CN105342594B - Single channel maternal abdominal fetal heart frequency robust estimation method for domestic monitoring - Google Patents

Single channel maternal abdominal fetal heart frequency robust estimation method for domestic monitoring Download PDF

Info

Publication number
CN105342594B
CN105342594B CN201510883842.0A CN201510883842A CN105342594B CN 105342594 B CN105342594 B CN 105342594B CN 201510883842 A CN201510883842 A CN 201510883842A CN 105342594 B CN105342594 B CN 105342594B
Authority
CN
China
Prior art keywords
msub
mrow
msubsup
fetal
mtd
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.)
Expired - Fee Related
Application number
CN201510883842.0A
Other languages
Chinese (zh)
Other versions
CN105342594A (en
Inventor
李国军
李俊兵
周晓娜
叶昌荣
包杨
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Chongqing Communication College of China PLA
Original Assignee
Chongqing Communication College of China PLA
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Chongqing Communication College of China PLA filed Critical Chongqing Communication College of China PLA
Priority to CN201510883842.0A priority Critical patent/CN105342594B/en
Publication of CN105342594A publication Critical patent/CN105342594A/en
Application granted granted Critical
Publication of CN105342594B publication Critical patent/CN105342594B/en
Expired - Fee Related legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Abstract

The invention provides a kind of single channel maternal abdominal fetal heart frequency robust estimation method for domestic monitoring, Holistic modeling is carried out using parent electrocardio as a part for Fetal ECG, stomach wall Fetal ECG signal condition spatial model is established from waveform morphology angle, the recurrence estimation of Fetal ECG R ripples and parent electrocardio waveform is realized with Strong tracking filter method, and peak value positioning is carried out to the Fetal ECG R ripples waveform of estimation, so as to obtain Fetal Instantaneous Heart Rate, this method is applied to calculating, fetal heart frequency is directly extracted on the limited local monitoring terminal of storage capacity, it is and simple to operate, pregnant woman completely can be with complete independently monitoring process, improve Consumer's Experience.

Description

Single channel maternal abdominal fetal heart frequency robust estimation method for domestic monitoring
Technical field
The present invention relates to fetal monitoring field, and in particular to a kind of single channel maternal abdominal fetus heart for domestic monitoring Rate robust estimation method.
Background technology
Perinatal period, pregnant woman need periodically to carry out routine foetus health inspection to hospital.For high risk pregnancy pregnant woman, to doctor Institute checks that number is then more frequent.This kind of pregnant woman goes on a journey inconvenience mostly, and is used as populous nation, and obstetrics of China outpatient is many It is more.Therefore, fetal heart frequency home monitoring system is developed, pregnant woman can stay indoors and realizes the dynamic monitor of fetus, to alleviate The pressure of obstetrics outpatient service doctor, high-risk patient trip risk is reduced, be the key subjects of current mother and baby's monitoring arts research.
At present, the fetal heart frequency based on doppler ultrasound (fetal heart rate, FHR) monitoring is that clinically fetus is good for The basic fundamental means of health inspection.However, the Fetal Heart Rate Monitoring mode based on supersonic sounding is very sensitive to the position of foetus, it is necessary to will visit Head is directed at heart of fetus valve in real time, it usually needs specialist is operated.And this mode belongs to active detection, to tire Youngster's health exist it is potentially hazardous, one-time continuous monitoring the time be usually not more than in half an hour.Above-mentioned factor, limit Doppler and surpass Application of the sound Fetal Heart Rate Monitoring in home telemonitoring.
The fetus heart is realized from pregnant woman's stomach wall collection heart of fetus electric signal (fetal electro-cardiogram, FECG) Rate is guarded, simple to operate due to completely noninvasive, can at any time be carried out in whole perinatal period, is the fetus for being best suited for remote domestic Noninvasive monitoring method.But Fetal ECG causes the FECG that stomach wall gathers to believe by being transmitted by a variety of muscle layer tissues to body surface Make an uproar than extremely low, the interference such as its frequency spectrum and parent electrocardio signal (Maternal ECG, MECG), electrode contact noise, myoelectricity noise The big portion of signal is overlapping, separation and Extraction Fetal ECG nomography complex.This cause current Embryo long-range guardianship also need to by Stomach wall FECG is transferred to monitor center and is analyzed and handled, then the fetal heart frequency of extraction is back into client.It is whole Individual monitoring process is, it is necessary to which real-time access to communication networks, system realize complexity, and resource consumption is big, and cost is high.And pregnant woman is completely sudden and violent It is exposed in electromagnetic environment, poor user experience, is not easy to be received by user.
Separation and Extraction goes out the warp that faint Fetal ECG composition is electricity physiological signal process field from maternal abdominal electrocardio Allusion quotation problem, it experienced from initial adaptive noise cancellation to singular value decomposition (SVD) i.e. pivot analysis (PCA), independent entry " total blindness " that (ICA) is representative is analyzed to separate, then to be recently proposed to constrain ICA, cycle meta analysis (PICA) is representative The research process of " half-blindness " separation.Move due to itself amplitude of stomach wall FECG faint (lower 10 to 30 times than parent interference) and by between fighting State changes (including heart rate, waveform morphology), once Fetal ECG is completely covered by parent QRS complex, existing separation method is usual Very poor robustness is shown, the deep fades of Fetal ECG is frequently resulted in, even disappears completely.Meanwhile existing extraction side Method needs to be acquired the redundant configuration of passage, causes complex operation, and pregnant woman needs layman's cooperation is lower could complete fetal monitoring Process.
The content of the invention
The application by providing a kind of single channel maternal abdominal fetal heart frequency robust estimation method for domestic monitoring, with Solves the noninvasive monitor system realization complexity of remote domestic fetus in the prior art, resource consumption is big, and poor robustness and operation are multiple Caused by miscellaneous pregnant woman can not complete independently monitoring technical problem.
In order to solve the above technical problems, the application is achieved using following technical scheme：
A kind of single channel maternal abdominal fetal heart frequency robust estimation method for domestic monitoring, comprise the following steps：
S1：Signal Pretreatment：The electrocardiosignal of stomach wall fetus is pre-processed, to remove the drift of the baseline in electrocardiosignal Shifting and Hz noise；
S2：The state-space model of single channel stomach wall Fetal ECG signal is built, initializes to obtain mould by systematic parameter The initial value of relevant parameter needed for type；
S3：Strong tracking waveform is estimated：According to the state-space model of Fetal ECG signal and the model parameter of extraction, use Strong tracking filter method, to realize the recurrence estimation of Fetal ECG R ripples and each waveform of parent electrocardio；
S4：Fetus R ripples position：Using recursive Kalman filtering method, R fluctuation state positioning is realized, so as to obtain fetus wink When heart rate；
Wherein, the method that the state-space model of single channel stomach wall Fetal ECG signal is built in step S2 is as follows：
First, the part using parent electrocardio as Fetal ECG, a kind of Fetal ECG signal dynamics of six states are built Model：
In formula,For the instantaneous phase of k moment Fetal ECGs,For the instantaneous heart rate of k moment Fetal ECGs,For The instantaneous phase of k moment parent electrocardios,For the instantaneous heart rate of k moment parent electrocardios, δ is sampling time interval, the k+1 moment Fetal ECG R ripples are modeled with a Gaussian function, are expressed as Rk+1,Point Not Wei Fetal ECG R ripples correspond to amplitude, width and the center of gaussian kernel function, k+1 moment parent electrocardio P ripples and T wavelength-divisions It is not modeled with two gaussian kernel functions, is expressed as Pk+1And Tk+1, k+1 moment parent electrocardios QRS complex uses three Gausses Kernel function is modeled, and is expressed as Ck+1, αi,kAnd bi,k, i ∈ { P1,P2,Q,R,S,T1,T2It is respectively each waveform of parent electrocardio The amplitude and width of corresponding gaussian kernel function,Respectively The center of gaussian kernel function is corresponded to for each waveform of parent electrocardio,For Gaussian noise；
The observation model for obtaining system by the Fetal ECG signal dynamics model is：
In formula,Noise is measured for Fetal ECG phase,Noise, S are measured for parent electrocardio phasekFor the fetus heart Electric signal measurement sample, υS,kTo measure noise；
Make XkRepresent state vector, YkRepresent observation vector, UkRepresent system noise, VkRepresent to measure noise, function f and g System equation and observational equation are described respectively, then the state-space model of stomach wall Fetal ECG signal is：
The present invention is estimated as target with fetal heart frequency, the position of fetus R ripples need to be only tracked, therefore, only with regard to Fetal ECG R ripples Modeling, described with single gaussian kernel function, but for stomach wall FECG, by the position of foetus, gestational age, uterine myoelectricity and electrode Larger change, especially P ripples, T ripples between by shooting generally be present in the influence of the factors such as configuration, parent electrocardio waveform.In order to The preferably dynamic change of tracking parent electrocardio waveform, the present invention is classified as P ripples, three parts of QRS complex waves and T ripples are entered respectively Row modeling, and expand in state vector, corresponding waveform parameter is constrained with this, so as to build a kind of Fetal ECG of six states Signal dynamics model.
Further, systematic parameter initialization is to carry out the premise of waveform estimation, that is, determines disposal and the side of state vector Difference and the average and variance of system noise and observation noise, in step S2 systematic parameter initialization include parent population parameter and initialize With fetus parameter initialization, wherein, parent population parameter initialization specifically include：R ripple spies are carried out to the electrocardiosignal after pretreatment Survey, obtain parent instantaneous heart rateAverage and variance, while electrocardiosignal is converted into phase field from time domain, so as to obtain The instantaneous phase of parent electrocardioThe height that the average heart of parent is clapped and the average heart is clapped is obtained with single channel singular value decomposition method This nuclear parameter；
Fetus parameter initialization specifically includes：Removed from the electrocardiosignal after pretreatment the parent electrocardio of extraction into Point, obtain Fetal ECG according to a preliminary estimate, then R ripple detections are carried out, obtain fetal heart frequencyAverage and variance, while by electrocardio Signal is converted to phase field from time domain, so as to obtain the instantaneous phase of Fetal ECGWith single channel singular value decomposition side Method obtains the Gauss nuclear parameter that the average heart of fetus is clapped and the average heart is clapped.
Because single-channel electrodes allocation plan needs to change with position of foetus situation, and the fetus heart that different electrode positions obtain Electrical parameter is widely different.Under normal circumstances, single channel Fetal Heart Rate Monitoring is carried out every time, is required for carrying out the initial of systematic parameter Change.But because the parameter initialization time is short, generally only about 10 hearts are needed to clap, so only existing 10 seconds or so in the incipient stage Time delay, after electrode configuration is stable, you can realize On-line Estimation, algorithm real-time is unaffected.Therefore, as a kind of preferable Technical scheme, the electrocardiogram (ECG) data of 10 seconds is chosen in step S22 in systematic parameter initialization procedure.
From the state-space model of stomach wall Fetal ECG signal, system dynamics model has obvious non-linear. When said system model has enough precision, EKF filter (Extended Kalman Filtering, EKF) can To provide more accurately state estimation.Therefore, in the estimation of adult's ecg wave form, tiny characteristic extraction and segmentation, EKF estimates Meter method obtains in-depth study and application.However, in the processing of stomach wall electrocardio, because fetal in utero rotation etc. causes stomach wall The perturbation of Fetal ECG model structure, cause Fetal ECG signal dynamics model and described non-linear stomach wall Electrocardiograph not It can match completely.In addition it is electric caused by the random sexual act of myoelectricity noise, parent caused by its exterior interference noise uterine contractile The non-stationaries such as pole contact noise.This requires that filtering system possesses very strong adaptive tracing ability, and EKF filter Electrocardiosignal method for dynamic estimation poor robustness under framework.On the one hand, the rough error that EKF is brought to observation model uncertainty Interference robust is very poor；On the other hand, when system reaches stable state, its gain tends to minimum, will now lose to being mutated shape The ability of tracking of state, it is difficult to the processing system model perturbed problem of itself；EKF filter algorithm is to initial parameter simultaneously Dependence it is also larger.
The uncertain problem of system model in estimating for Nonlinear Dynamic, the present invention use a kind of Strong tracking filter side Method (Strong Tracking Filter, STF), this method is by introducing the fading factor of time-varying, dynamic regulation gain matrix, So that wave filter has preferable robustness to model uncertainty and computation complexity is suitable with EKF, so as to improve to the heart The dynamic tracking capabilities of electrical waveform.
Further, Strong tracking filter method is specially in step S3：
For the nonlinear state equation of the state-space model of Fetal ECG signal, EKF filter recursive estimation Algorithm is：
In formula,For the priori estimates of state vector,For the posterior estimate of state vector, KkFor adjustable gain Battle array,For the covariance matrix of prior estimate,For the covariance matrix of Posterior estimator, RkFor observation noise variance, QkFor shape State noise variance, rkFor the information after renewal, i.e., new breath, Ak、Bk、Ck、DkFor Fetal ECG signal condition spatial model Jacobin matrixes,
In EKF filter recursive Estimation Algorithm, new breath vector r at different moments is forcedkIt is orthogonal, i.e., it is simultaneously full The following two equations of foot：
Try to achieve the adjustable gain battle array K for meeting above formulak, you can obtain Strong tracking filter method.
Compared with prior art, the technical scheme that the application provides, the technique effect or advantage having are：Realize single-pass The robust On-line Estimation of road fetal heart frequency, suitable for calculating, tire directly being extracted on the local monitoring terminal that storage capacity is limited Youngster's heart rate, while ensure service quality；On the other hand, the electrode configuration of single channel collection is simple to operate, and pregnant woman completely can be only It is vertical to complete monitoring process, improve Consumer's Experience.
Brief description of the drawings
The fetal heart frequency that Fig. 1 is the present invention estimates flow chart；
Fig. 2 is original stomach wall FECG and scalp FECG comparison diagram；
Fig. 3 is that maternal ecg estimates comparison diagram with scalp FECG EKF waveforms；
Fig. 4 is that maternal ecg estimates comparison diagram with scalp FECG strong tracking waveform；
Fig. 5 is that the scalp Fetal ECG R ripples positioning figure based on strong tracking filter is schemed with the positioning of stomach wall Fetal ECG R ripples Comparison diagram；
Fig. 6 is scalp Fetal ECG instantaneous heart rate figure and stomach wall Fetal ECG instantaneous heart rate based on strong tracking filter Figure comparison diagram；
Fig. 7 is stomach wall Fetal Instantaneous Heart Rate figure and scalp Fetal Instantaneous Heart Rate figure based on strong tracking filter；
Fig. 8 is stomach wall Fetal Instantaneous Heart Rate and scalp Fetal Instantaneous Heart Rate comparison diagram based on strong tracking filter；
Fig. 9 is stomach wall Fetal Instantaneous Heart Rate and scalp Fetal Instantaneous Heart Rate dependency graph based on strong tracking filter；
Figure 10 is stomach wall fetal heart frequency and scalp fetal heart frequency segmental averaging comparison diagram based on strong tracking filter；
Figure 11 is that the stomach wall fetal heart frequency based on strong tracking filter contrasts with scalp fetal heart frequency segmental averaging error Figure.
Embodiment
The embodiment of the present application is by providing a kind of single channel maternal abdominal fetal heart frequency Robust Estimation for domestic monitoring Method, realize complicated to solve the noninvasive monitor system of remote domestic fetus in the prior art, resource consumption is big, poor robustness and Caused by complex operation pregnant woman can not complete independently monitoring technical problem.
It is right below in conjunction with Figure of description and specific embodiment in order to be better understood from above-mentioned technical proposal Above-mentioned technical proposal is described in detail.
Embodiment
A kind of single channel maternal abdominal fetal heart frequency robust estimation method for domestic monitoring, as shown in figure 1, including such as Lower step：
S1：Signal Pretreatment：The electrocardiosignal of stomach wall fetus is pre-processed, to remove the drift of the baseline in electrocardiosignal Shifting and Hz noise；
S2：The state-space model of single channel stomach wall Fetal ECG signal is built, initializes to obtain mould by systematic parameter The initial value of relevant parameter needed for type；
S3：Strong tracking waveform is estimated：According to the state-space model of Fetal ECG signal and the model parameter of extraction, use Strong tracking filter method, to realize the recurrence estimation of Fetal ECG R ripples and each waveform of parent electrocardio；
S4：Fetus R ripples position：Using recursive Kalman filtering method, R fluctuation state positioning is realized, so as to obtain fetus wink When heart rate；
Wherein, the method that the state-space model of single channel stomach wall Fetal ECG signal is built in step S2 is as follows：
The present invention is estimated as target with fetal heart frequency, the position of fetus R ripples need to be only tracked, therefore, only with regard to Fetal ECG R ripples Modeling, described with single gaussian kernel function, but for stomach wall FECG, by the position of foetus, gestational age, uterine myoelectricity and electrode Larger change, especially P ripples, T ripples between by shooting generally be present in the influence of the factors such as configuration, parent electrocardio waveform.In order to The preferably dynamic change of tracking parent electrocardio waveform, the present invention is classified as P ripples, three parts of QRS complex waves and T ripples are entered respectively Row modeling, and expand in state vector, corresponding waveform parameter is constrained with this, so as to build a kind of Fetal ECG of six states Signal dynamics model：
In formula,For the instantaneous phase of k moment Fetal ECGs,For the instantaneous heart rate of k moment Fetal ECGs,For k The instantaneous phase of moment parent electrocardio,For the instantaneous heart rate of k moment parent electrocardios, δ is sampling time interval, the k+1 moment Fetal ECG R ripples are modeled with a Gaussian function, are expressed as Rk+1,Point Not Wei Fetal ECG R ripples correspond to amplitude, width and the center of gaussian kernel function, k+1 moment parent electrocardio P ripples and T wavelength-divisions It is not modeled with two gaussian kernel functions, is expressed as Pk+1And Tk+1, k+1 moment parent electrocardios QRS complex uses three Gausses Kernel function is modeled, and is expressed as Ck+1, αi,kAnd bi,k, i ∈ { P1,P2,Q,R,S,T1,T2It is respectively each waveform of parent electrocardio The amplitude and width of corresponding gaussian kernel function,Respectively The center of gaussian kernel function is corresponded to for each waveform of parent electrocardio,For Gaussian noise；
The observation model for obtaining system by the Fetal ECG signal dynamics model is：
In formula,Noise is measured for Fetal ECG phase,Noise, S are measured for parent electrocardio phasekFor the fetus heart Electric signal measurement sample, υS,kTo measure noise；
It can be seen that the observation model of system is linear model, and dynamic model is nonlinear model, makes XkExpression state is sweared Amount, YkRepresent observation vector, UkRepresent system noise, VkRepresent to measure noise, function f and g describe system equation and observation respectively Equation, then the state-space model of stomach wall Fetal ECG signal be：
Further, systematic parameter initialization is to carry out the premise of waveform estimation, that is, determines disposal and the side of state vector Difference and the average and variance of system noise and observation noise, it is contemplated that the terseness of parameter initialization algorithm, and initial parameter Required precision is not high, is intentionally clapped without being precisely separating out, therefore, single channel singular value decomposition SVD side is used in the present embodiment Method, parent electrocardio and Fetal ECG are extracted, its averaging model is obtained, using static Gaussian function modeling method, so as to obtain height This nuclear parameter estimate, in step S2 systematic parameter initialization include parent population parameter initialization and fetus parameter initialization, wherein, Parent population parameter initialization specifically includes：R ripple detections are carried out to the electrocardiosignal after pretreatment, obtain parent instantaneous heart rate Average and variance, while electrocardiosignal is converted into phase field from time domain, so as to obtain the instantaneous phase of parent electrocardio The Gauss nuclear parameter that the average heart of parent is clapped and the average heart is clapped is obtained with single channel singular value decomposition method；
Fetus parameter initialization specifically includes：Removed from the electrocardiosignal after pretreatment the parent electrocardio of extraction into Point, obtain Fetal ECG according to a preliminary estimate, then R ripple detections are carried out, obtain fetal heart frequencyAverage and variance, while by electrocardio Signal is converted to phase field from time domain, so as to obtain the instantaneous phase of Fetal ECGWith single channel singular value decomposition side Method obtains the Gauss nuclear parameter that the average heart of fetus is clapped and the average heart is clapped.
Because single-channel electrodes allocation plan needs to change with position of foetus situation, and the fetus heart that different electrode positions obtain Electrical parameter is widely different.Under normal circumstances, single channel Fetal Heart Rate Monitoring is carried out every time, is required for carrying out the initial of systematic parameter Change.But because the parameter initialization time is short, generally only about 10 hearts are needed to clap, so only existing 10 seconds or so in the incipient stage Time delay, after electrode configuration is stable, you can realize On-line Estimation, algorithm real-time is unaffected.Therefore, as a kind of preferable Technical scheme, the electrocardiogram (ECG) data of 10 seconds is chosen in step S22 in systematic parameter initialization procedure.
From the state-space model of stomach wall Fetal ECG signal, system dynamics model has obvious non-linear. When said system model has enough precision, EKF filter (Extended Kalman Filtering, EKF) can To provide more accurately state estimation.Therefore, in the estimation of adult's ecg wave form, tiny characteristic extraction and segmentation, EKF estimates Meter method obtains in-depth study and application.However, in the processing of stomach wall electrocardio, because fetal in utero rotation etc. causes stomach wall The perturbation of Fetal ECG model structure, cause Fetal ECG signal dynamics model and described non-linear stomach wall Electrocardiograph not It can match completely.In addition it is electric caused by the random sexual act of myoelectricity noise, parent caused by its exterior interference noise uterine contractile The non-stationaries such as pole contact noise.This requires that filtering system possesses very strong adaptive tracing ability, and EKF filter Electrocardiosignal method for dynamic estimation poor robustness under framework.On the one hand, the rough error that EKF is brought to observation model uncertainty Interference robust is very poor；On the other hand, when system reaches stable state, its gain tends to minimum, will now lose to being mutated shape The ability of tracking of state, it is difficult to the processing system model perturbed problem of itself；EKF filter algorithm is to initial parameter simultaneously Dependence it is also larger.
The uncertain problem of system model in estimating for Nonlinear Dynamic, the present invention use a kind of Strong tracking filter side Method (Strong Tracking Filter, STF), this method is by introducing the fading factor of time-varying, dynamic regulation gain matrix, So that wave filter has preferable robustness to model uncertainty and computation complexity is suitable with EKF, so as to improve to the heart The dynamic tracking capabilities of electrical waveform.
Further, Strong tracking filter method is specially in step S3：
For the nonlinear state equation of the state-space model of Fetal ECG signal, EKF filter recursive estimation Algorithm is：
In formula,For the priori estimates of state vector,For the posterior estimate of state vector, KkFor adjustable gain Battle array,For the covariance matrix of prior estimate,For the covariance matrix of Posterior estimator, RkFor observation noise variance, QkFor shape State noise variance, rkFor the information after renewal, i.e., new breath, Ak、Bk、Ck、DkFor Fetal ECG signal condition spatial model Jacobin matrixes,
In EKF filter recursive Estimation Algorithm, new breath vector r at different moments is forcedkIt is orthogonal, i.e., it is simultaneously full The following two equations of foot：
Try to achieve the adjustable gain battle array K for meeting above formulak, you can obtain Strong tracking filter method.
However, for nonlinear system, using orthogonality principle, it is difficult accurate meet to make residual sequence constantly orthogonal.To subtract Few amount of calculation, generally use approximate data.A kind of suboptimum strong tracking filter of the multiple time-varying fading factor of band is proposed in the present embodiment Ripple algorithm, past data are faded using the fading factor of time-varying, by the association side for adjusting state forecast error in real time Poor battle array, to weaken influence of the old data to current filter value.
Predicting covariance battle arrayCalculating process it is as follows：
Λ (k+1)=diag { λ1(k+1),…,λn(k+1)}
In formula, forgetting factor 0.95≤ρ≤0.995.
In order to further verify the remarkable result of the present invention, the present embodiment has carried out further experiment and analysis.
At present, the PostgreSQL database for being available for stomach wall Fetal ECG signal transacting to study has three：Belgian Univ Louvain The abdomen of Daily databases, the non-invasive Fetal ECG database of Uni de Valencia Estudi Genera of Spain and Polish University of Silesia Wall and direct Fetal ECG database.The first two database includes some stomach wall Fetal ECG data acquiring and recordings, is the most Conventional measured data, for verifying the efficiency of Fetal ECG separation and Extraction algorithm.But the two databases are not all to bid Accurate Fetal ECG waveform or Fetal Instantaneous Heart Rate.This causes the algorithm performance based on measured data to be analyzed also in naked eyes sight The level examined, the quantitative research of algorithm performance must be realized using artificial synthesized data.
Recently, the database that Polish University of Silesia provides carries out disclosure on Physionet, and the database is given first The Fetal ECG signal from fetal scalp synchronous recording is gone out, data acquisition is in 5 pregnant woman, without chest electrode, four stomach walls Electrode is configured centered on navel, and scalp electrode synchronizes collection, sample rate 1KHz, resolution ratio 16, pregnancy period 38-41 Week, record duration 5 minutes.Scalp Fetal ECG directly gathers from fetus body surface, can obtain the Fetal ECG ripple of complete display Shape, as shown in Fig. 2 can obtain accurate Fetal Instantaneous Heart Rate by positioning R ripples, this is checking stomach wall fetal heart frequency algorithm for estimating Provide a reference well.Therefore, the present embodiment utilizes stomach wall using scalp Fetal ECG data in the database as reference Electrocardiogram (ECG) data record just set forth herein fetal heart frequency algorithm for estimating progress quantitative evaluation.
The present embodiment chooses the first passage that Fetal ECG is most weak in No. four record and carries out test of heuristics.Fig. 2 (a), (b) It sets forth the scalp Fetal ECG waveform of original stomach wall Fetal ECG signal and synchronous acquisition.It can be seen that original abdomen There is serious myoelectricity noise and baseline drift in wall electrocardio, corresponding with the scalp Fetal ECG signal of synchronous acquisition, can see Fetal ECG to stomach wall collection is very faint, and the particularly the 3rd to the 6th fetus heart is clapped almost completely by myoelectricity noise and parent Electrocardio is flooded, as shown in circle in Fig. 2 (a).
EKF recursive Estimation Algorithms and Strong tracking filter first compared with the waveform estimation of parent electrocardio and Fetal ECG R ripples The performance of algorithm.Fig. 3, Fig. 4 respectively illustrate the comparison diagram of the estimation of stomach wall Fetal ECG EKF waveforms and the estimation of strong tracking waveform. Display parent electrocardio EKF waveform estimated result figures in Fig. 3 (a).It can see near Fig. 3 sample points 6000, due to the parent heart Electric QRS complex itself causes the waveform tracking of parent electrocardio and Fetal ECG to occur by the influence of fight change and myoelectricity noise Multiple similar impulse waveforms be present, R ripples positioned and very big interference, serious shadow be present in larger error, the fetus R ripples of EKF estimations Instantaneous Fetal Heart Rate estimation is rung (in figure shown in circle).On the other hand, strong tracking change of the estimation to parent electrocardio waveform and dry Disturbing noise has very strong robustness, Fetal ECG and parent electrocardio waveform can be accurately estimated, as shown in Fig. 4 circles.Together When, even if when fetus R ripples are covered by parent electrocardio T ripples or QRS complex waves completely, Strong tracking filter method proposed by the present invention can To efficiently separate parent electrocardio and Fetal ECG R ripples, as shown in Fig. 4 square frames.
Further, the present embodiment is started with quantitative evaluation EKF recursive Estimation Algorithms and strong tracking filter from fetal heart frequency Performance.Fig. 5 gives pair of the positioning of scalp Fetal ECG R ripples and the positioning of stomach wall Fetal ECG R ripples under strong tracking filter Than figure, it is seen that the two can be accurately positioned R crest values.Fig. 6 is scalp Fetal Instantaneous Heart Rate and abdomen under strong tracking filter The comparison diagram of wall Fetal Instantaneous Heart Rate, it will be seen from figure 6 that the stomach wall Fetal Instantaneous Heart Rate that single channel obtains surrounds scalp tire Youngster's instantaneous heart rate fluctuates up and down, and the two maximum difference is no more than 4 and clapped/point (beat per minute, BPM).Their the average heart Rate is sufficiently close to, and differs only by 0.0785BPM.
Single channel stomach wall Fetal ECG data and the scalp electrocardiogram (ECG) data of 1 minute is randomly selected in the database, is carried out Strong tracking fetal heart frequency is estimated.Remove the jitter section of front and back end, Fig. 7 (a) is the stomach wall tire that continuous 100 hearts are clapped in 1 minute Youngster's instantaneous heart rate estimation condition, Fig. 7 (b) are the scalp Fetal Instantaneous Heart Rate estimation condition that continuous 100 hearts are clapped in 1 minute, Fig. 8 For stomach wall Fetal Instantaneous Heart Rate and scalp Fetal Instantaneous Heart Rate comparison diagram.It can be seen that, generally, estimate from Fig. 7 (a) and Fig. 7 (b) The stomach wall fetal heart frequency of meter has tracked scalp fetal heart frequency well, but certain fluctuation be present, particularly the 40th heart clap with 60th heart fluctuates larger between clapping, and error is all within 5BPM.Fig. 9 is that the correlation of stomach wall Fetal Heart Rate and scalp Fetal Heart Rate is surveyed Examination, it is seen that stomach wall Fetal Heart Rate has correlation well with scalp Fetal Heart Rate.
Because stomach wall Fetal ECG instantaneous heart rate has certain fluctuation, fetal heart frequency in every 10 seconds is put down for this , the Fetal Heart Rate and scalp fetal heart frequency of stomach wall fetus segmental averaging are analyzed again.Choose 5 minutes single channel stomach walls Fetal ECG is tested with scalp Fetal ECG data.Figure 10 gives the experimental result of 5 minutes, divides as seen from Figure 10 After Duan Pingjun, stomach wall fetal heart frequency is more smooth, more close with scalp fetal heart frequency.Figure 11 gives scalp fetal heart frequency With the Error Graph of stomach wall fetal heart frequency, the error provided in Figure 11 is standard deviation.It can be seen that both segmental averaging errors all exist Within 2BPM.
Summary, from the point of view of experimental result in terms of fetus R ripples waveform tracking and fetal heart frequency estimation two, the present invention proposes Fetal ECG robust processing method effectively can reliably recover fetus R ripple waveforms from single channel maternal abdominal electrocardio, Realize being accurately positioned for fetus R ripples, the Fetal Instantaneous Heart Rate and segmental averaging heart rate of acquisition and the scalp fetus heart of synchronous acquisition Rate is sufficiently close to, and most errors are less than 2BPM, and worst error is no more than 5BPM, is supervised suitable for family's fetal heart frequency dynamic Shield.
In above-described embodiment of the application, by providing a kind of single channel maternal abdominal fetal heart frequency for domestic monitoring Robust estimation method, Holistic modeling is carried out using parent electrocardio as a part for Fetal ECG, abdomen is established from waveform morphology angle Wall Fetal ECG signal condition spatial model, Fetal ECG R ripples and parent electrocardio waveform are realized with Strong tracking filter method Recurrence estimation, and peak value positioning is carried out to the Fetal ECG R ripples waveform of estimation, so as to obtain Fetal Instantaneous Heart Rate.
It should be pointed out that it is limitation of the present invention that described above, which is not, the present invention is also not limited to the example above, What those skilled in the art were made in the essential scope of the present invention changes, is modified, adds or replaces, and also should Belong to protection scope of the present invention.

Claims (4)

  1. A kind of 1. single channel maternal abdominal fetal heart frequency robust estimation method for domestic monitoring, it is characterised in that including with Lower step：
    S1：Signal Pretreatment：The electrocardiosignal of stomach wall fetus is pre-processed, with remove the baseline drift in electrocardiosignal and Hz noise；
    S2：The state-space model of single channel stomach wall Fetal ECG signal is built, initializes to obtain model by systematic parameter Need the initial value of relevant parameter；
    S3：Strong tracking waveform is estimated：According to the state-space model of Fetal ECG signal and the model parameter of extraction, with by force with Track filtering method, to realize the recurrence estimation of Fetal ECG R ripples and each waveform of parent electrocardio；
    S4：Fetus R ripples position：Using recursive Kalman filtering method, R fluctuation state positioning is realized, so as to obtain the instantaneous heart of fetus Rate；
    Wherein, the method that the state-space model of single channel stomach wall Fetal ECG signal is built in step S2 is as follows：
    First, the part using parent electrocardio as Fetal ECG, a kind of Fetal ECG signal dynamics model of six states is built：
    In formula,For the instantaneous phase of k moment Fetal ECGs,For the instantaneous heart rate of k moment Fetal ECGs,For the k moment The instantaneous phase of parent electrocardio,For the instantaneous heart rate of k moment parent electrocardios, δ is sampling time interval, the k+1 moment fetus hearts Electric R ripples are modeled with a Gaussian function, are expressed as Rk+1, Respectively fetus Ecg-r wave corresponds to amplitude, width and the center of gaussian kernel function, and k+1 moment parent electrocardio P ripples and T ripples are respectively with two Gaussian kernel function is modeled, and is expressed as Pk+1And Tk+1, k+1 moment parent electrocardios QRS complex entered using three gaussian kernel functions Row modeling, is expressed as Ck+1, αi,kAnd bi,k, i ∈ { P1,P2,Q,R,S,T1,T2It is respectively that each waveform of parent electrocardio corresponds to Gauss The amplitude and width of kernel function,The respectively parent heart The each waveform of electricity corresponds to the center of gaussian kernel function,For Gaussian noise；
    The observation model for obtaining system by the Fetal ECG signal dynamics model is：
    In formula,Noise is measured for Fetal ECG phase,Noise, S are measured for parent electrocardio phasekFor Fetal ECG signal Measure sample, υS,kTo measure noise；
    Make XkRepresent state vector, YkRepresent observation vector, UkRepresent system noise, VkRepresent to measure noise, function f and g difference System equation and observational equation are described, then the state-space model of stomach wall Fetal ECG signal is：
  2. 2. the single channel maternal abdominal fetal heart frequency robust estimation method according to claim 1 for domestic monitoring, its It is characterised by, systematic parameter, which initializes, in step S2 includes parent population parameter initialization and fetus parameter initialization, wherein, parent ginseng Number initialization specifically includes：R ripple detections are carried out to the electrocardiosignal after pretreatment, obtain parent instantaneous heart rateAverage With variance, while electrocardiosignal is converted into phase field from time domain, so as to obtain the instantaneous phase of parent electrocardioWith list Passage singular value decomposition method obtains the Gauss nuclear parameter that the average heart of parent is clapped and the average heart is clapped；
    Fetus parameter initialization specifically includes：The parent electrocardio composition of extraction is removed from the electrocardiosignal after pretreatment, is obtained Obtain Fetal ECG according to a preliminary estimate, then carry out R ripple detections, obtain fetal heart frequencyAverage and variance, while by electrocardiosignal from Time domain is converted to phase field, so as to obtain the instantaneous phase of Fetal ECGObtained with single channel singular value decomposition method The Gauss nuclear parameter that the average heart of fetus is clapped and the average heart is clapped.
  3. 3. the single channel maternal abdominal fetal heart frequency robust estimation method according to claim 1 for domestic monitoring, its It is characterised by, Strong tracking filter method is specially in step S3：
    For the nonlinear state equation of the state-space model of Fetal ECG signal, EKF filter recursive Estimation Algorithm For：
    <mfenced open = "{" close = ""> <mtable> <mtr> <mtd> <mrow> <msubsup> <mover> <mi>X</mi> <mo>^</mo> </mover> <mrow> <mi>k</mi> <mo>+</mo> <mn>1</mn> </mrow> <mo>-</mo> </msubsup> <mo>=</mo> <mi>f</mi> <mrow> <mo>(</mo> <msubsup> <mover> <mi>X</mi> <mo>^</mo> </mover> <mi>k</mi> <mo>+</mo> </msubsup> <mo>)</mo> </mrow> </mrow> </mtd> </mtr> <mtr> <mtd> <mrow> <msub> <mi>K</mi> <mi>k</mi> </msub> <mo>=</mo> <msubsup> <mi>P</mi> <mi>k</mi> <mo>-</mo> </msubsup> <msubsup> <mi>C</mi> <mi>k</mi> <mi>T</mi> </msubsup> <msup> <mrow> <mo>&amp;lsqb;</mo> <msub> <mi>C</mi> <mi>k</mi> </msub> <msubsup> <mi>P</mi> <mi>k</mi> <mo>-</mo> </msubsup> <msubsup> <mi>C</mi> <mi>k</mi> <mi>T</mi> </msubsup> <mo>+</mo> <msub> <mi>D</mi> <mi>k</mi> </msub> <msub> <mi>R</mi> <mi>k</mi> </msub> <msubsup> <mi>D</mi> <mi>k</mi> <mi>T</mi> </msubsup> <mo>&amp;rsqb;</mo> </mrow> <mrow> <mo>-</mo> <mn>1</mn> </mrow> </msup> </mrow> </mtd> </mtr> <mtr> <mtd> <mrow> <msubsup> <mi>P</mi> <mrow> <mi>k</mi> <mo>+</mo> <mn>1</mn> </mrow> <mo>-</mo> </msubsup> <mo>=</mo> <msub> <mi>A</mi> <mi>k</mi> </msub> <msubsup> <mi>P</mi> <mi>k</mi> <mo>+</mo> </msubsup> <msubsup> <mi>A</mi> <mi>k</mi> <mi>T</mi> </msubsup> <mo>+</mo> <msub> <mi>B</mi> <mi>k</mi> </msub> <msub> <mi>Q</mi> <mi>k</mi> </msub> <msubsup> <mi>B</mi> <mi>k</mi> <mi>T</mi> </msubsup> </mrow> </mtd> </mtr> <mtr> <mtd> <mrow> <msub> <mi>r</mi> <mi>k</mi> </msub> <mo>=</mo> <msub> <mi>Y</mi> <mi>k</mi> </msub> <mo>-</mo> <mi>g</mi> <mrow> <mo>(</mo> <msubsup> <mover> <mi>X</mi> <mo>^</mo> </mover> <mi>k</mi> <mo>-</mo> </msubsup> <mo>)</mo> </mrow> </mrow> </mtd> </mtr> <mtr> <mtd> <mrow> <msubsup> <mover> <mi>X</mi> <mo>^</mo> </mover> <mi>k</mi> <mo>+</mo> </msubsup> <mo>=</mo> <msubsup> <mover> <mi>X</mi> <mo>^</mo> </mover> <mi>k</mi> <mo>-</mo> </msubsup> <mo>+</mo> <msub> <mi>K</mi> <mi>k</mi> </msub> <msub> <mi>r</mi> <mi>k</mi> </msub> </mrow> </mtd> </mtr> <mtr> <mtd> <mrow> <msubsup> <mi>P</mi> <mrow> <mi>k</mi> <mo>+</mo> <mn>1</mn> </mrow> <mo>+</mo> </msubsup> <mo>=</mo> <msubsup> <mi>P</mi> <mi>k</mi> <mo>-</mo> </msubsup> <mo>-</mo> <msub> <mi>K</mi> <mi>k</mi> </msub> <msub> <mi>C</mi> <mi>k</mi> </msub> <msubsup> <mi>P</mi> <mi>k</mi> <mo>-</mo> </msubsup> </mrow> </mtd> </mtr> </mtable> </mfenced>
    In formula,For the priori estimates of state vector,For the posterior estimate of state vector, KkFor adjustable gain battle array,For the covariance matrix of prior estimate,For the covariance matrix of Posterior estimator, RkFor observation noise variance, QkMade an uproar for state Sound variance, rkFor the information after renewal, i.e., new breath, Ak、Bk、Ck、DkFor the Jacobin squares of Fetal ECG signal condition spatial model Battle array,
    In EKF filter recursive Estimation Algorithm, new breath vector r at different moments is forcedkIt is orthogonal, i.e., meet simultaneously following Two equations：
    <mfenced open = "" close = ""> <mtable> <mtr> <mtd> <mfenced open = "{" close = ""> <mtable> <mtr> <mtd> <mrow> <mi>E</mi> <mo>&amp;lsqb;</mo> <msub> <mi>X</mi> <mi>k</mi> </msub> <mo>-</mo> <msub> <mover> <mi>X</mi> <mo>^</mo> </mover> <mi>k</mi> </msub> <mo>&amp;rsqb;</mo> <msup> <mrow> <mo>&amp;lsqb;</mo> <msub> <mi>X</mi> <mi>k</mi> </msub> <mo>-</mo> <msub> <mover> <mi>X</mi> <mo>^</mo> </mover> <mi>k</mi> </msub> <mo>&amp;rsqb;</mo> </mrow> <mi>T</mi> </msup> <mo>=</mo> <mi>min</mi> </mrow> </mtd> </mtr> <mtr> <mtd> <mrow> <mi>E</mi> <mo>&amp;lsqb;</mo> <msub> <mi>r</mi> <mrow> <mi>k</mi> <mo>+</mo> <mi>j</mi> </mrow> </msub> <msup> <msub> <mi>r</mi> <mi>k</mi> </msub> <mi>T</mi> </msup> <mo>&amp;rsqb;</mo> <mo>=</mo> <mn>0</mn> </mrow> </mtd> </mtr> </mtable> </mfenced> </mtd> <mtd> <mrow> <mi>k</mi> <mo>=</mo> <mn>0</mn> <mo>,</mo> <mn>1</mn> <mo>,</mo> <mn>2</mn> <mo>,</mo> <mn>...</mn> </mrow> </mtd> <mtd> <mrow> <mi>j</mi> <mo>=</mo> <mn>0</mn> <mo>,</mo> <mn>1</mn> <mo>,</mo> <mn>2</mn> <mo>,</mo> <mn>...</mn> </mrow> </mtd> </mtr> </mtable> </mfenced>
    Try to achieve the adjustable gain battle array K for meeting above formulak, you can obtain Strong tracking filter method.
  4. 4. the single channel maternal abdominal fetal heart frequency robust estimation method according to claim 1 for domestic monitoring, its It is characterised by, chooses the electrocardiogram (ECG) data of 10 seconds in step S2 in systematic parameter initialization procedure.
CN201510883842.0A 2015-12-04 2015-12-04 Single channel maternal abdominal fetal heart frequency robust estimation method for domestic monitoring Expired - Fee Related CN105342594B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201510883842.0A CN105342594B (en) 2015-12-04 2015-12-04 Single channel maternal abdominal fetal heart frequency robust estimation method for domestic monitoring

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201510883842.0A CN105342594B (en) 2015-12-04 2015-12-04 Single channel maternal abdominal fetal heart frequency robust estimation method for domestic monitoring

Publications (2)

Publication Number Publication Date
CN105342594A CN105342594A (en) 2016-02-24
CN105342594B true CN105342594B (en) 2018-01-30

Family

ID=55318775

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201510883842.0A Expired - Fee Related CN105342594B (en) 2015-12-04 2015-12-04 Single channel maternal abdominal fetal heart frequency robust estimation method for domestic monitoring

Country Status (1)

Country Link
CN (1) CN105342594B (en)

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107684423A (en) * 2016-08-05 2018-02-13 深圳先进技术研究院 A kind of Fetal ECG separation method and device
CN108836315A (en) * 2018-07-23 2018-11-20 智联时空(北京)科技有限公司 Intelligent waistband for Fetal ECG monitoring

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101790346A (en) * 2007-07-24 2010-07-28 皇家飞利浦电子股份有限公司 Method of monitoring a fetal heart rate
TW201114406A (en) * 2009-10-21 2011-05-01 Ind Tech Res Inst Apparatus and method for maternal-fetal surveillance
CN102670188A (en) * 2012-04-25 2012-09-19 重庆大学 Robust adaptive estimation method for single-guide heart rate detection of fetus
CN104382589A (en) * 2014-12-09 2015-03-04 南京大学 Fetal electrocardiogram separation extraction method based on partial resampling by segments
KR101504487B1 (en) * 2014-05-23 2015-03-23 광주과학기술원 Real Time System for Measuring Fetal Heart Rate
CN104462867A (en) * 2014-12-11 2015-03-25 中国人民解放军重庆通信学院 T wave dynamic modeling and multi channel estimation fusion method

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20130018273A1 (en) * 2011-07-15 2013-01-17 Tufts University Systems and methods for analysis of fetal heart rate data

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101790346A (en) * 2007-07-24 2010-07-28 皇家飞利浦电子股份有限公司 Method of monitoring a fetal heart rate
TW201114406A (en) * 2009-10-21 2011-05-01 Ind Tech Res Inst Apparatus and method for maternal-fetal surveillance
CN102670188A (en) * 2012-04-25 2012-09-19 重庆大学 Robust adaptive estimation method for single-guide heart rate detection of fetus
KR101504487B1 (en) * 2014-05-23 2015-03-23 광주과학기술원 Real Time System for Measuring Fetal Heart Rate
CN104382589A (en) * 2014-12-09 2015-03-04 南京大学 Fetal electrocardiogram separation extraction method based on partial resampling by segments
CN104462867A (en) * 2014-12-11 2015-03-25 中国人民解放军重庆通信学院 T wave dynamic modeling and multi channel estimation fusion method

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
徐雯，等.基于扩展卡尔曼和多项式网络的单导.《浙江省信号处理学会2015年年会会议论文集-信号处理与大数据》.2015,127-133. *

Also Published As

Publication number Publication date
CN105342594A (en) 2016-02-24

Similar Documents

Publication Publication Date Title
US20200335217A1 (en) Methods and systems using mathematical analysis and machine learning to diagnose disease
Andreotti et al. Robust fetal ECG extraction and detection from abdominal leads
Behar et al. Combining and benchmarking methods of foetal ECG extraction without maternal or scalp electrode data
Liu et al. A multi-step method with signal quality assessment and fine-tuning procedure to locate maternal and fetal QRS complexes from abdominal ECG recordings
Tavakolian et al. Estimation of hemodynamic parameters from seismocardiogram
US11234613B2 (en) Respiration estimation method and apparatus
US9451901B2 (en) Method and device for evaluation of myocardial damages based on the current density variations
Vaneghi et al. A comparative approach to ECG feature extraction methods
CN106175731A (en) The signal processing system of non-contact vital sign monitoring
Valderrama et al. An open source autocorrelation-based method for fetal heart rate estimation from one-dimensional Doppler ultrasound
Lipponen et al. Principal component model for maternal ECG extraction in fetal QRS detection
Jamshidian-Tehrani et al. Fetal ECG extraction from time-varying and low-rank noninvasive maternal abdominal recordings
Su et al. Recovery of the fetal electrocardiogram for morphological analysis from two trans-abdominal channels via optimal shrinkage
CN105342594B (en) Single channel maternal abdominal fetal heart frequency robust estimation method for domestic monitoring
CN116172539A (en) Vital sign detection method, system, equipment and medium based on machine learning
Keenan et al. Detection of fetal arrhythmias in non-invasive fetal ECG recordings using data-driven entropy profiling
Sæderup et al. Comparison of cardiotocography and fetal heart rate estimators based on non-invasive fetal ECG
Li et al. Robust adaptive fetal heart rate estimation for single-channel abdominal ECG recording
Wang et al. Fetal electrocardiogram extraction algorithm in noise: using BSE
Zamrath et al. Robust and computationally efficient approach for Heart Rate monitoring using photoplethysmographic signals during intensive physical excercise
Zhidong Noninvasive diagnosis of coronary artery disease based on instantaneous frequency of diastolic murmurs and SVM
Zhang et al. Separating fetal ECG from transabdominal electrical signal: An application of AE-UNet3+
Danese et al. A wearable intelligent system for the health of expectant mom's and of their children
Azad et al. Postural and longitudinal variability in seismocardiographic signals
Wan et al. A method for extracting fECG based on ICA algorithm

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant
CF01 Termination of patent right due to non-payment of annual fee
CF01 Termination of patent right due to non-payment of annual fee

Granted publication date: 20180130

Termination date: 20201204