CN116224325A - LFMCW radar and underdetermined blind source separation-based multi-person respiratory signal detection method and system - Google Patents
LFMCW radar and underdetermined blind source separation-based multi-person respiratory signal detection method and system Download PDFInfo
- Publication number
- CN116224325A CN116224325A CN202111463338.7A CN202111463338A CN116224325A CN 116224325 A CN116224325 A CN 116224325A CN 202111463338 A CN202111463338 A CN 202111463338A CN 116224325 A CN116224325 A CN 116224325A
- Authority
- CN
- China
- Prior art keywords
- matrix
- signals
- time
- signal
- frequency
- 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
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S13/00—Systems using the reflection or reradiation of radio waves, e.g. radar systems; Analogous systems using reflection or reradiation of waves whose nature or wavelength is irrelevant or unspecified
- G01S13/88—Radar or analogous systems specially adapted for specific applications
Landscapes
- Engineering & Computer Science (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- Physics & Mathematics (AREA)
- Electromagnetism (AREA)
- Computer Networks & Wireless Communication (AREA)
- General Physics & Mathematics (AREA)
- Radar Systems Or Details Thereof (AREA)
Abstract
The invention discloses a multi-person respiratory signal detection method and system based on separation of an LFMCW radar and an underdetermined blind source. The method comprises the following steps: firstly, carrying out data rearrangement on radar echo signals of two paths of receiving antennas, extracting phases of the rearranged echo signals, and filtering noise and interference in the phase signals through a band-pass filter to obtain two paths of mixed signals breathed by multiple persons; then, two paths of mixed signals are transformed into a time-frequency domain by using short-time Fourier transform, two paths of time-frequency points form a clustering plane, and an improved neighbor propagation algorithm based on the fusion of angle information and energy information is used for cluster analysis to obtain the estimation of the number of people and the mixed matrix; and finally, separating out respiratory signals in a time-frequency domain through an L1 norm minimization algorithm according to the mixing matrix, and recovering to a time domain through inverse short-time Fourier transform to obtain respiratory signals of multiple persons. The method is effective and feasible, has reliable performance, and can acquire the breathing signals of multiple persons with the same distance gate and very close distance only by using the double-receiving antenna LFMCW radar system.
Description
Technical Field
The invention belongs to the field of vital sign monitoring, and particularly relates to a non-contact type multi-person respiratory signal detection method and system based on an LFMCW radar and underdetermined blind source separation technology.
Background
In recent years, microwave-based radar systems have been widely used for the detection of vital sign signals. Compared with the traditional contact type respiration detection device, the non-contact type respiration detection device has the advantages of no stimulation to skin, no discomfort, no electrode and cable binding and the like. With the development of digital signal processing technology and the improvement of hardware system performance, the multi-antenna system can monitor a plurality of objects at the same time, and can be widely applied to a plurality of scenes such as ward, bedroom, search and rescue after disasters and the like.
At present, radar-based multi-target respiration monitoring technology has been studied in depth, and the implementation modes are mainly divided into the following three types:
one is to separate multiple targets from each other angularly based on the antenna beams directed to each human target. The method mainly utilizes the direction of arrival estimation and the digital beam forming technology to finish the positioning of a plurality of human targets and the detection of respiratory signals, but the method has the limitation of the angle resolution of a radar system, and under the scene that the plurality of human targets are very close to each other, the direction of arrival estimation cannot accurately estimate the angle information, and the digital beam forming technology is not suitable for the situation. Particularly in radar systems with a small number of receive antennas, a wide beam cannot distinguish between closely spaced human targets.
Secondly, separating a plurality of targets according to different distance doors where the human targets are located. This method mainly uses the characteristics of FMCW radar to distinguish a plurality of human targets from each other in distance, but it is difficult to separate a plurality of targets located at the same distance gate.
And thirdly, extracting respiratory signals of a plurality of human bodies based on a separation algorithm. The method has less research at home and abroad, and the separation algorithm mainly comprises a Variation Modal Decomposition (VMD) algorithm and a blind source separation algorithm. The VMD algorithm can realize that a single J receiving antenna separates respiratory signals of a plurality of targets, but the algorithm depends on various initial parameters, can obtain distinct results under different parameters, and has poor robustness when no priori knowledge of the parameters exists. The research based on the blind source separation algorithm is concentrated on the positive definite condition, namely the number of the receiving antennas is required to be equal to the number of human targets, and the method greatly improves the hardware cost of the radar system along with the increase of the number of people to be detected.
Disclosure of Invention
The invention aims to solve the problems in the prior art and provides a non-contact type multi-person respiratory signal detection method and system based on an LFMCW radar and underdetermined blind source separation technology. The underdetermined blind source separation algorithm is applied to multi-person respiratory signal detection for the first time, and an improved neighbor propagation algorithm based on fusion of angle information and energy information is provided, so that a mixing matrix is estimated more accurately, and the multi-person respiratory signal is recovered accurately.
The technical solution for realizing the purpose of the invention is as follows: a multi-person respiratory signal detection method based on LFMCW radar and underdetermined blind source separation comprises the following steps:
in the method, in the process of the invention,
for mixing signal vector representing two paths of mixing signals based on blind source separation model
For source signal vector->
s 1 ,s 2 ,…,s P The respiratory signals of P human bodies are obtained, wherein A is a mixed matrix of 2 rows and P columns;
wherein (X) 1 (k i ),X 2 (k i ) I) represents the cluster center, k of the i-th class on the cluster plane i Index value of the i-th cluster center;
Further, the step 1 specifically includes:
collecting echo signals s r (t) and local oscillator signal s t (t) performing down-conversion to obtain a difference frequency signal s b (t)。
Further, in step 2, the preprocessing is performed on the radar echo signals corresponding to each receiving antenna, which specifically includes:
step 2-1, carrying out rearrangement processing on the difference frequency signals according to the radar repetition period;
step 2-2, calculating the phase of the rearranged echo signals by using a differential cross multiplication algorithm;
and 2-3, filtering the phase signal to obtain a mixed signal.
Further, in the step 2-1, the rearrangement processing is performed on the difference frequency signal according to the radar repetition period, which specifically includes:
(1) the frequency modulation period is T r Is a frequency modulation signal of (1), and the sampling period of the ADC is T s At each ofM sampling points are collected in the frequency modulation period, echo data of N frequency modulation periods are collected, and then the difference frequency signals can be rearranged into a matrix R of N rows and M columns, and the matrix R is expressed as:
R[n,m]＝s b (t＝m·T s +n·T r )
where n is the row index of the matrix, expressed in T r The sampling number is the sampling period, and the value range is 1 to N; m is the column index of the matrix, expressed in T s The sampling number is the sampling period, and the value range is 1 to M;
(2) for matrix R [ n, m ]]N is counted as each line of fft Is fast fourier transformed to obtain a matrix R f [n,v]V is FFT index value, and the value range is 1 to N fft ；
(3) Calculating matrix R f [n,v]The energy of each column is calculated as follows:
the FFT index value V corresponding to the maximum energy value represents the distance gate number of the position where the target is located, and then the target distance r may be represented by the FFT index value V:
wherein K is the frequency modulation coefficient of the linear frequency modulation continuous wave radar, f s Is the sampling rate; the positioning of the target is realized through the method;
the slow time-difference frequency signal s (n) of the position of the target is expressed as:
s(n)＝R f (n,V)。
further, in step 2-2, the phase of the rearranged echo signal is calculated by using a differential cross multiplication algorithm, and the calculation formula is as follows:
in the method, in the process of the invention,
is the phase of a slow time-difference frequency signal s I (n) and s Q (n) is quadrature-arm IQ signal, n is frequency modulation period T r The number of samples for a sampling period.
Further, the filtering of the phase signal in step 2-3 is specifically:
the phase signal is filtered by a band-pass filter h (t) with the order of L-1 and the passband frequency of 0.15-0.55hz, the frequency range is a basic respiratory frequency band, and a mixed signal x (t) is obtained after the filtering, and the calculation formula is as follows:
further, the estimating the mixing matrix in step 3 specifically includes:
step 3-1, transforming the mixed signals breathed by multiple persons to a time-frequency domain, and thinning the mixed signals so as to facilitate cluster analysis; the specific process comprises the following steps:
(1) short-time Fourier transform is carried out on the multi-person respiration mixed signal x (t) corresponding to the two paths of receiving antennas:
X i (t,f)＝∫x i (τ)·g(t-τ)·e -j2πft dτt∈(0,L t -1),f∈(0,L f -1)
wherein g (t) is a Gaussian window function, x i (t) is the multi-person respiration mixed signal corresponding to the ith receiving antenna, X i (t, f) is a time-frequency matrix of the multi-person respiration mixed signal corresponding to the ith receiving antenna, L f Representing FFT point number for the number of matrix lines; l (L) t For the number of matrix columns, the number of time frames is represented;
(2) taking the positive frequency band of the time-frequency matrix, i.e. the frequency f takes a value of 0 to L f 2-1, matrix X i (t, f) rearrangement to one-dimensional vector X i (q) a rearrangement formula:
X i (q)＝X i (t,f)q＝t+f·L t
wherein q is the index number of the time-frequency vector, and the time-frequency vector corresponding to the two paths of receiving antennas is X 1 (q) and X 2 (q)，X 1 (q) and X 2 (q) forming a two-dimensional clustering plane defining a point X on the plane * (q)＝(X 1 (q),X 2 (q))；
Step 3-2, estimating the number of people to be detected and a mixing matrix based on an improved neighbor propagation algorithm of angle information and energy information fusion;
the improved neighbor propagation algorithm flow comprises the following steps:
(1) calculating a similarity matrix s (i, k), wherein the similarity matrix is used for measuring the point X on the clustering plane * (i) And point X * (k) The similarity matrix is defined as follows:
s(i,k)＝abs[X * (k)]·exp(-ρ·(1-cos(∠[X * (i),X * (k)])))
in the formula, abs [ X ] * (k)]Representing clustered planar points X * (k) Is a modulus of (2); angle [ X ] * (i),X * (k)]Representing point X * (i) And point X * (k) Is included in the plane of the first part; ρ is the attenuation coefficient of the nonlinear function, and the larger the ρ value is, the faster the attenuation is, and the influence of insufficient sparse points on clustering can be reduced;
(2) initializing an attraction degree matrix r (i, k) and a attribution degree matrix a (i, k) to 0; wherein the attraction matrix represents the suitability of data point k as the cluster center of data point i; the attribution degree matrix indicates whether the data point i selects the data point k as the clustering center thereof;
(3) updating the attraction matrix r (i, k)
r t+1 (i,k)＝s(i,k)-max j≠k [a t (i,j)+s(i,j)]
Wherein r is t+1 (i, k) represents the attraction degree of the next iteration, a t (i, k) represents the degree of attribution of the current iteration;
(4) updating the home degree matrix a (i, k)
(5) Weighting the current result of step (3) (4) and the last iteration result using an attenuation coefficient λ, typically taking the attenuation coefficient λ=0.5, the weighting formula is as follows:
r t+1 (i,k)＝λ·r t (i,k)+(1-λ)·r t+1 (i,k)
a t+1 (i,k)＝λ·a t (i,k)+(1-λ)·a t+1 (i,k)
(6) repeating the steps (3), (4) and (5) until the matrix converges or the maximum iteration number is reached, and ending the algorithm;
(7) determining data point X * (i) Is the cluster center point X of (2) * (k)＝(X 1 (k),X 2 (k) K) should satisfy the formula:
max{r(i,k)+a(i,k)}
number of cluster centers
The method is used for estimating the number of people to be detected, and a clustering center is a mixing matrix and is expressed as follows:
wherein X is * (k i )＝(X 1 (k i ),X 2 (k i ) I) represents the cluster center of the i-th class, k i Index value of the i-th cluster center;
is an estimate of the number of people to be tested.
Further, step 4 describes mixing the matrix according to the estimation
The method for recovering the respiratory signal specifically comprises the following steps:
step 4-1, obtaining an optimal solution of the underdetermined problem according to the mixed matrix; when the number of the receiving antennas is smaller than the number of people to be detected, the blind source separation problem is an underdetermined problem, the number of unknown sources is larger than the number of equations, the equations have no unique solution, constraint conditions are introduced, and if the L1 norm to be solved is minimum, the source recovery is converted into an optimization problem; the objective function may be expressed as:
where s (t, f) is a time-frequency domain estimate of each respiratory signal source, X (t, f) is a respiratory mix signal in the time-frequency domain,
a mixing matrix obtained by calculation in the step 3-2;
for a two receive antenna system, the L1 norm minimization algorithm steps are as follows:
(1) computing a mixing matrix
Is a full rank submatrix of 2 rows and 2 columns, altogether +.>
And is denoted as a k ，
(2) All possible solutions are found for a point X (t, f) in the sparse domain,
(3) find the L1 norm of each solution, take the smallest norm
As an optimal estimate of the source signal, namely:
(4) repeating the steps (1), 2 and 3) to obtain the optimal solution of all points in the sparse domain;
and 4-2, recovering the time-frequency domain respiration signals separated in the step 4-1 to a time domain, wherein the formula is as follows:
wherein h (τ -t) is a rectangular window,
is the recovered time domain signal.
A multi-person respiratory signal detection system based on LFMCW radar separation from underdetermined blind sources, the system comprising:
the radar echo signal acquisition module is used for detecting the chest movements of a plurality of human targets in the same range gate by using a double-receiving antenna LFMCW radar to obtain echo signals of two paths of receiving antennas;
the preprocessing module is used for respectively preprocessing radar echo signals corresponding to each receiving antenna to obtain two paths of mixed signals of multi-person respiration, and the mixed signals of the multi-person respiration in each path are the linear superposition of the respiratory signals of the multi-person; wherein the mixed signal is represented in matrix form as:
in the method, in the process of the invention,
for mixing signal vector representing two paths of mixing signals based on blind source separation model
For source signal vector->
s 1 ,s 2 ,…,s P The respiratory signals of P human bodies are obtained, wherein A is a mixed matrix of 2 rows and P columns;
a mixing matrix estimation module, configured to estimate the mixing matrix: using short-time Fourier transforms to vector mixed signals
Transforming to time-frequency domain to obtain time-frequency domain down-mixed signal vector +.>
Time-frequency point X 1 And X 2 Forming a clustering plane, performing cluster analysis by using an improved neighbor propagation algorithm based on the fusion of angle information and energy information to obtain the number of clusters and a cluster center, wherein the number of clusters is +.>
The method is characterized in that the method is used for estimating the number of people to be detected, the clustering center is used for estimating a mixing matrix, and the method is expressed as follows:
wherein (X) 1 (k i ),X 2 (k i ) I) represents the cluster center, k of the i-th class on the cluster plane i Index value of the i-th cluster center;
a source recovery module for mixing matrix according to estimation
Recovering the respiratory signal; under the underdetermined condition, when the target number is greater than the number of the receiving antennas, recovering and converting the respiratory signals into an optimization problem, introducing constraint conditions, solving the L1 norm minimization to obtain an optimal solution, and obtaining the estimation of the respiratory signals in the time-frequency domain>
Finally pair->
Performing inverse short time Fourier transform to obtain time domain estimate +.>
Compared with the prior art, the invention has the remarkable advantages that:
1) The LFMCW radar is utilized to realize non-contact respiration monitoring, which can penetrate through barriers such as clothes, bedclothes and the like, and compared with the traditional contact type monitoring, the operation is more convenient, the uncomfortable feeling of a human body can be reduced, and the limitation of the length of an electrode and a cable is avoided;
2) The multi-person respiratory signal detection method based on the underdetermined blind source separation algorithm has no limitation on the angle resolution of a radar system, and can be suitable for scenes with short distances of multiple persons, such as overnight sleep monitoring of couples, mother and infant and the like;
3) In the realization of underdetermined blind source separation algorithm, an improved neighbor propagation algorithm based on the fusion of angle information and energy information is provided to estimate a mixing matrix, the influence of insufficient sparse points with smaller energy and larger deviation between angles and clustering centers on a clustering result is reduced, and the estimation of the mixing matrix is more accurate;
4) The multi-person respiratory signal detection method based on the underdetermined blind source separation algorithm can detect and separate respiratory signals of more than 2 persons only by using double receiving antennas, and has the advantages of simple system structure, low cost, reliable performance and convenient implementation.
The invention is described in further detail below with reference to the accompanying drawings.
Drawings
FIG. 1 is a block flow diagram of a non-contact multi-person respiratory signal detection method based on underdetermined blind source separation of the present invention.
Fig. 2 is a diagram of a time domain waveform and a frequency domain waveform of a mixed signal after preprocessing two paths of receiving antennas, where fig. a) is a time domain signal obtained by preprocessing a first path of receiving antennas, fig. b) is a frequency domain signal obtained by preprocessing a first path of receiving antennas, fig. c) is a time domain signal obtained by preprocessing a second path of receiving antennas, and fig. d) is a frequency domain signal obtained by preprocessing a second path of receiving antennas in an embodiment.
FIG. 3 is a graph of clustering results using a modified neighbor propagation algorithm based on nonlinear projection in one embodiment.
Fig. 4 is a comparison of time domain waveforms of respiratory signals and respiratory band reference signals of three subjects recovered using the algorithm of the present invention in one embodiment, wherein fig. a) is a time domain waveform of respiratory signals and respiratory band signals recovered by a first subject, fig. b) is a time domain waveform of respiratory signals and respiratory band signals recovered by a second subject, and fig. 4=c) is a time domain waveform of respiratory signals and respiratory band signals recovered by a third subject.
Fig. 5 is a comparison of the frequency domain waveforms of the respiratory signal and the respiratory band reference signal of three subjects recovered using the algorithm of the present invention, in one embodiment, a) the frequency domain waveforms of the respiratory signal and the respiratory band signal recovered by the first subject, b) the frequency domain waveforms of the respiratory signal and the respiratory band signal recovered by the second subject, and c) the frequency domain waveforms of the respiratory signal and the respiratory band signal recovered by the third subject.
Detailed Description
In order to make the objects, technical solutions and advantages of the present application more apparent, the present application will be further described in detail with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the present application.
In one embodiment, the monitoring of the 3 person respiratory signal within the same range gate is accomplished using a dual receive antenna LFMCW radar system. With reference to fig. 1, a non-contact multi-person respiratory signal detection method based on LFMCW radar and underdetermined blind source separation technology is provided. The method comprises the following steps:
and step 1, detecting the chest cavity motion of 3 persons in the same range gate by using a double-receiving antenna LFMCW radar to obtain echo signals of two paths of receiving antennas.
and 2-1, carrying out rearrangement processing on the difference frequency signals according to the radar repetition period. According to the distance gate of the target, extracting the slow time difference frequency signal of the distance, wherein the specific process comprises the following steps:
the specific process comprises the following steps:
(1) the frequency modulation period is T r Is a frequency modulation signal of (1), and the sampling period of the ADC is T s Collecting M sampling points in each frequency modulation period, and collecting echo data of N frequency modulation periods, wherein the difference frequency signals can be rearranged into a matrix R of N rows and M columns, and the matrix R can be expressed as:
R[n,m]＝s b (t＝m·T s +n·T r )
where n is the row index of the matrix, expressed in T r The sampling number is the sampling period, and the value range is 1 to N; m is the column index of the matrix, expressed in T s The sampling number of the sampling period is 1 to M.
(2) For matrix R [ n, m ]]N is counted as each line of fft Is fast fourier transformed to obtain a matrix R f [n,v]V is FFT index value, and the value range is 1 to N fft 。
(3) Calculating matrix R f [n,v]The energy of each column is calculated as follows:
the FFT index value V corresponding to the maximum energy value represents the range gate number of the target position because of the difference frequency f of the linear frequency modulation continuous wave radar b Proportional to the echo delay τ, the target distance r may be represented by the FFT index value V:
wherein K is the frequency modulation coefficient of the linear frequency modulation continuous wave radar, f s Is the sampling rate. The positioning of the target is realized through the formula. The slow time-difference frequency signal s (n) of the location of the target can be expressed as:
s(n)＝R f (n,V)
and 2-2, calculating the phase of the slow time difference frequency signal. Chest motion signal modulation in phase of radar echo slow time signal, differential cross multiplication algorithm (DACM) based on quadrature arm IQ signal s I (n) and s Q (n) calculating the phase of the slowly-time-difference frequency signal
The differential cross multiplication algorithm has the following calculation formula:
wherein n is a frequency modulation period T r The number of samples for a sampling period.
And 2-3, filtering direct current components and clutter interference in the phase of the echo signal. The filter uses a band-pass filter h (t) with an order of L-1 and a passband frequency of 0.15-0.55hz, and the frequency range is the basic respiratory frequency band. The mixed signal x (t) is obtained after filtering, and the calculation formula is as follows:
the three-person respiration mixed signal of the two paths of receiving antennas is obtained after pretreatment and is shown in figure 2.
and 3-1, transforming the mixed signals breathed by multiple persons into a time frequency domain, and thinning the mixed signals for cluster analysis. The specific process comprises the following steps:
(1) short-time Fourier transform is carried out on the multi-person respiration mixed signal x (t) corresponding to the two paths of receiving antennas:
X i (t,f)＝∫x i (τ)·g(t-τ)·e -j2πft dτt∈(0,L t -1),f∈(0,L f -1)
wherein g (t) is a Gaussian window function, x i (t) is the multi-person respiration mixed signal corresponding to the ith receiving antenna, X i (t, f) is the ith receive antenna pairA time-frequency matrix of the mixed signal is breathed by a plurality of persons. L (L) f Representing FFT point number for the number of matrix lines; l (L) t The number of matrix columns is the number of time frames.
(2) Taking the positive frequency band of the time-frequency matrix, i.e. the frequency f takes a value of 0 to L f 2-1, matrix X i (t, f) rearrangement to one-dimensional vector X i (q) a rearrangement formula:
X i (q)＝X i (t,f)q＝t+f·L t
wherein q is the index number of the time-frequency vector, and the time-frequency vector corresponding to the two paths of receiving antennas is X 1 (q) and X 2 (q)。X 1 (q) and X 2 (q) forming a two-dimensional clustering plane. Define a point X on a plane * (q)＝(X 1 (q),X 2 (q))
And 3-2, estimating the number of people to be detected and the mixing matrix based on an improved neighbor propagation algorithm of the fusion of the angle information and the energy information.
The improved neighbor propagation algorithm flow comprises the following steps:
(1) calculating a similarity matrix s (i, k), wherein the similarity matrix is used for measuring the point X on the clustering plane * (i) And point X * (k) Is a similarity of (3). The similarity matrix is defined as follows:
s(i,k)＝abs[X * (k)]·exp(-ρ·(1-cos(∠[X * (i),X * (k)])))
in the formula, abs [ X ] * (k)]Representing clustered planar points X * (k) Is a modulus of (2); angle [ X ] * (i),X * (k)]Representing point X * (i) And point X * (k) Is included in the plane of the first part; ρ is the attenuation coefficient of the nonlinear function, and the larger the ρ value is, the faster the attenuation is, and the influence of insufficient sparse points on clustering can be reduced. Unlike the traditional neighbor propagation algorithm, which measures the similarity by calculating the Euclidean distance between points on a clustering plane, the method considers that the breathing signals are sparse in time-frequency domain, namely each breathing signal in the frequency domain determines a straight line on the clustering plane, so that the similarity is measured by adopting the included angle between the points, and the influence of insufficient sparse points on the clustering is reduced by using a nonlinear function; at the same time, the energy information of the time-frequency points is considered to reduce the low energy point pairInfluence of clustering.
(2) The attraction degree matrix r (i, k) and the attribution degree matrix a (i, k) are initialized to 0. Wherein the attraction matrix represents the suitability of data point k as the cluster center of data point i; the attribution matrix indicates whether data point i selects data point k as its cluster center.
(3) Updating the attraction matrix r (i, k)
r t+1 (i,k)＝s(i,k)-max j≠k [a t (i,j)+s(i,j)]
Wherein r is t+1 (i, k) represents the attraction degree of the next iteration, a t (i, k) represents the degree of attribution of the current iteration.
(4) Updating the home degree matrix a (i, k)
(5) To avoid the occurrence of numerical oscillations, the current result of step (3) (4) is weighted with the last iteration result using a damping coefficient λ, typically taking the damping coefficient λ=0.5, the weighting formula is as follows:
r t+1 (i,k)＝λ·r t (i,k)+(1-λ)·r t+1 (i,k)
a t+1 (i,k)＝λ·a t (i,k)+(1-λ)·a t+1 (i,k)
(6) repeating the steps (3), (4) and (5) until the matrix converges or the maximum iteration number is reached, and ending the algorithm.
(7) Determining data point X * (i) Is the cluster center point X of (2) * (k)＝(X 1 (k),X 2 (k) K) should satisfy the formula:
max{r(i,k)+a(i,k)}
number of cluster centers
The estimated number of people to be measured is obtained. The cluster center, i.e., the mixing matrix, can be expressed as:
wherein X is * (k i )＝(X 1 (k i ),X 2 (k i ) I) represents the cluster center of the i-th class, k i Index value of the i-th cluster center;
is an estimate of the number of people to be tested.
In connection with fig. 3, 3 cluster centers are obtained as estimates of the mixing matrix.
and 4-1, obtaining an optimal solution of the underdetermined problem according to the mixed matrix. When the number of the receiving antennas is smaller than the number of people to be detected, the blind source separation problem is an underdetermined problem, the number of unknown sources is larger than the number of equations, the equations have no unique solution, constraint conditions are introduced, and the source recovery is converted into an optimization problem if the L1 norm to be solved is minimum. The objective function may be expressed as:
where s (t, f) is a time-frequency domain estimate of each respiratory signal source. X (t, f) is a respiration mixed signal in a time-frequency domain,
and (3) calculating the obtained mixing matrix for the step 3-2.
For a two receive antenna system, the L1 norm minimization algorithm steps are as follows:
(1) computing a mixing matrix
Is a full rank sub-matrix of (c). The submatrix is 2 rows and 2 columns, which are all->
And is denoted as a k ，
(2) All possible solutions are found for a point X (t, f) in the sparse domain,
(3) find the L1 norm of each solution, take the smallest norm
As an optimal estimate of the source signal, i.e
(4) Repeating the steps (1), 2 and 3) to obtain the optimal solution of all points in the sparse domain;
and 4-2, recovering the time-frequency domain respiration signals separated in the step 4-1 to a time domain, wherein the using formula is as follows:
wherein h (τ -t) is a rectangular window,
is the recovered time domain signal. />
With reference to fig. 4 and 5, the recovered three-person respiratory signal is consistent with the respiratory belt signal, and the performance is reliable.
In summary, the non-contact type multi-person respiratory signal detection method based on the LFMCW radar and underdetermined blind source separation technology is used for detecting vital sign signals of a plurality of objects which are very close to each other, and breaks through the limitation of the angle resolution of a radar system. The system is simple, effective and feasible, has reliable performance, and can be widely applied to a plurality of scenes such as wards, bedrooms, search and rescue after disasters and the like.
The foregoing shows and describes the basic principles, main steps and advantages of the present invention. It will be understood by those skilled in the art that the present invention is not limited to the embodiments described above, and that the above embodiments and descriptions are merely illustrative of the principles of the present invention, and various changes and modifications may be made without departing from the spirit and scope of the invention, which is defined in the appended claims. The scope of the invention is defined by the appended claims and equivalents thereof.
Claims (9)
1. The multi-person respiratory signal detection method based on the separation of the LFMCW radar and the underdetermined blind source is characterized by comprising the following steps:
step 1, detecting chest cavity movements of a plurality of human targets in the same range gate by using a double-receiving antenna LFMCW radar to obtain echo signals of two paths of receiving antennas;
step 2, respectively preprocessing radar echo signals corresponding to each receiving antenna to obtain two paths of mixed signals of multi-person respiration, wherein each path of mixed signals of multi-person respiration is the linear superposition of the paths of multi-person respiration signals; wherein the mixed signal is represented in matrix form as:
in the method, in the process of the invention,
mixed signal vector for representing two mixed signals based on blind source separation model>
For source signal vector->
s 1 ,s 2 ,…,s P The respiratory signals of P human bodies are obtained, wherein A is a mixed matrix of 2 rows and P columns;
step 3, estimating the mixing matrix: using short-time Fourier transforms to vector mixed signals
Transforming to time-frequency domain to obtain time-frequency domain down-mixed signal vector +.>
Time-frequency point X 1 And X 2 Forming a clustering plane, performing cluster analysis by using an improved neighbor propagation algorithm based on the fusion of angle information and energy information to obtain the number of clusters and a cluster center, wherein the number of clusters is +.>
The method is characterized in that the method is used for estimating the number of people to be detected, the clustering center is used for estimating a mixing matrix, and the method is expressed as follows:
wherein (X) 1 (k i ),X 2 (k i ) I) represents the cluster center, k of the i-th class on the cluster plane i Index value of the i-th cluster center;
step 4, according to the estimated mixing matrix
Recovering the respiratory signal; under the underdetermined condition, when the target number is greater than the number of the receiving antennas, recovering and converting the respiratory signals into an optimization problem, introducing constraint conditions, solving the L1 norm minimization to obtain an optimal solution, and obtaining the estimation of the respiratory signals in the time-frequency domain>
Finally pair->
Performing inverse short time Fourier transform to obtain time domain estimate +.>
2. The method for detecting the respiratory signals of the multiple persons based on the separation of the LFMCW radar and the underdetermined blind source according to claim 1, wherein the step 1 specifically comprises:
collecting echo signals s r (t) and local oscillator signal s t (t) performing down-conversion to obtain a difference frequency signal s b (t)。
3. The method for detecting multi-person respiratory signals based on LFMCW radar and underdetermined blind source separation according to claim 1 or 2, wherein the preprocessing of radar echo signals corresponding to each receiving antenna in step 2 comprises:
step 2-1, carrying out rearrangement processing on the difference frequency signals according to the radar repetition period;
step 2-2, calculating the phase of the rearranged echo signals by using a differential cross multiplication algorithm;
and 2-3, filtering the phase signal to obtain a mixed signal.
4. The method for detecting multi-person respiratory signals based on LFMCW radar and underdetermined blind source separation according to claim 3, wherein the step 2-1 of rearranging the difference frequency signals according to the radar repetition period specifically comprises:
(1) the frequency modulation period is T r Is a frequency modulation signal of (1), and the sampling period of the ADC is T s Collecting M sampling points in each frequency modulation period, and collecting echo data of N frequency modulation periods, wherein the difference frequency signals can be rearranged into a matrix R of N rows and M columns, and the matrix R is expressed as:
R[n,m]＝s b (t＝m·T s +n·T r )
wherein n is a matrixLine index, expressed in T r The sampling number is the sampling period, and the value range is 1 to N; m is the column index of the matrix, expressed in T s The sampling number is the sampling period, and the value range is 1 to M;
(2) for matrix R [ n, m ]]N is counted as each line of fft Is fast fourier transformed to obtain a matrix R f [n,v]V is FFT index value, and the value range is 1 to N fft ；
(3) Calculating matrix R f [n,v]The energy of each column is calculated as follows:
the FFT index value V corresponding to the maximum energy value represents the distance gate number of the position where the target is located, and then the target distance r may be represented by the FFT index value V:
wherein K is the frequency modulation coefficient of the linear frequency modulation continuous wave radar, f s Is the sampling rate; the positioning of the target is realized through the method;
the slow time-difference frequency signal s (n) of the position of the target is expressed as:
s(n)＝R f (n,V)。
5. the method for detecting respiratory signals of multiple persons based on LFMCW radar and underdetermined blind source separation according to claim 4, wherein the phase of the rearranged echo signals is calculated by using a differential cross multiplication algorithm in step 2-2, and the calculation formula is as follows:
in the method, in the process of the invention,
is the phase of a slow time-difference frequency signal s I (n) and s Q (n) is quadrature-arm IQ signal, n is frequency modulation period T r The number of samples for a sampling period.
6. The method for detecting the respiratory signals of multiple persons based on the separation of the LFMCW radar and the underdetermined blind source according to claim 5, wherein the filtering of the phase signals in step 2-3 is specifically:
the phase signal is filtered by a band-pass filter h (t) with the order of L-1 and the passband frequency of 0.15-0.55hz, the frequency range is a basic respiratory frequency band, and a mixed signal x (t) is obtained after the filtering, and the calculation formula is as follows:
7. the method for detecting respiratory signals of multiple persons based on LFMCW radar and underdetermined blind source separation according to claim 6, wherein estimating the mixing matrix in step 3 specifically comprises:
step 3-1, transforming the mixed signals breathed by multiple persons to a time-frequency domain, and thinning the mixed signals so as to facilitate cluster analysis; the specific process comprises the following steps:
(1) short-time Fourier transform is carried out on the multi-person respiration mixed signal x (t) corresponding to the two paths of receiving antennas:
X i (t,f)＝∫x i (τ)·g(t-τ)·e -j2πft dτt∈(0,L t -1),f∈(0,L f -1)
wherein g (t) is a Gaussian window function, x i (t) is the multi-person respiration mixed signal corresponding to the ith receiving antenna, X i (t, f) is a time-frequency matrix of the multi-person respiration mixed signal corresponding to the ith receiving antenna, L f Representing FFT point number for the number of matrix lines; l (L) t For the number of matrix columns, the number of time frames is represented;
(2) taking the positive frequency band of the time-frequency matrix, i.e. the frequency f takes a value of 0 to L f 2-1, matrix X i (t, f) rearrangement to one-dimensional vector X i (q) a rearrangement formula:
X i (q)＝X i (t,f)q＝t+f·L t
wherein q is the index number of the time-frequency vector, and the time-frequency vector corresponding to the two paths of receiving antennas is X 1 (q) and X 2 (q)，X 1 (q) and X 2 (q) forming a two-dimensional clustering plane defining a point X on the plane * (q)＝(X 1 (q),X 2 (q))；
Step 3-2, estimating the number of people to be detected and a mixing matrix based on an improved neighbor propagation algorithm of angle information and energy information fusion;
the improved neighbor propagation algorithm flow comprises the following steps:
(1) calculating a similarity matrix s (i, k), wherein the similarity matrix is used for measuring the point X on the clustering plane * (i) And point X * (k) The similarity matrix is defined as follows:
s(i,k)＝abs[X * (k)]·exp(-ρ·(1-cos(∠[X * (i),X * (k)])))
in the formula, abs [ X ] * (k)]Representing clustered planar points X * (k) Is a modulus of (2); angle [ X ] * (i),X * (k)]Representing point X * (i) And point X * (k) Is included in the plane of the first part; ρ is the attenuation coefficient of the nonlinear function, and the larger the ρ value is, the faster the attenuation is, and the influence of insufficient sparse points on clustering can be reduced;
(2) initializing an attraction degree matrix r (i, k) and a attribution degree matrix a (i, k) to 0; wherein the attraction matrix represents the suitability of data point k as the cluster center of data point i; the attribution degree matrix indicates whether the data point i selects the data point k as the clustering center thereof;
(3) updating the attraction matrix r (i, k)
r t+1 (i，k)＝s(i，k)-max j≠k |a t (i，j)+s(i，j)|
Wherein r is t+1 (i, k) represents the attraction degree of the next iteration, a t (i, k) represents the degree of attribution of the current iteration;
(4) updating the home degree matrix a (i, k)
(5) Weighting the current result of step (3) (4) and the last iteration result using an attenuation coefficient λ, typically taking the attenuation coefficient λ=0.5, the weighting formula is as follows:
r t+1 (i,k)＝λ·r t (i,k)+(1-λ)·r t+1 (i,k)
a t+1 (i,k)＝λ·a t (i,k)+(1-λ)·a t+1 (i,k)
(6) repeating the steps (3), (4) and (5) until the matrix converges or the maximum iteration number is reached, and ending the algorithm;
(7) determining data point X * (i) Is the cluster center point X of (2) * (k)＝(X 1 (k),X 2 (k) K) should satisfy the formula:
max{r(i,k)+a(i,k)}
number of cluster centers
The method is used for estimating the number of people to be detected, and a clustering center is a mixing matrix and is expressed as follows:
wherein X is * (k i )＝(X 1 (k i ),X 2 (k i ) I) represents the cluster center of the i-th class, k i Index value of the i-th cluster center;
is an estimate of the number of people to be tested.
8. The method for detecting respiratory signals of multiple persons based on LFMCW radar separation from underdetermined blind sources as recited in claim 7 wherein the step 4 is based on an estimated mixing matrix
The method for recovering the respiratory signal specifically comprises the following steps:
step 4-1, obtaining an optimal solution of the underdetermined problem according to the mixed matrix; when the number of the receiving antennas is smaller than the number of people to be detected, the blind source separation problem is an underdetermined problem, the number of unknown sources is larger than the number of equations, the equations have no unique solution, constraint conditions are introduced, and if the L1 norm to be solved is minimum, the source recovery is converted into an optimization problem; the objective function may be expressed as:
where s (t, f) is a time-frequency domain estimate of each respiratory signal source, X (t, f) is a respiratory mix signal in the time-frequency domain,
a mixing matrix obtained by calculation in the step 3-2;
for a two receive antenna system, the L1 norm minimization algorithm steps are as follows:
(1) computing a mixing matrix
Is a full rank submatrix of 2 rows and 2 columns, altogether +.>
And is denoted as a k ，/>
(2) All possible solutions are found for a point X (t, f) in the sparse domain,
(3) find the L1 norm of each solution, take the smallest norm
As an optimal estimate of the source signal, namely:
(4) repeating the steps (1), 2 and 3) to obtain the optimal solution of all points in the sparse domain;
and 4-2, recovering the time-frequency domain respiration signals separated in the step 4-1 to a time domain, wherein the formula is as follows:
wherein h (τ -t) is a rectangular window,
is the recovered time domain signal.
9. A multi-person respiratory signal detection system based on LFMCW radar separation from underdetermined blind sources, the system comprising:
the radar echo signal acquisition module is used for detecting the chest movements of a plurality of human targets in the same range gate by using a double-receiving antenna LFMCW radar to obtain echo signals of two paths of receiving antennas;
the preprocessing module is used for respectively preprocessing radar echo signals corresponding to each receiving antenna to obtain two paths of mixed signals of multi-person respiration, and the mixed signals of the multi-person respiration in each path are the linear superposition of the respiratory signals of the multi-person; wherein the mixed signal is represented in matrix form as:
in the method, in the process of the invention,
mixed signal vector for representing two mixed signals based on blind source separation model>
For source signal vector->
s 1 ,s 2 ,…,s P The respiratory signals of P human bodies are obtained, wherein A is a mixed matrix of 2 rows and P columns;
a mixing matrix estimation module, configured to estimate the mixing matrix: using short-time Fourier transforms to vector mixed signals
Transforming to time-frequency domain to obtain time-frequency domain down-mixed signal vector +.>
Time-frequency point X 1 And X 2 Forming a clustering plane, performing cluster analysis by using an improved neighbor propagation algorithm based on the fusion of angle information and energy information to obtain the number of clusters and a cluster center, wherein the number of clusters is +.>
The method is characterized in that the method is used for estimating the number of people to be detected, the clustering center is used for estimating a mixing matrix, and the method is expressed as follows:
wherein (X) 1 (k i ),X 2 (k i ) I) the cluster center representing the i-th class on the cluster plane，k i Index value of the i-th cluster center;
a source recovery module for mixing matrix according to estimation
Recovering the respiratory signal; under the underdetermined condition, when the target number is greater than the number of the receiving antennas, recovering and converting the respiratory signals into an optimization problem, introducing constraint conditions, solving the L1 norm minimization to obtain an optimal solution, and obtaining the estimation of the respiratory signals in the time-frequency domain>
Finally pair->
Performing inverse short time Fourier transform to obtain time domain estimate +.>
/>
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111463338.7A CN116224325A (en) | 2021-12-02 | 2021-12-02 | LFMCW radar and underdetermined blind source separation-based multi-person respiratory signal detection method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111463338.7A CN116224325A (en) | 2021-12-02 | 2021-12-02 | LFMCW radar and underdetermined blind source separation-based multi-person respiratory signal detection method and system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116224325A true CN116224325A (en) | 2023-06-06 |
Family
ID=86571746
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111463338.7A Pending CN116224325A (en) | 2021-12-02 | 2021-12-02 | LFMCW radar and underdetermined blind source separation-based multi-person respiratory signal detection method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116224325A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117235476A (en) * | 2023-11-14 | 2023-12-15 | 利维智能（深圳）有限公司 | Fault detection method, device, computer equipment and storage medium |
-
2021
- 2021-12-02 CN CN202111463338.7A patent/CN116224325A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117235476A (en) * | 2023-11-14 | 2023-12-15 | 利维智能（深圳）有限公司 | Fault detection method, device, computer equipment and storage medium |
| CN117235476B (en) * | 2023-11-14 | 2024-02-13 | 利维智能（深圳）有限公司 | Fault detection method, device, computer equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11412937B2 (en) | Multi-person vital signs monitoring using millimeter wave (mm-wave) signals | |
| Islam et al. | Separation of respiratory signatures for multiple subjects using independent component analysis with the JADE algorithm | |
| CN110007366B (en) | Life searching method and system based on multi-sensor fusion | |
| CN113854981A (en) | Millimeter wave radar-based sports multi-person vital sign monitoring method | |
| Wang et al. | Noncontact heart rate measurement based on an improved convolutional sparse coding method using IR-UWB radar | |
| Wu et al. | A non-contact vital signs detection in a multi-channel 77GHz LFMCW radar system | |
| Koda et al. | Noncontact respiratory measurement for multiple people at arbitrary locations using array radar and respiratory-space clustering | |
| Pramudita et al. | Radar system for detecting respiration vital sign of live victim behind the wall | |
| CN115399747B (en) | Multi-person vital sign monitoring method based on millimeter wave radar | |
| Le et al. | Multivariate singular spectral analysis for heartbeat extraction in remote sensing of uwb impulse radar | |
| CN112882016B (en) | Multi-person vital sign detection method based on improved robust adaptive beam forming | |
| CN116224325A (en) | LFMCW radar and underdetermined blind source separation-based multi-person respiratory signal detection method and system | |
| Fioranelli et al. | Contactless radar sensing for health monitoring | |
| CN113341392B (en) | Human behavior classification method based on multi-station radar micro-Doppler motion direction finding | |
| Yamamoto et al. | Non-contact heartbeat detection by heartbeat signal reconstruction based on spectrogram analysis with convolutional LSTM | |
| Zhang et al. | A coarse-to-fine detection and localization method for multiple human subjects under through-wall condition using a new telescopic SIMO UWB radar | |
| Yang et al. | Body orientation and vital sign measurement with IR-UWB radar network | |
| CN110161491A (en) | A kind of ranging and respiratory rate estimation method for faint life entity | |
| Ding et al. | A fall detection method based on K-nearest neighbor algorithm with MIMO millimeter-wave radar | |
| Deng et al. | Efficient living subject localization and weak vital-sign signal enhancement using impulse radio based UWB radar | |
| CN111568396A (en) | V2iFi is based on vital sign monitoring technology in compact radio frequency induction's car | |
| CN113729678B (en) | Wireless sign measurement system based on millimeter wave radar | |
| CN115736886A (en) | Multi-target respiration rate parameter estimation method based on millimeter wave radar | |
| Pan et al. | A spectrum estimation approach for accurate heartbeat detection using Doppler radar based on combination of FTPR and TWV | |
| Sameera et al. | Heart rate detection using single-channel Doppler radar system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |