CN104188649A - Method for guaranteeing signal linear synthesis instantaneity in multipoint physiological electricity monitoring - Google Patents

Abstract

Translated fromChinese

一种多点生理电监测中保障信号线性合成实时性的方法，包括步骤，(1)根据需要布置好体表的多个测点、连接好监测系统、准备监测，(2)预监测记录一段多点生理电信号，用周期成分分析法或者独立分量分析法或者其他方法，求解相应的优化问题，得到相应的信号最优线性组合向量，(3)进入正式监测，用第(2)步确定的信号最优线性组合向量，对每一段新监测的多点生理电信号进行线性合成，以合成出需要的纯净信号，其特征是：把第(2)步预监测阶段确定的信号最优线性组合向量，直接用于了第(3)步正式监测过程中的信号线性合成，而第(3)步正式监测阶段不再求解优化问题，从而保证了整个正式监测阶段线性合成处理的实时性。

A method for ensuring the real-time performance of signal linear synthesis in multi-point physiological electrical monitoring, comprising the steps of (1) arranging multiple measuring points on the body surface according to needs, connecting the monitoring system, and preparing for monitoring; (2) pre-monitoring and recording a section For multi-point physiological electrical signals, use periodic component analysis method or independent component analysis method or other methods to solve the corresponding optimization problem, and obtain the corresponding signal optimal linear combination vector, (3) enter the formal monitoring, and use the step (2) to determine The optimal linear combination vector of the signals is used to linearly synthesize each section of newly monitored multi-point physiological electrical signals to synthesize the required pure signal, which is characterized in that: the optimal linear The combination vector is directly used in the linear synthesis of signals in the formal monitoring process of step (3), and the optimization problem is no longer solved in the formal monitoring phase of step (3), thus ensuring the real-time performance of the linear synthesis process in the entire formal monitoring phase.

Description

In the monitoring of multiple spot physiology electric, ensure a kind of method of the linear synthetic real-time of signal

Technical field

The application relates to a kind of method that ensures the linear synthetic real-time of signal in multiple spot physiology electric monitoring.

Dynamic monitoring electrocardiogram, dynamic monitoring electroencephalogram, dynamic monitoring electromyogram etc., the monitoring and measuring application of these Human Physiology signals of telecommunication, is all often to monitor at a plurality of points of body surface.The electro-physiological signals of being monitored by body surface each point is in fact all the mixed signal of various electro-physiological signals; For example, in the EEG signals of head detection, be often mixed with eye electricity, electrocardio and other interference and noise, the fetal rhythm signal of telecommunication detecting from anemia of pregnant woman's abdominal part, be mixed with very large parent electrocardiosignal, also have respiratory wave, myoelectricity and other interference and noise etc.People carry out linear combination by the multichannel electro-physiological signals in the monitoring of body surface multiple spot, can synthesize more clean simple electro-physiological signals, greatly improve signal to noise ratio.

Multiple spot electro-physiological signals is carried out to linear synthetic the determining of optimum linearity mix vector that relate to.So-called determine optimum linearity mix vector, as its name suggests, will solve exactly an optimization problem under given object function, and the protruding multiextremal optimization problem of right and wrong often, amount of calculation is very large.In the monitoring of multiple spot physiology electric, if each time period signal of monitoring is first solved to an optimization problem, determine corresponding optimum linearity mix vector, and then accordingly this time period signal is carried out to the synthetic processing of linearity, real-time is difficult to guarantee.At present, the linear composition algorithm of the various multiple spot electro-physiological signals of research, is all to utilize each time period signal to determine the optimum linearity mix vector of oneself, is suitable for off-line Non real-time processing; In actual multiple spot physiology electric monitoring system, there is not yet the linear composition algorithm of signal and be applied to real-time processing.

The application will propose a solution exactly, so that the linear synthetic processing of signal can be applied in multiple spot physiology electric monitoring system, goes, and processes in real time.

Background technology

Multiple spot electro-physiological signals is carried out to linearity and synthesize, have periodic component analytic process (Periodic Component Analysis), Independent component analysis (independent component analysis, ICA) etc.These methods, the optimization aim function while determining optimum linearity mix vector is not quite similar, but as previously mentioned, is all solving-optimizing problem the final optimum linearity mix vector of determining for concrete synthetic application.Such as, have from the M road mixed signal x of M some monitoring of anemia of pregnant woman's stomach wall_i(n) n=1, N, i=1 ..., M, N is the sampling number of monitoring time section, now to synthesize the clean Fetal ECG signal fECG in parent electrocardiosignal mECG He Yi road that a road is clean by this M road signal linearity, no matter how, finally all determine the optimum linearity mix vector W of a synthetic maternal ecg_moptimum linearity mix vector W with a synthetic fetal electrocardiogram_f.If note W_m=[w_m1, w_m2..., w_mM] W_f=[w_f1, w_f2..., w_fM], have:

$\mathrm{mECG}=[w_{m1},w_{m2},...,w_{\mathrm{mM}}]\left[\begin{array}{c}{x}_1\left(n\right)n=1,...,N\\ x_2\left(n\right)n=1,...,N\\ .\\ .\\ .\\ x_M\left(n\right)n=1,...,N\end{array}\right]---\left(1\right)$

$\mathrm{fECG}=[w_{f1},w_{f2},...,w_{\mathrm{fM}}]\left[\begin{array}{c}{x}_1\left(n\right)n=1,...,N\\ x_2\left(n\right)n=1,...,N\\ .\\ .\\ .\\ x_M\left(n\right)n=1,...,N\end{array}\right]---\left(2\right)$

Before address, the algorithm that existing multiple spot electro-physiological signals is linear synthetic, is all to utilize each time period signal to determine the optimum linearity mix vector of oneself, is suitable for carrying out off-line Non real-time processing; In actual multiple spot physiology electric real-time monitoring system, because amount of calculation is large, there is not yet these algorithms and be applied to real-time processing.Improve the real-time of system, can rely on the processing speed that improves computer hardware, also can concrete optimized algorithm be improved, but it seems at present, in a foreseeable future, these approach all cannot be dealt with problems at all.Must look for another way and propose a plan, the real-time to ensure that multiple spot electro-physiological signals is linear synthetic, can be applied in physiology electric real-time monitoring system and go.

List of references:

[1]Mohammed?Assam?Ouali，and?Kheireddine?Chafaa，“Separation?of?composite?maternal?ECG?using?SVD?decomposition，”Computer?Applications?Technology(ICCAT)，2013International?Conference?on，pp.1-4，Jan.2013.

[2]A.Van?Oosterom，“Spatial?filtering?of?the?fetal?electrocardiogram，”J.Perinat.Med，vol.14，pp.411-419，1986.

[3]Zhenwei?Shi，and?Changshui?Zhang，“Semi-blind?source?extraction?for?fetal?electrocardiogram?extraction?by?combining?non-Gaussianity?and?time-correlation，”Neurocomputing，vol.70，pp.1547-1581，2007.

[4]Reze?Sameni，Christian?Jutten，and?Mohammad?B.Shamsollahi，“Multichannel?Electrocardiogram?Decomposition?Using?Periodic?Component?Analysis，”IEEE?Bio.Med.Eng.，vol.55，No.8，pp.1935-1940，Aug.2008.

[5]Thato?Tsalaile，Reza?Sameni，Saeid?Sanei，et?al.，“Sequential?blind?source?extraction?for?quasi-periodic?signals?with?time-varying?period，”IEEE?Bio.Med.Eng.，vol.56，No.3，pp.646-655，Mar.2009.

[6]Aapo?Hyvarinen?and?Erkki?Oja，“A?fast?fixed-point?algorithm?for?independent?component?analysis，”Neural?Comput.vol.9，pp.1483-1492，1997.

[7]Jean-Francois?Cardoso，“Source?Separation?Using?Higher?Order?Moments，”ICASSP，proceedings，vol.4，May?1989.

[8]Wei?Lu，and?Jagath?C.Rajapakse，“Approach?and?Applications?of?Constrained?ICA，”IEEE?TNeuralNetwor.，vol.16，No.1，pp.203-212，Jan.2005.

[9]Wei?Lu，and?Jagath?C.Rajapakse，“ICA?with?Reference，”Neurocomputing，vol.69，pp.2244-2257，2006.

Summary of the invention

Goal of the invention.

The linear composition algorithm of existing multiple spot electro-physiological signals, is all based on each monitoring time segment signal itself, to determine the optimum linearity mix vector of this segment signal, is suitable for the multiple spot electro-physiological signals of monitoring to carry out the Non real-time processing of off-line.Object of the present invention is exactly to look for another way to propose a kind of scheme that can ensure the real-time of the linear synthetic processing of multiple spot electro-physiological signals, to meet actual physiology electric monitoring system to the linear synthetic demand of processing real-time of signal.

Technical scheme.

In a kind of multiple spot physiology electric monitoring, ensure the method for the linear synthetic real-time of signal, comprise step, (1) arrange as required a plurality of measuring points of body surface, connect monitoring system, prepare monitoring, (2) one section of multiple spot electro-physiological signals of premonitoring record, with periodic component analytic process or Independent component analysis or additive method, solve corresponding optimization problem, obtain corresponding signal optimum linearity mix vector, (3) enter formal monitoring, with the definite signal optimum linearity mix vector of (2) step, the multiple spot electro-physiological signals of the new monitoring of each section is carried out to linearity to be synthesized, to synthesize the purified signal of needs, it is characterized in that: definite signal optimum linearity mix vector of (2) step premonitoring stage, the signal being directly used in the formal observation process of (3) step is linear synthetic, and (3) step is formally monitored stage solving-optimizing problem no longer, thereby guaranteed the linear synthetic real-time of processing of whole formal monitoring stage.The FB(flow block) of method as shown in Figure 1.

The scheme that more than ensures the linear synthetic real-time of signal is not random conjecture, but research based on some basic laws and understanding propose.These rules comprise: the linear synthetic method of the multi-point signals such as (1) periodic component analysis, independent component analysis, can be separated for the blind source of multiple spot mixed signal, by linearity, synthesized and isolated each relatively pure source signal, these methods are confirmed from theory into action; (2) in theory, optimum linearity mix vector in the linear synthetic methods such as periodic component analysis, independent component analysis, by each signal source, to produce a characteristic to each monitoring point bang path of body surface to decide in essence, for the monitoring of multiple spot physiology electric, the bang path characteristic by physiology electric generation sources such as heart, brain, eyes to each monitoring point of body surface decides; (3), in multiple spot physiology electric observation process, the physiology electric generation sources such as heart, brain, eyes are generally that changeless, bang path is metastable to the position of each monitoring point of body surface.Therefore, definite optimum linearity mix vector of premonitoring stage being directly used in to linearity in formal observation process below synthetic is feasible from reason.This conclusion that inventor has also passed through lot of experiment validation, relevant paper is about to deliver.

Beneficial effect.

First, the application's scheme is obviously different from the existing method of utilizing current monitor signal section to determine this section of optimum linearity mix vector, therefore, has novelty.Second, the optimum composite vector that the premonitoring stage is determined is directly used in formal observation process below, make formal observation process no longer relate to solving-optimizing problem, only merely multi-point signal is carried out to linear weighted function summation, amount of calculation reduces greatly, can ensure utterly the real-time that signal is linear synthetic, therefore, this is a substantive progress.

Accompanying drawing 2 is online disclosed anemia of pregnant woman's stomach wall pyroelectric monitor data, Fig. 2 (a) is the electro-physiological signals data (0-10 section second) of (c) respectively corresponding 3 stomach wall measuring points (b), and they all can regard the mixed signal of parent electrocardio, fetal rhythm electricity and noise etc. as.For the sake of simplicity, here using that to synthesize clean parent heart power signal be example: inventor by 0-10 section second multi-point signal as premonitoring phase data, by independent component analysis method, determined that the optimum linearity mix vector of this segment data is Wm=[-0.4748,0.0523,-0.8786], this section of maternal ecg result of synthesizing with this vector is as shown in accompanying drawing 3 (a), obviously more pure than former stomach wall electricity mixed signal; Then, by definite optimum linearity mix vector Wm=[-0.4748 of premonitoring stage, 0.0523,-0.8786] being directly used in the linearity of each segment data afterwards synthesizes, obtained equally success, accompanying drawing 3 (b)-(h) show respectively the synthetic result of 20-30 section second, 70-80 section second, 120-130 section second, 170-180 section second, 220-230 section second, 270-280 section second and 300-310 section second, has all obtained the relatively parent electrocardio of " totally ".This example has illustrated the present invention program's effectiveness.

Accompanying drawing explanation

Fig. 1, ensures the linear block diagram that synthesizes Real-Time Performance of signal in multiple spot physiology electric monitoring of the present invention.

Fig. 2,3 physiology electric Monitoring Data of premonitoring stage body surface in embodiment

Fig. 3, the synthetic result schematic diagram of the some linearities of whole monitoring stage signal in embodiment

The specific embodiment (embodiment)

Utilize the data of online disclosed data base MIT Non-Invasive Fetal Electrocardiogram Database to carry out example as multiple spot Monitoring Data.In this data base, comprise altogether 55 groups of data, sample rate is 1kHz, and every group of data comprise that 2 road breasts lead signal and the 3 or 4 road stomach wall signals of telecommunication.For convenient, 3 stomach wall signals of telecommunication of the data that only to take out suite number be " ecgca748 " are specifically described, and see shown in accompanying drawing (2); They are all mixed signals, have comprised the noises such as parent electrocardio, fetal rhythm electricity and myoelectricity, except parent electrocardio, other compositions a little less than, in the present embodiment, intend by being example to the synthetic parent electrocardio that obtains a road " totally " of their linearity.

Total monitoring length of these group data is 318 seconds, frontly addresses, and by the step of this programme, usings its data of first 10 seconds as premonitoring phase data, and follow-up data, as formal monitoring phase data, is usingd every 10 seconds as a signal segment (marking altogether 30 sections).In the premonitoring stage, utilizing independent component analysis method to determine and obtaining the cardiac electrical optimum linearity mix vector of this segment data parent is Wm=[-0.4748,0.0523 ,-0.8786]; Then, by definite optimum linearity mix vector Wm=[-0.4748 of premonitoring stage, 0.0523 ,-0.8786] linearity that is directly used in data is afterwards synthetic.Accompanying drawing 3 (b)-(h) show the respectively synthetic result of 20-30 section second, 70-80 section second, 120-130 section second, 170-180 section second, 220-230 section second, 270-280 section second and 300-310 section second, it is all the parent electrocardio of comparison " totally ", synthesize and obtained success, the feasibility of method has been described.

Claims (1)

1. in a multiple spot physiology electric monitoring, ensure the method for the linear synthetic real-time of signal, comprise step, (1) arrange as required a plurality of measuring points of body surface, connect monitoring system, prepare monitoring, (2) one section of multiple spot electro-physiological signals of premonitoring record, with periodic component analytic process or Independent component analysis or additive method, solve corresponding optimization problem, obtain corresponding signal optimum linearity mix vector, (3) enter formal monitoring, with the definite signal optimum linearity mix vector of (2) step, the multiple spot electro-physiological signals of the new monitoring of each section is carried out to linearity to be synthesized, to synthesize the purified signal of needs, it is characterized in that: definite signal optimum linearity mix vector of (2) step premonitoring stage, the signal being directly used in the formal observation process of (3) step is linear synthetic, and (3) step is formally monitored stage solving-optimizing problem no longer, thereby guaranteed the linear synthetic real-time of processing of whole formal monitoring stage.