CN106137184A - Electrocardiosignal QRS complex detection method based on wavelet transformation - Google Patents

Abstract

The present invention relates to electrocardiosignal QRS complex detection method technical field, particularly to a kind of electrocardiosignal QRS complex detection method based on wavelet transformation, the steps include: to select a special QRS wave as To Template, the low frequency signal on the 4th yardstick and high-frequency signal after carrying out wavelet transformation and preserving conversion；Carrying out wavelet transformation after signal to be detected carries out pretreatment, obtain characteristic frequency spectrum radio-frequency component derivation and take absolute value, retain its maximum point, the maximum cycle acquisition QRS wave waveform in conjunction with QRS complex carries out wavelet transformation；Low frequency signal on 4th yardstick after conversion and template carried out amplitude related operation, the high-frequency signal on the 4th yardstick after conversion and template carried out phase place related operation；Compose weight computing again to respectively two kinds of computings, it may be judged whether coupling；It is an object of the invention to provide one the shortest, the electrocardiosignal QRS complex detection method based on wavelet transformation that accuracy rate is high.

Description

Electrocardiosignal QRS complex detection method based on wavelet transformation

Technical field

The present invention relates to electrocardiosignal QRS complex detection method technical field, become based on small echo particularly to one The electrocardiosignal QRS complex detection method changed.

Background technology

QRS wave: represent the potential change of sequences of ventricular depolarization, is the ripple of sequences of ventricular depolarization process appearance；QRS wave is ECG The most obvious part in waveform, the electric behavior of heart when it reflects ventricular systole, during the generation of therefore QRS wave Between and shape give current cardiac state in which information.In this sense, QRS wave detection is all The basis that electrocardiosignal automatically analyzes, traditional QRS wave testing goal is the identification of detection QRS wave, i.e. sentences Whether disconnected is QRS wave；The position of the QRS complex that it is paid close attention to, but the feature of each QRS wave can not be differentiated； If there being the electrocardiosignal in N number of QRS wave cycle, we want to find or phase similar to m-th QRS wave Same all waveforms, present method is artificial cognition, the longest, and accuracy rate is low.

Summary of the invention

It is an object of the invention to overcome the above-mentioned deficiency in the presence of prior art, it is provided that one is the shortest, The electrocardiosignal QRS complex detection method based on wavelet transformation that accuracy rate is high.

In order to realize foregoing invention purpose, the invention provides techniques below scheme:

A kind of electrocardiosignal QRS complex detection method based on wavelet transformation, the steps include:

A, one special QRS wave of selection are as To Template；

B, the QRS wave template selected is carried out wavelet transformation and preserves conversion after low frequency letter on the 4th yardstick Number and high-frequency signal；

C, signal to be detected is carried out pretreatment, the sample rate of signal to be detected is changed into preset range Sample rate, pretreated electrocardiosignal is carried out wavelet transformation, obtains characteristic frequency spectrum radio-frequency component and ask Lead and take absolute value, retaining the maximum point of derivation absolute value, according to the change of maximum point amplitude, in conjunction with The maximum cycle of QRS complex obtains signature waveform, extracts QRS wave；

D, the QRS wave finally extracted in step C is carried out wavelet transformation；On the 4th yardstick after conversion Low frequency signal and template transformation after low frequency signal carry out amplitude related operation, on the 4th yardstick after converting High-frequency signal and template transformation after high-frequency signal carry out phase place related operation；

E, compose weight factors to two kinds of computings respectively, two related operations be multiplied with weight factor do again and, Obtain the degree of correlation of QRS wave and the template QRS wave selected in step B, judge that measured signal obtains with this To QRS wave whether mate with template QRS wave.

Traditional QRS wave testing goal is the identification of detection QRS wave, i.e. determines whether QRS wave；It institute The position of the QRS complex paid close attention to, but the feature of each QRS wave can not be differentiated；Present techniques is mainly characterized by On the basis of determining the position of QRS wave, identify special QRS wave feature, the most first select a spy Levy waveform, then find out same or similar waveform.Present techniques substantially increases QRS wave respectively Efficiency, need the problem of plenty of time and manpower by artificial judgment and identification feature QRS before solving, And the erroneous judgement that artificial judgment causes；

Giving one example, have the electrocardiosignal in 10000 QRS wave cycles, we want to find and the 10th All waveforms that QRS wave is similar or identical, if at least needing within 24 hours, just can complete by artificial cognition, And just can be completed less than 1 minute by this technology only used time, and accuracy rate is more than 99%.

As the preferred version of the present invention, in step E, if degree of correlation reaches more than 95%, then it is assumed that The QRS wave that measured signal obtains mates with template QRS wave.

As the preferred version of the present invention, in step C, to characteristic frequency spectrum radio-frequency component derivation and take absolute value After, small-signal interference is rejected.

As the preferred version of the present invention, in step E, compose a weight factor, power to respectively two kinds of computings Weight span 0～1.

As the preferred version of the present invention, in step E, define amplitude associated weight W_pWith phase place associated weight W_S, and W_S+W_p=1.

Compared with prior art, beneficial effects of the present invention:

The shortest, accuracy rate is high.

Accompanying drawing illustrates:

Fig. 1 is the wavelet transform exploded view of electrocardiosignal.

Fig. 2 is ecg characteristics signal detection process figure.

Detailed description of the invention

Below in conjunction with embodiment and detailed description of the invention, the present invention is described in further detail.But should be by This is interpreted as that the scope of the above-mentioned theme of the present invention is only limitted to below example, all real based on present invention institute Existing technology belongs to the scope of the present invention.

Embodiment 1

A kind of electrocardiosignal QRS complex detection method based on wavelet transformation, the steps include:

A, one special QRS wave of selection are as To Template, and definition template qrs signal is M (t)；

B, to select QRS wave template carry out wavelet transformation, equation below 1:

SWT₄(M (t))=A₄(M(t))+D₄(M(t))；

Wherein, A₄(M (t)) is the low-frequency component after M (t) small echo changes；D₄(M (t)) is the height after M (t) small echo changes Frequently composition, the low frequency signal preserved after converting on the 4th yardstick and high-frequency signal；

C, signal to be detected is carried out pretreatment, the sample rate of signal to be detected is changed into preset range Sample rate (as it is shown in figure 1, ECG signal processing is exactly the sample rate of electrocardiosignal changing input, Sample rate will change over the sample rate in preset range；Purpose is the electrocardiosignal of the wavelet transformation making output It is the data of a fixed range sample rate, makes the frequency range of QRS wave fall in its characteristics of decomposition spectrum bands, Fix the decomposition scale of wavelet transformation, it is ensured that the spatial cache of wavelet decomposition is to fix little space, and And secure the complexity of calculating, as shown in Fig. 2 (a), it is simply that secure the electrocardiosignal of decomposition scale.Sampling During frequency transformation, extract change to be followed of counting, it is ensured that the electricity physiological signal being input to wavelet decomposition is to limit In the range of sample rate, simultaneously ensure signal undistorted.), pretreated electrocardiosignal is carried out small echo change Change and (acquired in the spectral range more slightly higher than QRS complex frequency range (7-27Hz) by wavelet transformation Characteristic frequency spectrum composition, is conducive to weakening tall and big P ripple and the interference of T ripple.Fig. 2 (b) is to securing decomposition chi The electrocardiosignal of degree carries out the characteristic frequency spectrum signal graph after wavelet transformation；After based on wavelet transformation, former electrocardio The R peak value of signal becomes the zero crossing of characteristic frequency spectrum signal, in the rising edge of the R ripple of former electrocardiosignal Section and trailing edge, form respectively two extreme points；The two extreme point is adjacent extreme point, and we claim For extreme point pair.Wavelet transformation has been widely used in the characteristic wave detection of electrocardiosignal, is mainly used in The time-frequency characteristic of wavelet transformation and multi-resolution characteristics.It is prior art about wavelet transformation, thus more The process of wavelet transformation does not repeats them here.), obtain characteristic frequency spectrum radio-frequency component derivation taking absolute value and (adopt With wavelet decomposition high-frequency characteristic spectrum component being done derivation conversion, after derivation conversion, make the mistake of characteristic frequency spectrum signal Zero point is transformed to maximum point and highlights, it is to avoid find the extreme value mistake to producing for location R ripple Disturb, and amount of calculation is greatly reduced.Fig. 2 (c) is to obtain oscillogram to after characteristic frequency spectrum signal derivation；Ask Waveguide transformation can be transformed to maximum point by needing the zero crossing between the extreme point pair that the waveform identified is corresponding, and Maximum point summit can be obtained continuous by the character of derivative.Meanwhile, derivation process can make the amplitude of radio-frequency component add By force, the amplitude of low-frequency component is decayed, and Q ripple, R ripple, S ripple belong to high frequency in effective band Point, and T ripple and P ripple belong to low-frequency component, after derivation, Q ripple, R ripple, S wave amplitude are strengthened, and T ripple Or the amplitude of P ripple can decay further, effectively inhibit tall and big P ripple and the interference of T ripple, after being conducive to Continuous identifying processing.The purpose that in the present embodiment takes absolute value derivative composition is to avoid producing because of waveform variations Raw erroneous judgement.In electrocardiosignal, because waveform variations often there are the feelings such as the big S of little R, the big little R of Q Condition, after wavelet transformation, owing to being affected by R-wave amplitude, the waveform after R ripple correspondent transform is the most weak, although Derivation can highlight R ripple, but it is the most weak to compare, and is difficult to accurately be detected, due to Q ripple, R ripple, S The pitch of waves, now can be by identifying Q ripple or S ripple approximate location R ripple, then in waveform variations from the least Time identify that Q ripple or S ripple just seem no less important.Therefore, derivative is taken absolute value, just can be at Q, R, S Finding a waveform of the amplitude maximum of corresponding maximum in ripple, traditional method is to realize this function 's.), small-signal interference is rejected, particularly as follows:

The data taking the short period carry out moving average calculating, obtain average and the baseline of small-signal, to little letter Number the effective amplification of mean set one, obtain the envelope of small-signal, to the data zero setting in envelope, And the average of small-signal is set to next signature waveform detection the rolling off the production line of threshold value, and (described small-signal is for by a small margin The random interfering signal of value, described big signal is exactly normal electrocardiosignal.Small-signal interference is rejected, Retain the bigger extreme point of described derivative absolute value.Need when there is no normal electrocardiosignal to reject small-signal Interference, it is necessary to the data taking the short period in the derivative absolute value of characteristic frequency spectrum composition carry out moving average, To determine average, lower limit effective for mean set one and amplification reject the interference of small-signal, this Method computation complexity is the least；The extreme point that thresholding is less in removing neighborhood, retains bigger extreme point；Pick Except the smooth random disturbance signal of small magnitude, electrocardiosignal is frequently accompanied by small magnitude stationary random signal and baseline The interference such as drift, present stationary random signal and the shifted signal superposition slowly of small magnitude, and interference is normal QRS wave detection accuracy rate.Especially when not having normal electrocardiosignal to input, the inspection of occasional mistake Measure QRS wave.Study carefully its main cause, be because the interference of small magnitude stationary random signal, reject small magnitude and put down After steady stochastic signal, detection would not be interfered by baseline drift.According to small magnitude stationary random signal Feature, the data taking the short period in the derivative absolute value of characteristic frequency spectrum composition carry out moving average calculating. Calculated average and the baseline of small magnitude stationary random signal, to mean set once effectively amplification Obtain the envelope of small magnitude smooth random disturbance signal, the data zero setting between envelope can be rejected slightly Value stationary random signal.The average simultaneously obtained can be used as the lower limit of next signature waveform detection threshold value.This Method computation complexity is the least, real-time is high)；

Retain the maximum point of derivation absolute value, according to the change of maximum point amplitude, in conjunction with QRS complex Maximum cycle obtain signature waveform, extract QRS wave, particularly as follows:

The threshold value of electrocardiosignal is calculated by self-training method or Adaptive Thresholding；Remove the waveform less than threshold value Data, available QRS complex shape；Maximum extreme point (described maximum in the detection continuous print QRS cycle Extreme point is R ripple or S ripple or the crest of Q ripple in QRS signature waveform, and signature waveform detection is to Fig. 2 (d) The process of middle transform data, obtains signature waveform according to the change of amplitude.Detection to signature waveform is one The combination of series simple computation method, mainly has the methods such as self-training method, adaptive threshold, herein in connection with The maximum cycle of QRS complex.Self-training method is obtained by a minimum duration electrocardiosignal training Max-thresholds；Adaptive threshold is that the max-thresholds obtained according to two sections of self-training methods of continuous adjacent calculates averagely, Obtain present threshold value.Present threshold value needs to combine the bottom threshold that previous step (in four) obtains be modified and obtain Obtained more accurate threshold value, and used the method for thresholding to remove the Wave data less than threshold value, i.e. can get QRS Composite wave-shape.But the QRS complex shape now obtained corresponds to one section of significant wave graphic data, can be in conjunction with maximum In the cycle, find out in the maximum extreme point in the continuous print QRS cycle, i.e. the QRS signature waveform corresponding to detection R ripple or S ripple or the crest of Q ripple, the local maximum point as shown in Fig. 2 (d).)；

Defining this qrs signal is Qrs (t)；

D, apply described formula 1 that Qrs (t) carries out wavelet transformation, equation below 2:

SWT₄(Qrs (t))=A₄(Qrs(t))+D₄(Qrs(t))；

Low frequency signal after low frequency signal on Qrs (t) the 4th yardstick is converted to template M (t) carries out the relevant fortune of amplitude Calculate, equation below 3:

If

X (t)=A₄(M(t))；

Y (t)=A₄(Qrs(t))；

Then

$\begin{array}{c}\mathrm{ScopeCorr}={\mathrm{Cor}}_{\mathrm{scope}}(X\left(t\right),Y\left(t\right))\\ =\frac{\mathrm{cov}(X\left(t\right),Y\left(t\right))}{\sqrt{D\left(X\left(t\right)\right)}\·\sqrt{D\left(Y\left(t\right)\right)}}\end{array}$

Wherein, Corr_scope(X, Y) represents amplitude related operation, and ScopeCorr represents amplitude dependency, cov (X, Y) Representing covariance computing, D (X) represents expectation computing；

It is relevant that high-frequency signal after high-frequency signal on Qrs (t) the 4th yardstick is converted to template M (t) carries out phase place Computing, equation below 4:

If

Then

The phase place of X (t) is

The phase place of Y (t) is

Then phase correlation

$\begin{array}{c}\mathrm{PhasicCorr}={\mathrm{Corr}}_{\mathrm{phasic}}(D_4\left(M\left(t\right)\right),D_4\left(\mathrm{Qrs}\left(t\right)\right))\\ =\frac{\mathrm{cov}(P_x\left(t\right),P_y\left(t\right))}{\sqrt{D\left(P_x\left(t\right)\right)}\·\sqrt{D\left(P_y\left(t\right)\right)}}\end{array}$

Wherein, Corr_phasicX () represents phase place related operation, PhasicCorr represents phase correlation, cov (X, Y) table Showing covariance computing, D (X) represents expectation computing；

E, compose weight factors to two kinds of computings respectively, two related operations be multiplied with weight factor do again and, Obtain the degree of correlation of QRS wave and the template QRS wave selected in step B, judge the QRS of selection with this Whether ripple mates with template QRS wave, particularly as follows:

Definition amplitude associated weight W_p(weight span 0～1) and phase place associated weight W_S(weight value Scope 0～1), and W_S+W_p=1, then the QRS wave QRS wave that measured signal obtains is whole with template QRS wave Body correlation calculations equation below 5:

WholeCorr=ScopeCorr*W_s+PhasicCorr*W_p；

If degree of correlation WholeCorr more than 95%, thinks the QRS wave and template QRS that measured signal obtains Ripple mates.

Claims (5)

1. an electrocardiosignal QRS complex detection method based on wavelet transformation, the steps include:

A, one special QRS wave of selection are as To Template；

B, the QRS wave template selected is carried out wavelet transformation and preserves conversion after low frequency letter on the 4th yardstick Number and high-frequency signal；

C, signal to be detected is carried out pretreatment, the sample rate of signal to be detected is changed into preset range Sample rate, pretreated electrocardiosignal is carried out wavelet transformation, obtains characteristic frequency spectrum radio-frequency component and ask Lead and take absolute value, retaining the maximum point of derivation absolute value, according to the change of maximum point amplitude, in conjunction with The maximum cycle of QRS complex obtains signature waveform, extracts QRS wave；

D, the QRS wave finally extracted in step C is carried out wavelet transformation；On the 4th yardstick after conversion Low frequency signal and template transformation after low frequency signal carry out amplitude related operation, on the 4th yardstick after converting High-frequency signal and template transformation after high-frequency signal carry out phase place related operation；

E, compose weight factors to two kinds of computings respectively, two related operations be multiplied with weight factor do again and, Try to achieve QRS wave and the degree of correlation of template QRS wave that measured signal obtains, judge whether coupling with this.

Electrocardiosignal QRS complex detection method based on wavelet transformation the most according to claim 1, its It is characterised by, in step E, if degree of correlation reaches more than 95%, then it is assumed that the QRS that measured signal obtains Ripple mates with template QRS wave.

Electrocardiosignal QRS complex detection method based on wavelet transformation the most according to claim 1, its It is characterised by, in step C, after characteristic frequency spectrum radio-frequency component derivation taking absolute value, small-signal is disturbed Reject.

Electrocardiosignal QRS complex detection method based on wavelet transformation the most according to claim 2, its It is characterised by, in step E, composes a weight factor, weight span 0～1 to respectively two kinds of computings.

Electrocardiosignal QRS complex detection method based on wavelet transformation the most according to claim 4, its It is characterised by, in step E, defines amplitude associated weight W_pWith phase place associated weight W_S, and W_S+W_p=1.