CN103876730B - Blind extraction method for electrocatdiogram of mother and electrocardiogram of fetus based on second-order statistical properties - Google Patents

Abstract

Maternal and fetal ECG blind extraction method of the present invention discloses a method based on second-order statistical characteristics to solve the smaller mixed-signal to noise ratio, and the problem can not be better to extract in the extraction process work offline. Step of the present invention are: (1) obtaining electrophysiological mixed signal; (2) pretreatment; (3) selecting the maximum way mixed signal to noise ratio; (4) the estimated period; (5) for optimal ECG signal vector; (6) extracting ECG. The method of the present invention compared to the prior art maternal and fetal ECG extraction, to ensure the accuracy of the signal extraction, while also having a real-time online extraction, extraction efficiency advantages of the present invention may be used to collect from the machine to a heart monitor maternal ECG signal extracted maternal and fetal ECG.

Description

Blind extraction method of second-order statistical characteristics of the mother and fetus ECG-based

FIELD

[0001] The present invention relates to signal processing technologies, and further relates to a biomedical signal processing technologies based on an electric signal extraction blind mother and fetus second order statistical properties of the heart methods. The present invention may be used to collect ECG signals from maternal ECG monitoring unit to extract the maternal and fetal ECG.

Background technique

[0002] human electrocardiogram (ECG) is a reflection of objective indicators of human physiological activities. Obtained from the pregnant mother generally comprises the body surface electrical signal measuring ECG (MECG), mixed FECG signal (to FECG) and other body electrical signal (e.g., EMG EMG, etc.) and an observation of the background noise. Measuring the electrical signal acquired in the body of pregnant women, since the strength of large mother ships ECG signal using conventional filtering techniques can extract mother ECG. Fetus ECG quite weak, which is less than one-tenth the strength of the mother's ECG signal, and the spectrum of Fetal ECG ECG mother overlap each other, it is generally difficult to extract the filtering technique a clear stable Fetal electric signal.

[0003] Guangdong University have patented technology, "blind fetal ECG extraction method based on frequency transformation" (patent application number: 2011101444877 authorizing Publication No. CN 102160787B) discloses a frequency transform based on the fetus ECG blind extraction method. The obtained mixed signal acquisition method comprising the mother and fetus electrophysiological signals, selection signals mixed maternal and fetal ECG relatively sparse period, using the fuzzy function obtained relatively sparse period into the time frequency domain, and then using control generalized Rayleigh entropy function configured to extract the fetal ECG. The patented technology uses mixed maternal and fetal ECG signal is relatively sparse domain characteristics to solve the maternal and fetal ECG spectral overlap difficult to isolate the problem, but there are still shortcomings that the method works in offline mode and can not reflect fetal ECG signal varying characteristics.

[0004] Shanghai Maritime University proposed patent application (patent application number: 2013107297368, publication No. CN 103627796A) "generalized eigenvalue maximize adaptive blind extraction of fetal ECG based on the" discloses a fetal ECG adaptive blind signal extraction method. The obtained mixed signal acquisition method comprising the mother and fetus electrophysiological signals, real-time adaptive algorithm, obtaining eigenvectors and eigenvalues ​​mixed signals with different delays calculated covariance matrix is ​​set within the period range, and selects the maximum eigenvalue corresponding separation vector as a blind feature vector to extract the fetal ECG. Although this method using the second-order statistical properties of the signal, simplify computational complexity, improve the fetal ECG signal blind separation efficiency, but there are still shortcomings, the method of ECG cycle estimate is not precise enough, and because of over-reliance mixed maternal and fetal ECG signal acquisition to a different delay of the covariance matrix can be diagonalized, leading to a small collection mixed signal to noise ratio, poor extraction effect.

SUMMARY

[0005] The object of the present invention is to overcome the above disadvantages of the conventional methods, a method is proposed blind cardiac electrical signal based on second order statistical properties of the extracted, and the mother for adaptive real-time extraction of the fetal ECG.

[0006] The basic idea of ​​the invention is to first obtain the multiplexed electrical signal preprocessing physiological mixed, then the mixed signal all the way to the maximum SNR is selected from the multiple electrophysiological mixed signal acquisition, and then estimating the Mixture periodic signal, the mixed signal vector is extracted from the most optimum separation path, the final extract using the optimum separation vector ECG signals from the mixed signal preprocessing.

[0007] The specific steps of the present invention are as follows:

[0008] (1) the mixed signal acquired electrophysiological:

[0009] acquiring physiological signals mixed maternal abdomen and chest electrical multiplex synchronous collection;

[0010] (2) Pretreatment:

[0011] (2a) using a finite length digital low pass filter, the mixing of multiple electrophysiological EMG signal interference and noise filtering, to obtain the multiplexed electrical interference and noise filtered EMG physiological mixed signal;

[0012] (2b) using the finite length digital notch filter, the multi-channel electrical interference and noise filter physiological EMG mixed 50Hz frequency interference signal filtered to give multiple electrophysiological preprocessed downmix signal;

[0013] (3) Each SNR obtained mixed physiological electrical signals:

[0014] (3a) after pre-mixed electrophysiological Each ECG signal as a signal in the mother, fetus ECG as noise, signal to noise ratio to obtain the mother and fetus ECG ECG signals;

[0015] (3b) The pretreated Each electrophysiological mixed signal as a signal of fetal ECG, ECG mother as noise, signal to noise ratio obtained Fetal ECG and maternal ECG signals;

[0016] (4) estimation period:

[0017] (4a) selecting mother and fetus ECG ECG way maximum signal to noise ratio electrophysiological mixed signal, in which two adjacent mother ECG R wave interval, as a mother ECG cycle pre-estimated value;

[0018] (4b) using the autocorrelation coefficients formula to calculate the autocorrelation coefficient mother and fetus ECG ECG way maximum signal to noise ratio electrophysiological mixed signal;

[0019] (4c) the pre-estimate the mother's heart electrical cycles error of plus or minus 0.2 seconds as the search period of the search from the maximum absolute value of the correlation coefficient, the maximum value from the absolute value of the correlation coefficient time as a mother's heart electrical cycles accurate estimate;

[0020] (4d) selected Fetal ECG and maternal ECG way maximum signal to noise ratio electrophysiological mixed signal, wherein the fetal ECG two adjacent R wave interval, as the period of the fetal ECG pre-estimated value;

[0021] (4e) autocorrelation coefficients using autocorrelation coefficient formula, the fetal ECG and maternal ECG way maximum signal to noise ratio electrophysiological mixed signal;

[0022] (4f) pre-estimate the fetal heart electrical cycles of error of plus or minus 0.15 seconds as a search time, the search from the maximum absolute value of the correlation coefficient, the maximum value from the absolute value of the correlation coefficient as the time period of the fetal heart electrical accurate estimate;

[0023] (5) to obtain the optimal ECG vector:

[0024] (5a) using the steepest descent method, to obtain optimal mother ECG vector;

[0025] (5b) correction using Newton's method to obtain optimal fetal ECG vector;

[0026] (6) extracting ECG:

[0027] (6a) using the projection method, using the optimum vector ECG signal from the mother multiple electrophysiological preprocessed downmix signal extracted ECG mother;

[0028] (6b) using the projection method, using the optimum vector physiological fetal ECG signal from the multiplexer mixed electrical preprocessed to extract the fetal ECG.

[0029] The prior art and the present invention has the following advantages:

[0030] First, since the present invention is by mixing multiple electrophysiological signal acquired was pretreated, reacquisition maternal and fetal ECG optimum vector Electrophysiological mixed signal preprocessing, to overcome the existing eLECTRICAL smaller SNR physiological mixed signal extraction is preferably not less than maternal and fetal ECG signal, such that the present invention is mixed an additional condition signal to noise ratio without electrophysiological size.

[0031] Second, since the present invention is the electrical physiological signal by the mixed second order statistical properties of the mother and the fetus obtain the optimal ECG vectors, it overcomes the disadvantages of the prior art work off-line, to achieve the mother and fetus ECG online real-time signal is extracted, so that the high maternal and fetal ECG extraction efficiency, to better reflect when maternal and fetal ECG signal varying characteristics.

[0032] Third, since the present invention is the maximum signal to noise ratio by periodic physiological electrical signals mixed maternal and fetal ECG signal prior to pre-estimate, then set accurately estimate the error, the prior art overcomes the mother and fetus ECG cycle estimate is not precise enough deficiencies, so that the effect of the present invention is to extract the mother and fetus ECG is more favorable.

BRIEF DESCRIPTION

[0033] FIG. 1 is a flowchart of the present invention;

8 is electrically [0034] FIG. 2 of the present invention, the input signal waveform diagram physiological mixed;

[0035] FIG. 3 is a schematic diagram of the present invention using the autocorrelation coefficients obtained in the first signal path;

[0036] FIG. 4 is a schematic diagram of the present invention using the autocorrelation coefficients obtained in the first signal path 8;

[0037] FIG. 5 extracts the parent ECG waveform is a comparison chart using the four methods of the present invention and the prior art;

[0038] FIG. 6 is four methods of the present invention and the prior art to extract the fetal ECG waveforms can be compared to FIG.

Detailed ways

[0039] DRAWINGS further description of the invention:

[0040] Referring to FIG. 1, the specific embodiment of the present invention, the following steps:

[0041] Step 1, the input signal electrophysiological mixed.

[0042] Example embodiments of the present invention is the input 8 electrophysiological mixed signal, mixing the signal synchronized electrophysiological collected from the 3-way mix electrophysiological signal path 5 electrical physiological precursor mix signal and the chest of the maternal abdomen. Referring to the drawings in Figure 2 the input signal waveform diagram electrophysiological mixed eight, in FIG. 2 the abscissa represents time, the ordinate indicates the signal amplitude, FIG. 2 (a), (b), (c), (d) , (e) mixing a physiological signal acquired from the 5-way electrical parent chest, FIG. 2 (f), (g), (h) mixing a physiological signal from the 3-way power acquisition maternal abdomen. As can be seen from Figure 2, 5-way mixed electrophysiological signal from the primary mother chest ECG, and the power from the 3-way physiological abdomen mixed signal weakly fetal ECG.

[0043] Step 2 pretreatment.

[0044] The first step, based Hamming window of 8-order finite length digital low pass filter, for mixing the acquired multiplexed signal to interference and noise filtering EMG, EMG interference and noise filtered off to give the mixed multiplex signal;

[0045] The second step, using the notch filter 40 based on the order of finite length digital Hamming window, the multi-channel noise filter and EMG interference signal mixed 50Hz frequency interference filtered off, the obtained multi pretreatment mixed-signal path.

[0046] Step 3, to give the mixed signal to noise ratio for each channel of the electrical physiological.

[0047] The first step, after the pre-mixed electrophysiological Each ECG signal as a signal in the mother, fetus ECG as noise, signal to noise ratio to obtain the mother and fetus ECG ECG signals;

[0048] The second step, after the pre-mixed signal for each channel electrophysiological as fetal ECG signal, the ECG signal as noise mother, fetus ECG signal to noise ratio and to obtain the mother's ECG.

[0049] Step 4, estimation period.

[0050] ECG cycle mother estimate α, specific steps comprising:

[0051] The first step, selecting the maximum of the way the mother and fetus ECG ECG signal from the multiplexed signal to noise mixed after the pre-mixed signal, the mixed signal that is referred to at time t X (t) , in the X (t) two adjacent mother ECG R-wave interval as the estimated value pre mother electrical cycles of the heart. ECG cycle mother obtained from x (t) pre-estimated value α 0. 8 s;

[0052] The second step, x (t) is the autocorrelation coefficients calculated as follows:

[0054] wherein, r (δ) represents the delay time for the mother and fetus ECG ECG signal to noise ratio of the maximum autocorrelation coefficients physiological way when mixed electrical signal δ, δ represents the mother and fetus ECG ECG the ratio of maximum signal delay time of the mixed signal corresponding electrophysiological way time t, x (t) represents the mother and fetus ECG ECG way maximum signal to noise ratio electrophysiological mixed signal, t represents the mother and ECG fetal ECG way maximum signal to noise ratio mixed electrophysiological signal sampling timing, J (·) dt represent the integral operation of the sampling time t, [·] represents 2 squaring operation;

[0055] The third step, in a Q-square. [Delta] 2 San San a Q + square. 2 search time r (δ) of the maximum value of time for α = 0.74 seconds, [alpha] is a mother ECG cycle an accurate estimate.

[0056] estimated fetal ECG period β, the specific steps include:

[0057] The first step, selecting the maximum way Fetal ECG and maternal ECG signals from the multiplexed signal to noise mixed after the pre-mixed signal, the mixed signal that is referred to at time t y (t) . In the y (t) fetal ECG two adjacent R wave interval as the pre-estimated value of fetal ECG cycle. From y (t) roughly estimated fetal ECG R-wave period β〇 = 〇.5 seconds;

[0058] The second step, calculating y (t) autocorrelation function

[0060] where, f (ε) Table no delay FECG left foot mesh number and mother ECG left foot mesh number left foot mesh noise at time [epsilon] Physiological autocorrelation coefficients of the mixed signal is smaller than the maximum way power, ε represents the fetal heart mother ECG electrical signal and the maximum signal to noise ratio electrophysiological way mixed signal delay time relative to the time t, y (η) represents fetal ECG and maternal ECG way maximum signal to noise ratio electrophysiological mixed signal, η represents fetal ECG and maternal ECG signal to noise ratio of the maximum way mixed electrophysiological signal sampling timing, J (·) dn represents a sampling of the integral operation time η, [·] represents 2 squaring operation;

[0061] The third step, in f3Q-0. 15 [epsilon] San San βα + 0. 15 search time f (ε) is the maximum value of the moment β = 0.448 seconds, the heart electrical cycles fetal beta] Exact estimated value.

[0062] Step 5, to obtain optimal ECG vectors.

[0063] steepest descent method of approximation of the optimal vector ECG mother mother's heart for an electric signal vector minimized, thereby obtaining an optimal ECG signal vector e mother.

[0064] according to the following formula, to obtain the optimal vector ECG mother mother ECG vector approximation degree:

[0065] min J (w) = wrR (0) w - log [wTR (a) w] w

[0066] wherein, J (w) represents the degree of approximation to the optimal vector mother mother ECG heart electrical signal vector, w denotes the mother ECG vectors, 1ΐ + ί represents a minimum operation on the mother's ECG vector W , R (O) and R (a) respectively represent the mother and fetus ECG ECG way maximum signal to noise ratio electrophysiological signal delay are mixed and square autocorrelation matrix of [alpha], α represents the mother ECG cycle accurate estimates l〇g [·] represents a logarithmic operation, (*) [tau] represents the transpose operation.

[0067] Optimal mother ECG signal vector e is calculated according to the following recursive steps:

[0068] The first step, the mother ECG initialization vector w (0) = [1,1, ..., 1] τ, initialization x (t) are the delay autocorrelation matrix RlPRa square matrix, and α is zero;

[0069] The second step, according to the following formulas, updates the correlation matrix Rci:

[0070] Ra (k) = λ Ra (k ~ l) + X (k) xT (ka)

[0071] wherein, IT (k) denotes the k-th iteration x (t) a time delay of the autocorrelation matrix, k represents the number of iterations of the mother's heart electrical signal vector, Rci (k_l) denotes the k-th iteration iota the x (t) is a delay autocorrelation matrix, x (k) represents the time when the mother ECG signals and other noise ratio k of the maximum way mixed signal, x (k_ a) represents the time when a K_ mother ECG signals and other mixed signal to noise ratio all the way to the maximum, a forgetting factor [lambda], the present examples take λ = 0.96, (· "denotes a transpose operation;

[0072] The third step, according to the following formulas, updates the correlation matrix R °:

[0073] R0 (k) = λ R0 (k ~ l) + X (k) xT (k)

[0074] wherein, R ° (k) denotes the k-th iteration x (t) is the time delay of the autocorrelation matrix 0, k represents the number of iterations of the mother's heart electrical signal vector, R (kl) is the k-times iota iteration x (t) is the time delay of the autocorrelation matrix 0, x (k) represents the time for mothers ECG signals and other mixed signal to noise ratio all the way to the maximum when k, λ is a forgetting factor, the present examples take λ = 1.96 billion, (· "table does not transpose operation;

[0075] The fourth step is calculated according to the following formula mother ECG, vector iteration:

[0076] w (k + 1) = w (k) - μ [R (0) w (k) - [w (k) tR (a) w (k)] 1R (a) w (k)]

[0077] wherein, w (k + l) represents the k + 1 th iteration mother ECG signal vector, w (k) represents the mother of the ECG vector iteration k, k represents the iteration mother's heart an electrical signal vector number, μ denotes the iteration step the mother's heart long electrical signal vector, R (O) and R (a) respectively represent the mother and fetus ECG ECG way maximum signal to noise ratio electrophysiological mixed signal delay are square and a self-correlation matrix, a represents an accurate estimate of the mother's heart electrical cycles, (•) 1 denotes a transpose operation, (· Γ1 represents inverse operation;

[0078] The fifth step, so that k = k + Ι, repeat the second step to the fifth step until convergence.

[0079] The Newton's method of approximation of the correction fetal ECG electrical signal vector and the optimum vector minimizing the fetal heart, so as to obtain optimal fetal ECG vector f.

[0080] according to the following formula, to obtain fetal ECG vectors with the best approximation of the fetal ECG vectors of:

[0081] min jL (u) = urV (0) u - log [uTV (, ii) u] U

[0082] wherein the approximation degree L (u) Table fetal ECG vectors are not the best fetal heart electrical signal vector, u the vector table is not a fetal electrocardiogram, mUillM expressed on fetal ECG minimize vector u operation, V (0) and V (β) represent fetal ECG and maternal ECG signal to noise ratio of the maximum signal delay along electrophysiological mixed beta] are square and the autocorrelation matrix, β represents fetal ECG an accurate estimate of the period, l〇g [·] represents a logarithmic operation (*) 7 denotes a transpose operation.

[0083] Optimal fetal ECG vector f recursive calculation as follows:

[0084] The first step, fetal ECG initialization vector W (square) = [1,1, ..., 1] τ, initialization y (t) is the delay of the inverse matrix of the autocorrelation matrix Q 0 and y (t ) β delay autocorrelation matrix V e is a unit matrix;

[0085] The second step, according to the following formula, update matrices Ve:

[0086] Vp (η) = λ Vp (η-1) + χ (η) χτ (η-β)

[0087] wherein, Rfi (η) denotes the iteration [eta] y (t) is the time delay of the autocorrelation matrix β, τι electrical signal vector represents the number of iterations of the mother's heart, Re (η -1) represents η -1 iterations y (t) is the time delay of the autocorrelation matrix beta], XU) table is not the maximum time during fetal ECG signals and other noise ratio rI way mixed signal, x (rI-β) is not time table FECG η-β and the time of the maximum signal to noise ratio all the way to the other mixed signal, [lambda] is the forgetting factor, the present examples take λ = 0. 96, (·) 1¾ illustrating a transpose operation;

[0088] The third step, according to the following formula, update matrices Q:

[0090] wherein, QU) represents the first iteration [eta] y (t) is the delay of the inverse matrix of the autocorrelation matrix 0, η represents the number of iterations of the mother's heart electrical vectors, Q (nl) represents the iteration nl y (t) is the delay of the inverse matrix of the autocorrelation matrix 0, χ (η) represents the time for the fetus ECG signals and other noise ratio of the maximum way mixed signal [eta], [lambda] is the forgetting factor [lambda] of the present example taking . = square 96, (·) τ denotes a transpose operation;

[0091] The fourth step, the fetal ECG is calculated according to the following formula vector, iteration:

[0093] wherein, U (n + i) n + represents fetal ECG vector iteration i, uU) represents fetal ECG vector iteration of [eta], [eta] represents the number of iterations fetal heart electrical signal vector, v (o) and ν (β) represent fetal ECG signals and other mixed signal to noise ratio all the way to the maximum delay beta] are square and the autocorrelation matrix, β represents an accurate estimate of fetal heart electrical cycles, Q represents fetal ECG signals and other mixed signal to noise ratio all the way to the maximum delay of the inverse square matrix of the autocorrelation matrix, (· "denotes a transpose operation, (·) represents inverse operation 4, (*) indicates finding _2 inverse square operation;

[0094] The fifth step, so that η = η + ι, repeat the second step to the fifth step until convergence.

[0095] Step 6, extracting ECG.

[0096] The projection method, using the optimum vector ECG mother mother extracted ECG physiological mixed signal from the multiplexer electrically pretreated. According to the formula, mothers get ECG:

[0097] h (t) = eTd (t)

[0098] where, h (t) represents the time t mother extracted ECG, e represents the optimal mother ECG vector, d (t) represents the time t after the mixed signal preprocessing, (·) τ denotes transpose transpose operation.

[0099] The projection method, using the optimum vector physiological fetal ECG signal from the multiplexer mixed electrical preprocessed to extract the fetal ECG. According to the following formula, to obtain fetal ECG:

[0100] ζ (t) = fTd (t)

[0101] wherein, z (t) represents the time t Fetal ECG extraction, f denotes the optimal fetal ECG vector, d (t) represents the pre-mixed signal at time t, (·) τ denotes transpose transpose operation.

[0102] The following figures further described in conjunction with the effects of the present invention.

[0103] 1. The simulation conditions

[0104] System simulation run of the present invention is a Intel (R) Core (TM) i7-2600CPU 65003. 40GHz, 32 bit Windows operating system, using the simulation software MATLAB (R2008a).

[0105] 2. Content and simulation results analysis

[0106] 8 electrical input of FIG. 2 in the drawings physiological mixed signal of the first channel signal is mixed electrophysiological accurate estimate of electrical cycles simulated search mother's heart, the pre-set value of the estimated period a 0.8 sec search period 0.6 <δ <1.0, the autocorrelation coefficients to obtain a first mixed signal electrophysiological way as shown in the drawings is a schematic view 3.

[0107] accurate estimate of 8 electrophysiological mixed signal simulation fetal electrocardiogram cycle search electrophysiological eight inputs of mixing 2 shown in the figures in the signal, the pre-set period estimated value = square β〇 .5 seconds, a search period 0.35S ε <0.45, to obtain a first mixed signal electrophysiological 8 is a schematic view of the autocorrelation coefficients as illustrated in Figure 4.

[0108] 8 electrophysiological four kinds of mixed-signal simulation method of the present invention and the prior art shown in FIG. 2 in input to the drawings extraction mother and fetus ECG, ECG waveform extraction mother comparison is shown in Figure 5 , the ECG waveform of the fetus extracted comparison is shown in Figure 6.

The first electrical path shown in [0109] Referring to FIG 3 mixed physiological signal autocorrelation coefficients schematic, in Figure 3 the ordinate represents the size of the self correlation coefficients, the abscissa represents time. In the time range indicated by the abscissa 〇.6 <δ <1.0 can be seen in FIG label point location of the maximum autocorrelation coefficient, the marked points in FIG abscissa α = 0.74 seconds mother ECG an accurate estimate period.

The first electrical path 8 shown in [0110] Referring to FIG. 4 Physiological mixed signal autocorrelation coefficients schematic in FIG. 4 the ordinate represents the size of the self correlation coefficients, the abscissa represents time. The maximum autocorrelation coefficients obtained in the time range indicated by the abscissa 〇.35 <ε <0.45 was noted in the figures location points, the points labeled in FIG abscissa β = 0.448 seconds Fetal ECG cycle the accurate estimate.

FECG [0111] Referring to Figure 5, four methods of the present invention and the prior art simulated ECG waveform extraction mother FIG comparison, referring to Figure 6, four methods of the present invention and the prior art emulation extracted FIG comparison signal waveform. FIG 5 (a) and 6 (a) is a graph meter, respectively with AKBarros and A.Cichocki in the article "Extraction of specific signals with temporal structure" (Neural Comput., Vol. 13, no.9, pp. 1995 --2003, 2001) singular value decomposition disclosed in the batch method, maternal and fetal ECG waveform extracted. In FIG. 5 (b) and 6 (b) are employed curve X. -L. Li and X. -D. Zhang in the article "Sequential blind extraction adopting second-order statistics" (IEEE Signal Process. Lett., .. vol 14, no 1, pp 58 -. in 61,2007) disclosed in second order statistics based joint diagonalization sequential method of extracting a waveform of maternal and fetal ECG. (C) a graph in FIG 5 (c) and FIG. 6 are employed W. Liu, DP Mandic, and A. Cichocki in the article "Analysis and online realization of the CCA approach for blind source separation" (IEEE Trans. Neural Netw ., 18, no 5, pp 1505 vol -.. 1510, 2007) published in the fall classic CCA guidelines based on adaptive gradient method to extract the waveform maternal and fetal ECG. (D), graph 5 (d) and FIG. 6 are employed A. Cichocki and R. Thawonmas in the article "On-line algorithm for blind signal extraction of arbitrarily distributed but temporally correlated sources using second order statistics" (Neural Process ... Lett, vol 12, no 1, pp 91 -.. 98, the minimum prediction error based on an adaptive extraction quasi 2000) published extracted waveform of maternal and fetal ECG. FIG. 5 (e) and 6 (e) is a waveform of the curves mother and fetus ECG of the present invention is extracted.

Simulation of the extracted ECG waveform mother FIG Comparative [0112] Referring to the drawings the four methods of the present invention and the prior art shown in FIG. 5. FIG 5 (a), (b), (c), (d), (e) the abscissa represents time and the ordinate represents signal amplitude. Can be seen by comparing the waveforms of the other four mother ECG extraction method is not clear enough, and there is a distortion, the method of the present invention, the extracted ECG mother well suppressed fetal ECG and noise.

Four methods of the present invention and prior art [0113] FIG. 6 with reference to the accompanying drawings simulation extracted fetal ECG waveforms can be compared to FIG. FIG 6 (a), (b), (c), (d), (e) the abscissa represents time and the ordinate represents signal amplitude. Can be seen by comparing the waveforms of the other four methods to extract the fetal ECG is not clear enough, the fetal ECG noisy, and the doping mother ECG, the method of the present invention to extract the fetal ECG better suppress maternal ECG signal and noise.

Claims (7)

CLAIMS 1. A method of extracting the mother and fetus ECG second order statistical properties based blind, comprising the steps of: (1) the mixed signal acquired electrophysiological: acquiring maternal abdomen and chest multiple electrophysiological signal synchronized acquisition mixed; (2) pretreatment: (2a) using a finite length digital low pass filter, the mixing of multiple electrophysiological EMG signal interference and noise filtering, to obtain the multiplexed electrical interference and noise filtered EMG physiological mixed signal; (2B ) using finite length digital notch filter, the multi-channel electrical interference and noise filter physiological EMG mixed 50Hz frequency interference signal filtered off, to obtain multiple physiological electrical preprocessed downmix signal; (3) each obtained physiological mixed signal to noise ratio of the electric path: (3a) each channel mixed signal electrophysiological mother pretreatment ECG signal as a fetal ECG as noise, to give the mother and fetus ECG ECG letter noise ratio; (3B) each channel mixed signal electrophysiological pretreated as fetal ECG signal, the ECG signal as noise mother, fetus ECG signal to noise ratio and to obtain the mother's ECG; (4) estimation Period: (4a) selecting mother and fetus ECG ECG way maximum signal to noise ratio electrophysiological mixed signal, in which two adjacent mother ECG R wave interval, as a mother heart electrical cycles pre-estimated value; (4B) using the autocorrelation coefficients formula to calculate the autocorrelation coefficient mother and fetus ECG ECG way maximum signal to noise ratio electrophysiological mixed signal; (4C) a pre-estimate the mother's heart electrical cycles error value minus 0.2 seconds as the search period, the absolute value of the maximum autocorrelation coefficient of the search, the mother takes the maximum value as the center of electrical cycles accurate estimate of the absolute value of the self correlation coefficients; (4D) selected fetal ECG and maternal ECG way maximum signal to noise ratio electrophysiological mixed signal, wherein the fetal ECG two adjacent R wave interval, as the pre-estimate the fetal heart electrical cycles; ( 4e) using autocorrelation coefficient formula, the maximum autocorrelation coefficient of the mixed signal along electrical physiological fetal ECG and maternal ECG signal to noise ratio; (4F) pre fetal heart electrical cycles Error count value plus or minus 0.15 seconds as a search time, the search from the maximum absolute value of the correlation coefficient, and takes the maximum value as the fetal heart electrical cycles accurate estimate of the absolute value of the self correlation coefficients; (5 ) for optimal ECG signal vector: (5a) by steepest descent method, for optimal mother ECG signal vector; (5b) by the modified Newton method for optimal fetal ECG signal vector; (6) extract ECG: (6a) using the projection method, using the optimum vector ECG mother mother extracted ECG physiological mixed signal from the multiplexer electrically pretreated; (6B) using the projection method, the best use of fetal ECG extraction fetal ECG vector physiological mixed signal from the multiplexer electrically pretreated.

2. The method of claim extracting statistical properties of the mother and the fetal ECG blind second order based on 1 wherein said autocorrelation coefficients of the equations in step (4b) as follows:

Wherein, r (S) represents the delay time for the mother and fetus ECG ECG signal to noise ratio of the maximum autocorrelation coefficients of the mixed signal along electrical physiological when S, S represents the mother and fetus ECG ECG signal to noise than the maximum delay time of the mixed signal corresponding electrophysiological way time t, X (t) represents the mother and fetus ECG ECG way maximum signal to noise ratio electrophysiological mixed signal, t represents the mother and fetus ECG ECG the ratio of maximum signal is the mixed signal electrophysiological way sampling time, J (•) dt represent the integral operation of the sampling time t, [•] 2 represents the squaring operation.

3. The method of blind extraction mother and fetus ECG based on second order statistical properties, characterized according to claim 1, step (4E) autocorrelation coefficients according to the following formula:

Wherein, f (e) represents fetal ECG and maternal ECG signal to noise ratio at the maximum delay time e electrophysiological way mixed signal autocorrelation coefficients, e represents the fetal and maternal ECG heart electrical noise channel than the maximum delay time of the mixed signal corresponding electrophysiological way time t, y (n) represents the fetal ECG and maternal ECG way maximum signal to noise ratio electrophysiological mixed signal, n represents fetal ECG and maternal ECG the ratio of maximum signal is the mixed signal electrophysiological way sampling time, J (•) dn represents the integral operation of the sampling time n, [•] 2 represents the squaring operation.

4. Extraction of the mother and the fetal ECG blind statistical characteristics based on second order, characterized in that said claim 1, the step steepest descent method in the step (5a) as follows: First, according to the following formula give the mother the best ECG signal vector and vector mother ECG approximation degree:

Wherein, J (w) represents the degree of approximation to the optimal vector mother mother ECG heart electrical signal vector, w denotes the mother heart

Affinity ECG and fetal ECG signal to noise ratio of the maximum electrophysiological way signal delay are mixed and a square of the autocorrelation matrix, a represents an accurate estimate of the mother's heart electrical cycles, l〇g [•] represents logarithmic operation, (•) T denotes a transpose operation; the second step, according to the following equation, iterative mother ECG vector: w (k + 1) = w (k) -y [R (O) w ( k) - [w (k) tR (a) w (k)] 1R (a) w (k)] where, w (k + L) represents the mother ECG vector k + 1 iteration, w ( k) represents the k th iteration mother ECG signal vector, k represents the number of iterations mother heart electrical signal vector, y denotes the vector iteration step the mother's heart electrical length, R (O) and RU) represent the mother ECG fetal ECG signal to noise ratio and the maximum way electrophysiological signal delay are mixed and a square of the autocorrelation matrix, a represents an accurate estimate of the mother's heart electrical cycles, (*) T denotes a transpose operation, (• r1 represents the inverse operation; a third step iteration until convergence mother ECG, ECG mother optimal vector.

The extraction mother and second order statistical properties of fetal ECG based blind, characterized according to claim 1, step (5b) in the Newton's method of correcting the following steps: a first step according to the following formula, get approximation of the fetal ECG signal vector and optimal electrical signal vector fetal heart:

Wherein, L (u) represents the fetal ECG vector optimal approximation of the fetal heart electrical signal vectors, u represents the fetus

Fetal ECG and maternal ECG signal to noise ratio of the maximum delay of the mixed signal electrophysiological way are square autocorrelation matrix and e, e represents an accurate estimate of electrical cycles fetal heart, i0g [•] represents taking the operands, (•) T denotes a transpose operation; a second step of calculating the fetal ECG vector according to the following formula, iteration:

Wherein, U (n + 1) n + i represents the iteration fetal ECG vectors, U (n) represents the fetal ECG vector iteration n, n is the number of iterations electrical signal vector table is not Fetal, v (o) and v (| 3) respectively, the table does not fetal ECG and maternal ECG signal to noise ratio of the maximum delay of the mixed signal electrophysiological way are square autocorrelation matrix and e, e represents fetal ECG an accurate estimate of the period, (•) T denotes a transpose operation, (•) <showing inverse operation, (•) j represents the square of the negation operation; a third step iteration until convergence fetal ECG, the most preferably fetal ECG vectors.

6. The method of blind extraction mother and fetus ECG based on second order statistical properties, characterized in that said claim 1, the step (6a) in the projection method refers to the following formula, to obtain mother ECG: h (t) = eTd (t) where, h⑴ represents mother ECG extraction time t, e represents the optimal mother ECG vector, d⑴ electrophysiological mixed signal represents time t after pretreatment, (•) T It represents a transpose operation.

7. The method of extracting statistical properties of the mother and the fetal ECG second order blind basis, wherein in said claim 1, the step (6b) in the projection method refers to the following formula, to obtain fetal ECG: z (t) = fTd (t) where, z (t) represents the time t fetal ECG extraction obtained, f • fetal ECG represents the optimal vector, d (t) represents the time t electrically pretreated physiological mixed signal, (•) T represents a transpose operation.