CN103110422B - Breath and heartbeat real-time separating method based on biological radar detection - Google Patents

Abstract

The invention discloses a breath and heartbeat real-time separating method based on biological radar detection. Breath and heartbeat signals can be separated from body moving signals in real time after the breath and body moving signals which are detected by a biological radar are pre-processed, harmonic wave of the breath signals is detected, self-adaptive harmonic wave is counteracted and the like, non-contacting and real-time detection on the breath and heartbeat signals are achieved and thereby requirements of biological signal real-time monitoring on clinical patients (burning or infection), senior person in families and other patients of chronic diseases are met.

Description

The breathing and the heart beating real-time separation method that based on bioradar, detect

Technical field

The present invention relates to a kind of bioradar detection technique of utilizing in physiology signal non-contact detecting, by breathing with heart beating, carry out separated method in real time.

Background technology

Along with the needs of biomedical engineering, radar, electronics, computer technology and military affairs, medical science, social development, many Chinese scholars propose the new concept of a kind of physiological signal Detection Techniques---bioradar (Bioradar).This technological incorporation Radar Technology, biomedical engineering technology are in one, penetrable nonmetal medium (clothes, hospital gauze etc.), without any need for electrode or sensor contact life entity, can in larger distance, detect the physiological signal (as breathed, body is moving etc.) of human body, realize the non-contact detecting of physiological signal.Therefore, this technology can be widely used in the occasions such as clinical monitoring, home health care.

Bioradar generally comprises front end, antenna, signal condition hardware and back end signal processing software part, under the prerequisite of the hardware performance index optimums such as front end, signal conditioner, how separated breathing and heartbeat signal in real time in back end signal is processed, to bioradar, applying in clinical physiological signal or home health care has conclusive effect, therefore be subject to the attention of countries in the world height, research worker has proposed diverse ways and has detected the real-time separated of breathing and heart beating for bioradar.According to adopted method, mainly can be divided into: (1) time-domain digital filter method; (2) frequency domain filtering method; (3) wavelet decomposition and Reconstruction Method.Wherein, time-domain filtering method is to adopt FIR(finite impulse response) digital filter directly arranges two digital filters by the moving signal of body, the cut-off frequency of the wave digital lowpass filter that respiration channel is corresponding is 0.5Hz, and the digital band-pass filter lower-cut-off frequency that heart beating passage is corresponding is that 0.6Hz and upper cut-off frequency are 3.3Hz.The advantage of the method be algorithm simple, be easy to realize, but there are two defects: the firstth, instantaneous heartbeat signal can not obtain in real time, main cause is that FIR wave filter exists certain signal time delay when design; The secondth, will cause the partial information of heartbeat signal lose, main cause be approach filter cutoff frequency heartbeat signal also will be by filtering; Above two defects make the isolated breathing of FIR digital filtering method and heartbeat signal, can not be applied to the requirement (requiring physiological parameter real-time and accuracy) of clinical physiological signal monitoring.Frequency domain filtering method is that breathing and the moving signal of body that bioradar is detected carry out FFT(fast Fourier transform) calculate, then according to the spectrum width of breathing, frequency domain filter is set the frequency spectrum of the moving signal of body is carried out to filtering, therefrom isolate the frequency spectrum of heartbeat signal, the last time domain waveform that therefrom extracts again heartbeat signal through contrary FFT conversion.The advantage of the method is from the moving signal of body, to isolate heartbeat signal exactly, and directly obtains instantaneous heart rate value.But also there are two problems: the first, be to meet and to calculate fast FFT direct transform and inverse transformation, to need people be zero padding or reduce data length, easily produces data redundancy and increase operand, thereby reduce, breathes the real-time separated with heartbeat signal; Second: when the fundamental frequency of the higher hamonic wave of breathing and heartbeat signal is overlapping, cannot be from the moving frequency spectrum of body effective filtering breath signal, this situation is very many in clinical cardiovascular patient, so the method can not meet the monitoring requirement of clinical physiological signal.Wavelet method is the frequency band according to breath signal, selects suitable wavelet packet to carry out Time Domain Decomposition the moving signal of body, then by the signal zero setting of breathing in frequency band, and other signal components of residue is carried out to wavelet reconstruction, finally synthetic heartbeat signal.The advantage of the method is that capacity of resisting disturbance is strong, and can be by all information completelies ground reconstruct of heartbeat signal, but is difficult to select wavelet packet during wavelet decomposition, and algorithm is complicated, operand is large, can not realize and breathing and heartbeat signal real-time separated.Therefore, the method does not meet the monitoring requirement of clinical physiological signal yet.

Summary of the invention

The object of the invention is in physiology signal non-contact detecting, particularly utilize in the process of bioradar detection, the real-time separation method of the simple breathing of a kind of algorithm and heart beating is provided, can from the moving signal of body, isolate in real time breathing and heartbeat signal, thereby meet the requirement of clinical patient (burn, infection), the old man of family and other chronics' physiological signal Real-Time Monitoring.

For reaching above object, the present invention takes following technical scheme to be achieved:

The breathing and the heart beating real-time separation method that based on bioradar, detect, is characterized in that, comprises the steps:

(1) the moving signal of breath signal bioradar being detected and body carries out respectively pretreatment, breathe and the frequency range of the moving signal of body is limited in 5Hz, be limited to-1V of the amplitude of signal extremely+1V, pretreatment comprises digital filtering and normalized;

(2) pretreated breath signal is through breathing harmonic wave detection module, therefrom extract the higher hamonic wave more than secondary of breath signal, comprise that breath signal first-harmonic is estimated and breath signal harmonic wave is synthetic, wherein, breath signal first-harmonic estimates to adopt auto-correlation algorithm to solve fundamental frequency; The synthetic Gauss-Newton algorithm that adopts of breath signal harmonic wave;

(3) output of breathing harmonic wave detection module is sent into self adaptation harmonic cancellation module and as its reference-input signal; The moving signal of pretreated body is sent into self adaptation harmonic cancellation module and as its original input signal, by adaptive algorithm, constantly adjust and upgrade the parameter of sef-adapting filter, the difference of the harmonic components of the respiratory component in the moving signal of body and the breath signal of reference input square hour, the now output of sef-adapting filter is exactly heartbeat signal.

In said method, the described digital filtering of step (1) adopts Butterworth iir filter; Normalized is carried out as follows: y=(x-Min)/(Max-Min), wherein x is the breath signal after digital filtering, and y is signal after normalization, and Max and Min represent respectively to breathe after filtering and body moves maximum and the minima of signal.

The idiographic flow that the described breathing harmonic wave of step (2) detects is as follows:

The first step: the fundamental frequency that solves breath signal

Choose suitable window function breath signal is carried out to segment processing, to asking auto-correlation function in each data segment, then auto-correlation function is carried out to power Spectral Estimation, the fundamental frequency f that the corresponding Frequency point of maximum of power spectrum energy is breath signal ₀, then from front to back, successively breath signal is carried out windowing, self correlation, asks the processing such as power spectrum, to f ₀constantly upgrade;

Second step: build breath signal higher hamonic wave mathematical model

The model definition of the movement of thorax that respiratory movement is caused is:

$S(a,n,{\mathrm{\ω}}_0)=a_{2l+1}+\underset{l=1}{\overset{L}{\mathrm{\Σ}}}{a}_{2l-1}\cos ({\mathrm{l\ω}}_{0}m+a_{2l})$

In formula: a represents breath signal, n represents sampling instant, ω ₀for the first-harmonic angular frequency of breath signal, f ₀the fundamental frequency that represents breath signal, f _srepresent sample frequency, l ω ₀for the frequency of l subharmonic, a _ifor [a ₁, a ₂... .a _2l+1] ^tvector, a _2lrepresent the phase place of l subharmonic, a _2l-1represent the amplitude of l subharmonic, a _2l+1represent the DC component of l subharmonic;

The 3rd step: utilize Gauss-Newton algorithm to solve a in second step mathematical model _ivalue, constantly by the f upgrading ₀value is carried out interative computation, when meeting the optimal solution of Gauss-Newton, obtains the higher hamonic wave function of synthetic breath signal.

The described adaptive algorithm of step (3) adopts LMS algorithm, step size mu=0.00002, and it is as follows that the exponent number of wave filter is got 20. concrete calculation procedures:

The filter factor vector of definition time n=0 is initial value W (0), then carries out iteration:

A, by the estimated value of the filter filtering coefficient vector of current time n, input breath signal harmonic vector x (n) and body move signal d (n), error signal:

$e\left(n\right)=d\left(n\right)-x^H\left(n\right)\hat{w}\left(n\right)$

B, again by the renewal valuation of recurrence method calculating filter coefficient vector:

$\hat{w}(n+1)=\hat{w}\left(n\right)+\mathrm{\μe}\left(n\right)x\left(n\right)$

C, time index n is increased to 1, then start step a, repeat above-mentioned calculating, arrive steady-state algorithm always and stop.

Compare with heart beating separation method with the breathing of existing non-contact detecting, the present invention has the following advantages:

1, real-time is good: in the separation process of breathing and heartbeat signal, there is no the time delay of signal, the real-time of breathing and heart beating is all better than frequency domain filtering and wavelet decomposition reconstructing method.

2, the suitability is strong: no matter body moves whether the heartbeat signal frequency comprising in signal is the stack of breath signal higher hamonic wave, all can effectively from the moving signal of body, isolate heartbeat signal, the suitability is all better than time-domain filtering, frequency domain filtering and wavelet decomposition reconstructing method.

Accompanying drawing explanation

Below in conjunction with accompanying drawing and the specific embodiment, the present invention is described in further detail.

Fig. 1 breathes during bioradar of the present invention detects and the functional block diagram of heart beating real-time separation method.In figure: input signal is breath signal and the moving signal of body that biological detections of radar arrives human body, and output signal is isolated heartbeat signal from the moving signal of body.

Fig. 2 is the functional structure chart of Fig. 1 self adaptation harmonic cancellation module.

Fig. 3 is the human body heartbeat signal that adopts the inventive method and extract and electrocardiosignal by comparison.Wherein (a) figure is breath signal, (b) figure is the moving signal of body, (c) figure is isolated heartbeat signal, and (d) figure is the electrocardiosignal of monitoring simultaneously, (e)-(h) be respectively breath signal, body moving signal, isolated heartbeat signal and the corresponding frequency spectrum of electrocardiosignal.

Fig. 4 is the heart rate comparison of computational results of 16 experimental subjecies.Wherein, dotted line represents the heart rate calculating from the human body heartbeat signal of the inventive method extraction; Solid line represents the heart rate calculating from the electrocardiosignal of monitoring simultaneously.

The specific embodiment

As shown in Figure 1, a kind of breathing of bioradar detection and real-time separation method of heart beating of utilizing, comprises two-way pre-processing module (breath signal pretreatment, the moving Signal Pretreatment of body), a breathing harmonic wave detection module and a self adaptation harmonic cancellation module.Pre-processing module by bioradar non-contact detecting to breathing and the moving signal of body carry out pretreatment; Pretreated breath signal is through breathing harmonic wave detection module, therefrom extracts the harmonic wave more than secondary of breath signal, directly sends into self adaptation harmonic cancellation module and as its reference-input signal; The moving signal of pretreated body directly sends into self adaptation harmonic cancellation module and as its original input signal, its output signal is heartbeat signal, realizes from the moving signal of body and isolates breath signal.

Wherein, pre-processing module comprises digital filtering and normalized, can will breathe and the frequency range of the moving signal of body be limited in 5Hz, be limited to-1V of the amplitude of signal extremely+1V.Digital filtering adopts Butterworth iir filter, and the design parameter of wave filter adopts Matlab software to calculate; Normalized is scaled to [1 by the amplitude of breathing and the moving signal of body, 1] interval, normalized is carried out according to following formula: y=(x-Min)/(Max-Min), wherein x is the breath signal after digital filtering, y is signal after normalization, and Max and Min represent respectively to breathe after filtering and body moves maximum and the minima of signal.

Breathe harmonic wave detection module and synthesized and formed by the estimation of breath signal first-harmonic and breath signal harmonic wave, breath signal first-harmonic estimates to adopt autocorrelation technique, the synthetic employing of breath signal harmonic wave gauss-newton method.The process that breath signal harmonic wave detects is as follows:

The first step: the fundamental frequency that solves breath signal.

Choose suitable window function breath signal is carried out to segment processing, to asking auto-correlation function in each data segment, then auto-correlation function is carried out to power Spectral Estimation, the fundamental frequency f that the corresponding Frequency point of maximum of power spectrum energy is breath signal ₀.Then from front to back, successively breath signal is carried out windowing, self correlation, asks the processing such as power spectrum, to f ₀constantly upgrade.

Second step: the mathematical model that builds breath signal higher hamonic wave.

The model definition of the movement of thorax that respiratory movement is caused is:

$S(a,n,{\mathrm{\ω}}_0)=a_{2l+1}+\underset{l=1}{\overset{L}{\mathrm{\Σ}}}{a}_{2l-1}\cos ({\mathrm{l\ω}}_{0}m+a_{2l})$

In formula: a represents breath signal, n represents sampling instant, ω ₀for the first-harmonic angular frequency of breath signal ( f ₀the fundamental frequency that represents breath signal, f _srepresent sample frequency), l ω ₀for the frequency of l subharmonic, a _ifor [a ₁, a ₂... .a _2l+1] ^tvector, a _2lrepresent the phase place of l subharmonic, a _2l-1represent the amplitude of l subharmonic, a _2l+1represent the DC component of l subharmonic.

The 3rd step: utilize gauss-newton method to synthesize breath signal higher hamonic wave.

Adopt gauss-newton method to solve a in above-mentioned mathematical model _ivalue, constantly by the f upgrading ₀value is carried out interative computation, when meeting the optimal solution of Gauss-Newton, obtains the higher hamonic wave function of synthetic breath signal.

With reference to figure 2, self adaptation harmonic cancellation module will be breathed the reference input of the output of harmonic wave detection module as sef-adapting filter (adaptive noise canceller), original input using the moving signal of pretreated body as sef-adapting filter, by adaptive algorithm, constantly adjust and upgrade the parameter of sef-adapting filter, but the difference of the harmonic components of the respiratory component in the moving signal of body and the breath signal of reference input square hour, the now output of sef-adapting filter is exactly heartbeat signal.Adaptive algorithm adopts simple LMS(least mean-square error) algorithm, step size mu=0.00002, it is as follows that the exponent number of wave filter is got 20. its calculation procedures: the filter factor vector of definition time n=0 is initial value W (0), then carries out iteration:

(1) by the estimated value of the filter filtering coefficient vector of current time n, input breath signal harmonic vector x (n) and body move signal d (n), error signal:

$e\left(n\right)=d\left(n\right)-x^H\left(n\right)\hat{w}\left(n\right)$

In formula: the weight coefficient that represents wave filter, T represents that the matrix that signal is formed carries out transposition computing.

(2) again by the renewal valuation of recurrence method calculating filter coefficient vector:

$\hat{w}(n+1)=\hat{w}\left(n\right)+\mathrm{\μe}\left(n\right)x\left(n\right)$

In formula: μ represents the time step of adaptive updates wave filter weight coefficient.

(3) time index n is increased to 1, then start step (1), repeat above-mentioned calculating, arrive steady-state algorithm always and stop.

As shown in Figure 3, from frequency spectrum, can find out, crest frequency corresponding in breath signal power spectrum is 0.2188Hz, and now corresponding breathing rate is 13 beats/min; Crest frequency corresponding in heartbeat signal power spectrum is 1.0Hz, and now corresponding heart rate is 60 beats/min; Crest frequency corresponding in electrocardiosignal power spectrum is 1.02Hz, and now corresponding heart rate is 61 beats/min.From result, can find out, it is very little with the heart rate error obtaining from electrocardiosignal that separation method provided by the invention obtains heart rate.

As shown in Figure 4, as can be seen from the figure the heart rate of 16 experimental subjecies is closely similar.Experimental result shows, breathing provided by the invention and heart beating real-time separation method are effective, feasible completely.

Claims (4)

1. the breathing and the heart beating real-time separation method that based on bioradar, detect, is characterized in that, comprises the steps:

(1) the moving signal of breath signal bioradar being detected and body carries out respectively pretreatment, breathe and the frequency range of the moving signal of body is limited in 5Hz, be limited to-1V of the amplitude of signal extremely+1V, pretreatment comprises digital filtering and normalized;

(2) pretreated breath signal is through breathing harmonic wave detection module, therefrom extract the higher hamonic wave more than secondary of breath signal, comprise that breath signal first-harmonic is estimated and breath signal harmonic wave is synthetic, wherein, breath signal first-harmonic estimates to adopt auto-correlation algorithm to solve fundamental frequency; The synthetic Gauss-Newton algorithm that adopts of breath signal harmonic wave;

(3) output of breathing harmonic wave detection module is sent into self adaptation harmonic cancellation module and as its reference-input signal; The moving signal of pretreated body is sent into self adaptation harmonic cancellation module and as its original input signal, by adaptive algorithm, constantly adjust and upgrade the parameter of sef-adapting filter, the difference of the harmonic components of the respiratory component in the moving signal of body and the breath signal of reference input square hour, the now output of sef-adapting filter is exactly heartbeat signal.

2. breathing and the heart beating real-time separation method detecting based on bioradar as claimed in claim 1, it is characterized in that, the described normalized of step (1) is carried out as follows: y=(x-Min)/(Max-Min), wherein x is breath signal or the moving signal of body after digital filtering, y is signal after normalization, and Max and Min represent respectively to breathe after filtering or maximum and the minima of the moving signal of body.

3. breathing and the heart beating real-time separation method detecting based on bioradar as claimed in claim 1, is characterized in that, the idiographic flow that the described breathing harmonic wave of step (2) detects is as follows:

The first step: the fundamental frequency that solves breath signal

Choose suitable window function breath signal is carried out to segment processing, to asking auto-correlation function in each data segment, then auto-correlation function is carried out to power Spectral Estimation, the fundamental frequency f that the corresponding Frequency point of maximum of power spectrum energy is breath signal ₀, then from front to back, successively breath signal is carried out windowing, self correlation, asks power spectrum to process, to f ₀constantly upgrade;

Second step: build breath signal higher hamonic wave mathematical model

The model definition of the movement of thorax that respiratory movement is caused is:

$S(a,n,{\mathrm{\ω}}_0)=a_{2l+1}+\underset{l=1}{\overset{l}{\mathrm{\Σ}}}{a}_{2l-1}\cos ({\mathrm{l\ω}}_{0}n+a_{2l})$

In formula: a represents breath signal, n represents sampling instant, ω ₀for the first-harmonic angular frequency of breath signal, f ₀the fundamental frequency that represents breath signal, f _srepresent sample frequency, l ω ₀for the frequency of l subharmonic, a _ifor [a ₁, a ₂... .a _2l+1] ^tvector, i=1,2 ..., 2l+1, a _2lrepresent the phase place of l subharmonic, a _2l-1represent the amplitude of l subharmonic, a _2l+1represent the DC component of l subharmonic;

The 3rd step: utilize Gauss-Newton algorithm to solve a in second step mathematical model _ivalue, constantly by the f upgrading ₀value is carried out interative computation, when meeting the optimal solution of Gauss-Newton, obtains the higher hamonic wave function of synthetic breath signal.

4. breathing and the heart beating real-time separation method detecting based on bioradar as claimed in claim 1, it is characterized in that, the described adaptive algorithm of step (3) adopts LMS algorithm, step size mu=0.00002, and it is as follows that the exponent number of wave filter is got 20. concrete calculation procedures:

The filter factor vector of definition time n=0 is initial value W (0), then carries out iteration:

A, by the estimated value of the filter filtering coefficient vector of current time n, input breath signal harmonic vector x (n) and body move signal d (n), error signal:

$e\left(n\right)=d\left(n\right)-x^H\left(n\right)\hat{w}\left(n\right)$

B, again by the renewal valuation of recurrence method calculating filter coefficient vector:

$\hat{w}(n+1)=\hat{w}\left(n\right)+\mathrm{\μe}\left(n\right)x\left(n\right)$

C, time index n is increased to 1, then start step a, repeat above-mentioned calculating, arrive steady-state algorithm always and stop.