat frequency and the spatial position information of each target human body to a monitoring server or a user side in real time.

According to the millimeter wave-based non-contact far-field multi-human body respiration and heart rate monitoring method, distance sets and angle sets are constructed according to micro-distance spectrum data and micro-angle spectrum data, then the respective respiration frequencies of multi-target human bodies in each distance set and each angle set are calculated based on wavelet transformation, and the respective heartbeat frequencies of the multi-target human bodies in each distance set and each angle set are calculated based on a clustering density algorithm; therefore, on the premise of ensuring the detection accuracy, the respiration and the heart rate of the human body can be monitored in real time at a long distance of more than five meters. The spatial position of each target human body is determined through the analysis of a neural network model, and the signal data of each target human body is classified and recognized by adopting a pattern recognition classifier; therefore, a plurality of targets can be effectively distinguished, identified and monitored simultaneously. And the interference of a plurality of complex signals in an actual monitoring scene is solved by preprocessing, shielding sensing signals of static objects in space and sensing signals without Doppler effect, removing animal related signals in the millimeter wave sensing signals and removing all non-periodic interference signals caused by limb movement and walking of a human body. The method has the advantages of remote monitoring, simultaneous monitoring of multiple target human bodies, high accuracy of human heart rate and respiration monitoring results, great referential significance of the monitoring results and the like.

According to the millimeter wave-based non-contact far-field multi-human-body respiration and heart rate monitoring method, in order to effectively detect far-field (more than five meters) weak signals, a time-division multiplexing MINO antenna structure with multiple groups of transmitting and receiving is used, so that the weak signal picking capacity and azimuth angle resolution of the system are improved. Referring to fig. 16, the millimeter wave sensing device includes a microstrip array antenna, a microprocessor, a millimeter wave chip, a transmitting end multi-channel rf switch, a receiving end multi-channel rf switch, and a communication module, where the microstrip array antenna includes at least one transmitting array and at least one receiving array; the output ends of the multi-channel radio frequency switches of the transmitting end are connected with the transmitting arrays, and the input ends of the multi-channel radio frequency switches of the transmitting end are respectively connected with the microprocessor and the millimeter wave chip; the input end of the receiving end multi-channel radio frequency switch is connected with each receiving array, and the output end of the receiving end multi-channel radio frequency switch is connected with the millimeter wave chip; the microprocessor is in communication connection with the millimeter wave chip, and the microprocessor is in communication connection with the monitoring server and the user side through the communication module.

Specifically, the receiving array is arranged by taking a half wavelength λ/2 of the used frequency as a mutual distance between the microstrips, and the transmitting array is arranged by taking a wavelength λ of the used frequency as a mutual distance between the microstrips. The multi-channel radio frequency switch at the transmitting end is controlled by a microprocessor signal to determine a controlled transmitting array; the transmitting arrays 1 to 4 are scanned in sequence, each transmitting array and the corresponding receiving data are used as a set, and 4 sets of data sets are obtained after each transmitting end scanning is completed. The radio frequency amplifier in the millimeter wave chip transmits the millimeter wave with specific frequency from the transmitting array. The receiving array antenna transmits the received millimeter wave signals to a low-noise amplifier in the millimeter wave chip by a receiving end multi-channel radio frequency switch for receiving. And the millimeter wave chip processes the received millimeter wave reflection signal, receives the millimeter wave reflection signal by the microprocessor and enters the next step of algorithm processing.

Referring to fig. 2, the preprocessing the received millimeter wave sensing signal specifically includes:

Specifically, by using a continuous frequency modulation signal FMCW of a millimeter wave sensor, autocorrelation function and cross-correlation function analysis are performed on signal data x (t)(s) (t) + n (t) (s (t) is an actual signal, and n (t)) is noise in four sets of each scanning of reflected point cloud signal data, so that a real millimeter wave sensing signal s (t) is screened out, and environmental noise interference, multipath interference and radio frequency crosstalk between reflection and reception are eliminated through a digital filtering algorithm to improve the signal-to-noise ratio (SNR). And sequentially carrying out digital filtering processing, spatial multipath interference and crosstalk elimination and secondary elimination on possible environmental noise and harmonic noise on the time domain digital signal. The weak Doppler effect detection in the monitoring space can further improve the identification accuracy and the human body counting accuracy of the target human body position in the monitoring space by analyzing the human body micromotion algorithm through autocorrelation and cross-correlation.

Therefore, the interference of a plurality of complex signals in an actual monitoring scene can be solved, and the accuracy and the referential performance of the target human heart rate respiration monitoring result are ensured.

Referring to fig. 3, performing micro doppler operation on the preprocessed millimeter wave sensing signals, and demodulating the micro doppler data obtained by the operation into micro distance spectrum data and micro angle spectrum data, specifically including:

Specifically, point cloud data is analyzed and modeled to obtain an environment model of environment static objects (walls, ceilings, furniture, household large-scale appliances, televisions, computers and the like), and the static objects in a detection space (FOV) are removed. As the measurement object is an ultralow frequency signal within 10Hz, a low pass filter algorithm LPF is introduced to obtain all low frequency signal data within 10Hz, the low pass filter LPF is carried out on the original point cloud data, and I-Q demodulation is analyzed to obtain an amplitude RSS frequency spectrum and a phase phi frequency spectrum. Here the amplitude is a function of distance and the phase is a function of angle. The micro-distance spectrum and the angle spectrum are converted by the amplitude RSS spectrum and the phase phi spectrum to determine the polar coordinates of each human body signal data.

Therefore, the interference of static objects in the monitoring space is eliminated, the interference of a plurality of complex signals existing in an actual monitoring scene can be further solved, and the accuracy and the referential performance of the heart rate and respiration monitoring result of the target human body are ensured.

Referring to fig. 4, removing animal-related signals in the millimeter wave sensing signals specifically includes:

Specifically, the RSS amplitude and angular frequency ω of the signal data in each set of each scan are analyzed, RSS thresholds and ω thresholds are set, and all periodic signal data meeting the thresholds are retained. And performing energy spectrum integral summation of amplitude RSS on the analyzed signal data of each period, and comparing the energy spectrum integral summation with a preset reflection amplitude RSS energy spectrum integral summation value of an adult to eliminate the numerical interference of animals such as pets and the like.

Therefore, interference of animals such as pets in the monitoring space is eliminated, interference of a plurality of complex signals existing in an actual monitoring scene can be further solved, and accuracy and referential of the heart rate and respiration monitoring result of the target human body are guaranteed.

Referring to fig. 5, the process of I-Q demodulating the micro-doppler data is as follows:

Data f (n) and sin (ω) are multiplied by a first multiplier_an), multiplying, filtering by adopting a low-pass filtering algorithm LPF after multiplying, and extracting data by adopting a data extraction algorithm in integral multiple after filtering to obtain data I (n);

Referring to fig. 6, the constructing a distance set and an angle set based on the separated micro-distance spectrum data and micro-angle spectrum data specifically includes:

Specifically, Hamming function (Hamming) weighting operation is carried out on the I-Q demodulation data so as to inhibit the influence of side lobe data and improve the signal-to-noise ratio of main lobe data. And performing FFT inverse transformation on the demodulated data to form a function D of the distance pulse and a function phi of the angle pulse. All data of the function points are subjected to moving object discrimination (MTI) filtering. Since the heart rate signal is superimposed on the respiration signal, the two are at different frequencies. The respiratory frequency is in the range of 0.1Hz-0.5Hz, and the heartbeat frequency is in the range of 0.8Hz-4 Hz. The two types of signals are separated by a digital band-pass filter (4th-order IIR (infinite impulse response) wideband-quad digital filter). And weighting the distance function and the angle function by a Hamming (Hamming) function to perform distance FFT and angle FFT. Different sets of distances and angles are constructed. Within each set is different human heart rate respiration data. The human body data of different distances are in different distance sets; the human body data within the same distance set will be distinguished by different angle sets.

Therefore, the signal-to-noise ratio of demodulation data is improved, the interference of a plurality of complex signals existing in an actual monitoring scene can be further solved, and the accuracy and the referential performance of the target human heart rate and respiration monitoring result are ensured.

Referring to fig. 7, a digital band-pass filter is used to separate the respiration signal and the heartbeat signal, the digital band-pass filter includes a first adder and a second adder, an input signal x (n) is input to an input end of the first adder, an output end of the first adder outputs a filtered signal Y '(n), the filtered signal Y' (n) is input to an input end of the second adder, an output end of the second adder outputs a signal Y (n), and an algorithm of the digital band-pass filter is as follows:

In the millimeter wave-based non-contact far-field multi-human-body respiration and heart rate monitoring method, the respective respiratory frequencies of the multiple human bodies in each distance set and each angle set are calculated based on wavelet transformation, and the respective heartbeat frequencies of the multiple human bodies in each distance set and each angle set are calculated based on a clustering density algorithm, and the method specifically includes the following steps:

Referring to fig. 8, the calculating the respective heartbeat frequency of multiple human bodies in each distance set and each angle set based on the clustering density algorithm specifically includes:

Specifically, an RNN/LSTM neural network model is used to determine the position of each human body within the FOV of the detection space. And recording each data peak value in the distance set, and eliminating all aperiodic interference signals caused by the limb movement and walking of the human body according to the periodicity of the peak values in a specific time window T. And (4) carrying out classification and identification on the signal data by applying a KNN pattern recognition classifier algorithm. In the time window T, a wavelet transform WT algorithm is applied to calculate periodic peak values in accordance with the respiratory frequency spectrum, and human respiratory frequency in a corresponding distance set and an angle set is obtained. Because the heartbeat signal is far weaker than the respiration signal and is easy to be interfered by various factors, for peak value calculation in specific time T, a simple method for calculating a periodic peak value is not adopted, but a cluster density algorithm (DBSCAN) is adopted, different peak value data sets of FFT (fast Fourier transform) after heart rate band-pass digital filtering are classified, the maximum and minimum peak value density sets are removed, the middle peak value density set is taken as a reference, the peak value period in the set is calculated, and the human body heartbeat frequency in different distance sets and angle sets is obtained.

Therefore, all aperiodic interference signals caused by limb movement and walking of the human body are removed, the interference of a plurality of complex signals existing in an actual monitoring scene can be solved, and the accuracy and the referential performance of the heart rate and respiration monitoring result of the target human body are further improved.

In specific implementation, one or more millimeter wave sensors access a local server or a cloud server through a local area network, a wireless local area network or an operator network. The obtained various information of the human body is connected into the server through signal communication, and the server is connected with the user operation and maintenance platform and the user mobile equipment.

Referring to fig. 9, the method for constructing the static environment model in the space environment is as follows:

Specifically, an initial physical value of the spatial static object is analyzed and calculated according to the preprocessed initial millimeter wave sensing signal, and a spatial environment static object model is constructed based on the initial physical value. Specifically, the initial physical values include the distance, angle, and signal strength amplitude RSS of the initial millimeter wave sensing signal of the environmental still in the monitored space. The environmental still includes sofas, electrical appliances, walls, ceilings, furniture, computers, televisions and the like. Therefore, the static object data in the monitoring space can be stored in advance, and the static objects in the environment in the monitoring space are shielded, so that the identification accuracy of the target human body position and the accuracy of counting a plurality of human bodies are improved.

Referring to fig. 10, the MTI moving object screening digital filtering is adopted, and the specific screening process includes:

acquiring a time domain digital signal x (T) at time T, and filtering and outputting y (T) ═ x (T) - (1-K) w (T), wherein the delay T weighting function w (T) is the delay w (T) ═ v (T-T) of the mixing function v (T), and the mixing function v (T) ═ y (T) + w (T).

Therefore, a plurality of human body targets can be discriminated, and therefore the recognition accuracy of each human body position and the accuracy of counting a plurality of human bodies are improved.

Referring to fig. 11, the neural network adopts a long-short term memory recurrent neural network LSTM, and the learning process of the LSTM neural network is as follows:

s101: acquiring a learning type data vector Xk at the moment k, using the learning type data vector Xk as an input layer, and determining an input weight vector U by a weight value W;

s102: determining a vector function Sk ═ f (Uk · Xk + W · Sk-1) of the hidden layer at the moment k; wherein Uk represents an input weight vector at the moment k, and Sk-1 represents a vector function of a hidden layer at the moment k-1;

s103: setting an output weight vector V, and determining a vector function Ok of an output layer at the moment k as g (V & Sk);

s104: after the learning of the data at time k is completed, steps S101 to S103 are executed in a loop.

Therefore, the positions of all target human bodies to be identified in the monitoring space can be effectively analyzed through the long-term and short-term memory recurrent neural network LSTM, the spatial position identification accuracy is high, and the identification result is stable.

Referring to fig. 12, the pattern recognition classification learner employs a KNN classifier, and the recognition process of the KNN classifier is as follows:

acquiring a data set in a subset space, and calculating data samples based on an Euclidean distance function;

acquiring k nearest training samples in the data samples, and performing weighted average on the k samples based on the distance;

and selecting the category with the most occurrence in the k samples, taking the obtained weighted average as a corresponding category, and outputting the identified category as the corresponding posture.

Therefore, the KNN classifier is used for classifying and aggregating a plurality of target human bodies, micro-motion human bodies, space static objects and environmental interferents, so that the number of the target human bodies and the micro-motion human bodies is determined, and the counting accuracy of the target human bodies is improved.

Referring to fig. 13, performing signal digital filtering on the time domain digital signal specifically includes:

s201: setting digital filtering parameters, and carrying out anti-interference mean digital filtering on the time domain digital signals of the millimeter wave sensing signals;

s202: predicting data at the K +1 th moment by the data at the K th moment, and estimating a prediction error at the K +1 th moment by the prediction error at the K th moment;

s203: calculating Kalman gain according to the data at the K moment and the prediction data at the K +1 moment, calculating the optimal estimation value of the data, and calculating the prediction error of the current moment K;

s204: step S202 and step S203 are looped.

In this way, the interference signal in the millimeter wave monitoring signal can be preliminarily filtered.

Referring to fig. 14, the performing spatial multi-path interference crosstalk cancellation on the time domain digital signal specifically includes:

s301: is obtained whenTime domain digital signal S of millimeter wave received after transmission of previous moment K_KCalculating the weight Q of the current time K_K；

S302: acquiring time domain digital signal S of millimeter wave transmitted at K moment and received by K +1_K+1Calculating the weight Q at the time K +1_K+1；：

S303: generating a multipath interference cancellation amount: Δ S ═ S_K·Q_K-S_K+1·Q_K+1And calculating effective data after interference cancellation: s ═ S_K-ΔS；

S304: and (6) looping the steps S301 to S303 until all data converge.

In this way, the interference signal and the crosstalk signal in the millimeter wave monitoring signal can be effectively eliminated.

Referring to fig. 15, the spatial noise processing is performed on the time domain digital signal of the millimeter wave monitoring signal, and the specific method is as follows:

carrying out autocorrelation digital noise signal monitoring and cross-correlation digital noise signal monitoring on the time domain digital signal of the millimeter wave monitoring signal, and screening out a digital noise signal;

and calculating the phase difference time domain of the digital noise signal, introducing the digital noise signal into a delayer, introducing the output signal of the delayer and the antecedent noise signal into a multiplier, introducing the output signal of the multiplier into an integrator, introducing the output signal of the integrator into a digital FIR filter, and outputting a digital noise function.

In this way, the noise signal in the millimeter wave monitor signal can be effectively eliminated. For millimeter wave signals collected in a monitoring space, the preprocessing of the millimeter wave signals is sequentially subjected to the digital filtering, the spatial multipath interference elimination, the spatial noise processing and the like, so that interference signals, noise signals, obstacle signals and the like in the millimeter wave monitoring signals can be effectively eliminated. And then powerful data support is provided for subsequent target identification and target state analysis, the sensitivity and the agility of target identification are ensured, the accuracy and the reliability of a target monitoring result are improved, and the anti-interference capability is strong.

Referring to fig. 17, the millimeter wave-based non-contact far-field multi-human body respiration and heart rate monitoring method has the following specific principle flow:

firstly, an operator controls the scanning of a microstrip antenna emission group through a monitoring server or a user terminal, the intermediate frequency received signal IF after mixing, ADC sampling, primary signal filtering of the signal, autocorrelation function and cross-correlation function analysis of data scanned in each group, noise model modeling, denoising elimination and multipath interference and crosstalk elimination filtering. Therefore, the millimeter wave sensing equipment adopts the time-sharing multiplexing MINO antenna structure with multiple groups of transmitting and receiving, the weak signal pickup capability and the azimuth resolution of the system can be improved, and the effective detection of the far-field (more than five meters) weak signals can be realized. And the interference of a plurality of complex signals existing in an actual monitoring scene can be solved through digital filtering processing, spatial multipath interference crosstalk elimination and secondary elimination of possible environmental noise and harmonic noise, and the accuracy and the referential of a target human heart rate and respiration monitoring result are ensured.

Then, Doppler and micro Doppler operations are carried out on the data, an environment static object model is established, static object and Doppler-effect-free data are shielded, I-Q demodulation is carried out on the micro Doppler data, an analytic amplitude and an analytic phase are obtained, the amplitude phase is converted into a distance and an angle, the polar coordinates of human body signal data are determined, a threshold value is set for the amplitude and the angular frequency of the analytic data, all period data in the threshold value are reserved, integral summation is carried out on the analytic amplitude energy spectrum, and pet interference data are removed. Therefore, the interference of animals such as static objects and pets in the monitoring space is eliminated, the signal-to-noise ratio of demodulation data is improved, the interference of a plurality of complex signals existing in an actual monitoring scene can be further solved, and the accuracy and the referential performance of the target human heart rate and respiration monitoring result are ensured.

Then, carrying out Hamming function weighting on the demodulated data, carrying out FFT inverse transformation on the demodulated phase data, carrying out a distance pulse domain function and an angle pulse domain function, carrying out a moving target discrimination filter MTI on each data, carrying out Hamming function weighting processing on each data, and carrying out distance FFT and angle FFT on each data so as to obtain distance data sets 1-n and angle data sets 1-n, and obtain peak data of a human body 1 and peak data of a human body n; determining the position of a human body through an RNN/LSTM machine learning model, retaining periodic peak data, eliminating aperiodic peak data caused by sudden movement and sudden ambulation of limbs, and classifying signal data by a KNN classifier; and calculating effective periodic peak values in a T time window to estimate the respiratory frequency of the human body based on the respiratory bandpass digital filtering of 0.1Hz-0.5Hz and the heartbeat bandpass digital filtering of 0.8Hz-4Hz through a wavelet transform algorithm WT, and calculating the peak value density of the human body heartbeat channel by applying a clustering density algorithm DBSCAN. And (5) estimating the periodic peak number in the time window T of the density set by adopting the median density value to obtain the heartbeat frequency. Therefore, all aperiodic interference signals caused by limb movement and walking of the human body are removed, the interference of a plurality of complex signals existing in an actual monitoring scene can be solved, and the accuracy and the referential performance of the heart rate and respiration monitoring result of the target human body are further improved.

And finally, transmitting the respiratory frequency, the heartbeat frequency and the spatial position information of each target human body to a monitoring server or a user side in real time.

According to the millimeter wave-based non-contact far-field multi-human-body respiration and heart rate monitoring method, the distance sets and the angle sets are constructed, then the respective respiration frequencies of the multi-target human bodies in each distance set and each angle set are calculated based on wavelet transformation, and the respective heartbeat frequencies of the multi-target human bodies in each distance set and each angle set are calculated based on a clustering density algorithm; therefore, the breathing and the heart rate of the human body can be monitored in real time at a long distance of more than five meters on the premise of ensuring the detection accuracy. And the spatial position of each target human body is determined through the analysis of a neural network model, the signal data of each target human body is classified and identified by adopting a mode identification classifier, and the target human body is screened by adopting MTI moving target screening digital filtering, so that the aim of effectively distinguishing, identifying and monitoring a plurality of targets is realized. And the interference of a plurality of complex signals in an actual monitoring scene is solved by preprocessing, shielding sensing signals of static objects in space and sensing signals without Doppler effect, eliminating animal related signals in the millimeter wave sensing signals and eliminating all non-periodic interference signals caused by limb movement and walking of a human body. The method has the advantages of remote monitoring, simultaneous monitoring of multiple target human bodies, high accuracy of human heart rate and respiration monitoring results, great referential significance of the monitoring results and the like.

Finally, the above embodiments are only used for illustrating the technical solutions of the present invention and not for limiting, although the present invention is described in detail with reference to the embodiments, it should be understood by those skilled in the art that modifications or equivalent substitutions can be made on the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all of them should be covered in the protection scope of the present invention.

Claims

1. A millimeter wave-based non-contact far-field multi-human-body respiration heart rate monitoring method is characterized by comprising the following steps:

2. The millimeter wave-based non-contact far-field multi-human-body respiration and heart rate monitoring method according to claim 1, wherein the preprocessing of the received millimeter wave sensing signal specifically comprises:

3. The millimeter wave-based non-contact far-field multi-human-body respiration and heart rate monitoring method according to claim 1, wherein the micro-doppler operation is performed on the preprocessed millimeter wave sensing signals, and the micro-doppler data obtained by the operation is demodulated into micro-distance spectrum data and micro-angle spectrum data, and specifically comprises:

4. The millimeter wave-based non-contact far-field multi-human-body respiration and heart rate monitoring method according to claim 3, wherein the elimination of animal-related signals in the millimeter wave sensing signals specifically comprises:

5. The millimeter wave-based non-contact far-field multi-human body respiration and heart rate monitoring method according to claim 4, wherein the I-Q demodulation process of the micro-Doppler data is as follows:

6. The millimeter wave-based non-contact far-field multi-human-body respiration and heart rate monitoring method according to claim 1, wherein the distance set and the angle set are constructed based on the separated micro-distance spectrum data and micro-angle spectrum data, and specifically comprises:

7. The millimeter wave based non-contact far-field multi-human body respiration and heart rate monitoring method according to claim 6, wherein a digital band-pass filter is used to separate the respiration signal and the heart rate signal, the digital band-pass filter includes a first adder and a second adder, the input signal X (n) is input to the input end of the first adder, the output end of the first adder outputs the filtered signal Y '(n), the filtered signal Y' (n) is input to the input end of the second adder, the output end of the second adder outputs the signal Y (n), and the algorithm of the digital band-pass filter is as follows:

8. The millimeter wave-based non-contact far-field multi-human respiration and heart rate monitoring method according to claim 1, wherein the method for calculating the respective respiratory frequencies of the multiple human bodies in each distance set and each angle set based on the wavelet transform and calculating the respective heart beat frequencies of the multiple human bodies in each distance set and each angle set based on a clustering density algorithm specifically comprises:

9. The millimeter wave-based non-contact far-field multi-human respiration and heart rate monitoring method according to claim 8, wherein the calculating of the respective heart rate frequencies of the multiple humans in each distance set and angle set based on a cluster density algorithm specifically comprises:

10. The millimeter wave-based non-contact far-field multi-human body respiration and heart rate monitoring method according to claim 3, wherein the spatial environment static model is constructed by the following steps: