CN114680840A - Non-contact vital sign monitoring method based on different durations - Google Patents
Non-contact vital sign monitoring method based on different durations Download PDFInfo
- Publication number
- CN114680840A CN114680840A CN202011586333.9A CN202011586333A CN114680840A CN 114680840 A CN114680840 A CN 114680840A CN 202011586333 A CN202011586333 A CN 202011586333A CN 114680840 A CN114680840 A CN 114680840A
- Authority
- CN
- China
- Prior art keywords
- heartbeat
- data
- detection
- respiration
- respiratory
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000012544 monitoring process Methods 0.000 title claims abstract description 39
- 238000000034 method Methods 0.000 title claims abstract description 34
- 230000029058 respiratory gaseous exchange Effects 0.000 claims abstract description 151
- 238000001514 detection method Methods 0.000 claims abstract description 150
- 230000011218 segmentation Effects 0.000 claims abstract description 10
- 238000000605 extraction Methods 0.000 claims abstract description 8
- 238000007781 pre-processing Methods 0.000 claims abstract description 8
- 230000000241 respiratory effect Effects 0.000 claims description 66
- 230000007958 sleep Effects 0.000 claims description 32
- 230000001121 heart beat frequency Effects 0.000 claims description 27
- 230000006399 behavior Effects 0.000 claims description 26
- 230000007774 longterm Effects 0.000 claims description 20
- 238000000354 decomposition reaction Methods 0.000 claims description 19
- 238000005070 sampling Methods 0.000 claims description 19
- 238000001228 spectrum Methods 0.000 claims description 18
- 238000010187 selection method Methods 0.000 claims description 17
- 239000013598 vector Substances 0.000 claims description 17
- 230000036387 respiratory rate Effects 0.000 claims description 12
- 238000004364 calculation method Methods 0.000 claims description 6
- 238000001914 filtration Methods 0.000 claims description 6
- 230000002159 abnormal effect Effects 0.000 claims description 5
- 238000006467 substitution reaction Methods 0.000 claims description 4
- NAWXUBYGYWOOIX-SFHVURJKSA-N (2s)-2-[[4-[2-(2,4-diaminoquinazolin-6-yl)ethyl]benzoyl]amino]-4-methylidenepentanedioic acid Chemical compound C1=CC2=NC(N)=NC(N)=C2C=C1CCC1=CC=C(C(=O)N[C@@H](CC(=C)C(O)=O)C(O)=O)C=C1 NAWXUBYGYWOOIX-SFHVURJKSA-N 0.000 claims description 3
- 238000003556 assay Methods 0.000 claims description 3
- 238000013075 data extraction Methods 0.000 claims description 3
- 230000000717 retained effect Effects 0.000 claims description 3
- 230000009466 transformation Effects 0.000 claims description 3
- 238000013519 translation Methods 0.000 claims description 3
- 230000036541 health Effects 0.000 description 20
- 230000036544 posture Effects 0.000 description 18
- 238000002474 experimental method Methods 0.000 description 10
- 238000005516 engineering process Methods 0.000 description 8
- 238000010586 diagram Methods 0.000 description 6
- 210000000038 chest Anatomy 0.000 description 5
- 230000000694 effects Effects 0.000 description 5
- 230000005540 biological transmission Effects 0.000 description 4
- 230000008859 change Effects 0.000 description 4
- 238000012360 testing method Methods 0.000 description 4
- 208000024172 Cardiovascular disease Diseases 0.000 description 3
- 201000010099 disease Diseases 0.000 description 3
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 3
- 230000037081 physical activity Effects 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 208000017667 Chronic Disease Diseases 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 239000012634 fragment Substances 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 230000008447 perception Effects 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 208000014085 Chronic respiratory disease Diseases 0.000 description 1
- 206010020772 Hypertension Diseases 0.000 description 1
- 206010037211 Psychomotor hyperactivity Diseases 0.000 description 1
- 206010049418 Sudden Cardiac Death Diseases 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000003542 behavioural effect Effects 0.000 description 1
- 230000000747 cardiac effect Effects 0.000 description 1
- 239000000969 carrier Substances 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 208000029078 coronary artery disease Diseases 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 230000006806 disease prevention Effects 0.000 description 1
- 238000006073 displacement reaction Methods 0.000 description 1
- 238000005265 energy consumption Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 230000001788 irregular Effects 0.000 description 1
- 239000003550 marker Substances 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000000737 periodic effect Effects 0.000 description 1
- 230000008092 positive effect Effects 0.000 description 1
- 238000011897 real-time detection Methods 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- 230000003860 sleep quality Effects 0.000 description 1
- 210000000115 thoracic cavity Anatomy 0.000 description 1
Images
Classifications
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/02—Detecting, measuring or recording pulse, heart rate, blood pressure or blood flow; Combined pulse/heart-rate/blood pressure determination; Evaluating a cardiovascular condition not otherwise provided for, e.g. using combinations of techniques provided for in this group with electrocardiography or electroauscultation; Heart catheters for measuring blood pressure
- A61B5/0205—Simultaneously evaluating both cardiovascular conditions and different types of body conditions, e.g. heart and respiratory condition
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7203—Signal processing specially adapted for physiological signals or for diagnostic purposes for noise prevention, reduction or removal
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7235—Details of waveform analysis
- A61B5/7253—Details of waveform analysis characterised by using transforms
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7235—Details of waveform analysis
- A61B5/7253—Details of waveform analysis characterised by using transforms
- A61B5/7257—Details of waveform analysis characterised by using transforms using Fourier transforms
Landscapes
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Physiology (AREA)
- Heart & Thoracic Surgery (AREA)
- Molecular Biology (AREA)
- Veterinary Medicine (AREA)
- Public Health (AREA)
- Biophysics (AREA)
- Pathology (AREA)
- Biomedical Technology (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Signal Processing (AREA)
- Surgery (AREA)
- Animal Behavior & Ethology (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Psychiatry (AREA)
- Cardiology (AREA)
- Pulmonology (AREA)
- Mathematical Physics (AREA)
- Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)
Abstract
The invention relates to a non-contact vital sign monitoring method based on different time lengths, which comprises the following steps: (1) preprocessing data; (2) extracting data; (3) breathing and heartbeat subcarrier selection; (4) based on different durations of breath and heartbeat detection. Aiming at the problem that the breathing heartbeat data after segmentation and extraction has a long time or a short time, a detection scheme based on different time lengths is provided and divided into long-time detection and short-time detection, a peak value method detection method comprising a false peak removal algorithm is adopted for long-time breathing, and an FFT detection method comprising a parabola interpolation method is adopted for short-time breathing and heartbeat.
Description
Technical Field
The invention belongs to the technical field of computers, and particularly relates to a non-contact vital sign monitoring method based on different time lengths.
Background
With the increasing concern of people on their health, the earliest disease treatment is gradually changed into real-time health evaluation and early disease prevention. Due to the accelerated pace of life and the increased pressure of life, the long-term sleep quality is low, so that people are in a sub-health state, chronic diseases are induced or various diseases, especially life-threatening cardiovascular diseases, are directly caused in the past for a long time. The respiratory rate and the heart rate during sleeping are the main basis for judging cardiovascular diseases, and the problems of irregular heart rate and the like can induce coronary heart disease, hypertension or sudden cardiac death. Therefore, night health monitoring of modern science and technology is adopted, and potential chronic diseases and respiratory diseases are found and cardiovascular diseases are prevented by monitoring sleep breathing and heartbeat; the monitoring of the sleeping behavior can judge the sleeping quality at night according to the times of turning over activities, time intervals, changes of sleeping postures and the like. Therefore, the long-term night health monitoring of people in a home environment has positive significance for preventing diseases, improving sub-health and improving life quality of different people.
Due to the defects of technologies such as a camera, wearable equipment and an infrared technology, the popularization of night health monitoring in a family environment is greatly influenced, and under the background, researchers provide a human body perception technology based on wireless signals.
The sensing technology based on wireless signals senses human body activities by using the change characteristics of the wireless signals generated by the human body activities in the transmission process, and mainly comprises radar-based and WiFi-based sensing technologies. The health monitoring based on radar adopts Doppler radar to sense human body activity by measuring signal frequency shift generated by electromagnetic waves reflected by human body, and the application of the radar in a home environment is hindered due to the fact that radar equipment is expensive. The health monitoring based on WiFi also realizes human body perception according to the change of human body reflected electromagnetic waves, but WiFi equipment has the characteristics of wide distribution, low cost, low energy consumption, easiness in operation, easiness in deployment and the like, and is more suitable for realizing non-contact health monitoring.
The night health monitoring of the human body is realized by detecting the respiratory rate and the heart rate of the human body and identifying the night sleeping behavior. According to different detection modes, the existing health monitoring can be divided into a contact type and a non-contact type.
In the contact health monitoring, the common contact detection methods include an impedance detection method, a special bed and mattress detection method, a wearable device detection method and the like. In contact health monitoring, sensor equipment with low power consumption, high sensitivity and high measurement accuracy is mostly used for collecting relevant data of a detected object. Not only is the cost high, but also a plurality of sensors are required to directly contact the human body, and the sensors are restricted or limited to a certain degree and cause discomfort to the human body. In the non-contact health monitoring, the sensor device is not required to be worn or contacted, discomfort is not brought to a subject, and the non-contact health monitoring system is suitable for long-time monitoring. Common non-contact detection methods include infrared detection, machine vision detection, radio signal detection, and the like. The detection method based on the radio signal realizes the health monitoring without contact, long distance, light influence and sensitive information. However, the detection method based on ultra-wideband and continuous wave radar is expensive and not suitable for health monitoring of ordinary families.
With the rapid development of wireless communication technology in recent years, WiFi devices have been widely popularized in the global scope, and have many advantages in the aspects of cost, usability, universality and the like, so that the sensing technology based on WiFi signals is rapidly a research hotspot. In 2014, Halperin et al of Washington university in USA successfully extracted CSI information with finer granularity from Intel5300 wireless network card, and a large amount of health monitoring research based on non-contact behavior recognition and vital sign detection of commercial WiFi equipment CSI information appears. In 2019, Yu Gu et al used a pair of WiFi devices and omnidirectional antennas to achieve real-time detection of breathing and heartbeat in different sleep postures, with accuracy rates of 96.636% and 94.215% respectively. In 2020, Abdelwahed Khamis et al propose a WiFi-based system for detecting respiratory cycle in real time and also establish a model of the relationship between the chest displacement and the phase difference change related to respiration.
Most health monitoring systems based on WiFi sensing only sense the breathing heartbeat of a human body in different sleeping postures or only sense the behavior of the human body, and no health monitoring system senses the breathing heartbeat and also senses the behavior in the sleeping process. Night health monitoring systems that combine breathing heartbeat with behavioral awareness remain a focus of research.
Through searching, no patent publication related to the present patent application has been found.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provides a non-contact vital sign monitoring method based on different time lengths.
The technical problem to be solved by the invention is realized by adopting the following technical scheme:
a non-contact vital sign monitoring method based on different time lengths comprises the following steps:
(1) data pre-processing
Removing abnormal value
Removing the outliers by adopting a Hampel filtering algorithm based on a sliding window:
the Hampel filtering algorithm measures the distance of a sample in data from the median using Mean Absolute Deviation (MAD), finds outliers present in the data based on the offset distance, and replaces the outliers with more representative values.
② segmentation algorithm
The standard deviation of the CSI amplitude in the window is calculated through a sliding window with a fixed length, when the window slides to a sleep behavior, the calculated standard deviation is large, and when the window slides to a breathing heartbeat, the calculated standard deviation is small. The segmentation algorithm comprises the following specific steps:
(2) data extraction
Extracting amplitudes of respiration and heartbeat
And carrying out corresponding decomposition reconstruction on the respiratory heartbeat data by adopting DWT. The formula of the wavelet transform is as follows:
where the scale α corresponds to frequency and the amount of translation τ corresponds to time. The frequency range after wavelet decomposition is related to the sampling frequency, the sampling frequency is halved after each layer of decomposition, the sampling frequency of the input signal is assumed to be Fs, and the frequency ranges of the approximation coefficient and the detail coefficient of the Nth layer after wavelet N-layer decomposition are respectively formula (2) and formula (3):
assuming that N-layer decomposition is performed, the frequency range of the first layer approximation coefficients is 0-12.5 Hz, the frequency range of the first layer detail coefficients is 12.5-25 Hz, and the frequency ranges of the approximation coefficients and the detail coefficients after more layers of decomposition are shown in the following table:
② phase extraction of respiration and heartbeat
The extraction of the phase is to eliminate the unknown terms by performing linear transformation on the original phase, and the preprocessing algorithm of the phase is shown in the following table:
(3) sub-carrier selection for respiration and heartbeat
Sub-carrier wave selecting method
The periodicity of the signal included in the subcarrier is determined, and the subcarrier having a higher periodicity is selected as the detection data. The periodicity of the respiratory signal is judged by calculating the respiratory signal-to-noise ratio of each subcarrier; the method for judging the periodicity of the heartbeat signal is to calculate the heartbeat signal-to-noise ratio of each subcarrier. The larger the values of BNR and HNR, the higher the periodicity of the signal, and the more likely that the subcarrier contains a better respiration signal and a heartbeat signal. The formulas for BNR and HNR are as follows:
wherein BNR is respiratory signal-to-noise ratio, B _ EmaxAs respiratory energy, B _ EiTotal energy after FFT is made for the respiratory data; HNR is the Heartbeat signal-to-noise ratio, H _ EmaxFor heart beat energy, H _ EiAnd (4) performing total energy after FFT on the heartbeat data, wherein n is the data length of the FFT.
Amplitude and phase subcarrier selection
BNR or HNR of 30 subcarriers of amplitude and phase information of the respiration or heartbeat data are respectively calculated, the respiration or heartbeat respectively selects three BNR or HNR with larger amplitude and phase, then average values of the three BNR or HNR with larger amplitude and phase are respectively calculated, the sizes of the two average values are compared, and finally three subcarriers with larger amplitude or phase are selected. If the average value of the amplitude information BNR or HNR is large, three subcarriers of the amplitude information are used, otherwise, three subcarriers of the phase information are used.
(4) Breath and heartbeat detection based on different durations
Detection scheme based on different time lengths
After the sleep behaviors and the respiratory heartbeat are segmented and extracted, the time interval between the two sleep behaviors is long or short, so that the segmented and extracted respiratory and heartbeat data are also long or short, the respiration and heartbeat data with the time interval being more than or equal to 12 seconds are defined as long-time respiration and heartbeat data, and the respiration and heartbeat data with the time interval being less than 12 seconds are defined as short-time respiration and heartbeat data. The detection method of the respiration rate adopts a peak detection method and an FFT detection method; the detection method of the heart rate adopts an FFT detection method.
Long term respiration and heartbeat detection
1) Long term breath detection
a. Peak detection method
By detecting the time interval between two adjacent peaks in the respiration data, a threshold value of 1.67 is set for the minimum time interval between two adjacent peaks in the detection of the peaks. The false peak removing algorithm is a window of 0.5 second forward and 0.5 second backward, for a total of 1 second, centered on the detected peak value, and detects whether a value larger than the peak value exists in the window. If present, the peak should be removed as a false peak, and if not, the peak should be retained as a true peak.
After removing false peaks, obtaining all peak values in the respiratory data, further obtaining time intervals of two adjacent peak values, and setting time interval vectors of all peak values of a certain subcarrier as s ═ { s (1), s (2.. s (n) }, and ordering:
when x is equal to e, y takes the minimum value, and e is the breathing cycle of the subcarrier, that is, x, which can make y take the minimum value, is the breathing cycle of the subcarrier. And carrying out weighted average on the breathing cycles of the three subcarriers to obtain a final breathing cycle. Let the BNR vectors of three subcarriers be b ═ { b (1), b (2), b (3) }, the breathing cycle vectors of three subcarriers be e ═ { e (1), e (2), e (3) }, and the final breathing cycle T is:
thus the respiration rate VbreathingComprises the following steps:
b. the result of the detection
In order to verify the feasibility of the peak detection method adopted by long-term respiration, a BNR subcarrier selection method is adopted. The detected respiratory cycles and the final weighted average respiratory cycle are shown in the table below;
the results of the detection of 12 second breath data by the peak detection method are shown in the following table:
2) long-term heartbeat detection
FFT detection method
According to the invention, the sampling frequency Fs is 50Hz, the FFT length is the data length, and after FFT is carried out on heartbeat data, the maximum peak value and the index thereof within the heartbeat frequency range of 1-2 Hz are searched. The frequency resolution is improved by adopting a parabola interpolation method, so that the accuracy of estimating the heartbeat frequency is improved, and the parabola is used for replacing the original curve of the maximum peak value and the adjacent two values. The general formula of the parabola is:
where the point p on the parabola abscissa is the interpolated position, b is the amplitude or phase, and a is the curvature. Interpolation, i.e. fitting the maximum in the parabola, leads to p by the following calculation. The position of the maximum peak value in the frequency spectrum on the horizontal axis is 0, the positions of two adjacent values are-1 and 1 respectively, and the following steps are included:
y (-1) ═ alpha formula (10)
y (0) ═ beta formula (11)
y (1) ═ gamma type (12)
These three samples are written in the form of an interpolated parabola:
α＝ap2+2ap + a + b type (13)
β＝ap2+ b type (14)
γ＝ap2-2ap + a + b formula (15)
Equation (15) is subtracted from equation (13) to derive equation (16):
alpha-gamma is 4ap type (16)
Equation (17) is derived from equation (16):
substituting formula (17) into formula (13) to obtain formula (18):
formula (19) is obtained from formula (18):
substitution of equation (19) for equation (17), i.e., the position p of the plug:
k denotes the index of the maximum peak in the spectrum in the whole data, then k + p is the index of the maximum in the fitted parabola in the whole data. And finally, fitting the maximum value of the parabola as an estimation peak value, solving the frequency corresponding to the value as the estimated heartbeat frequency of the heartbeat data, wherein the estimated heartbeat frequency is expressed by the formula (21):
wherein Fs is the sampling frequency, and N is the FFT size, i.e., the number of sampling points.
According to the parabolic interpolation method, the estimated peak value p of three superior subcarriers in the heartbeat data is obtained, the heartbeat frequency of each subcarrier is obtained through calculation of a formula (28), and the final heartbeat frequency can be obtained by carrying out weighted average on the heartbeat frequencies of the three subcarriers. Let h be { h (1), h (2), h (3) } for HNR vectors of three subcarriers, F be { F (1), F (2), F (3) } for heartbeat frequency vectors of three subcarriers, and the final heartbeat frequency F is:
thus heart rate VheartbeatComprises the following steps:
b. The result of the detection
In order to verify the feasibility of the FFT detection method adopted by the long-time heartbeat data, an HNR subcarrier selection method is adopted. The estimated heart rate and the final weighted average heart rate are shown in the following tables, respectively:
to verify the feasibility of using the FFT assay for 12 seconds of heartbeat data. A section of heartbeat data of 12 seconds is taken, and three superior subcarriers of the data amplitude are selected by adopting an HNR subcarrier selection method.
Short-time breath and heartbeat detection
1) Short term breath detection
Short breaths the breath rate was measured for t seconds using FFT detection. In order to verify the feasibility of the scheme, considering the case of the slowest respiratory rate, the detection results of 4.8-10.8 seconds of respiratory data and 6 seconds of respiratory data with the respiratory rate between 10-36bpm are shown in the following table:
2) short-term heartbeat detection
Short-term heartbeats are detected by FFT detection for a t second heart rate. Taking 6 seconds of heartbeat data as an example, the spectrum of three superior subcarriers of the data amplitude after FFT is selected by an HNR subcarrier selection method, and the detection result is shown in the following table.
The invention has the advantages and positive effects that:
aiming at the problem that the respiratory heartbeat data after segmentation and extraction has long and short lengths, a detection scheme based on different time lengths is provided and divided into long-time detection and short-time detection, a peak value method detection method comprising a false peak removal algorithm is adopted for long-time respiration, an FFT detection method comprising a parabola interpolation method is adopted for short-time respiration and heartbeat,
drawings
FIG. 1 illustrates CSI amplitude information of sleep data in accordance with the present invention; the method comprises the steps of (a) original CSI amplitude information, (b) CSI amplitude information with an abnormal value removed;
FIG. 2 is a segmentation of sleep behavior and respiratory heartbeat in accordance with the present invention, wherein (a) sleep behavior and respiratory heartbeat data; (b) a sleep behavior marker;
FIG. 3 illustrates the sleep behavior and respiratory heartbeat divided in the present invention; wherein, (a) the first segment of respiratory heartbeat data, (b) the second segment of respiratory heartbeat data, (c) the third segment of respiratory heartbeat data, (d) the fourth segment of respiratory heartbeat data, (e) the first sleep behavior, (f) the second sleep behavior; (g) a third sleep behavior;
FIG. 4 is amplitude information of respiration and heartbeat in accordance with the present invention; wherein, (a) amplitude information of a third segment of breaths, (b) amplitude information of a third segment of heartbeats;
fig. 5 is the raw phase of 30 subcarriers for three antennas in the present invention;
FIG. 6 is a polar plot of raw phase and phase after pre-processing in the present invention;
FIG. 7 is a phase information of respiration and heartbeat in accordance with the present invention; wherein, (a) phase information of a third segment of breath, (b) phase information of a third segment of heartbeat;
FIG. 8 is a BNR and HNR of breath and heartbeat amplitude values in accordance with the present invention; wherein (a) BNR is the third segment respiratory amplitude, (b) HNR is the third segment heartbeat amplitude;
FIG. 9 illustrates BNR and HNR for respiratory and cardiac phases in accordance with the present invention; wherein, (a) BNR in the third segment respiratory phase, (b) HNR in the third segment heartbeat phase;
FIG. 10 is a block diagram of the subcarrier selection process for amplitude and phase of respiration or heartbeat in accordance with the present invention;
FIG. 11 is a BNR, HNR comparison of amplitude and phase of respiration and heartbeat in accordance with the present invention;
FIG. 12 is a graph of the preferred sub-carriers of the amplitude and phase of the respiration in the present invention; wherein, (a) the superior subcarrier of the third segment of respiration amplitude, (b) the superior subcarrier of the third segment of respiration phase;
FIG. 13 is a diagram of the preferred subcarriers of the heartbeat amplitude and phase in accordance with the present invention; wherein, (a) the superior subcarrier of the third segment of heartbeat amplitude, (b) the superior subcarrier of the third segment of heartbeat phase;
FIG. 14 illustrates a detection scheme of the present invention for different durations;
FIG. 15 is a block diagram of a peak detection method according to the present invention;
FIG. 16 is a graph of peaks in respiration data according to the invention; wherein (a) a respiratory cycle in the respiratory data, (b) a false peak in the respiratory data;
FIG. 17 shows false and true peaks of sleep breathing in accordance with the present invention; wherein (a) a false peak, (b) a true peak;
FIG. 18 is a graph of 12 seconds of breath data in accordance with the present invention;
FIG. 19 is a block diagram of a flow chart of an FFT detection method according to the present invention;
FIG. 20 is a frequency spectrum of third stage sleep heartbeat phase data in accordance with the present invention; wherein, (a) spectrum of 99.88-129.88 seconds, (b) spectrum of 129.88-159.88 seconds, (c) spectrum of 159.88-189.9 seconds;
FIG. 21 is a fitted parabola of the maximum peak and two adjacent values in the present invention;
FIG. 22 is a graph of the spectrum of 12 seconds of heartbeat data and amplitude in accordance with the present invention; wherein (a)12 seconds of heartbeat data, (b) a spectrum of 12 seconds of heartbeat data;
FIG. 23 is a graph of breath data for 6 seconds in accordance with the present invention; wherein (a) the breathing rate is 10bpm and (b) the breathing rate is between 10 and 36 bpm;
FIG. 24 is a graph of the spectrum of short term respiratory data in accordance with the present invention; wherein, (a) a respiratory spectrum of 10bpm, and (b) a respiratory spectrum of 10-36 bpm;
FIG. 25 is a 6 second heartbeat data and frequency spectrum of the present invention; wherein, (a)6 seconds of heartbeat data, (b) a spectrum of 6 seconds of heartbeat;
FIG. 26 is a hardware device of the present invention; the system comprises (a) an Intel5300 wireless network card and a receiving antenna, and (b) a TP-LINK wireless router;
FIG. 27 is a plan view of an experimental scenario in accordance with the present invention; wherein, (a) a laboratory plan, (b) a dormitory plan;
FIG. 28 is a block diagram of an experimental procedure in accordance with the present invention;
FIG. 29 is the raw amplitude information in the present invention; wherein, (a) the amplitude of antenna 1, (b) the amplitude of antenna 2, (c) the amplitude of antenna 3;
FIG. 30 shows the detection accuracy in different experimental scenarios of the present invention;
FIG. 31 breathing at different transceiver distances; wherein (a) 2m breath, (b) 3m breath, (c) 4m breath, (d) 5m breath
FIG. 32 is a diagram of heartbeats at different transceiver distances in the present invention; wherein (a) a heartbeat of 2 meters, (b) a heartbeat of 3 meters, (c) a heartbeat of 4 meters, (d) a heartbeat of 5 meters;
FIG. 33 shows the detection accuracy of different transceiving distances in the present invention;
FIG. 34 illustrates breathing in different sleeping positions according to the present invention; wherein, (a) breathing in a supine position, (b) breathing in a left lateral position, (c) breathing in a right lateral position, (d) breathing in a prone position;
FIG. 35 shows the heartbeat in different sleeping positions according to the present invention; wherein, (a) a supine heartbeat, (b) a left lateral heartbeat, (c) a right lateral heartbeat, and (d) a prone heartbeat;
FIG. 36 shows the detection accuracy of different sleeping postures in the present invention
Detailed Description
The present invention will be further described with reference to specific examples, which are intended to be illustrative, not limiting and are not intended to limit the scope of the invention.
Structures not specifically described in detail herein are to be understood as conventional in the art.
The invention provides a non-contact vital sign monitoring method based on different time lengths, which is specifically prepared and detected as follows:
1.1 data preprocessing
1.1.1 outlier removal
Due to the change of the WiFi transmission rate, the reason of the network card and the like, abnormal values exist in the night sleep data, and the sensing detection result is influenced. Therefore, a Hampel filtering algorithm based on a sliding window is adopted to remove the abnormal values.
The Hampel filtering algorithm measures the distance of a sample from a median in data by Mean Absolute Deviation (MAD), finds outliers present in the data based on the offset distance, and replaces the outliers with more representative values.
1.1.2 segmentation Algorithm
The standard deviation of the CSI amplitude in the window is calculated through a sliding window with a fixed length, when the window slides to a sleep behavior, the calculated standard deviation is large, when the window slides to a respiratory heartbeat, the calculated standard deviation is small, and by setting a threshold, CSI fragments with the standard deviation larger than the threshold are divided, namely the sleep behavior, and the rest fragments are the respiratory heartbeat, so that the division of the sleep behavior and the respiratory heartbeat is realized. The segmentation algorithm comprises the following specific steps:
1.2 data extraction
1.2.1 amplitude extraction of respiration and heartbeat
And carrying out corresponding decomposition reconstruction on the respiratory heartbeat data by adopting DWT. The formula of the wavelet transform is as follows:
where the scale α corresponds to frequency and the amount of translation τ corresponds to time. The multi-scale decomposition of DWT decomposes the signal f (t) into low and high frequency parts by means of low and high pass filters. The wavelet function determines the characteristics of the low-pass filter, the scale function determines the characteristics of the high-pass filter, and the decomposed coefficients consist of two parts: a low frequency (approximate) coefficient vector and a high frequency (detail) coefficient vector.
The frequency range after wavelet decomposition is related to the sampling frequency, the sampling frequency is halved after each layer of decomposition, the sampling frequency of the input signal is assumed to be Fs, and the frequency ranges of the approximation coefficient and the detail coefficient of the Nth layer after wavelet N-layer decomposition are respectively formula (3) and formula (4):
assuming that N-layer decomposition is performed, the frequency range of the first layer approximation coefficients is 0-12.5 Hz, the frequency range of the first layer detail coefficients is 12.5-25 Hz, and the frequency ranges of the approximation coefficients and the detail coefficients after more layers of decomposition are shown in the following table:
1.2.2 phase extraction of respiration and Heartbeat
These unknowns are eliminated by linear transformation of the original phase, the pre-processing algorithm for the phase is shown in the following table:
1.3 sub-Carrier selection for respiration and Heartbeat
1.3.1 subcarrier selection method
The subcarriers with different CSI have different center frequencies and wavelengths, each subcarrier has different signal strength and phase, and thus different subcarriers have different amplitude and phase information. And the CSI variation caused by respiration and heartbeat is periodic variation, the higher the CSI periodicity of the subcarrier is, the more sensitive the subcarrier is to respiration and heartbeat signals, and the higher the periodicity of the signals contained in the subcarrier is judged, and the subcarrier with the higher periodicity is selected as detection data.
The periodicity of the respiratory signal is determined by calculating the respiratory signal-to-Noise Ratio (BNR) of each subcarrier. The periodicity of the Heartbeat signal is determined by calculating the Heartbeat-to-Noise Ratio (HNR) of each subcarrier. The larger the values of BNR and HNR, the higher the periodicity of the signal, the more likely the subcarrier contains a better respiration signal and a heartbeat signal. The formulas for BNR and HNR are as follows:
wherein BNR is respiratory signal-to-noise ratio, B _ EmaxFor respiratory energy, B _ EiTotal energy after FFT is made for the respiratory data; HNR is the Heartbeat signal-to-noise ratio, H _ EmaxFor heart beat energy, H _ EiAnd (4) performing total energy after FFT on the heartbeat data, wherein n is the data length of the FFT.
1.3.2 amplitude and phase subcarrier selection
BNR or HNR of 30 subcarriers of amplitude and phase information of the respiration or heartbeat data are respectively calculated, and three BNR or HNR with larger amplitude and phase are respectively selected for respiration or heartbeat. And respectively calculating the average values of three BNR or HNR with amplitude and phase, comparing the two average values, and finally selecting three subcarriers with larger average values and amplitude or phase. If the average value of the amplitude information BNR or HNR is large, three subcarriers of the amplitude information are used, otherwise, three subcarriers of the phase information are used.
According to the method, the third segment of respiration selects three superior subcarriers of the amplitude information as detection data. The third heartbeat selects three better subcarriers of the phase information as detection data. Defining the selection method of the sub-carrier with the better breathing amplitude and phase as a BNR sub-carrier selection method; and defining the selection method of the subcarrier with the better heartbeat amplitude and phase as an HNR subcarrier selection method.
1.4 breath and heartbeat detection based on different durations
1.4.1 detection scheme based on different durations
After the sleep behaviors and the respiratory heartbeat are segmented and extracted, the time interval between the two sleep behaviors is long or short, so that the segmented and extracted respiratory and heartbeat data are also long or short, the respiration and heartbeat data with the time interval being more than or equal to 12 seconds are defined as long-time respiration and heartbeat data, and the respiration and heartbeat data with the time interval being less than 12 seconds are defined as short-time respiration and heartbeat data. Considering the real-time nature of the detection, for a long breath heartbeat, the respiration rate and the heart rate are calculated every 30 seconds. For short breath beats, the minimum time to detect the respiration rate and heart rate is also considered. The breathing rate of a human body ranges from 10bpm to 36bpm, and the breathing cycle is 1.67 seconds to 6 seconds; the human body heart rate range is 60-120 bpm, the heartbeat cycle is 0.5-1 second, the human body slowest breathing cycle and the slowest heartbeat cycle are respectively 6 seconds and 0.5 second, but the time of 0.5 second is too short, so only the shortest time for detecting the breathing rate is considered. The peak value detection method needs at least two peak values to detect the respiration rate, the respiration rate needs to be detected for 12 seconds approximately twice, and the FFT detection method can detect the respiration rate for 6 seconds only by one respiration period, so when t is more than or equal to 6 and less than 12 seconds, the FFT detection method is adopted to detect the respiration rate.
The detection method of the respiration rate adopts a peak detection method and an FFT detection method. The heart rate detection method adopts an FFT detection method, and the heart rate does not adopt a peak detection method because the heart rate changes rapidly and the time interval of the peak value detected by the peak detection method is not uniform, so that the detected heart rate has larger error. To analyze the algorithm complexity of the two detection methods of respiration, the time complexity and the space complexity of the two detection methods are illustrated by comparing the time and the physical memory used when the two detection methods are run. Taking the third section of respiratory data extracted by segmentation as an example, respectively adopting a peak detection method and an FFT detection method to carry out respiratory detection, wherein the algorithm complexity is compared with that shown in the following table, and in order to ensure the real-time performance of respiratory rate detection, when t is more than or equal to 12 seconds, the peak detection method is adopted to calculate the respiratory rate.
Suppose the data length of the respiration heartbeat is N, the duration is t, the unit is second, the integer of t/30 is N, and the remainder is m. For short breath and heartbeat data, the breath rate and heart rate are detected for t seconds. For long-time breathing heartbeat data duration, if the duration is 12-30 seconds, detecting the breathing rate and the heart rate of t seconds; if the time is more than 30 seconds and is an integral multiple of 30 seconds, detecting n respiration rates and heart rates of 30 seconds; if it is greater than 30 seconds and not an integer multiple of 30 seconds, (n-1) respiration rates and heart rates of 30 seconds and (30+ m) are detected. For example, if a segment of breath and heartbeat data is 75 seconds long, two 30 second breath rates and heart rates can be calculated, and the remaining 15 seconds of data are added to the second 30 seconds, i.e., the 75 second segment of data requires the first 30 seconds and the second 45 seconds of breath rates and heart rates to be calculated.
1.4.2 Long term breath and Heartbeat detection
1) Long term breath detection
a. Peak detection method
By detecting the time interval between two adjacent peaks in the respiration data, a threshold value of 1.67 is set for the minimum time interval between two adjacent peaks in the detection of the peaks.
The false peak removing algorithm is a window of 0.5 second forward and 0.5 second backward, for a total of 1 second, centered on the detected peak value, and detects whether a value larger than the peak value exists in the window. If present, the peak is a false peak that should be removed, and if not, the peak is a true peak that should be retained.
After removing false peaks, all peak values in the respiration data are obtained, and further, the time interval between two adjacent peak values is obtained, and the time interval vector of all peak values of a certain subcarrier is set as { s (1), s (2).. s (n) }, so that:
when x is equal to e, y takes the minimum value, and e is the breathing cycle of the subcarrier, that is, x, which can make y take the minimum value, is the breathing cycle of the subcarrier. And carrying out weighted average on the breathing cycles of the three subcarriers to obtain a final breathing cycle. Let b be { b (1), b (2), b (3) } for BNR vectors of three subcarriers, e be { e (1), e (2), e (3) } for breathing cycle vectors of three subcarriers, and the final breathing cycle T be:
thus the respiration rate VbreathingComprises the following steps:
b. the result of the detection
To verify the feasibility of using the peak detection method for long breaths. And selecting three superior subcarriers of the third section of respiration amplitude by adopting a BNR subcarrier selection method. The respiration rates and the final weighted average respiration rates are shown in the following tables, respectively:
to verify the feasibility of using the peak detection method for 12 seconds of breathing data at the slowest breathing rate of 10 bpm. Artificially controlling the respiration rate is approximately 10bpm of respiration data.
The results of the detection of 12 second breath data by the peak detection method are shown in the following table:
2) long-term heartbeat detection
FFT detection method
According to the invention, the sampling frequency Fs is 50Hz, the FFT length is the data length, and after FFT is carried out on heartbeat data, the maximum peak value and the index thereof within the heartbeat frequency range of 1-2 Hz are searched.
The frequency resolution is improved by adopting a parabola interpolation method, so that the accuracy of estimating the heartbeat frequency is improved, and the parabola is used for replacing the original curve of the maximum peak value and the adjacent two values. The general formula of the parabola is:
where the point p on the parabola abscissa is the interpolated position, b is the amplitude or phase, and a is the curvature.
Interpolation, i.e. fitting the maximum in the parabola, leads to p by the following calculation. The position of the maximum peak value in the frequency spectrum on the horizontal axis is 0, the positions of two adjacent values are-1 and 1 respectively, and the following steps are included:
y (-1) ═ alpha formula (10)
y (0) ═ beta type (11)
y (1) ═ gamma type (12)
These three samples are written in the form of an interpolated parabola:
α＝ap2+2ap + a + b type (13)
β＝ap2+ b type (14)
γ＝ap2-2ap + a + b formula (15))
Equation (15) is subtracted from equation (13) to derive equation (16):
alpha-gamma is 4ap type (16)
Equation (17) is derived from equation (16):
substituting formula (17) into formula (13) to obtain formula (18):
formula (19) is obtained from formula (18):
substitution of equation (19) for equation (17), i.e., the position p of the plug:
k denotes the index of the maximum peak in the spectrum in the whole data, then k + p is the index of the maximum in the fitted parabola in the whole data. And finally, fitting the maximum value of the parabola to be used as an estimation peak value, solving the frequency corresponding to the value to be the estimated heartbeat frequency of the heartbeat data, and enabling the estimated heartbeat frequency to be the formula (21):
where Fs is the sampling frequency and N is the FFT size, i.e., the number of sample points (starting at 1).
According to the parabolic interpolation method, the estimated peak value p of three superior subcarriers in the heartbeat data is obtained, the heartbeat frequency of each subcarrier is obtained through calculation of a formula (21), and the final heartbeat frequency can be obtained by carrying out weighted average on the heartbeat frequencies of the three subcarriers. Let h be { h (1), h (2), h (3) } for HNR vectors of three subcarriers, F be { F (1), F (2), F (3) } for heartbeat frequency vectors of three subcarriers, and the final heartbeat frequency F is:
thus heart rate VheartbeatComprises the following steps:
b. The result of the detection
And the feasibility of adopting an FFT (fast Fourier transform) detection method for the long-time heartbeat data is verified. Through an HNR subcarrier selection method, three superior subcarriers of the third section of heartbeat phase are selected, and the estimated heart rate and the final weighted average heart rate are respectively shown in the following table:
to verify the feasibility of using the FFT assay for 12 seconds of heartbeat data. A section of heartbeat data of 12 seconds is taken, and three superior subcarriers of the data amplitude are selected by adopting an HNR subcarrier selection method.
1.4.3 short-term breath and heartbeat detection
1) Short term breath detection
Short-term (t is more than or equal to 6 and less than 12) respiration is detected by an FFT detection method at the respiration rate of t seconds. In order to verify the feasibility of the scheme, considering the case of the slowest respiratory rate, the detection results of 4.8-10.8 seconds of respiratory data and 6 seconds of respiratory data with the respiratory rate between 10-36bpm are shown in the following table:
2) short-term heartbeat detection
And detecting the heart rate of t seconds by using an FFT (fast Fourier transform) detection method for short-time (t is more than or equal to 6 and less than 12) heartbeats. Taking 6 seconds of heartbeat data as an example, by an HNR subcarrier selection method, three superior subcarriers of the data amplitude are selected, and after FFT, the detection results are shown in the following table:
2.1 Experimental Equipment and Experimental Environment
2.1.1 Experimental Equipment
The Intel5300 wireless network card, the TP-LINK wireless router and the Linux 802.11 CSITools open source software package are used for collecting night sleep behaviors and respiration and heartbeat data of the human body. The TP-LINK wireless router comprises 3 antennas with 2.4GHz and 1 antenna with 5GHz, the computer terminal is connected with the WiFi of the router, the router is used as a transmitting end to send data packets to the wireless network card, and the Linux 802.11 CSITools can analyze CSI data packets from the Intel5300 wireless network card and acquire CSI information. A computer terminal equipped with a wireless network card is used as a Detection Point (DP), and the receiving end is connected with three 12dB gain receiving antennas, as shown in fig. 26 (a). The wireless router serves as an Access Point (AP), and as shown in fig. 26(b), the wireless router uses a 5GHz transmitting antenna, so that the transmitting end and the receiving end form a 1 × 3 MIMO system array.
Two experimental scenes are provided, and a plan view of the laboratory is shown in fig. 27(a), in which tables, chairs, cabinets and other sundries are provided, and the subject is located between the transmitting device TX and the receiving device RX; fig. 27(b) shows a plan view of a dormitory, in which a bed, a table, a chair, a cabinet, and other sundries are present, the subject is sleeping on the bed, and the transmitting device TX and the receiving device RX are present on both sides of the human body. The difference of the horizontal heights of the transceiving equipment and the human body is about 0.5m, and the distances between the transceiving equipment are 2, 3, 4 and 5 m.
2.1.2 Experimental procedures and data
And downloading and installing a Linux 802.11 CSI Tools software package at a computer terminal provided with the Ubuntu16.04 system. And the terminal is connected with the WiFi of the wireless router, and the receiving end and the transmitting end are connected by configuring the network and starting the wireless network card through commands. And then setting the gateway address, sampling frequency, transmission mode and the like of the wireless router. To detect information changes in a short time, the sampling frequency is set to 50Hz, i.e. one data packet is sent every 0.02 seconds. The experimental flow chart is shown in fig. 28.
And reading the monitored sleep behavior and the monitored respiration heartbeat data and analyzing CSI amplitude and phase information. The format of the obtained CSI data is a complex matrix of 30 × 3 by using a 1 × 3 MIMO system array, where each row corresponds to one subcarrier and each column corresponds to one receiving antenna. The original amplitude information of the antenna 1, the antenna 2 and the antenna 3 is plotted as a curve of the monitoring data with time, as shown in fig. 29(a) (b) (c), the three-dimensional graph of the original amplitude information shows that the three antennas are obviously different, and the information of each subcarrier of each antenna is also different.
According to the above experimental procedure, the sleep behavior and the respiration heartbeat are respectively monitored in the two experimental scenes. In the experimental stage, no other interference exists between the transceiver, the monitoring time of a laboratory is short, and the dormitory environment monitoring time is long. Meanwhile, a RestOn intelligent sleep monitor Z400TWP is used for monitoring respiration and heartbeat, the respiration rate and the heart rate detected by the monitor are used as real values, and the respiration rate and the heart rate detected by the monitor are integers, so that the respiration rate and the heart rate detected by the monitor are rounded and compared with the real values, and the accuracy rate is evaluated.
2.2 analysis of the results of respiration and Heartbeat detection
2.2.1 Effect of Experimental scenarios on test results
In two different experimental scenarios, repeated experiments of long-term and short-term respiration and heartbeat detection are performed respectively, the experimental results are shown in fig. 30, and when the distances between the transceiver devices are both 2m, the detection accuracy rates of the long-term respiration and heartbeat in the laboratory scenario are 96.1% and 95.5%, respectively. Because the amplitude and phase information of the CSI are fully utilized, not only is the threshold value used for peak value detection, but also false peaks are removed aiming at long-term respiration detection, and the detection accuracy rate of the long-term respiration is improved. Aiming at the detection of short-term respiration and heart rate, the FFT detection method comprising a parabola interpolation method is adopted, so that the detection accuracy is improved by 2%. The detection accuracy rates of the long-term respiration and heartbeat in the dormitory scene are 95.2% and 94.9%, respectively, which are lower than the accuracy rate in the laboratory scene because the dormitory scene is easily interfered by other roommates and more sundries.
The detection accuracy rates for short breaths and heartbeats in the laboratory setting are 95.5% and 95.1%, respectively, and the detection accuracy rates for short breaths and heartbeats in the dormitory setting are 94.5% and 94.1%, respectively, since the short detection is a shorter time to estimate the respiratory rate and heart rate for one minute, the accuracy rates for short breaths and heart rates are slightly lower than for long periods. Experiments show that the estimation error of 90% of the experimental respiration data is less than or equal to 2bpm in both long-term and short-term, and the estimation error of 85% of the experimental heartbeat data is less than or equal to 2bpm in both long-term and short-term.
2.2.2 Effect of Transceiver distance on detection results
In a laboratory scenario, repeated experiments of breath detection were performed at distances between different transceivers, and as shown in fig. 31, amplitude variation graphs of 60 second breath of the 30 th subcarrier in the supine sleeping position are respectively shown when the transceivers are at distances of 2m, 3m, 4m, and 5 m. It can be seen that no matter the short-distance 2m or the long-distance 5m, the respiratory data can be extracted, and the respiratory waveform is obvious; when the distances between the transceiver devices are 2m, 3m and 4m, the amplitude fluctuation of respiration is large, and the fluctuation is small at 5 m; when the distance between the transceiver is 2m, the fluctuation size is uniform, and when the distance is 3m, 4m and 5m, the fluctuation size is not uniform when the distance is 2m, and the fluctuation size is uniform.
Repeated experiments of heartbeat detection are respectively carried out under the distances among different transceiver devices, as shown in fig. 32, which are amplitude variation graphs of 60-second heartbeat of 30 th subcarrier in the supine sleeping position when the transceiver distances are 2m, 3m, 4m and 5m respectively. It can be seen that the heartbeat data can be extracted no matter the distance is 2m or 5m, and obvious heartbeat waveforms exist; when the distance between the transceiver devices is 2m and 3m, the amplitude fluctuation of the heartbeat is large, and the fluctuation is small at 4m and 5 m.
In summary, as the distance between the transceiver devices increases, the smaller the amplitude fluctuation of the respiration and heartbeat data, the less sensitive the respiration and heartbeat data to the human body, and therefore, the distance between the transceiver devices should be reasonably selected.
In the case where the distances between the transceivers were 2m, 3m, 4m, and 5m, respectively, the subject kept in a supine sleep position for 5 minutes each time, and outputted a respiration rate and a heart rate every 30 seconds, for a total of 30 repeated experiments. As shown in fig. 33, it can be seen that when the distance between the transceiver devices is 2m, the detection accuracy of respiration and heartbeat is the highest, and is 96.1% and 95.5%, respectively. With the increase of the distance between the transceiver devices, the detection accuracy of respiration and heartbeat is reduced because a relatively short distance can generate a relatively strong CSI signal for weak respiration and heartbeat activity of a human body, and a relatively high respiration and heartbeat signal-to-noise ratio can also be achieved. Similarly, a longer distance makes the received respiration and heartbeat signals weaker. When the distance is 5 meters, the CSI fluctuation of respiration and heartbeat is small, which affects detection performance.
2.2.3 Effect of sleep posture on test results
There are 4 main sleeping positions: when lying on the back, the chest of the person faces the transceiver, and the transmission signal is directly reflected to the receiver through the chest of the person; when the user lies on the left side or the right side, the side part of the chest cavity of the user faces the transceiver, and the transmitted signal is reflected to the receiver through one side of the chest of the user; when lying prone, the person's back faces the transceiver device, and breathing causes less back variation than chest variation.
Repeated experiments of breath detection were performed in different sleeping postures, as shown in fig. 34, which are graphs of amplitude changes of 60 second breaths of 30 th subcarrier in supine, left-side, right-side, and prone sleeping postures, respectively. It can be seen that the breathing data can be extracted from any sleeping posture, and the breathing waveform is obvious; when the sleeping posture is supine, the respiratory amplitude fluctuates uniformly, and the respiratory amplitudes in the other three sleeping postures fluctuate non-uniformly; when the sleeping positions are prone, the fluctuation of the respiratory amplitude is smaller than that of the respiratory amplitudes in the other three sleeping positions.
Repeated experiments of heartbeat detection are respectively carried out under different sleeping postures, and as shown in fig. 35, amplitude variation graphs of 60-second heartbeat of 30 th subcarrier in supine, left-side, right-side and prone sleeping postures are respectively carried out. It can be seen that the heartbeat data can be extracted regardless of the sleeping posture, the heartbeat waveform is obvious, and the fluctuation is uniform; the heartbeat amplitude fluctuation of the right side lying and the prone lying is smaller than the fluctuation of the back lying and the left side lying.
The test was continued for 5 minutes each time with the subject in different sleeping positions, outputting a respiration rate and heart rate every 30 seconds for a total of 30 replicates. The effect of different sleep postures on the test results, as shown in fig. 36, can be seen that when the sleep posture of the subject is supine, the respiratory rate and heart rate are the highest in accuracy, 96.3% and 95.8%, respectively; the sleeping postures of the left side and the right side are lower in accuracy than the sleeping posture of the back lying, and the accuracy in the prone lying is the lowest of the four sleeping postures.
Although the embodiments of the present invention have been disclosed for illustrative purposes, those skilled in the art will appreciate that: various substitutions, changes and modifications are possible without departing from the spirit and scope of the invention and the appended claims, and therefore the scope of the invention is not limited to the embodiments disclosed.
Claims (4)
1. A non-contact vital sign monitoring method based on different time lengths comprises the following steps:
(1) data pre-processing
Removing abnormal value
Removing the outliers by adopting a Hampel filtering algorithm based on a sliding window:
the Hampel filtering algorithm measures the distance of a sample from a median in data by Mean Absolute Deviation (MAD), finds outliers present in the data based on the offset distance, and replaces the outliers with more representative values.
② segmentation algorithm
The standard deviation of the CSI amplitude in the window is calculated through a sliding window with a fixed length, when the window slides to a sleep behavior, the calculated standard deviation is large, and when the window slides to a breathing heartbeat, the calculated standard deviation is small. The segmentation algorithm comprises the following specific steps:
(2) data extraction
Extracting respiratory and heartbeat amplitude values
And carrying out corresponding decomposition reconstruction on the respiratory heartbeat data by adopting DWT. The formula of the wavelet transform is as follows:
where the scale α corresponds to frequency and the amount of translation τ corresponds to time. The frequency range after wavelet decomposition is related to the sampling frequency, the sampling frequency is halved after each layer of decomposition, the sampling frequency of the input signal is assumed to be Fs, and the frequency ranges of the approximation coefficient and the detail coefficient of the Nth layer after wavelet N-layer decomposition are respectively formula (2) and formula (3):
assuming that N-layer decomposition is performed, the frequency range of the first layer approximation coefficients is 0-12.5 Hz, the frequency range of the first layer detail coefficients is 12.5-25 Hz, and the frequency ranges of the approximation coefficients and the detail coefficients after more layers of decomposition are shown in the following table:
② phase extraction of respiration and heartbeat
The extraction of the phase is to eliminate the unknown terms by performing linear transformation on the original phase, and the preprocessing algorithm of the phase is shown in the following table:
(3) sub-carrier selection for respiration and heartbeat
Sub-carrier wave selecting method
The periodicity of the signal included in the subcarrier is determined, and the subcarrier having a higher periodicity is selected as the detection data. The periodicity of the respiratory signal is judged by calculating the respiratory signal-to-noise ratio of each subcarrier; the method for judging the periodicity of the heartbeat signal is to calculate the heartbeat signal-to-noise ratio of each subcarrier. The larger the values of BNR and HNR, the higher the periodicity of the signal, the more likely the subcarrier contains a better respiration signal and a heartbeat signal. The formulas for BNR and HNR are as follows:
wherein BNR is respiratory signal-to-noise ratio, B _ EmaxFor respiratory energy, B _ EiTotal energy after FFT is made for the respiratory data; HNR is the Heartbeat signal-to-noise ratio, H _ EmaxFor heart beat energy, H _ EiAnd (4) performing total energy after FFT on the heartbeat data, wherein n is the data length of the FFT.
Amplitude and phase subcarrier selection
BNR or HNR of 30 subcarriers of amplitude and phase information of the respiration or heartbeat data are respectively calculated, the respiration or heartbeat respectively selects three BNR or HNR with larger amplitude and phase, then average values of the three BNR or HNR with larger amplitude and phase are respectively calculated, the sizes of the two average values are compared, and finally three subcarriers with larger amplitude or phase are selected. If the average value of the amplitude information BNR or HNR is large, three subcarriers of the amplitude information are used, otherwise, three subcarriers of the phase information are used.
(4) Breath and heartbeat detection based on different durations
Detection scheme based on different time lengths
After the sleep behaviors and the respiratory heartbeat are segmented and extracted, the time interval between the two sleep behaviors is long or short, so that the segmented and extracted respiratory and heartbeat data are also long or short, the respiration and heartbeat data with the time interval being more than or equal to 12 seconds are defined as long-time respiration and heartbeat data, and the respiration and heartbeat data with the time interval being less than 12 seconds are defined as short-time respiration and heartbeat data. The method for detecting the respiration rate adopts a peak value detection method and an FFT detection method; the detection method of the heart rate adopts an FFT detection method.
Long term respiration and heartbeat detection
1) Long term breath detection
a. Peak detection method
By detecting the time interval between two adjacent peaks in the respiration data, a threshold value T is set for the minimum time interval between two adjacent peaks in the detection of the peaks. The false peak removing algorithm is a window of 0.5 second forward and 0.5 second backward, for a total of 1 second, centered on the detected peak value, and detects whether a value larger than the peak value exists in the window. If present, the peak is a false peak that should be removed, and if not, the peak is a true peak that should be retained.
After removing false peaks, obtaining all peak values in the respiratory data, further obtaining time intervals of two adjacent peak values, and setting time interval vectors of all peak values of a certain subcarrier as s ═ { s (1), s (2.. s (n) }, and ordering:
when x is equal to e, y takes the minimum value, and e is the breathing cycle of the subcarrier, that is, x, which can make y take the minimum value, is the breathing cycle of the subcarrier. And carrying out weighted average on the breathing cycles of the three subcarriers to obtain a final breathing cycle. Let b be { b (1), b (2), b (3) } for BNR vectors of three subcarriers, e be { e (1), e (2), e (3) } for breathing cycle vectors of three subcarriers, and the final breathing cycle T be:
thus the respiration rate VbreathingComprises the following steps:
b. the result of the detection
In order to verify the feasibility of the peak detection method adopted by long-term respiration, a BNR subcarrier selection method is adopted. The detected respiratory cycles and the final weighted average respiratory cycle are shown in the following table:
the results of the detection of 12 second breath data by peak detection are shown in the following table:
2) long-term heartbeat detection
FFT detection method
The length of the FFT is the data length, and after the FFT is carried out on the heartbeat data, the maximum peak value and the index thereof within the heartbeat frequency range of 1-2 Hz are searched. The frequency resolution is improved by adopting a parabola interpolation method, so that the accuracy of estimating the heartbeat frequency is improved, and the parabola is used for replacing the original curve of the maximum peak value and the adjacent two values. The general formula of the parabola is:
where the point p on the parabola abscissa is the interpolated position, b is the amplitude or phase, and a is the curvature. Interpolation, i.e. fitting the maximum in the parabola, leads to p by the following calculation. The position of the maximum peak value in the frequency spectrum on the horizontal axis is 0, the positions of two adjacent values are-1 and 1 respectively, and the following steps are included:
y (-1) ═ alpha formula (10)
y (0) ═ beta type (11)
y (1) ═ gamma type (12)
These three samples are written in the form of an interpolated parabola:
α＝ap2+2ap + a + b type (13)
β＝ap2+ b type (14)
γ＝ap2-2ap + a + b formula (15)
Equation (15) is subtracted from equation (13) to derive equation (16):
alpha-gamma is 4ap type (16)
Equation (17) is derived from equation (16):
substituting formula (17) into formula (13) to obtain formula (18):
formula (19) is obtained from formula (18):
substitution of equation (19) for equation (17), i.e., the position p of the plug:
k denotes the index of the maximum peak in the spectrum in the whole data, then k + p is the index of the maximum in the fitted parabola in the whole data. And finally, fitting the maximum value of the parabola as an estimation peak value, solving the frequency corresponding to the value as the estimated heartbeat frequency of the heartbeat data, wherein the estimated heartbeat frequency is expressed by the formula (21):
wherein Fs is the sampling frequency, and N is the FFT size, i.e., the number of sampling points.
According to the parabolic interpolation method, the estimated peak value p of three superior subcarriers in the heartbeat data is obtained, the heartbeat frequency of each subcarrier is obtained through calculation of a formula (21), and the final heartbeat frequency can be obtained by carrying out weighted average on the heartbeat frequencies of the three subcarriers. Let h be { h (1), h (2), h (3) } for HNR vectors of three subcarriers, F be { F (1), F (2), F (3) } for heartbeat frequency vectors of three subcarriers, and the final heartbeat frequency F is:
thus heart rate VheartbeatComprises the following steps:
Vheartbeat60. F type (23)
b. The result of the detection
In order to verify the feasibility of the FFT detection method adopted by the long-time heartbeat data, an HNR subcarrier selection method is adopted. The estimated heart rate and the final weighted average heart rate are shown in the following tables, respectively:
to verify the feasibility of using the FFT assay for 12 seconds of heartbeat data. Taking a segment of heartbeat data of 12 seconds, adopting an HNR subcarrier selection method, and selecting three superior subcarriers of the data amplitude for detection, wherein the result is shown as follows:
short-time breath and heartbeat detection
1) Short term breath detection
Short breaths the breath rate was measured for t seconds using FFT detection. In order to verify the feasibility of the scheme, considering the case of the slowest respiratory rate, the detection results of 4.8-10.8 seconds of respiratory data and 6 seconds of respiratory data with the respiratory rate between 10-36bpm are shown in the following table:
2) short-term heartbeat detection
Short-term heartbeats are detected by FFT detection for a t second heart rate. Taking 6 seconds of heartbeat data as an example, three superior subcarriers of the data amplitude are selected by an HNR subcarrier selection method, and the FFT detection results are shown in the following table:
2. the non-contact vital sign monitoring method of claim 1, wherein the non-contact vital sign monitoring method comprises: the threshold value γ is set to 1.
3. The non-contact vital sign monitoring method of claim 1, wherein the non-contact vital sign monitoring method comprises: the sampling frequency was 50 Hz.
4. The non-contact vital sign monitoring method of claim 1, wherein the non-contact vital sign monitoring method comprises: the time interval threshold T is 1.67.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011586333.9A CN114680840A (en) | 2020-12-29 | 2020-12-29 | Non-contact vital sign monitoring method based on different durations |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011586333.9A CN114680840A (en) | 2020-12-29 | 2020-12-29 | Non-contact vital sign monitoring method based on different durations |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114680840A true CN114680840A (en) | 2022-07-01 |
Family
ID=82129515
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011586333.9A Pending CN114680840A (en) | 2020-12-29 | 2020-12-29 | Non-contact vital sign monitoring method based on different durations |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114680840A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116449353A (en) * | 2023-06-20 | 2023-07-18 | 精华隆智慧感知科技（深圳）股份有限公司 | Human body existence detection method, device, equipment and storage medium in sleep process |
-
2020
- 2020-12-29 CN CN202011586333.9A patent/CN114680840A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116449353A (en) * | 2023-06-20 | 2023-07-18 | 精华隆智慧感知科技（深圳）股份有限公司 | Human body existence detection method, device, equipment and storage medium in sleep process |
| CN116449353B (en) * | 2023-06-20 | 2023-08-15 | 精华隆智慧感知科技（深圳）股份有限公司 | Human body existence detection method, device, equipment and storage medium in sleep process |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111481184B (en) | Multi-target respiration heart rate monitoring method and system based on millimeter wave radar technology | |
| CN106725488B (en) | Wireless field intensity respiration detection method and device and respiration detector | |
| Ravichandran et al. | WiBreathe: Estimating respiration rate using wireless signals in natural settings in the home | |
| Wang et al. | PhaseBeat: Exploiting CSI phase data for vital sign monitoring with commodity WiFi devices | |
| CN109480787B (en) | Non-contact sleep monitoring equipment based on ultra-wideband radar and sleep staging method | |
| US20140378809A1 (en) | Systems and methods for extracting physiological characteristics using frequency harmonics | |
| US9532735B2 (en) | Apparatus and method for wireless monitoring using ultra-wideband frequencies | |
| CN109998549A (en) | A kind of human body respiration detection method based on WiFi channel state information | |
| Zhao et al. | CRH: A contactless respiration and heartbeat monitoring system with COTS RFID tags | |
| JP6310401B2 (en) | Method, system and computer program for processing a signal representing a physiological rhythm | |
| CN112386236A (en) | Vital sign monitoring method based on channel state phase information | |
| US20220218224A1 (en) | Sleep tracking and vital sign monitoring using low power radio waves | |
| WO1998043071A1 (en) | Method and apparatus for arbitrating to obtain best estimates for blood constituent values and rejecting harmonics | |
| CN105249925B (en) | A kind of traditional Chinese medical pulse manifestation collecting device and noise reduction system and noise-reduction method | |
| CN111568425B (en) | Non-contact type multi-person respiration detection method | |
| Liu et al. | Human biometric signals monitoring based on wifi channel state information using deep learning | |
| Han et al. | UWB radar for non-contact heart rate variability monitoring and mental state classification | |
| Li et al. | Mo-sleep: Unobtrusive sleep and movement monitoring via Wi-Fi signal | |
| WO2019080728A1 (en) | Method for detecting heart rate abnormality on the basis of wi-fi signal | |
| CN111568417A (en) | Non-contact type respiration real-time monitoring method | |
| Uysal et al. | RF-based noncontact respiratory rate monitoring with parametric spectral estimation | |
| CN114305355B (en) | Breathing heartbeat detection method, system and device based on millimeter wave radar | |
| CN115474901A (en) | Non-contact living state monitoring method and system based on wireless radio frequency signals | |
| CN114680840A (en) | Non-contact vital sign monitoring method based on different durations | |
| Shibao et al. | Heart-rate monitoring of moving persons using 79 GHz ultra-wideband radar sensor |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| WD01 | Invention patent application deemed withdrawn after publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20220701 |