key: cord-0761598-apwrjy7g
authors: Guo, Kai; Liu, Chang; Zhao, Shasha; Lu, Jingxin; Zhang, Senhao; Yang, Hongbo
title: Design of a Millimeter-Wave Radar Remote Monitoring System for the Elderly Living Alone Using WIFI Communication
date: 2021-11-26
journal: Sensors (Basel)
DOI: 10.3390/s21237893
sha: a91c20eb257a027fa81c2ea4cacd7a8619251439
doc_id: 761598
cord_uid: apwrjy7g

In response to the current demand for the remote monitoring of older people living alone, a non-contact human vital signs monitoring system based on millimeter wave radar has gradually become the object of research. This paper mainly carried out research regarding the detection method to obtain human breathing and heartbeat signals using a frequency modulated continuous wave system. We completed a portable millimeter-wave radar module for wireless communication. The radar module was a small size and had a WIFI communication interface, so we only needed to provide a power cord for the radar module. The breathing and heartbeat signals were detected and separated by FIR digital filter and the wavelet transform method. By building a cloud computing framework, we realized remote and senseless monitoring of the vital signs for older people living alone. Experiments were also carried out to compare the performance difference between the system and the common contact detection system. The experimental results showed that the life parameter detection system based on the millimeter wave sensor has strong real-time performance and accuracy.

Most existing measuring instruments require physical contact; they need to be attached to the patient for measurement and monitoring. Thus, these measuring instruments are not convenient for patients who require continuous monitoring for a long time. Moreover, in light of the current COVID-19 pandemic, non-contact vital signs monitoring equipment could become more important. Non-contact monitoring could help minimize the spread of the virus, and the use of non-contact monitoring methods could ensure the safety of health care personnel. Therefore, remote, non-contact health monitoring instruments are urgently needed [1] .

As the elderly population continues to rise, the population model has gradually shown an "inverted pyramid" pattern. In view of the situation where there is no one to take care of the elderly living alone at home, monitoring the elderly and their daily routines will become particularly important [2] . The elderly population needs care, and their living conditions require real-time monitoring using professional equipment.

Therefore, there is a great social value to study indoor safety monitoring terminals that can detect the daily activity routines of the elderly and ensure the daily safety of the elderly living alone [3] .

Millimeter wave is a new type of non-contact life signal detection method, which can detect the relevant life parameter signals produced by the human body due to heart and lung activities [4] . Compared to traditional ECG and pulse detectors, it has the used differential and cross multiply (DACM) algorithm to effectively solve the problem of arctangent demodulation phase interruption, thus improving the stability of physical sign detection. In 2015, Wu et al. proposed a spectrum-weighted accumulative method to improve the SNR of received signals. In 2018, Liang et al. proposed a vital sign monitoring system based on UWB radar, which utilizes the short-time Fourier Transform (STFT), Additionally, through the collection of empirical mode decomposition (EEMD Ensemble Empirical Mode Decomposition) spectrum to detect vital signs parameters, the method to solve the disadvantages of the EMD method, improved the detection accuracy [18] .

This paper mainly studies the detection and separation methods of human breathing and heartbeat signals under the FM continuous-wave system [19] .

The software and hardware design of the millimeter wave sensor system was mainly completed, and the FIR digital filter and the wavelet transform method were used to detect and separate breathing and heartbeat signals. On this basis, a method of suppressing respiratory harmonics based on adaptive filters was proposed for the interference of respiratory harmonics in the heartbeat signal. Through the transplantation of algorithms and the construction of software systems, we used cloud services to realize remote online calculation of vital signs parameters. With the smaller size of the radar module, vital signs monitoring can be realized without touch.

Breathing can cause chest wall displacement; the displacement was about 1~12 mm, and the breathing frequency range was 0.1~0.5 Hz; the heartbeat can cause the chest wall displacement to change 1~2 mm, and the heartbeat frequency range was 0.8~2.5 Hz [20] (see Table 1 ). [20] [21] [22] .

Breathing signal 1-12 mm 0.1-0.5 Hz

Heartbeat signal 1-2 mm 0. Breathing and heartbeat movements can cause tiny vibrations in the chest wall, and millimeter wave radar is able to detect this tiny displacement through the phase change of the signal [21] . So it can be used for predicting respiratory rate and heart rate [20] (see Table 1 ).

Traditional medical testing methods used thoracic cavity palpation and observation methods to detect breathing, but subjects may intentionally change their breathing rate and pattern when they perceive the measurement. Therefore, the use of non-contact methods to measure breathing frequency has great practical value [22] .

Millimeter wave is a special class of radar technology that uses short-wavelength electromagnetic waves. The electromagnetic wave signal emitted by the radar system is reflected by other objects in the emission path, and by capturing the reflected signal, the radar system can determine information such as the distance, speed, and angle of the object. Since the target echo distance information of the LFMCW system radar can be simply processed using the Fast Fourier Transform (FFT). This was the most widely used modulation method, and has been used for the most in-depth research on chirp continuous wave radar. The LFMCW radar system emits linear FM pulse signals and captures the signals reflected by objects in its emission path. Figure 1 is a simplified block diagram of the LFMCW radar system. The system working principle was as follows: first the signal source generated a linear FM pulse which was transmitted by the transmitting antenna; the object The "mixer" was used to mix the receiving end (RX) and transmitting end (TX) signals to generate an intermediate frequency (IF) signal. The output of the mixer contained two kinds of signals, which were the sum and the difference between the Rx and Tx chirp frequencies. There was also a low-pass filter used to limit the signal, allowing only signals with a difference in frequency to pass.

The role of the mixer was to combine two input signals into a signal with a new frequency for two sinusoidal input signals and ： = sin( + ∅ )

The instantaneous frequency of the output signal was the difference of the instantaneous frequency of the two input signals, and the phase is equal to the difference of the phase of the two input signals:

In the FMCW radar system, the frequency of the emission signal increases linearly over time, and this type of signal was also known as the linear FM pulse signal. Figure 2 was a function of the amplitude of the linear FM pulse varying over time. The "mixer" was used to mix the receiving end (RX) and transmitting end (TX) signals to generate an intermediate frequency (IF) signal. The output of the mixer contained two kinds of signals, which were the sum and the difference between the Rx and Tx chirp frequencies. There was also a low-pass filter used to limit the signal, allowing only signals with a difference in frequency to pass.

The role of the mixer was to combine two input signals into a signal with a new frequency for two sinusoidal input signals x 1 and x 2 :

The instantaneous frequency of the output signal x out was the difference of the instantaneous frequency of the two input signals, and the phase is equal to the difference of the phase of the two input signals:

In the FMCW radar system, the frequency of the emission signal increases linearly over time, and this type of signal was also known as the linear FM pulse signal. Figure 2 was a function of the amplitude of the linear FM pulse varying over time. Figure 2 showed the radar waveform used for life signal detection in this paper. The radar waveform used was the sawtooth wave, T c was the sawtooth wave period, and B was the frequency modulation bandwidth. Within a sawtooth wave period, the radar signal emitted could be expressed as:

In Formulas (2)-(4), f c was the center frequency of the radar transmitted signal, γ was the slope of sawtooth wave, and ϕ was the initial phase of the transmitted signal. The echo reflected after the radar signal irradiated on the chest wall of a human body can be expressed as: Figure 2 showed the radar waveform used for life signal detection in this paper. The radar waveform used was the sawtooth wave,T was the sawtooth wave period, and B was the frequency modulation bandwidth. Within a sawtooth wave period, the radar signal emitted could be expressed as:

In Formulas (2)-(4), f was the center frequency of the radar transmitted signal,γ was the slope of sawtooth wave, and φ was the initial phase of the transmitted signal. The echo reflected after the radar signal irradiated on the chest wall of a human body can be expressed as:

In Formulas (2)-(5), represented the speed of light and was the echo signal amplitude, a value related to the distance from the target to the radar and the radar reflection cross-sectional area (RCS), which was inversely proportional to the distance from the target to the radar and inversely proportional to the RCS. The RCS value of radar was related to the shape, size, material and other factors of the measured target.2 ( )/C is the delay of radar echo signal, the mathematical expression of ( ) is:

In Formulas (2)-(6), was the distance from the radar antenna to the target pleural motion center, and was the slow time increasing with the repetitions of the sawtooth wave. Radar echo signal was mixed with local oscillator signal to obtain an intermediate frequency signal: In Formulas (2)-(5), C represented the speed of light and σ was the echo signal amplitude, a value related to the distance from the target to the radar and the radar reflection cross-sectional area (RCS), which was inversely proportional to the distance from the target to the radar and inversely proportional to the RCS. The RCS value of radar was related to the shape, size, material and other factors of the measured target. 2R 1 (τ)/C is the delay of radar echo signal, the mathematical expression of R 1 (τ) is:

In Formulas (2)-(6), d 0 was the distance from the radar antenna to the target pleural motion center, and τ was the slow time increasing with the repetitions of the sawtooth wave. Radar echo signal was mixed with local oscillator signal to obtain an intermediate frequency signal:

According to Formulas (2)-(8), the frequency of the intermediate frequency signal was f b , and the phase was ϕ b . Since the value of 4πγR 2 1 (τ) C 2 , 4πγR 2 1 (τ)/C 2 , was very small and can be ignored. Therefore, the phase ϕ b of the intermediate frequency signal could be further simplified as

According to Equations (2)-(8) and (2)-(10), it could be seen that both frequency f b and phase ϕ b of radar intermediate frequency signal S IF (t) contain displacement information R 1 (τ) of detecting target chest cavity. IF frequency f b could be obtained by FFT of radar IF signal, but the maximum frequency resolution that FFT can achieve at this time was 1/T c . Put into the formula, it can be shown that the maximum range resolution that FFT of radar IF signal can obtain was C/2B, because the chest displacement caused by respiration and heartbeat was very small. Therefore, it was difficult to obtain the chest displacement change of the measured target directly through the radar intermediate frequency signal

Use radar intermediate-frequency signal measured target to get the chest cavity displacement change law of each of the sawtooth sampling points as a row of the matrix, you can get a matrix (m for sawtooth wave number, n sampling points of each tooth), Vital signals related to breathing and heartbeat rates can then be obtained. The overall framework of the millimeter wave radar hardware is shown in Figure 3 . The system constructed a complete modular circuit based on STM32f103 chip. The circuit mainly includes:

(1) Power supply module;

(2) TI C3220 WIFI communication module;

(3) IWR6843 mm-wave sensor; (4) UART serial port interface for communication with the millimeter wave sensor module command system; TI CC3220 WIFI communicates with the millimeter wave sensor module data transmission interface to obtain raw data of the millimeter wave radar for back end data processing.

According to Equations (2)-(8) and (2)-(10), it could be seen that both frequency f and phase φ of radar intermediate frequency signal S (t) contain displacement information R (τ) of detecting target chest cavity. IF frequency f could be obtained by FFT of radar IF signal, but the maximum frequency resolution that FFT can achieve at this time was 1/T . Put into the formula, it can be shown that the maximum range resolution that FFT of radar IF signal can obtain was C/2B, because the chest displacement caused by respiration and heartbeat was very small. Therefore, it was difficult to obtain the chest displacement change of the measured target directly through the radar intermediate frequency signal frequency f .

Use radar intermediate-frequency signal measured target to get the chest cavity displacement change law of each of the sawtooth sampling points as a row of the matrix, you can get a matrix (m for sawtooth wave number, n sampling points of each tooth), Vital signals related to breathing and heartbeat rates can then be obtained.

The overall framework of the millimeter wave radar hardware is shown in Figure 3 . The system constructed a complete modular circuit based on STM32f103 chip. The circuit mainly includes:

(1) Power supply module;

(2) TI C3220 WIFI communication module;

(3) IWR6843 mm-wave sensor; (4) UART serial port interface for communication with the millimeter wave sensor module command system; TI CC3220 WIFI communicates with the millimeter wave sensor module data transmission interface to obtain raw data of the millimeter wave radar for back end data processing. 

The transceiver circuit was built based on the transceiver integrated chip IWR6843 of TI Company. The IWR6843 was an integrated single-chip mmwave sensor based on FMCW radar technology capable of operation in the 60-GHz to 64-GHz band. It was built 

The transceiver circuit was built based on the transceiver integrated chip IWR6843 of TI Company. The IWR6843 was an integrated single-chip mmwave sensor based on FMCW radar technology capable of operation in the 60-GHz to 64-GHz band. It was built with TI 45 nm RFCMOS process and enabled unprecedented levels of integration in an extremely small form factor. The IWR6843 was an ideal solution for low power, self-monitored, ultra-accurate radar systems in the industrial space. It had three transmitting channels and four receiving channels. The antenna design is shown in Figure 4 . with TI 45 nm RFCMOS process and enabled unprecedented levels of integration in an extremely small form factor. The IWR6843 was an ideal solution for low power, self-monitored, ultra-accurate radar systems in the industrial space. It had three transmitting channels and four receiving channels. The antenna design is shown in Figure 4 . The emission system consists of three parallel emission links, each with an independent binary phase as well as amplitude control (see Figure 5 ). The maximum output power for each emission link was 12 dBm, amplitude noise up to −145 dBc/Hz. The emission system consists of three parallel emission links, each with an independent binary phase as well as amplitude control (see Figure 5 ). The maximum output power for each emission link was 12 dBm, amplitude noise up to −145 dBc/Hz. The emission system consists of three parallel emission links, each with an independent binary phase as well as amplitude control (see Figure 5 ). The maximum output power for each emission link was 12 dBm, amplitude noise up to −145 dBc/Hz. The receiving system consists of four parallel channels, each including a low noise amplifier (LNA), mixer, intermediate frequency filter, and ADC (see Figure 6 ). All four receiving channels can be run simultaneously. Compared with the traditional mixer, the orthogonal mixing can effectively inhibit the mirror interference, while dividing the IF signals into real and virtual parts, thus reducing the ADC sampling rate. The receiver coefficient of noise was 15 dB, IF gain range of 24~48 dB, step 2dB, IF bandwidth of 5 MHz, the maximum sampling rate of 12.5 MHz. The receiving system consists of four parallel channels, each including a low noise amplifier (LNA), mixer, intermediate frequency filter, and ADC (see Figure 6 ). All four receiving channels can be run simultaneously. Compared with the traditional mixer, the orthogonal mixing can effectively inhibit the mirror interference, while dividing the IF signals into real and virtual parts, thus reducing the ADC sampling rate. The receiver coefficient of noise was 15 dB, IF gain range of 24~48 dB, step 2dB, IF bandwidth of 5 MHz, the maximum sampling rate of 12.5 MHz. The emission system consists of three parallel emission links, each with an independent binary phase as well as amplitude control (see Figure 5 ). The maximum output power for each emission link was 12 dBm, amplitude noise up to −145 dBc/Hz. The receiving system consists of four parallel channels, each including a low noise amplifier (LNA), mixer, intermediate frequency filter, and ADC (see Figure 6 ). All four receiving channels can be run simultaneously. Compared with the traditional mixer, the orthogonal mixing can effectively inhibit the mirror interference, while dividing the IF signals into real and virtual parts, thus reducing the ADC sampling rate. The receiver coefficient of noise was 15 dB, IF gain range of 24~48 dB, step 2dB, IF bandwidth of 5 MHz, the maximum sampling rate of 12.5 MHz. The peripheral circuit included power management, external FLASH, high-speed interface, and other peripheral modules. The power management scheme was used to divide the input voltage into multiple channels by switching power supply, and further reduce the voltage by cascade low-voltage differential linear regulator (LDO). At the same time, π type filter was added between different levels to reduce noise. Finally, four outputs were formed, corresponding to the IO voltage of the main chip, the analog power supply voltage, the RF transceiver voltage and the core and SRAM voltage.

We customized the radar core board from Changsha Ruigan Company, which provided the radar core board with customized interface based on our needs. We designed and welded the WIFI data board of the radar by ourselves, and there is no similar sample on the market. The final data transmission circuit board based on WIFI communication is shown in Figures 7 and 8 .

We customized the radar core board from Changsha Ruigan vided the radar core board with customized interface based on o and welded the WIFI data board of the radar by ourselves, and th on the market. The final data transmission circuit board based on shown in Figures 7 and 8 . After we connect the core board and the Wi-Fi communicat the overall WIFI communication-based millimeter wave radar vit tem is shown in Figure 9 . The size of the whole hardware system w 

We customized the radar core board from Changsha Ruigan Company, which pr vided the radar core board with customized interface based on our needs. We design and welded the WIFI data board of the radar by ourselves, and there is no similar samp on the market. The final data transmission circuit board based on WIFI communication shown in Figures 7 and 8 . After we connect the core board and the Wi-Fi communication board we designe the overall WIFI communication-based millimeter wave radar vital signs monitoring sy tem is shown in Figure 9 . The size of the whole hardware system was 34.0 × 32.0 × 10 m Figure 7 is the schematic diagram of the circuit, the WIFI circuit board and the radar board were connected by connecting pins, on which power supply and UART serial port transmission were integrated.

After we connect the core board and the Wi-Fi communication board we designed, the overall WIFI communication-based millimeter wave radar vital signs monitoring system is shown in Figure 9 . The size of the whole hardware system was 34.0 × 32.0 × 10 mm. The hardware mainly includes the core board and the WIFI data communication board. Firmware is an embedded program running on a circuit board, not software for algorithm calculations, so we generally treat firmware as part of the hardware. Two different hardware will have two different firmwares, the main contributions we made in the design and production are shown in Table 2 . The hardware mainly includes the core board and the WIFI data communication board. Firmware is an embedded program running on a circuit board, not software for algorithm calculations, so we generally treat firmware as part of the hardware. Two different hardware will have two different firmwares, the main contributions we made in the design and production are shown in Table 2 . From the above table, we mainly designed and produced the WIFI communication board, and the firmware of the two boards in the system was debugged, modified, and finalized by us.

In the LFMCW system, the frequency of the emission signal varies linearly over time. This periodic frequency scan was commonly referred to as chirp [23] [24] [25] [26] [27] , as a linear FM pulse.

The frame period was set to 50ms, and each frame contained two chirp structures per frame. During the data processing, the chirp of each frame was sampled, and then the FFT calculation of the distance dimension extracts the phase information of the distance unit where the object was being measured, which contains the chest wall vibration data, so reciprocated. Thus, for the original chest wall displacement data, the sampling rate was associated with the frame period, and the sampling rate of the original phase data was 20 Hz. In addition, according to the chip data manual, the gain range of the receiver was 24-48 dB, step 2 dB. Since this paper needs to detect minor shifts and has a low SNR, the intermediate frequency reception gain was set to a maximum of 48 dB. A schematic diagram of the emission waveform after the configuration is shown in Figure 10 . Sampling according to each frame cycle in the figure above, first ADC samples the median frequency signal according to the overall algorithm processing process, then distance dimension FFT processing, the target distance is calculated, and then a series of sub- Figure 10 . Emission waveform diagram [9] [10] [11] .

Sampling according to each frame cycle in the figure above, first ADC samples the median frequency signal according to the overall algorithm processing process, then distance dimension FFT processing, the target distance is calculated, and then a series of subsequent algorithm processing detection and separation of human respiratory and heartbeat signals [24] [25] [26] [27] .

The overall software workflow is shown in Figure 11 , and the software system architecture based on cloud services is shown in Figure 12 .

Object Range Bin Extract phase and unwrap for the object range bin Further processing Figure 10 . Emission waveform diagram [9] [10] [11] .

Sampling according to each frame cycle in the figure above, first ADC samples the median frequency signal according to the overall algorithm processing process, then distance dimension FFT processing, the target distance is calculated, and then a series of subsequent algorithm processing detection and separation of human respiratory and heartbeat signals [24] [25] [26] [27] .

The overall software workflow is shown in Figure 11 , and the software system architecture based on cloud services is shown in Figure 12 . Figure 11 . Software flow chart of millimeter wave radar module.

Our system used a cloud service software architecture. The device did not need to be directly connected to the computer via a data cable. The data was sent directly to the server via Wi-Fi, and the server's program calculated the monitored data in real time by unpacking and computing the socket data. The data was saved on the server, and the software obtained and displayed the parameters in real time through POST and GET methods. The software system architecture diagram based on cloud service is shown in Figure 12 . Our cloud server used Alibaba Cloud's ECS server, and the server uses Ubuntu 18.04 system. The algorithm was transplanted and packaged into a dynamic link library of the Linux system (.so file). The netty framework was used on the server side to realize real-time data unpacking and algorithm calculation and store it in the database. The server used the RESTFUL framework of node.js, and the computer software obtained the calculated data through GET and post methods. As shown in Figure 12 , the algorithm must be run on a remote server for the system to implement cloud computing. Since the server system was Linux without an operation interface, we needed to transplant the algorithm. Considering the data was transmitted to the server through a socket, the socket framework used by the server was called netty, which was in java environment, so we need to package the algorithm of the C++ language. The C++ language algorithm needed to replace and rewrite the MFP-specific functions, and then define all functions as "extern 'C' int __declspec(dllexport)" so that the packaged. So, the file can be called externally. We needed to transfer the transplanted algorithm file into the computing server environment used and package the C++ language algorithm into the Linux platform dynamic link library by running the "g++ linuxData-Calculate.cpp -fPIC -shared -o libDataCal_1.so" command under bash.

After the algorithm was transplanted, it combined the netty framework to realize real-time unpacking and calculation of the data, and the result was stored in the server. The display software could obtain real-time data through the RestFUL API. Our system used a cloud service software architecture. The device did not need to be directly connected to the computer via a data cable. The data was sent directly to the server via Wi-Fi, and the server's program calculated the monitored data in real time by unpacking and computing the socket data. The data was saved on the server, and the software obtained and displayed the parameters in real time through POST and GET methods. The software system architecture diagram based on cloud service is shown in Figure 12 . Our cloud server used Alibaba Cloud's ECS server, and the server uses Ubuntu 18.04 system. The algorithm was transplanted and packaged into a dynamic link library of the Linux system (.so file). The netty framework was used on the server side to realize real-time data unpacking and algorithm calculation and store it in the database. The server used the RESTFUL framework of node.js, and the computer software obtained the calculated data through GET and post methods.

As shown in Figure 12 , the algorithm must be run on a remote server for the system to implement cloud computing. Since the server system was Linux without an operation interface, we needed to transplant the algorithm. Considering the data was transmitted to the server through a socket, the socket framework used by the server was called netty, which was in java environment, so we need to package the algorithm of the C++ language. The C++ language algorithm needed to replace and rewrite the MFP-specific functions, and then define all functions as "extern 'C' int __declspec(dllexport)" so that the packaged. So, the file can be called externally. We needed to transfer the transplanted algorithm file into the computing server environment used and package the C++ language algorithm into the Linux platform dynamic link library by running the "g++ linuxDataCalculate.cpp -fPIC -shared -o libDataCal_1.so" command under bash.

After the algorithm was transplanted, it combined the netty framework to realize real-time unpacking and calculation of the data, and the result was stored in the server. The display software could obtain real-time data through the RestFUL API.

The software functions designed based on this algorithm are roughly introduced as follows:

(1) Control the start, pause, and end of data collection and transmission functions;

(2) Real-time display of time-domain waveforms of breathing and heartbeat signals;

(3) Spectral analysis of breathing and heartbeat signals (4) Estimate and display the respiratory rate and heart rate numerically.

The PC-terminal display interface design was shown in Figure 13 . As shown in Figure 12 , the algorithm must be run on a remote server for the system to implement cloud computing. Since the server system was Linux without an operation interface, we needed to transplant the algorithm. Considering the data was transmitted to the server through a socket, the socket framework used by the server was called netty, which was in java environment, so we need to package the algorithm of the C++ language. The C++ language algorithm needed to replace and rewrite the MFP-specific functions, and then define all functions as "extern 'C' int __declspec(dllexport)" so that the packaged. So, the file can be called externally. We needed to transfer the transplanted algorithm file into the computing server environment used and package the C++ language algorithm into the Linux platform dynamic link library by running the "g++ linuxData-Calculate.cpp -fPIC -shared -o libDataCal_1.so" command under bash.

After the algorithm was transplanted, it combined the netty framework to realize real-time unpacking and calculation of the data, and the result was stored in the server. The display software could obtain real-time data through the RestFUL API.

The software functions designed based on this algorithm are roughly introduced as follows:

(1) Control the start, pause, and end of data collection and transmission functions;

(2) Real-time display of time-domain waveforms of breathing and heartbeat signals;

(3) Spectral analysis of breathing and heartbeat signals (4) Estimate and display the respiratory rate and heart rate numerically.

The PC-terminal display interface design was shown in Figure 13 . In Section 2.4, the hardware and software design of the 60 GHz millimeter wave sensor was completed. The transceiver circuit was built based on the TI transceiver integrated chip IWR6843, and the transceiver system was composed of antennas. Design of peripheral circuits, including power supply, high-speed data interface, etc., constitutes a complete system. Then, the software configuration framework was designed according to the communication protocol format of the chip, and the RF and digital front-end configuration was carried out. The LFMCW sawtooth wave transmitting waveform is created, and the ADC sampling mode is configured to sample the front-end data, which makes preliminary preparation for the subsequent signal processing. Figure 14 shows the systematic process of the detection of the human heartbeat and respiratory signals. According to the whole algorithm processing process, the middle frequency signal was first sampled by ADC, and then the range dimension FFT processing to calculate the target distance. Next, FFT calculation was performed for each chirp to find the distance unit corresponding to the target. Then, the phase information corresponding to the distance unit was extracted for subsequent algorithm processing.

the ADC sampling mode is configured to sample the front-end data, which makes preliminary preparation for the subsequent signal processing. Figure 14 shows the systematic process of the detection of the human heartbeat and respiratory signals. According to the whole algorithm processing process, the middle frequency signal was first sampled by ADC, and then the range dimension FFT processing to calculate the target distance. Next, FFT calculation was performed for each chirp to find the distance unit corresponding to the target. Then, the phase information corresponding to the distance unit was extracted for subsequent algorithm processing. (1) The two band pass filters were set to 0.1-0.6 Hz and 0.8-4.0 Hz corresponding to the respiratory and heartbeat bands, respectively, where the respiratory filter start frequency was 0.1 H z to filter out the interference of DC noise; (2) Filters were designed using the fir1 function of order 41 and the window function selected a Hamming window; (3) The raw data was filtered using the filter function.

The FIR filter can effectively isolate the breathing signal. Although it has some noise cancellation effect on the heartbeat signal, the heartbeat frequency information cannot be extracted from it, so the chest wall displacement signal will be processed by wavelet transform to isolate the heartbeat signal. Wavelet analysis was performed using a series of functions in MATLAB's wavelet transformation toolkit, and the overall procedure was as follows:

(1) The appropriate wavelet basis function and decomposition layers are selected and one-dimensional discrete wavelet decomposition of the signal using the wavedec function; (2) Use approximate coefficient and detail coefficient to extract the decomposition though appcoef and detcoef functions; (3) The waverec function was reconstructed using the extracted approximation and detail coefficients to obtain the component sizes corresponding to the signal at different frequencies.

Section 2.4 briefly describes the signal detection and separation methods of human heartbeat and breathing. Section 3 will further verify the accuracy and stability of the system, and we conducted comparative experiments with traditional contact devices. (1) The two band pass filters were set to 0.1-0.6 Hz and 0.8-4.0 Hz corresponding to the respiratory and heartbeat bands, respectively, where the respiratory filter start frequency was 0.1 H z to filter out the interference of DC noise; (2) Filters were designed using the fir1 function of order 41 and the window function selected a Hamming window; (3) The raw data was filtered using the filter function.

The FIR filter can effectively isolate the breathing signal. Although it has some noise cancellation effect on the heartbeat signal, the heartbeat frequency information cannot be extracted from it, so the chest wall displacement signal will be processed by wavelet transform to isolate the heartbeat signal. Wavelet analysis was performed using a series of functions in MATLAB's wavelet transformation toolkit, and the overall procedure was as follows:

(1) The appropriate wavelet basis function and decomposition layers are selected and one-dimensional discrete wavelet decomposition of the signal using the wavedec function; (2) Use approximate coefficient and detail coefficient to extract the decomposition though appcoef and detcoef functions; (3) The waverec function was reconstructed using the extracted approximation and detail coefficients to obtain the component sizes corresponding to the signal at different frequencies.

Section 2.4 briefly describes the signal detection and separation methods of human heartbeat and breathing. Section 3 will further verify the accuracy and stability of the system, and we conducted comparative experiments with traditional contact devices.

This experiment was carried out in a laboratory environment, as shown in Figure 15 . Subjects were two healthy women aged 28 years and 26 years old, measured under normal respiration. The subjects were within normal detection range, facing the millimeter-wave sensor and chest facing the sensor. Since this experiment was non-contact detection, the measured environment and human body's movement were greatly disturbed, so it should be kept stationary as far as possible. There should be no metallic object between the sensor antenna and the subject.

Subjects were two healthy women aged 28 years and 26 years old, measured unde respiration. The subjects were within normal detection range, facing the millime sensor and chest facing the sensor. Since this experiment was non-contact detec measured environment and human body's movement were greatly disturbed, so be kept stationary as far as possible. There should be no metallic object between th antenna and the subject. The debugging program was downloaded to the IWR6843 through the softw Flash provided by TI. The tester sat in front of the radar and remained as still as during the test. The radar test distance can reach about 2.5 m. Because the single ECG module of the comparison device limited the measurement distance, the r distance is about 1 m. Figure 16 shows the measurements of the tester while mai normal breathing. 

The previous article completed the separation of breathing and heartbeat sig studied the influence of their harmonics. The relevant adaptive filtering algor been proposed for simulation and experimental verification. Furthermore, to v The debugging program was downloaded to the IWR6843 through the software UniFlash provided by TI. The tester sat in front of the radar and remained as still as possible during the test. The radar test distance can reach about 2.5 m. Because the singlechannel ECG module of the comparison device limited the measurement distance, the radar test distance is about 1 m. Figure 16 shows the measurements of the tester while maintaining normal breathing. measured environment and human body's movement were greatly disturbed, so it sh be kept stationary as far as possible. There should be no metallic object between the se antenna and the subject. The debugging program was downloaded to the IWR6843 through the software Flash provided by TI. The tester sat in front of the radar and remained as still as pos during the test. The radar test distance can reach about 2.5 m. Because the single-cha ECG module of the comparison device limited the measurement distance, the rada distance is about 1 m. Figure 16 shows the measurements of the tester while mainta normal breathing. 

The previous article completed the separation of breathing and heartbeat signal studied the influence of their harmonics. The relevant adaptive filtering algorithm been proposed for simulation and experimental verification. Furthermore, to verif 

The previous article completed the separation of breathing and heartbeat signals and studied the influence of their harmonics. The relevant adaptive filtering algorithm has been proposed for simulation and experimental verification. Furthermore, to verify the system's accuracy and stability, a comparative experiment was conducted with the conventional contact equipment. The selected contact device was Single channel ECG module. The investigation adopted the method of control variables to achieve synchronous detection. Two healthy subjects were measured in the normal breathing state, and two devices, millimeter wave sensor, and single channel ECG module were used for the experiment. Millimeter wave sensor will be detected by an algorithm based on the adaptive filter. A complete detection cycle was selected as a set of experimental data. In addition, respiration rate could not be detected in this experiment, so only heart rate data was compared. First of all, as shown in Figure 17 , the comparison experiment was set. Subjects were asked to hold the electrodes on both sides of the single channel ECG module with both hands under normal breathing state to measure heart rate, and at the same time, heart rate was measured with the millimeter wave sensor.

ule. The investigation adopted the method of control variable detection. Two healthy subjects were measured in the norma devices, millimeter wave sensor, and single channel ECG mod periment. Millimeter wave sensor will be detected by an algori filter. A complete detection cycle was selected as a set of expe respiration rate could not be detected in this experiment, so onl pared. First of all, as shown in Figure 17 , the comparison exp were asked to hold the electrodes on both sides of the single c both hands under normal breathing state to measure heart ra heart rate was measured with the millimeter wave sensor. Then, through the comparative experiment of millimeter w nel ECG module in Figure 18 , ECG signal and heartbeat signal ously. The figure shows the ECG and millimeter wave signals collected by the single lead ECG equipment. The heart rate calculated by millimeter wave radar was 63, and the heart rate calculated by medical single channel ECG device was 64.

Continuous collection and testing 60 times of heart rate ratio; two comparative experiments (see Figures 19 and 20) : The figure shows the ECG and millimeter wave signals collected by the single lead ECG equipment. The heart rate calculated by millimeter wave radar was 63, and the heart rate calculated by medical single channel ECG device was 64.

Continuous collection and testing 60 times of heart rate ratio; two comparative experiments (see Figures 19 and 20) : The figure shows the ECG and millimeter wave signals collected by the single lead ECG equipment. The heart rate calculated by millimeter wave radar was 63, and the hear rate calculated by medical single channel ECG device was 64.

Continuous collection and testing 60 times of heart rate ratio; two comparative ex periments (see Figures 19 and 20) : We use a medical contact respiratory acquisition device called sense-u to conduct comparative experiments with millimeter wave radar. We recorded the respiratory rate values of the two devices every 30 s, collected 30 times, and drew the comparison curve. The experimental and comparative results are shown in Figure 21 . 

As can be seen from the experimental results of tester 1 and tester 2 in FIG. 19 and  FIG. 20 , two healthy subjects were measured under normal breathing conditions, and millimeter-wave sensor and single channel ECG module were used for experiments simultaneously. The two heart rate data were overlapped and compared, and the error was within 6-7 bpm. You can see that the error fluctuation is slight. In the breathing comparison experiment in Figure 21 , there was a certain gap between the contact breathing frequency We use a medical contact respiratory acquisition device called sense-u to conduct comparative experiments with millimeter wave radar. We recorded the respiratory rate values of the two devices every 30 s, collected 30 times, and drew the comparison curve. The experimental and comparative results are shown in Figure 21 . We use a medical contact respiratory acquisition device called sense-u to conduct comparative experiments with millimeter wave radar. We recorded the respiratory rate values of the two devices every 30 s, collected 30 times, and drew the comparison curve. The experimental and comparative results are shown in Figure 21 . 

As can be seen from the experimental results of tester 1 and tester 2 in FIG. 19 and  FIG. 20 , two healthy subjects were measured under normal breathing conditions, and millimeter-wave sensor and single channel ECG module were used for experiments simultaneously. The two heart rate data were overlapped and compared, and the error was within 6-7 bpm. You can see that the error fluctuation is slight. In the breathing comparison experiment in Figure 21 , there was a certain gap between the contact breathing frequency 

As can be seen from the experimental results of tester 1 and tester 2 in FIG. 19 and FIG.  20 , two healthy subjects were measured under normal breathing conditions, and millimeterwave sensor and single channel ECG module were used for experiments simultaneously. The two heart rate data were overlapped and compared, and the error was within 6-7 bpm. You can see that the error fluctuation is slight. In the breathing comparison experiment in Figure 21 , there was a certain gap between the contact breathing frequency and the non-contact breathing frequency, but the basic trend was the same. The measurement results of the millimeter wave sensor based on the adaptive filtering algorithm were close to those of the Single channel ECG module, which proved the measurement accuracy and robustness of the system in this paper.

This paper designed a millimeter-wave radar remote monitoring system for the elderly living alone, based on WIFI communication; a real-time and efficient life detection system. This paper conducted a software and hardware construction of a millimeterwave sensor system and detected and separated it using an FIR digital filter and wavelet transform method for respiratory and heartbeat signals. The design of the system has good penetrating ability, can pass through obstacles such as clothes and bedding, is small in size, easy to integrate, and has higher efficiency. At the same time, combined with cloud computing, explore the possibility of remote vital signs monitoring by millimeter wave radar.

We hope to use our knowledge and technology and apply it to help disadvantaged groups in society. The system designed in this paper can realize real-time and efficient remote monitoring of vital signs (heart rate and breathing rate) of elderlies living alone, preventing them from having accidents without care or in guardianship loopholes. 

The data presented in this study are available on request from the corresponding author.

Noncontact Physiological Dynamics Detection Using Low-power Digital-IF Doppler Radar

Non-Contact Vital Signs Monitor

Wearable Millimeter-Wave Device for Contactless Measurement of Arterial Pulses

Surface Chest Motion Decomposition for Cardiovascular Monitoring

Millimeter-Wave Radars-on-Chip Enabling Next-Generation Cyberphysical Infrastructures

Human tracking and identification through a millimeter wave radar

Adaptive Virtual Waveform Design for Millimeter-Wave Joint Communication-Radar

Acceleration of FPGA based Convolutional Neural Network for Human Activity Classification using Millimeter-Wave Radar

Noninvasive microwave measurement of respiration

Measurement of human heartbeat and respiration signals using phase detection radar

Range correlation and I/Q performance benefits in single-chip silicon Doppler radars for non-contact cardiopulmonary monitoring. Microw. Theory Tech

Non-contact respiratory monitoring system using a ceilingattached microwave antenna

Contact-free measurement of heart rate, respiration rate, and body movements during sleep

Experimental Study of UWB Pulse Radar for Life Detection

Non-invasive detection of cardiac and respiratory rates from stepped frequency continuous wave radar measurements using the state space method

Vital signs monitoring of multiple people using a FMCW millimeter-wave sensor

Spectrum-averaged Harmonic Path (SHAPA) algorithm for non-contact vital sign monitoring with ultra-wideband (UWB) radar

Phase-Based Methods for Heart Rate Detection Using UWB Impulse Doppler Radar

Assessment of CODAR SeaSonde and WERA HF Radars in Mapping Surface Currents on the West Florida Shelf

A System-Theoretic View on Breathing Detection Using Chirp Sequence Modulated Radar Sensors

Discussion on the Mechanism of Vital Sign Detection with CW Radars

UWB radar signal processing in measurement of heartbeat features

The Theory and Design of Chirp Radars

Millimeter-Wave Radar Cane: A Blind People Aid with Moving Human Recognition Capabilities

Vital Sign Detection and Radar Self-Motion Cancellation through Clutter Identification

Vital Sign Signal Extraction Method Based on Permutation Entropy and EEMD Algorithm for Ultra-Wideband Radar

The Overview of Human Localization and Vital Sign Signal Measurement Using Handheld IR-UWB Through-Wall Radar

Acknowledgments: This research was supported in part by the Project "National key R&D Program of China" (2017YFB1304103).

The authors declare no conflict of interest.