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

Abstract

The invention discloses a blind extraction method of maternal and fetal electrocardiogram based on second-order statistical characteristics, which solves the problem that mixed signals with small signal-to-noise ratio cannot be extracted well and the extraction process works in an offline mode. The steps of the present invention are: (1) obtaining the electrophysiological mixed signal; (2) preprocessing; (3) selecting the one-way mixed signal with the largest signal-to-noise ratio; (4) estimating the period; (5) obtaining the optimal electrocardiogram Signal vector; (6) Extract ECG signal. Compared with the method for extracting the electrocardiogram signals of mothers and fetuses in the prior art, the present invention not only ensures the accuracy of the extracted signals, but also has the advantages of real-time online extraction and high extraction efficiency. The maternal and fetal ECG signals were extracted from the maternal ECG signals.

Description

Blind Extraction Method of Maternal and Fetal ECG Based on Second-order Statistical Characteristics

technical field

The present invention relates to the technical field of signal processing, and further relates to a method for blindly extracting maternal and fetal electrocardiographic signals based on second-order statistical characteristics in the technical field of biomedical signal processing. The present invention can be used for extracting the mother's and fetus's electrocardiogram signals from the mother's electrocardiogram signals collected by the electrocardiogram monitoring machine.

Background technique

The human electrocardiogram (ECG) is an objective indicator reflecting the physiological activities of the human body. The electrical signals measured from the body surface of pregnant women generally include maternal electrocardiogram (MECG), fetal electrocardiogram (FECG), other electrical signals of the human body (such as electromyographic signal EMG, etc.) and a mixed signal of observed background noise. Among the obtained electric signals measured on the body surface of pregnant women, since the strength of the mother's electrocardiogram is generally relatively high, conventional filtering techniques can be used to extract the mother's electrocardiogram. The fetal ECG signal is quite weak, its strength is less than one-tenth of the mother's ECG signal, and the frequency spectrum of the fetal ECG signal and the mother's ECG signal overlap each other, so it is difficult to extract a clear and stable fetal ECG signal by general filtering technology. electric signal.

Guangdong University of Technology has a patented technology "A Method for Blind Extraction of Fetal ECG Based on Time-Frequency Transformation" (Patent Application No.: 201110144487.7, Authorized Publication No. CN 102160787B) which discloses a method for blind extraction of fetal ECG based on Time-Frequency Transformation . This method acquires a mixed signal containing mother and fetal electrophysiological signals, selects a relatively sparse period of time in the mixed signal, and transforms the obtained relatively sparse period of time into the time-frequency domain by using a fuzzy function, and then uses The generalized Rayleigh entropy constructs the contrast function to extract the fetal ECG signal. This patented technology utilizes the characteristics of relatively sparse time domain of maternal and fetal ECG mixed signals to solve the problem that the frequency spectrum of maternal and fetal ECG signals overlaps and is difficult to separate. Time-varying properties of fetal ECG signals.

The patent application "A method for adaptive blind extraction of fetal ECG signals based on the maximization of generalized eigenvalues" (patent application number: 201310729736.8, publication number CN 103627796A) proposed by Shanghai Maritime University discloses an adaptive blind extraction of fetal ECG signals method. This method acquires a mixed signal including maternal and fetal electrophysiological signals, and based on an adaptive real-time algorithm, calculates the eigenvalues and eigenvectors of different delay covariance matrices of the mixed signal within a set period range, and selects the eigenvalue corresponding to the largest eigenvalue. The eigenvectors are blindly separated vectors to extract fetal ECG signals. Although this method uses the second-order statistical characteristics of the signal to simplify the computational complexity and improve the efficiency of blind separation of fetal ECG signals, there are still shortcomings that the method is not accurate enough for the period estimation of ECG signals, and due to over-reliance on Due to the diagonalization of different delay covariance matrices of the collected maternal and fetal ECG mixed signals, the extraction effect is not good for the collected mixed signals with small signal-to-noise ratio.

Contents of the invention

The purpose of the present invention is to overcome the shortcomings of the above-mentioned existing methods, and proposes a method for blind extraction of electrocardiographic signals based on second-order statistical characteristics, which realizes adaptive real-time extraction of maternal and fetal electrocardiographic signals.

The basic idea of realizing the present invention is to firstly preprocess the obtained multi-channel electrophysiological mixed signals, and then select a mixed signal with the largest signal-to-noise ratio from the obtained multi-channel electrophysiological mixed signals, and then estimate the period of the mixed signal , extract the optimal separation vector from the mixed signal, and finally use the optimal separation vector to extract the ECG signal from the preprocessed mixed signal.

Concrete steps of the present invention are as follows:

(1) Obtain electrophysiological mixed signals:

Obtain synchronously acquired multi-channel electrophysiological mixed signals of the mother's chest and abdomen;

(2) Pretreatment:

(2a) Using a finite-length digital low-pass filter to filter out myoelectric interference and noise on the multi-channel electrophysiological mixed signal, and obtain the multi-channel electrophysiological mixed signal after filtering out the myoelectric interference and noise;

(2b) Using a finite-length digital notch filter, perform 50 Hz power frequency interference filtering on the multi-channel electrophysiological mixed signal after filtering the myoelectric interference and noise, and obtain the pre-processed multi-channel electrophysiological mixed signal;

(3) Obtain the signal-to-noise ratio of each electrophysiological mixed signal:

(3a) taking the mother's electrocardiogram in each electrophysiological mixed signal after preprocessing as a signal, and the fetal electrocardiogram as noise to obtain the signal-to-noise ratio of the mother's electrocardiogram and the fetal electrocardiogram;

(3b) taking the fetal ECG signal in each electrophysiological mixed signal after preprocessing as a signal, and the mother's ECG signal as noise to obtain the signal-to-noise ratio of the fetal ECG signal and the mother's ECG signal;

(4) Estimation cycle:

(4a) Select an electrophysiological mixed signal with the largest signal-to-noise ratio between the maternal ECG signal and the fetal ECG signal, and use the time interval between two adjacent R waves of the maternal ECG signal as a pre-estimation of the period of the maternal ECG signal value;

(4b) using the autocorrelation coefficient formula to calculate the autocorrelation coefficient of the electrophysiological mixed signal with the largest signal-to-noise ratio between the mother's electrocardiogram and the fetus' electrophysiological signal;

(4c) The estimated value of the mother's ECG signal period fluctuates up and down by 0.2 seconds as the search time period, searches for the maximum value of the absolute value of the autocorrelation coefficient, and takes the moment when the absolute value of the autocorrelation coefficient takes the maximum value as the mother's ECG An accurate estimate of the period of the signal;

(4d) Select an electrophysiological mixed signal with the largest signal-to-noise ratio between the fetal ECG signal and the maternal ECG signal, and use the time interval between two adjacent R waves of the fetal ECG signal as a pre-estimation of the fetal ECG signal cycle value;

(4e) using the autocorrelation coefficient formula to calculate the autocorrelation coefficient of the electrophysiological mixed signal with the largest signal-to-noise ratio between the fetal ECG signal and the mother's ECG signal;

(4f) The up-and-down error of the pre-estimated value of the fetal ECG signal cycle is 0.15 seconds as the search time period, and the maximum value of the absolute value of the autocorrelation coefficient is searched for, and the moment when the absolute value of the autocorrelation coefficient is maximum is taken as the fetal ECG An accurate estimate of the period of the signal;

(5) Obtain the optimal ECG signal vector:

(5a) adopting the steepest descent method to obtain the optimal mother's ECG signal vector;

(5b) Using the modified Newton method to obtain the optimal fetal ECG signal vector;

(6) Extract the ECG signal:

(6a) Using the projection method, using the optimal mother's ECG signal vector to extract the mother's ECG signal from the preprocessed multi-channel electrophysiological mixed signal;

(6b) Using the projection method, using the optimal fetal ECG signal vector to extract the fetal ECG signal from the preprocessed multi-channel electrophysiological mixed signal.

Compared with the prior art, the present invention has the following advantages:

First, since the present invention preprocesses the obtained multi-channel electrophysiological mixed signals, and then obtains the optimal vectors of the mother and fetal ECG signals from the preprocessed electrophysiological mixed signals, it overcomes the problem of the prior art. The electrophysiological mixed signal with a small noise ratio cannot better extract the electrocardiographic signals of the mother and the fetus, so that the present invention does not need the additional condition of the signal-to-noise ratio of the electrophysiological mixed signal.

Second, because the present invention obtains the optimal maternal and fetal ECG signal vectors through the second-order statistical characteristics of the electrophysiological mixed signal, it overcomes the shortcomings of the prior art in offline mode and realizes the real-time monitoring of the maternal and fetal ECG signals. The online extraction makes the extraction of the maternal and fetal electrocardiographic signals more efficient, and can better reflect the time-varying characteristics of the maternal and fetal electrocardiographic signals.

Third, because the present invention pre-estimates the period of the mother and fetus' electrocardiogram in the electrophysiological mixed signal with the largest signal-to-noise ratio, and then sets the error for accurate estimation, which overcomes the prior art's estimation of the mother's and fetus' electrocardiogram. Insufficiency of inaccurate cycle estimation makes the present invention more effective in extracting mother and fetus electrocardiogram signals.

Description of drawings

Fig. 1 is a flowchart of the present invention;

Fig. 2 is 8 road electrophysiological mixed signal waveform diagrams that the present invention inputs;

Fig. 3 is a schematic diagram of the autocorrelation coefficient of the No. 1 signal obtained by adopting the present invention;

Fig. 4 is the schematic diagram of the autocorrelation coefficient of the 8th road signal obtained by adopting the present invention;

Fig. 5 is to adopt four kinds of methods of the present invention and prior art to extract maternal electrocardiogram waveform contrast figure;

Fig. 6 is a comparison diagram of fetal ECG signal waveforms extracted by four methods of the present invention and the prior art.

detailed description

The present invention will be further described below in conjunction with accompanying drawing:

With reference to Fig. 1, concrete implementation steps of the present invention are as follows:

Step 1, input electrophysiological mixed signal.

In the embodiment of the present invention, 8 electrophysiological mixed signals are input, and the electrophysiological mixed signals are synchronously collected from 5 electrophysiological mixed signals of the mother's chest and 3 electrophysiological mixed signals of the mother's abdomen. With reference to the 8-way electrophysiological mixed signal waveform diagram shown in Figure 2, the abscissa in Fig. 2 represents the time, and the ordinate represents the signal amplitude, Fig. 2 (a), (b), (c), (d) , (e) are 5-channel electrophysiological mixed signals collected from the mother's chest, and Fig. 2(f), (g), (h) are 3-channel electrophysiological mixed signals collected from the mother's abdomen. It can be seen from Figure 2 that the 5-channel electrophysiological mixed signals from the chest are mainly maternal ECG signals, while the 3-channel electrophysiological mixed signals from the abdomen contain weak fetal ECG signals.

Step 2, preprocessing.

The first step is to use an 8th-order finite-length digital low-pass filter based on the Hamming window to filter out myoelectric interference and noise on the obtained multi-channel mixed signal, and obtain a multi-channel mixed signal after filtering out myoelectric interference and noise. Signal;

The second step is to use a 40-order finite-length digital notch filter based on the Hamming window to filter out 50 Hz power frequency interference on the multi-channel mixed signal after the EMG and noise filtering, and obtain the pre-processed multi-channel mixed signal .

In step 3, the signal-to-noise ratio of each electrophysiological mixed signal is obtained.

In the first step, the mother's ECG signal in each electrophysiological mixed signal after preprocessing is used as a signal, and the fetal ECG signal is used as noise to obtain the signal-to-noise ratio of the mother's ECG signal and the fetal ECG signal;

In the second step, the fetal ECG signal in each electrophysiological mixed signal after preprocessing is used as the signal, and the maternal ECG signal is used as noise to obtain the signal-to-noise ratio of the fetal ECG signal and the maternal ECG signal.

Step 4, estimate the period.

Estimate the period α of the mother's ECG signal, and the specific steps include:

The first step is to select the mixed signal with the largest signal-to-noise ratio between the mother’s ECG signal and the fetal ECG signal from the preprocessed multi-channel mixed signal, and record the mixed signal at time t as x(t), and set x The time interval between two adjacent R waves of the mother's ECG signal in (t) is used as the pre-estimated value of the period of the mother's ECG signal. Obtain the pre-estimated value α ₀ =0.8 second of mother's electrocardiogram cycle from x(t);

In the second step, the formula for calculating the autocorrelation coefficient of x(t) is as follows:

$\mathrm{<span\; class="notranslate">\; r</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&delta;</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{{<span\; class="notranslate">\; \&Integral;</span>}_{\mathrm{<span\; class="notranslate">\; t</span>}}\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; \&delta;</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; dt</span>}}{{<span\; class="notranslate">\; \&Integral;</span>}_{\mathrm{<span\; class="notranslate">\; t</span>}}{\mathrm{<span\; class="notranslate">\; x</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; dt</span>}}$

Among them, r(δ) represents the autocorrelation coefficient of the electrophysiological mixed signal with the largest signal-to-noise ratio of the maternal ECG signal and fetal ECG signal when the delay time is δ, and δ represents the signal-to-noise ratio of the maternal ECG signal and fetal ECG signal The delay time of the electrophysiological mixed signal with the largest ratio relative to time t, x(t) represents the electrophysiological mixed signal with the largest signal-to-noise ratio between the mother’s ECG signal and the fetal ECG signal, and t represents the mother’s ECG signal and the fetal ECG signal The sampling time of one electrophysiological mixed signal with the largest signal-to-noise ratio, ∫(·)dt represents the integration operation on the sampling time t, and [·] ² represents the square operation;

The third step is to search for the maximum value of r(δ) at α=0.74 seconds within the time range of α ₀ −0.2≤δ≤α ₀ +0.2, then α is an accurate estimated value of the period of the mother's ECG signal.

Estimating the period β of the fetal ECG signal, the specific steps include:

The first step is to select a mixed signal with the largest signal-to-noise ratio between the fetal ECG signal and the maternal ECG signal from the preprocessed multiple mixed signals, and record the mixed signal at time t as y(t). The time interval between two adjacent R waves of the fetal ECG signal in y(t) is used as the pre-estimated value of the fetal ECG signal period. Roughly estimating from y(t) that the period of the fetal electrocardiographic signal R wave is β ₀ =0.5 second;

The second step is to calculate the autocorrelation function of y(t)

$\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&epsiv;</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{{<span\; class="notranslate">\; \&Integral;</span>}_{\mathrm{<span\; class="notranslate">\; t</span>}}\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; \&epsiv;</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; dt</span>}}{{<span\; class="notranslate">\; \&Integral;</span>}_{\mathrm{<span\; class="notranslate">\; t</span>}}{\mathrm{<span\; class="notranslate">\; the\; y</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; dt</span>}}$

Among them, f(ε) represents the autocorrelation coefficient of the electrophysiological mixed signal with the largest signal-to-noise ratio between the fetal ECG signal and the mother’s ECG signal when the delay time is ε, and ε represents the signal-to-noise ratio of the fetal ECG signal and the mother’s ECG signal The delay time of the electrophysiological mixed signal with the largest ratio relative to time t, y(n) represents the electrophysiological mixed signal with the largest signal-to-noise ratio between the fetal ECG signal and the mother’s ECG signal, and n represents the fetal ECG signal and the mother’s ECG signal The sampling time of one electrophysiological mixed signal with the largest signal-to-noise ratio, ∫(·)dn represents the integral operation of sampling time n, and [·] ² represents the square operation;

The third step is to search for the maximum value of f(ε) at the moment of β=0.448 seconds within the time range of β ₀ −0.15≤ε≤β ₀ +0.15, and then β is an accurate estimated value of the period of the fetal ECG signal.

Step 5, obtaining the optimal ECG signal vector.

The method of steepest descent is used to minimize the approximation degree between the mother's ECG signal vector and the optimal mother's ECG signal vector, so as to obtain the optimal vector e of the mother's ECG signal.

According to the following formula, the approximation degree between the mother's ECG signal vector and the optimal mother's ECG signal vector is obtained:

$\underset{\mathrm{<span\; class="notranslate">\; w</span>}}{\mathrm{<span\; class="notranslate">\; min</span>}}\mathrm{<span\; class="notranslate">\; J</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; w</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; w</span>}}^{\mathrm{<span\; class="notranslate">\; T</span>}}\mathrm{<span\; class="notranslate">\; R</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 0</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; w</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; log</span>}<span\; class="notranslate">\; [</span>{\mathrm{<span\; class="notranslate">\; w</span>}}^{\mathrm{<span\; class="notranslate">\; T</span>}}\mathrm{<span\; class="notranslate">\; R</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&alpha;</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; w</span>}<span\; class="notranslate">\; ]</span>$

Among them, J(w) represents the approximation degree between the mother's ECG signal vector and the optimal mother's ECG signal vector, w represents the mother's ECG signal vector, Represents the minimization operation on the mother’s ECG signal vector w, R(0) and R(α) respectively represent the time delay of the electrophysiological mixed signal with the largest signal-to-noise ratio of the mother’s ECG signal and the fetal ECG signal, respectively 0 and α The autocorrelation matrix of , α represents the accurate estimate of the period of the mother's ECG signal, log[·] represents the logarithmic operation, (·) ^T represents the transpose operation.

The optimal maternal ECG signal vector e is recursively calculated as follows:

The first step is to initialize the mother’s ECG signal vector w(0)=[1,1,…,1] ^T , and initialize the autocorrelation matrix R ⁰ and R ^α with the time delay of x(t) being 0 and α respectively as zero matrix ;

In the second step, update the correlation matrix R ^α according to the following formula:

R ^α (k)=λR ^α (k-1)+x(k)x ^T (k-α)

Among them, R ^α (k) represents the autocorrelation matrix with x(t) delay of α in the kth iteration, k represents the number of iterations of the mother’s ECG signal vector, and R ^α (k-1) represents the k-1th time The iterative x(t) time delay is an autocorrelation matrix of α, x(k) represents the mother’s ECG and other mixed signals with the largest signal-to-noise ratio at time k, and x(k-α) represents the time is k Mother's electrocardiogram at the time of -α and other mixed signals with the largest signal-to-noise ratio, λ is the forgetting factor, and this example takes λ=0.96, ( ) ^T represents the transposition operation;

The third step is to update the correlation matrix R ⁰ according to the following formula:

R ⁰ (k)=λR ⁰ (k-1)+x(k)x ^T (k)

Among them, R ⁰ (k) represents the autocorrelation matrix with x(t) delay of 0 for the kth iteration, k represents the number of iterations of the mother’s ECG signal vector, and R(k-1) represents the k-1th iteration x(t) is an autocorrelation matrix with a time delay of 0, x(k) represents the mother’s ECG and other mixed signals with the largest signal-to-noise ratio at time k, and λ is the forgetting factor. In this example, λ=0.96 , (·) ^T represents the transpose operation;

The fourth step is to iteratively calculate the mother's ECG signal vector according to the following formula:

w(k+1)=w(k)-μ[R(0)w(k)-[w(k) ^T R(α)w(k)] ^-1 R(α)w(k)]

Among them, w(k+1) represents the mother’s ECG signal vector of the k+1 iteration, w(k) represents the mother’s ECG signal vector of the k-th iteration, k represents the number of iterations of the mother’s ECG signal vector, μ Represents the iterative step size of the mother's ECG signal vector, R(0) and R(α) respectively represent the self-time delay of the electrophysiological mixed signal with the largest signal-to-noise ratio of the mother's ECG signal and fetal ECG signal, respectively. Correlation matrix, α represents the accurate estimate of the period of the mother's ECG signal, (·) ^T represents the transpose operation, (·) ^-1 represents the inverse operation;

In the fifth step, set k=k+1, and repeat the second to fifth steps until convergence.

The modified Newton method is used to minimize the degree of approximation between the fetal ECG signal vector and the optimal fetal ECG signal vector, so as to obtain the optimal fetal ECG signal vector f.

According to the following formula, the approximation degree between the fetal ECG signal vector and the optimal fetal ECG signal vector is obtained:

$\underset{\mathrm{<span\; class="notranslate">\; u</span>}}{\mathrm{<span\; class="notranslate">\; min</span>}}\mathrm{<span\; class="notranslate">\; L</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; u</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; u</span>}}^{\mathrm{<span\; class="notranslate">\; T</span>}}\mathrm{<span\; class="notranslate">\; V</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 0</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; u</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; log</span>}<span\; class="notranslate">\; [</span>{\mathrm{<span\; class="notranslate">\; u</span>}}^{\mathrm{<span\; class="notranslate">\; T</span>}}\mathrm{<span\; class="notranslate">\; V</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&beta;</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; u</span>}<span\; class="notranslate">\; ]</span>$

Among them, L(u) represents the approximation degree between the fetal ECG signal vector and the optimal fetal ECG signal vector, u represents the fetal ECG signal vector, Represents the minimization operation on the fetal ECG signal vector u, V(0) and V(β) respectively represent the time delay of the electrophysiological mixed signal with the largest signal-to-noise ratio of the fetal ECG signal and the mother’s ECG signal, respectively 0 and β The autocorrelation matrix of , β represents the precise estimation value of the fetal ECG signal period, log[·] represents the logarithmic operation, (·) ^T represents the transpose operation.

The optimal fetal ECG signal vector f is recursively calculated as follows:

The first step is to initialize the fetal ECG signal vector w(0)=[1,1,…,1] ^T , and initialize the inverse matrix Q of the autocorrelation matrix whose y(t) time delay is 0 and the y(t) time delay The autocorrelation matrix V of ^β is the identity matrix;

In the second step, the matrix V ^β is updated according to the following formula:

V ^β (η)=λV ^β (η-1)+x(η)x ^T (η-β)

Wherein, R ^β (η) represents the autocorrelation matrix whose y(t) time delay of the nth iteration is β, and η represents the number of iterations of the mother's electrocardiographic signal vector, and R ^β (η-1) represents the η-1 time The iterative y(t) time delay is the autocorrelation matrix of β, x(η) represents the fetal ECG and other signals with the largest signal-to-noise ratio when the time is η, and x(η-β) represents the time as η Fetal ECG and other signal-to-noise ratio mixed signals with the largest signal-to-noise ratio at the time of -β, λ is the forgetting factor, this example takes λ=0.96, ( ) ^T represents the transposition operation;

The third step is to update the matrix Q according to the following formula:

$\mathrm{<span\; class="notranslate">\; Q</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&eta;</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; \&lambda;</span>}}<span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; Q</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&eta;</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\frac{\mathrm{<span\; class="notranslate">\; Q</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&eta;</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&eta;</span>}<span\; class="notranslate">\; )</span>{\mathrm{<span\; class="notranslate">\; x</span>}}^{\mathrm{<span\; class="notranslate">\; T</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&eta;</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; Q</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&eta;</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>}{\mathrm{<span\; class="notranslate">\; \&lambda;</span>}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; x</span>}}^{\mathrm{<span\; class="notranslate">\; T</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&eta;</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; Q</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&eta;</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&eta;</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; ]</span>$

Wherein, Q(n) represents the inverse matrix of the autocorrelation matrix whose y(t) time delay of the nth iteration is 0, and n represents the number of iterations of the mother's electrocardiographic signal vector, and Q(n-1) represents the nth-1 The y(t) time delay of the second iteration is the inverse matrix of the autocorrelation matrix of 0, x(η) represents the fetal ECG and other signal-to-noise ratios when the time is η The mixed signal of the maximum, λ is the forgetting factor, this Example gets λ=0.96, ( ) ^T represents transposition operation;

The fourth step is to iteratively calculate the fetal ECG signal vector according to the following formula:

$\mathrm{<span\; class="notranslate">\; u</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&eta;</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{<span\; class="notranslate">\; [</span>{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; u</span>}{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&eta;</span>}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; T</span>}}\mathrm{<span\; class="notranslate">\; V</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&beta;</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; u</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&eta;</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; )</span>}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 2</span>}{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; u</span>}{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&eta;</span>}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; T</span>}}\mathrm{<span\; class="notranslate">\; V</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 0</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; u</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&eta;</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; )</span>}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}}\mathrm{<span\; class="notranslate">\; u</span>}{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&eta;</span>}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; T</span>}}\mathrm{<span\; class="notranslate">\; V</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&beta;</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; u</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&eta;</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ]</span>{\mathrm{<span\; class="notranslate">\; Q</span>}}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}\mathrm{<span\; class="notranslate">\; V</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&beta;</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; u</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&eta;</span>}<span\; class="notranslate">\; )</span>}{\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 2</span>}{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; u</span>}{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&eta;</span>}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; T</span>}}\mathrm{<span\; class="notranslate">\; V</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 0</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; u</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&eta;</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; )</span>}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}}\mathrm{<span\; class="notranslate">\; u</span>}{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&eta;</span>}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; T</span>}}\mathrm{<span\; class="notranslate">\; V</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&beta;</span>}<span\; class="notranslate">\; )</span>{\mathrm{<span\; class="notranslate">\; Q</span>}}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}\mathrm{<span\; class="notranslate">\; V</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&beta;</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; u</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&eta;</span>}<span\; class="notranslate">\; )</span>}$

Wherein, u (n+1) represents the fetal electrocardiographic signal vector of the n+1 iteration, u (n) represents the fetal electrocardiographic signal vector of the n iteration, and n represents the number of iterations of the fetal electrocardiographic signal vector, V (0) and V(β) represent the autocorrelation matrix of the mixed signal with the maximum signal-to-noise ratio of fetal ECG and other signals, respectively, with 0 and β autocorrelation matrix. Fetal ECG and other signal-to-noise ratios are the inverse matrix of the autocorrelation matrix with a mixed signal delay of 0, (·) ^T represents the transpose operation, (·) ^-1 represents the inverse operation, (·) ^-2 Indicates the negative quadratic operation;

In the fifth step, set η=η+1, and repeat the second to fifth steps until convergence.

Step 6, extracting ECG signals.

Using the projection method, the mother's ECG signal is extracted from the preprocessed multi-channel electrophysiological mixed signal by using the optimal maternal ECG signal vector. According to the following formula, the mother's ECG signal is obtained:

h(t)=e ^T d(t)

Among them, h(t) represents the maternal ECG signal extracted at time t, e represents the optimal maternal ECG signal vector, d(t) represents the mixed signal after preprocessing at time t, and (·) ^T represents the transposition operation.

Using the projection method, the optimal fetal ECG signal vector is used to extract the fetal ECG signal from the preprocessed multi-channel electrophysiological mixed signal. According to the following formula, the fetal ECG signal is obtained:

z(t)=f ^T d(t)

Among them, z(t) represents the fetal ECG signal extracted at time t, f represents the optimal fetal ECG signal vector, d(t) represents the mixed signal after preprocessing at time t, and (·) ^T represents the transposition operation.

The effects of the present invention will be further described below in conjunction with the accompanying drawings.

1. Simulation conditions

The emulation operation system of the present invention is Intel (R) Core (TM) i7-2600CPU 6503.40GHz, 32 Windows operating systems, emulation software adopts MATLAB (R2008a).

2. Simulation content and result analysis

Carry out simulation search for the accurate estimated value of the mother's electrocardiogram cycle in the 8-way electrophysiological mixed signal of input shown in accompanying drawing 2, set the pre-estimated cycle value α ₀ =0.8 second, search In the time period 0.6≤δ≤1.0, the schematic diagram of the autocorrelation coefficient of the first electrophysiological mixed signal is shown in Figure 3.

The 8th path electrophysiological mixed signal of the input shown in accompanying drawing 2 is simulated and searched for an accurate estimated value of the fetal electrophysiological signal cycle, and the pre-estimated cycle value β ₀ =0.5 second is set, and the search The time period 0.35≤ε≤0.45, the schematic diagram of the autocorrelation coefficient of the eighth channel of electrophysiological mixed signal is shown in Fig. 4 .

Adopt four kinds of methods of the present invention and prior art to the 8 road electrophysiological mixed signal simulations of input shown in accompanying drawing 2 and extract mother and fetal electrocardiogram signal, extract mother electrocardiogram signal waveform contrast figure as shown in Figure 5, The comparison chart of extracted fetal ECG signal waveform is shown in Figure 6.

Referring to the schematic diagram of the autocorrelation coefficient of the first electrophysiological mixed signal shown in FIG. 3 , the ordinate in FIG. 3 represents the magnitude of the autocorrelation coefficient, and the abscissa represents time. In the time range 0.6≤δ≤1.0 represented by the abscissa, it can be seen that the maximum value of the autocorrelation coefficient is the location of the marked point in the figure, and the abscissa of the marked point in the figure is α=0.74 seconds is the period of the mother’s ECG signal exact estimate.

Referring to the schematic diagram of the autocorrelation coefficient of the eighth electrophysiological mixed signal shown in FIG. 4 , the ordinate in FIG. 4 represents the magnitude of the autocorrelation coefficient, and the abscissa represents time. In the time range 0.35≤ε≤0.45 represented by the abscissa, the maximum value of the autocorrelation coefficient can be obtained as the location of the marked point in the figure, and the abscissa of the marked point in the figure β=0.448 seconds is the exact period of the fetal ECG signal cycle. estimated value.

With reference to accompanying drawing 5, the mother's electrocardiogram signal waveform contrast chart that the present invention and four kinds of methods of prior art emulation extracts, with reference to accompanying drawing 6, the present invention and prior art four kinds of method emulation extraction fetal electrocardiogram waveform comparisons picture. The curves in Figure 5(a) and Figure 6(a) are respectively adopted by A.K.Barros and A.Cichocki in the article "Extraction of specific signals with temporal structure" (Neural Comput., vol.13, no.9, pp.1995 –2003, 2001) disclosed the singular value decomposition batch method to extract the waveforms of maternal and fetal ECG signals. The curves in Figure 5(b) and Figure 6(b) are respectively adopted by X.-L.Li and X.-D.Zhang in the article "Sequential blind extraction adopting second-order statistics" (IEEE Signal Process. Lett., The sequential joint diagonalization method based on second-order statistics disclosed in vol.14, no.1, pp.58–61, 2007) extracts the waveforms of maternal and fetal ECG signals. The curves in Figure 5(c) and Figure 6(c) are respectively adopted by W.Liu, D.P.Mandic, and A.Cichocki in the article "Analysis and online realization of the CCA approach for blind source separation" (IEEE Trans.Neural Netw. , vol.18, no.5, pp.1505–1510, 2007) published in the adaptive gradient descent method based on the classic CCA criterion, the waveforms of the extracted maternal and fetal ECG signals. The curves in Figure 5(d) and Figure 6(d) are respectively adopted by A.Cichockiand R.Thawonmas in the article "On-line algorithm for blind signal extraction of arbitrarily distributed but temporarily correlated sources using second order statistics" (NeuralProcess.Lett. , vol.12, no.1, pp.91–98, 2000) published in the adaptive extraction method based on the minimum prediction error criterion, the waveforms of the extracted maternal and fetal ECG signals. The curves in Fig. 5(e) and Fig. 6(e) are respectively the waveforms of the maternal and fetal electrocardiographic signals extracted by the present invention.

Referring to the comparison diagram of the waveforms of the mother's electrocardiogram extracted by the simulation of the present invention and the four methods of the prior art shown in accompanying drawing 5 . The abscissa in Fig. 5(a), (b), (c), (d), and (e) represents time, and the ordinate represents signal amplitude. It can be seen from the comparison that the waveforms of the mother's electrocardiogram extracted by the other four methods are not clear enough and have distortion, and the mother's electrocardiogram extracted by the method of the present invention suppresses the fetal electrocardiogram and noise well.

Refer to the comparison diagram of fetal electrocardiogram signal waveforms extracted by simulation and extraction by four methods of the present invention and the prior art shown in accompanying drawing 6 . The abscissa in Fig. 6 (a), (b), (c), (d), and (e) represents time, and the ordinate represents signal amplitude. It can be seen by comparison that the waveforms of the fetal electrocardiographic signals extracted by the other four methods are not clear enough, the fetal electrocardiographic signal is noisy, and is doped with the mother's electrocardiographic signal. The fetal electrocardiographic signal extracted by the method of the present invention is better suppressed. maternal ECG signal and noise.

Claims (7)

1., based on mother and the Fetal ECG Blind extracting method of second-order statistics, comprise the steps:

(1) electro physiology mixed signal is obtained:

Obtain parent chest and the abdominal part Multi-path electricity physiology mixed signal of synchronous acquisition;

(2) pretreatment:

(2a) use limited long wave digital lowpass filter, myoelectricity interference and noise filtering are carried out to Multi-path electricity physiology mixed signal, obtain the Multi-path electricity physiology mixed signal after the interference of filtering myoelectricity and noise;

(2b) use limited long digital trap, the filtering of 50Hz Hz noise is carried out to the Multi-path electricity physiology mixed signal after myoelectricity interference and noise filtering, obtains pretreated Multi-path electricity physiology mixed signal;

(3) signal to noise ratio of every road electro physiology mixed signal is obtained:

(3a) using in the electro physiology mixed signal of pretreatment Hou Mei road, mother's electrocardiosignal is as signal, and Fetal ECG signal, as noise, obtains the signal to noise ratio of mother's electrocardiosignal and Fetal ECG signal;

(3b) using in the electro physiology mixed signal of pretreatment Hou Mei road, Fetal ECG signal is as signal, and mother's electrocardiosignal, as noise, obtains the signal to noise ratio of Fetal ECG signal and mother's electrocardiosignal;

(4) cycle estimator:

(4a) select mother's electrocardiosignal and the maximum road electro physiology mixed signal of Fetal ECG Signal-to-Noise, by the interval of wherein adjacent two R ripples of mother's electrocardiosignal, estimate evaluation as mother's electrocardiosignal cycle;

(4b) adopt autocorrelation coefficient formula, calculate the autocorrelation coefficient of mother's electrocardiosignal and the maximum road electro physiology mixed signal of Fetal ECG Signal-to-Noise;

(4c) evaluation of estimating in mother's electrocardiosignal cycle is fluctuated error 0.2 second as section search time, the maximum of the absolute value of search autocorrelation coefficient, gets the fine estimation of moment as mother's electrocardiosignal cycle of maximum using the absolute value of autocorrelation coefficient;

(4d) select Fetal ECG signal and the maximum road electro physiology mixed signal of mother's electrocardiosignal signal to noise ratio, by the interval of wherein adjacent two R ripples of Fetal ECG signal, estimate evaluation as the Fetal ECG signal period;

(4e) adopt autocorrelation coefficient formula, calculate the autocorrelation coefficient of Fetal ECG signal and the maximum road electro physiology mixed signal of mother's electrocardiosignal signal to noise ratio;

(4f) evaluation of estimating of Fetal ECG signal period is fluctuated error 0.15 second as section search time, the maximum of the absolute value of search autocorrelation coefficient, gets the fine estimation of moment as the Fetal ECG signal period of maximum using the absolute value of autocorrelation coefficient;

(5) optimum electrocardiosignal vector is obtained:

(5a) adopt steepest descending method, obtain optimum mother's electrocardiosignal vector;

(5b) adopt correction Newton method, obtain optimum Fetal ECG signal vector;

(6) electrocardiosignal is extracted:

(6a) adopt projecting method, use optimum mother's electrocardiosignal vector to extract mother's electrocardiosignal from pretreated Multi-path electricity physiology mixed signal;

(6b) adopt projecting method, use optimum Fetal ECG signal vector to extract Fetal ECG signal from pretreated Multi-path electricity physiology mixed signal.

2. mother based on second-order statistics according to claim 1 and Fetal ECG Blind extracting method, it is characterized in that, the autocorrelation coefficient formula described in step (4b) is as follows:

$r\left(\mathrm{\&delta;}\right)=\frac{{\&Integral;}_{t}x\left(t\right)x(t-\mathrm{\&delta;})\mathrm{dt}}{{\&Integral;}_t{x}^2\left(t\right)\mathrm{dt}}$

Wherein, the autocorrelation coefficient of mother's electrocardiosignal when r (δ) represents that time delay is δ and the maximum road electro physiology mixed signal of Fetal ECG Signal-to-Noise, δ represents the time delay of the relative t of a road electro physiology mixed signal that mother's electrocardiosignal and Fetal ECG Signal-to-Noise are maximum, x (t) represents mother's electrocardiosignal and the maximum road electro physiology mixed signal of Fetal ECG Signal-to-Noise, t represents the sampling instant of mother's electrocardiosignal and the maximum road electro physiology mixed signal of Fetal ECG Signal-to-Noise, ∫ () dt represents the integration operation to sampling instant t, [] ²represent squared operation.

3. mother based on second-order statistics according to claim 1 and Fetal ECG Blind extracting method, it is characterized in that, the autocorrelation coefficient formula described in step (4e) is as follows:

$f\left(\mathrm{\&epsiv;}\right)=\frac{{\&Integral;}_{n}y\left(n\right)y(n-\mathrm{\&epsiv;})\mathrm{dn}}{{\&Integral;}_n{y}^2\left(n\right)\mathrm{dn}}$

Wherein, the autocorrelation coefficient of Fetal ECG signal when f (ε) represents that time delay is ε and the maximum road electro physiology mixed signal of mother's electrocardiosignal signal to noise ratio, ε represents the time delay of the relative t of a road electro physiology mixed signal that Fetal ECG signal and mother's electrocardiosignal signal to noise ratio are maximum, y (n) represents Fetal ECG signal and the maximum road electro physiology mixed signal of mother's electrocardiosignal signal to noise ratio, n represents the sampling instant of Fetal ECG signal and the maximum road electro physiology mixed signal of mother's electrocardiosignal signal to noise ratio, ∫ () dn represents the integration operation to sampling instant n, [] ²represent squared operation.

4. mother based on second-order statistics according to claim 1 and Fetal ECG Blind extracting method, it is characterized in that, the step of the steepest descending method described in step (5a) is as follows:

The first step, according to the following formula, obtains the degree of approximation of mother's electrocardiosignal vector and optimum mother's electrocardiosignal vector:

$\underset{w}{\min }J\left(w\right)=w^{T}R\left(0\right)w-\log \left[w^{T}R\left(\mathrm{\&alpha;}\right)w\right]$

Wherein, J (w) represents the degree of approximation of mother's electrocardiosignal vector and optimum mother's electrocardiosignal vector, and w represents mother's electrocardiosignal vector, expression minimizes operation about mother's electrocardiosignal vector w's, R (0) and R (α) represents that mother's electrocardiosignal and the maximum road electro physiology mixed signal time delay of Fetal ECG Signal-to-Noise are respectively the autocorrelation matrix of 0 and α respectively, α represents the fine estimation in mother's electrocardiosignal cycle, log [] represents operation of taking the logarithm, () ^trepresent matrix transpose operation;

Second step, according to the following formula, iterative computation mother electrocardiosignal vector:

w(k+1)＝w(k)-μ[R(0)w(k)-[w(k) ^TR(α)w(k)] ^-1R(α)w(k)]

Wherein, w (k+1) represents mother's electrocardiosignal vector of kth+1 iteration, w (k) represents mother's electrocardiosignal vector of kth time iteration, k represents the iterations of mother's electrocardiosignal vector, μ represents the iteration step length of mother's electrocardiosignal vector, R (0) and R (α) represents that mother's electrocardiosignal and the maximum road electro physiology mixed signal time delay of Fetal ECG Signal-to-Noise are respectively the autocorrelation matrix of 0 and α respectively, α represents the fine estimation in mother's electrocardiosignal cycle, () ^trepresent matrix transpose operation, () ^-1represent inversion operation;

3rd step, iteration mother electrocardiosignal, until convergence, obtains optimum mother's electrocardiosignal vector.

5. mother based on second-order statistics according to claim 1 and Fetal ECG Blind extracting method, is characterized in that, the correction Newton method step described in step (5b) is as follows:

The first step, according to the following formula, obtains the degree of approximation of Fetal ECG signal vector and optimum Fetal ECG signal vector:

$\underset{u}{\min }L\left(u\right)=u^{T}V\left(0\right)u-\log \left[u^{T}V\left(\mathrm{\&beta;}\right)u\right]$

Wherein, L (u) represents the degree of approximation of Fetal ECG signal vector and optimum Fetal ECG signal vector, and u represents Fetal ECG signal vector, represent and minimize operation about Fetal ECG signal vector u, V (0) and V (β) represents that Fetal ECG signal and the maximum road electro physiology mixed signal time delay of mother's electrocardiosignal signal to noise ratio are respectively the autocorrelation matrix of 0 and β respectively, β represents the fine estimation of Fetal ECG signal period, log [] represents operation of taking the logarithm, () ^trepresent matrix transpose operation;

Second step, according to the following formula, iterative computation Fetal ECG signal vector:

$u(\mathrm{\&eta;}+1)=\frac{[{\left(u{\left(\mathrm{\&eta;}\right)}^{T}V\left(\mathrm{\&beta;}\right)u\left(\mathrm{\&eta;}\right)\right)}^{-1}+2{\left(u{\left(\mathrm{\&eta;}\right)}^{T}V\left(0\right)u\left(\mathrm{\&eta;}\right)\right)}^{-2}u{\left(\mathrm{\&eta;}\right)}^{T}V\left(\mathrm{\&beta;}\right)u\left(\mathrm{\&eta;}\right)]{\left(V\left(0\right)\right)}^{-1}V\left(\mathrm{\&beta;}\right)u\left(\mathrm{\&eta;}\right)}{1+2{\left(u{\left(\mathrm{\&eta;}\right)}^{T}V\left(0\right)u\left(\mathrm{\&eta;}\right)\right)}^{-2}u{\left(\mathrm{\&eta;}\right)}^{T}V\left(\mathrm{\&beta;}\right){\left(V\left(0\right)\right)}^{-1}V\left(\mathrm{\&beta;}\right)u\left(\mathrm{\&eta;}\right)}$

Wherein, u (η+1) represents the Fetal ECG signal vector of η+1 iteration, u (η) represents the Fetal ECG signal vector of the η time iteration, η represents the iterations of Fetal ECG signal vector, V (0) and V (β) represents that Fetal ECG signal and the maximum road electro physiology mixed signal time delay of mother's electrocardiosignal signal to noise ratio are respectively the autocorrelation matrix of 0 and β respectively, β represents the fine estimation of Fetal ECG signal period, () ^trepresent matrix transpose operation, () ^-1represent inversion operation, () ^-2represent the operation of negate quadratic power;

3rd step, iteration Fetal ECG signal, until convergence, obtains optimum Fetal ECG signal vector.

6. mother based on second-order statistics according to claim 1 and Fetal ECG Blind extracting method, it is characterized in that, the projecting method described in step (6a) refers to according to the following formula, obtains mother's electrocardiosignal:

h(t)＝e ^Td(t)

Wherein, h (t) represents mother's electrocardiosignal vector that mother's electrocardiosignal that t is extracted, e represent optimum, and d (t) represents the pretreated electro physiology mixed signal of t, () ^trepresent matrix transpose operation.

7. mother based on second-order statistics according to claim 1 and Fetal ECG Blind extracting method, it is characterized in that, the projecting method described in step (6b) refers to according to the following formula, obtains Fetal ECG signal:

z(t)＝f ^Td(t)

Wherein, z (t) represents that t extracts the Fetal ECG signal obtained, and f represents optimum Fetal ECG signal vector, and d (t) represents the pretreated electro physiology mixed signal of t, () ^trepresent matrix transpose operation.