CN102240208A - Electrocardiosignal denoising wavelet algorithm implementable in single chip microcomputer - Google Patents

Abstract

The invention discloses an electrocardiosignal denoising wavelet algorithm implementable in a single chip microcomputer, which is particularly suitable for a pocket-sized portable mobile multi-parameter monitoring instrument. The electrocardiosignal denoising wavelet algorithm is characterized in that a method for removing three kinds of noise of an electrocardiosignal through a lifting wavelet and median filtering is proposed. The construction of the lifting wavelet comprises three processes of: (1) decomposition: even/odd decomposition is adopted; (2) predication: an odd number sequence is predicated by using an even number sequence; and (3) updating: the even number sequence is updated by using the odd number sequence predicated in the previous step. A wavelet function is reconstructed, a proper wavelet basis is selected, an original signal is decomposed into three layers through the lifting wavelet, a low frequency signal at the third layer is reserved, baseline lift in the signal is then removed by using the median filtering, and finally, the electrocardiosignal subsequent to wave filtering is reconstructed by using a lifting wavelet reconstruction algorithm. The electrocardiosignal denoising wavelet algorithm disclosed by the invention has the advantages that: the structure is simple, the amount of in-situ computation is low, the memory space is saved, and inverse transformation can be directly implemented by reversion.

Description

A kind of electrocardio denoising wavelet algorithm that can in single-chip microcomputer, realize

Technical field

The present invention relates to a kind of electrocardio denoising wavelet algorithm that can in single-chip microcomputer, realize, be specially adapted to pocket portable mobile multi-parameter monitor.

Background technology

Electrocardio-activity is the basic activity of heart, the existing history that goes up a century of Electrocardiographic clinical practice, although science and technology development makes more and more modernization of clinical diagnosis technology, but electrocardiogram is still the main supplementary means of clinical diagnosis cardiovascular disease so far, especially still has unique effect at aspects such as ischemic heart desease, various ARR diagnosis.

Electrocardiosignal is a kind of faint bioelectrical signals, signal amplitude is less than 5mv, frequency range is between 0.05～250HZ, its energy mainly concentrates between 0.25～35HZ, again owing in the process that electrocardiosignal is handled, can sneak into various noises and interference unavoidably, thereby be easy to cause the distortion of electrocardiosignal, even covered characteristic information in the original electrocardiographicdigital waveform

Wherein main noise jamming has following three kinds:

1) the power frequency interfering frequency is 50HZ, shows as sine wave clocklike, and it is irregular that the lighter can cause electrocardiogram, can cause electrocardio when disturbing greatly and be beyond recognition.Be to produce owing to surrounding has alternating-current installation/AC installation, patient's limbs contact ferrum bed, reasons such as lead line loose contact or fracture.

2) the myoelectricity interfering frequency is many shows as a series of irregular tiny prickles in electrocardiogram between 10～300HZ, easily mistaken diagnosis is the atrial fibrillation ripple.Be owing to reasons such as room temperature is low excessively, the spiritual overstrain of examinee, battery lead plate and contact skin tension produce.

3) baseline drift frequency is many below 1HZ, shows as that the electrocardiogram baseline swings up and down or lifting suddenly, and is influential to accurate judgement S_T section.Be owing to reasons such as the detected person breathes shakiness, lead line tractive tension, battery lead plate and contact skin are bad produce.

The time-frequency locating features of traditional wavelet has extraordinary effect in the noise of removing electrocardiosignal, but because it is based on convolution algorithm, and will be when removing baseline drift with about signal decomposition to ten layer, therefore amount of calculation is big, the computation complexity height, demand height to memory space is unfavorable for real-time processing.

Summary of the invention

The present invention is directed to the drawback that existing electrocardio denoising method exists, proposed a kind of new solution: the algorithm that uses a kind of Lifting Wavelet and medium filtering to combine is removed three kinds of noises in the electrocardiosignal.Lifting Wavelet is a kind of implementation method of wavelet transformation more fast and effectively, it does not rely on Fourier transformation, inherited the multiresolution characteristic of first generation small echo, adopt Lifting Wavelet that primary signal is decomposed into three layers, keep trilaminar low frequency signal, with the baseline drift in the medium filtering removal signal, utilize Lifting Wavelet restructing algorithm reconstruct filtering electrocardiosignal afterwards at last then.

Compared with prior art, the beneficial effect that the present invention has is: simple in structure, operand is low, former bit arithmetic, save memory space, inverse transformation and can directly reverse and realize etc., and Daubechies is verified, and any wavelet transform can be decomposed into a series of simple lifting step.

In the implementation procedure of Lifting Wavelet algorithm, combine the advantage of two kinds of small echos, decompose with different small echos in the different numbers of plies, ground floor haar small echo, be because the haar wavelet algorithm is simple relatively, and certain effect is arranged for the removal of baseline drift, second layer bior4.4 small echo is used the sym8 small echo for the 3rd layer, and power frequency is disturbed and the myoelectricity interference has effect preferably for removing simultaneously.

Based on the above results, can draw:

(1) characteristics that Lifting Wavelet is simple in structure, operand is low are fit to chip microcontroller.

(2) Lifting Wavelet can better be removed multiple interference noise at different decomposition layers with different small echos.

(3) do medium filtering in the low frequency signal after decomposition, reduced to calculate and counted, help reducing amount of calculation.

Description of drawings

Fig. 1 is the structure of Lifting Wavelet.

The acquisition system block diagram that Fig. 2 adopts for experiment.

Fig. 3 is for the original electrocardiographicdigital signal and contain noise signal.

Fig. 4 adopts medium filtering to go baseline drift after the 3rd layer of decomposition.

Three kinds of denoising methods of Fig. 5 are gone the mixed noise contrast.

Three kinds of denoising methods of Fig. 6 contrast actual electrocardiosignal denoising.

The specific embodiment

1. the structure of Lifting Wavelet

The structure of Lifting Wavelet is divided into three processes, comprises decomposition, prediction and renewal, and Fig. 1 is the concrete block diagram of the structure of Lifting Wavelet.

1) decomposes

Primary signal S is resolved into two parts.Multiple decomposition method can be arranged, and the simplest is that odd even is decomposed, and what use in this research is exactly this method, according to the subscript of signal with signal decomposition is

With , wherein j represents the j layer signal, the small echo that this decomposition produces is called lazy small echo.Formula is as follows:

$(S_{j,2\mathrm{<figure-callout\; id="0"\; label="k"\; filenames="HSA00000114874100021.png,HSA00000114874100032.png"\; state="\{\{state\}\}">k</figure-callout>}}^0,S_{j,2k+1}^0)=\mathrm{Split}\left(S_{j+1}\right)$

2) prediction

Go to predict the odd number sequence with the even number sequence.Detailed process is earlier wave filter P to be acted on the even number sequence to obtain P (S _{J, 2k}), then with the difference of odd sequence and predictive value as new odd number sequence.In forecasting process, we will make predictive value according to the characteristics of signal itself

As far as possible near predicted value

The sequence that obtains so just can comprise information still less, so be very important in the selection of this process median filter P.Concrete formula is as follows:

$S_{j,2k+1}=S_{j,2k+1}^0-P\left(S_{j,2k}^0\right)$

3) upgrade

Use the odd sequence that one-step prediction crosses and remove to upgrade even sequence.Detailed process is to construct an operator U earlier, makes it act on odd sequence and obtains U (S _{J, 2k+1}), the even number sequence that the summation of itself and even sequence is obtained upgrading then.The purpose of upgrading is that some global property of original signal collection is held in its subclass relaying continuation of insurance.Formula is as follows:

$S_{j,2k}=S_{j,2\mathrm{<figure-callout\; id="0"\; label="k"\; filenames="HSA00000114874100021.png,HSA00000114874100032.png"\; state="\{\{state\}\}">k</figure-callout>}}^0+U\left(S_{j,2k+1}\right)$

The forward transform process that above-mentioned three steps are Lifting Wavelet only needs to change the plus-minus symbol and just can obtain its transformation by reciprocal direction process.As can be seen, the implementation procedure of lifting wavelet transform is fairly simple.

2. experimental technique

2.1 data acquisition platform

The self-control experiment porch is adopted in this experiment, and the acquisition system performance meets national standard " singly leading and multichannel electrocardiograph YY1139-2000 ".This acquisition system is made up of several modules such as signal input, signal processing, A/D conversion and transmission, power supplys, sees Fig. 2.

Electrocardiosignal is through the electrode and the line input acquisition system of leading, and system delivers to signal processing after electrocardiosignal is done necessary lead selection, voltage-limiting protection.The processing of electrocardiosignal comprises isolation, amplification, filtering and level adjustment, electrocardiosignal after treatment, 12 A/D conversions that carried by master chip are converted to digital signal, and storage.The electrocardiosignal that collects is passed to host computer through serial ports, handles for experiment and uses.

In this experiment, the MPS450 multiparameter analog meter of producing with Fluke Biological Science Co., Ltd produces electrocardiosignal, is obtained by above-mentioned acquisition system collection.Noise and interfering signal all carry function by analog meter and produce, and the acquisition system sample frequency is 560Hz, sees Fig. 3.

2.2 decompose determining of the number of plies

Though the frequency range of electrocardiosignal is between 0.05～250HZ, its energy mainly concentrates between 0.25～35HZ, therefore just can obtain more complete electrocardiosignal as long as keep the signal of this frequency band.The sample frequency of the electrocardiosignal that we use in experiment is 560HZ, according to sampling thheorem, the signal highest frequency that efficient recovery can be arranged is 280HZ, again because each layer wavelet decomposition is the characteristics of two frequency bands with signal halves, therefore we carry out 3 layers of decomposition with Lifting Wavelet to the electrocardiosignal that contains noise, the frequency of the 3rd layer of low frequency signal that obtains like this can satisfy the integrity of electrocardiosignal between 0～35HZ.

2.3 the selection of wavelet basis

In different applications, choosing of wavelet basis is different, but mainly follows following four principles:

(1) symmetry is very useful for avoiding phase shift in Flame Image Process.

(2) vanishing moment is very useful for compression.

(3) regularity obtains preferably for the reconstruct of signal or image that smooth effect is very useful.

(4) tightly propping up property localization ability is stronger, helps determining the catastrophe point of signal.

Different wavelet basiss is different for the Signal Processing result, and the also difference that has is very big, is vital for choosing of wavelet basis therefore.In the process of electrocardiosignal denoising, we have selected two kinds of small echos for use by repeatedly comparison and experiment, are respectively sym8 and bior4.4.

(Fig. 3 (b, c)) tests in a large number, find that the sym8 small echo disturbs denoising effect to be better than the bior4.4 small echo for the power frequency of cardiac electrical signal, and the bior4.4 small echo goes the myoelectricity effects of jamming to be better than sym8, sees Table 1 to initial data.

Three kinds of contrasts that remove the mode of making an uproar of table 1

	The Sym8 small echo	The Bior4.4 small echo	Lifting Wavelet
				Go power frequency to disturb (SNR/MSE)	32.73/63.17	30.61/67.66	32.40/63.92
Go myoelectricity to disturb (SNR/MSE)	68.08/30.44	69.14/29.91	68.37/30.19
				Remove mixed noise (SNR/MSE)	24.58/85.21	24.46/85.41	24.64/85.00
Actual electrocardio denoising (SNR/MSE)	68.32/40.93	67.78/41.25	67.84/41.23
				Analog electrocardiogram signal running time (S)	0.4658	0.5142	0.1770
Actual electrocardiosignal running time (S)	0.4728	0.5455	0.1430

2.4 medium filtering is removed baseline drift

Mainly below 1Hz, if utilize the method for small echo to remove, time and space complexity are excessive for the baseline drift composition of electrocardiosignal.Medium filtering is modal method of going baseline drift, has algorithm simple, realizes advantage flexibly.

Medium filtering is exactly for a segment signal, length is the one piece of data of N before getting a bit, this segment data is sorted, the value of any is as the value of this point in the middle of getting then, whole segment signal is all as above handled, just obtained the drift composition of signal, deducting the drift composition with primary signal at last is exactly the signal of removing after the baseline drift.

Electrocardiosignal through 3 layers of wavelet decomposition after data volume significantly reduce, become originally 1/8, remove baseline drift with medium filtering then, can effectively reduce amount of calculation.We carry out proof of algorithm with the data that contain baseline drift, see Fig. 3 (d), according to experimental result, the signal after the wavelet decomposition are implemented medium filtering, utilize the electrocardiosignal of signal reconstruction after the filtering, go baseline drift satisfactory for result, and signal distortion is low, sees Fig. 4.

Based on above analysis result, we adopt ground floor to decompose and use the haar small echo, and the second layer decomposes uses the bior4.4 small echo, and the 3rd layer of decomposition constitutes Lifting Wavelet with the sym8 small echo, adopts the method for medium filtering after the 3rd layer of decomposition, removes the electrocardiosignal noise.After adopting above denoising method and traditional small echo (sym8 and bior4.4) denoising, medium filtering goes the processing method of baseline to do contrast again, handle one section electrocardiosignal of mixing noise that contains that has added power frequency interference, myoelectricity interference and baseline drift, see Fig. 3 (e).

Effectiveness for further verification algorithm uses this denoising method that one section actual electrocardiosignal is done denoising, investigates its denoising effect, sees Fig. 6.

Analyze as can be known by Fig. 5, Fig. 6 and table 1, Lifting Wavelet, Sym8 small echo and Bior4.4 small echo all have good effect for three kinds of noises removing electrocardiosignal, but Lifting Wavelet has remarkable advantages on speed, almost be 1/3 of other two kinds of small echo running times, this with regard to having solved the small echo computing because its time complexity and can't be applied to actual problem.And we are in the removal for baseline drift, do not use the method for wavelet decomposition to about ten rank to remove, but adopted present everybody relatively approve remove the reasonable medium filtering of baseline drift, this has further reduced amount of calculation again, make our algorithm in single-chip microcomputer, to realize, good effect is arranged for the processing of live signal.

In addition to the implementation, the present invention can also have other embodiments.All employings are equal to the technical scheme of replacement or equivalent transformation formation, all drop in the protection domain of requirement of the present invention.

Claims (5)

1. the electrocardio denoising wavelet algorithm that can realize in single-chip microcomputer is characterized in that: proposed a kind of method of using Lifting Wavelet and medium filtering to remove three kinds of noises of electrocardiosignal.The reconstruct wavelet function, select suitable wavelet basis, adopting Lifting Wavelet is three layers with the original electrocardiographicdigital signal decomposition, in the different numbers of plies with different wavelet decomposition, keep trilaminar low frequency signal, with the baseline drift in the medium filtering removal signal, utilize Lifting Wavelet restructing algorithm reconstruct filtering electrocardiosignal afterwards at last then.

2. a kind of electrocardio denoising wavelet algorithm that can realize in single-chip microcomputer according to claim 1, it is characterized in that: Lifting Wavelet and medium filtering combine.

3. a kind of electrocardio denoising wavelet algorithm that can realize in single-chip microcomputer according to claim 1 is characterized in that: the Lifting Wavelet structure is divided into three processes:

1) decomposes, adopt odd even to decompose;

2) prediction goes to predict the odd number sequence with the even number sequence;

3) upgrade, use the odd sequence that one-step prediction crosses and remove to upgrade even sequence.

4. a kind of electrocardio denoising wavelet algorithm that can in single-chip microcomputer, realize according to claim 1, it is characterized in that: ground floor decomposes uses the haar small echo, the second layer decomposes uses the bior4.4 small echo, the 3rd layer of decomposition constitutes Lifting Wavelet with the sym8 small echo, after the 3rd layer of decomposition, adopt the method for medium filtering, remove the electrocardiosignal noise.

5. a kind of electrocardio denoising wavelet algorithm that can in single-chip microcomputer, realize according to claim 1, it is characterized in that: following four principles are mainly followed in the selection of wavelet basis:

1) symmetry.

2) vanishing moment.

3) regularity.

4) propping up property tightly.

Images