Method for Correcting Baseline Drift of Dynamic ECG Signal with Sudden Jump Noise

专利摘要:
A method for correcting baseline drift of dynamic ECG signal with sudden jump noise is provided. The problem of ECG signal deformation, which is caused by inaccurate baseline extraction when processing the region using conventional filter method, is improved by accurately detecting the sudden jump regions involved in the ECG signals and processing them separately. The method of the present invention is suitable for baseline drift correction processing of various ECG signals containing baseline drift, especially for wearable dynamic ECG signals. The waveform of the baseline drift correction signal obtained by the method of the present invention is significantly better than that obtained by other methods.
公开号:NL2025147A
申请号:NL2025147
申请日:2020-03-17
公开日:2020-05-06
发明作者:Xie Xiaoyun；Wang Yinglong；Liu Hui；Shu Minglei；Zhou Shuwang；Liu Zhaoyang
申请人:Shandong Computer Science Ct Nat Supercomputer Ct Jinan；
IPC主号:

专利说明:

Deze publicatie komt overeen met de oorspronkelijk ingediende stukken.
Method for Correcting Baseline Drift of Dynamic ECG Signal with Sudden Jump Noise
Technical Field
The present invention relates to the technical field of electrocardiography (ECG) signal noise processing, in particular to a method for correcting baseline drift of a dynamic ECG signal with sudden jump noise.
Background
Baseline drift is a common noise in ECG signals with the frequency less than 1 Hz. Baseline correction is very important for accurately diagnosing ECG abnormalities such as ST segment elevation or depression. Currently, the commonly used baseline correction methods include high-pass filtering, median filtering, wavelet transform, spline interpolation, mathematical morphology filtering, adaptive filtering an others.
As dynamic ECG monitoring equipment such as Holter is widely used, especially miniaturized wearable and portable ECG monitoring terminals develop rapidly in recent years, motion interference has become one of the main noise sources of ECG signals, and baseline drift has been increasingly serious. The noise has increased amplitude and frequency concurrently, which overlaps with normal ECG signals to some extent. In addition, unfortunately, the impedance between human skin and electrode may rapidly increase or decrease with the change of displacement under big movements, resulting in a sudden jump in the amplitude of the collected signal within a short time of less than 100 ms. The change of the amplitude can reach more than 0.5 mV (as shown in areas A, B, and C of FIG. 1).
However, none of the above-mentioned baseline correction methods concentrates on local features of signal and processes them accordingly. During baseline extraction, the sudden jump region cannot be sensitively and effectively acquired, resulting in a failure in the correction of certain severe baseline drift in the sudden jump region. Median filtering method is the only one that can detect the location where the sudden jump occurs and fix it accordingly. However, the median filtering method cannot accurately positioning the sudden jump region or fails to detect the overall sudden jump region for processing. Therefore, accurately identifying and effectively eliminating sudden jump region in dynamic ECG signals is a major challenge in baseline drift correction of the dynamic ECG signals.
Summary
In order to overcome the shortcomings of the above prior art, the present invention provides a method for correcting baseline drift of dynamic ECG signal, which can accurately positioning a sudden jump region and obtain optimal baseline drift correction results by separately processing the sudden jump region and the mild change region.
The technical solution adopted by the present invention to overcome the technical defects are as follows.
A method for correcting baseline drift of dynamic ECG signal with sudden jump noise, including the steps of
a) detecting, by a computer, a sudden jump region in the ECG signal to obtain a mild change region and the sudden jump region in the ECG signal;
b) processing the sudden jump region separately to extract the baseline bn u in the sudden jump region;
c) extracting the baseline bn c in the mild change region, and smoothing the extracted baseline by moving average method;
d) merging the baseline bn u extracted from the sudden jump region with the baseline bn_c extracted from the mild change region to obtain the initial baseline bn and corrected data, and smoothing an error between connection points of the sudden jump region and the mild change region by a filter combination.
Further, step a) is implemented by the following sub-steps:
a-1) filtering, by the computer, the ECG signal y including baseline drift and sudden jump noise in the ECG signals using median filters with a width of 200 ms and a width of 600 ms respectively to obtain the baseline signal y' where R wave influence is removed;
a-2) performing a signal length transformation process on the baseline signal y' at a signal width of 30 ms via the formula
to obtain the signal L_: where P wave and T wave are depressed and sudden jump region is obvious, wherein 1+=30 ms, / ⁼ {θ, 1,2,...Αξ,,}, AC is the number of windows in the signal, ^— ƒ/< Jfy-1 , ƒ is a sampling frequency of the signal, yk is the amplitude of the sampling point in the ECG signal y, yk-i is the amplitude of the (k-l)^th sampling point in the ECG signal y;
a-3) defining the threshold thr min, wherein thr _/nm=l.l*mim(Li), adding the region of the signal L, that satisfies Li>thr min to a candidate set, and the region of the signal L_t that does not satisfy L>ihr min to the mild change region, calculating Shannon entropy H of the signal L, after the transformation, calculating Shannon entropy H, and amplitude range Ampi of the i^th candidate region, and defining the amplitude change threshold as thr a;
a-4) traversing each region in the candidate set, adding regions that meet the conditions of H>H and Ampi>thr a to the sudden jump region, and adding regions that do not meet the conditions of Hi>H and Ampi> thr a to the mild change region.
Further, in step b), the median filtered baseline signal y' is smoothed by the moving average method to obtain the baseline bw_u in the sudden jump region.
Further, in step c), by traversing the sudden jump region and separately processing each sudden jump region, the baseline in this region is defined as (-thr r ) *noisy signals in this region, and the processed baseline is defined as bw c, _ AmP_j wherein thr r is the zoom ratio, and thr r---—.
— ³ — 1 Λ jfe T
Further, step d) is implemented by the following sub-steps:
d-1) obtaining the initial smoothed baseline bwl by the formula bwl — ~[(bw ° S) · S + (bw · S) ° S] wherein <S’=(). 1 *ƒ), the ° in the 2 formula is the morphological opening operation, * is the morphological closing operation;
d-2) smoothing the initial smoothed baseline bwl by a smoothing filter method where sample point having a width of wj is calculated via the formula to obtain the final baseline bw', and obtaining the corrected signal via the formulay-bw'.
Further, in step a-3), the Shannon entropy H is calculated via the formula H~ y₌₀ p, *log₂p i , wherein the signal is divided into n segments according to its amplitude, 2,...ƒ<} , and Pj is the probability value of the transformed signal belonging to the Jj'^! segment; the Shannon entropy H of the i^th candidate region is calculated via the formula ' j-bPLj *log₂
P_Lj , wherein the signal is divided into k segments according to its amplitude, Pi, is the probability value of the j^ih segment in the candidate region ƒ; the amplitude range Amp, is calculated via the formula
Amp, = abs(y„ - y_iw), «herein_y„ is the value of the first element of the i^th region, y_m is the value of the last element of the i^lh region.
The present invention has the following advantages. The problem of ECG signal deformation, which is caused by inaccurate baseline extraction when processing the region using conventional filter method, is improved by accurately detecting the sudden jump regions involved in the ECG signals and processing them separately. The method of the present invention is suitable for baseline drift correction processing of various ECG signals containing baseline drift, especially for wearable dynamic ECG signals. The waveform of the baseline drift correction signal obtained by the method of the present invention is significantly better than that obtained by other methods.
Brief Description of the Drawings
FIG. 1 is a schematic diagram showing the sudden jump region in the ECG signal;
FIG. 2 is a flow chart of detecting the sudden jump region of the present invention;
FIG. 3 is a flow chart of extracting baseline from the mild change region and the sudden jump region of the present invention;
FIG. 4 is a flow chart of smoothing the baseline of the present invention; and
FIG. 5 is a flow chart showing the method of the present invention.
Detailed Description of the Embodiments
The present invention is further described hereinafter with reference to FIG. 1 and FIG. 2.
A method for correcting baseline drift of dynamic ECG signal with sudden jump noise, including the steps of:
a) detecting, by a computer, a sudden jump region in the ECG signal to obtain a mild change region and the sudden jump region in the ECG signal;
b) processing the sudden jump region separately to extract the baseline bw u in the sudden jump region;
c) extracting the baseline bw c in the mild change region, and smoothing the extracted baseline by moving average method;
d) merging the baseline bw u extracted from the sudden jump region with the baseline bw_c extracted from the mild change region to obtain the initial baseline bw and corrected data, and smoothing an error between connection points of the sudden jump region and the mild change region by a filter combination.
The problem of ECG signal deformation, which is caused by inaccurate baseline extraction when processing the region using conventional filter method, is improved by accurately detecting the sudden jump regions involved in the ECG signals and processing them separately. The method of the present invention is suitable for baseline drift correction processing of various ECG signals containing baseline drift, especially for wearable dynamic ECG signals. The waveform of the baseline drift correction signal obtained by the method of the present invention is significantly better than that obtained by other methods.
Further, as shown in FIG. 2, step a) is implemented by the following sub-steps:
a-1) filtering, by the computer, the ECG signal y including baseline drift and sudden jump noise in the ECG signals using median filters with the width of 200 ms and the width of 600 ms respectively to obtain the baseline signal y' where R wave influence is removed;
a-2) performing a signal length transformation process on the baseline signal y'
to obtain the signal L, where P wave and T wave are depressed and sudden jump region is obvious, wherein u^,=30 ms, i ⁼ {θ, 1,2,...N_w J N_w is the number of windows in the signal, = y_k — y_k_^ , f_s is a sampling frequency of the signal, yi< is the amplitude of the k¹¹' sampling point in the ECG signal y, yk-i is the amplitude of the (k-l)^,h sampling point in the ECG signal v;
a-3) defining the threshold thr min, wherein thr m/w=l.l*mim(Li), adding the region of the signal L. that satisfies Li>thr min to a candidate set, and adding the region of the signal L, that does not satisfy Li>thr min to the mild change region, calculating Shannon entropy H of the signal Z, after the transformation, calculating Shannon entropy H_s and amplitude range Amp, of the i^th candidate region, and defining the amplitude change threshold as thr a;
a-4) traversing each region in the candidate set, adding regions that meet the conditions of Hi>H and Ampi>thr a to the sudden jump region, and adding regions that do not meet the conditions of Hi>H and Ampp thr a to the mild change region.
Further, as shown in FIG. 3, in step b), the median filtered baseline signal y' is smoothed by the moving average method to obtain the baseline bw it in the sudden jump region.
Further, as shown in FIG. 3, in step c), by traversing the sudden jump region and separately processing each sudden jump region, the baseline in this region is defined as (1-thr r) *noisy signals in this region, and the processed baseline is defined as bw c, wherein thr r is the zoom ratio, and thr r— ———.
— ’ — ’ — 1 n rfi t
Further, as shown in FIG. 4, step d) is implemented by the following sub-steps: d-1) obtaining the initial smoothed baseline bwl by the formula bwl — —-[(bw ° S) · S + (bw · S) ° S] wherein 6-0.1¾ the ° in the formula is the morphological opening operation, ’ is the morphological closing operation; bw © 5 — bw 'V S ^s *T Ö, that is, etching the baseline bw using S and then expanding the baseline bw, bw -;§= ( bw ® 0 S, that is, expanding the baseline bw and then etching the baseline bw.
d-2) smoothing the initial smoothed baseline bwl by a smoothing filter method where sample point having a width of wy is calculated via the formula ~ _m=o^3V1[/7] / wf _t0 obtain the final baseline bw', and obtaining the corrected signal via the formulay-bw'.
Further, in step a-3), the Shannon entropy Miscalculated via the formula i=r>Pj Pj , wherein the signal is divided into segments according to its amplitude, /={0,1,2, , and //is the probability value of the transformed signal belonging to the ƒ^h segment; the Shannon entropy H of the i^th candidate region is calculated via the formula
Hj ' j-bPLj *log₂
Plj , wherein the signal is divided into k segments according to its amplitude, Pi, is the probability value of the j^ih segment in the candidate region ƒ; the amplitude range Amp, is calculated via the formula
Amp, = abs(y, - y_iw), «herein_y„ is the value of the first element of the i^th region, y_m is the value of the last element of the i^lh region.

权利要求:
Claims (6)
[1]
A method for correcting baseline deviation of a dynamic ECG signal with abrupt jump noise, comprising the following steps:
a) detecting an abrupt jump region in the ECG signal by a computer to obtain a mild change region and the abrupt jump region in the ECG signal;
b) separately processing the abrupt jump area to derive the baseline bw u in the abrupt jump area;
c) deriving the baseline bw c in the mild change region, and smoothing the derived baseline by a moving average method;
d) merging the baseline bw u derived from the abrupt jump area with the baseline bw c derived from the mild change area to obtain the initial baseline bw and corrected data, and smoothing an error between junctions of the abrupt jump area and the mild change area by a filter combination.
[2]
The method for correcting the baseline deviation of a dynamic ECG abrupt jump noise signal according to claim 1, wherein step a) is implemented by the following substeps:
a-1) filtering, by the computer, of the ECG signal v including the baseline deviation and abrupt jump noise in the ECG signals using median filters with a width of 200 ms and a width of 600 ms around the baseline signal y obtain where R wave influence is removed;
a-2) performing a signal length transformation process on the baseline signal y 'at a signal width of 30 ms via the formula
F ² to it

[3]
The method for correcting the baseline deviation of a dynamic ECG abrupt jump noise signal according to claim 2, wherein in step b) the median filtered baseline signal y 'is smoothed by the moving average method to obtain the baseline bw u in the abrupt jump area.
[4]
The method for correcting the baseline deviation from the abrupt jump noise dynamic ECG signal according to claim 2, wherein in step c), by traversing the abrupt jump area and processing each abrupt jump area separately, the baseline is defined in this area if (1 -thr r) * noise signals in this area and the processed baseline is defined as bw c, thr r is the magnification ratio and thr
[5]
A method for correcting the baseline deviation of a dynamic ECG abrupt jump noise signal according to claim 2, wherein step d) is implemented by the following substeps:
d-1) obtaining the initial smoothed baseline bwl with the formula bwl = | [(bwoS) · S + (bw · S) oS], _{where 5 = the} o in the formula is the morphological opening operation, * the morphological closing operation is;
d-2) smoothing the initial smoothed baseline bwl by a smoothing filter method where the sample point with a width of w / is calculated by the formula
l) W = Σ Xo ^èwl [«l I ^W f to obtain the final baseline bw ', and obtain the corrected signal via the formula y-bw'.
[6]
A method for correcting the baseline deviation of a dynamic ECG signal with abrupt jump noise according to claim 2, wherein in step a-3) the Shannon entropy H is calculated via the formula y = o P, * l ° g ₂ P i, where the signal is divided into n segments, according to its amplitude, 7 = {O, 1,2, ..} A-, and P is the probability value of the transformed signal to the j belongs ^{to the} segment; the Shannon entropy H of the i ^th candidate region is calculated using the formula h 'i ^ * log ₂ oPLj
Plj, in which the signal is divided into k segments according to its amplitude, the probability value P i, j of the ^{j, the} segment is in the candidate area i; Ampi the amplitude range is calculated through the formula Amp, = abs (y _a, where yn is the value of the first element of the i ^th area, v, _w is the value of the last element of the i ^th region.

类似技术:

---

公开号 | 公开日 | 专利标题

---

NL2025147A|2020-05-06|Method for Correcting Baseline Drift of Dynamic ECG Signal with Sudden Jump Noise

---

Sun et al.2002|ECG signal conditioning by morphological filtering

---

Chawla et al.2008|RETRACTED: A new statistical PCA–ICA algorithm for location of R-peaks in ECG

---

Choi et al.2010|Development of ECG beat segmentation method by combining lowpass filter and irregular R–R interval checkup strategy

---

Sabherwal et al.2017|Automatic detection of the R peaks in single-lead ECG signal

---

Singh et al.2018|An efficient ECG denoising technique based on non-local means estimation and modified empirical mode decomposition

---

Chen et al.2019|Removal of power line interference from ECG signals using adaptive notch filters of sharp resolution

---

Verma et al.2013|An integration of improved median and morphological filtering techniques for electrocardiogram signal processing

---

Xin et al.2014|ECG baseline wander correction based on mean-median filter and empirical mode decomposition

---

Dora et al.2016|Robust ECG artifact removal from EEG using continuous wavelet transformation and linear regression

---

Lin et al.2013|Discrete-wavelet-transform-based noise reduction and R wave detection for ECG signals

---

Bing et al.2020|Noise reduction in ECG signal using an effective hybrid scheme

---

Nguyen et al.2020|Artifact elimination in ECG signal using wavelet transform

---

Rakshit et al.2017|An improved EMD based ECG denoising method using adaptive switching mean filter

---

Janušauskas et al.2013|Ensemble empirical mode decomposition based feature enhancement of cardio signals

---

Hadji et al.2016|Wavelet-based performance in denoising ecg signal

---

Mishra et al.2016|ECG signal denoising using time-frequency based filtering approach

---

WO2018023697A1|2018-02-08|Fetal electrocardiosignal separation method and device

---

Nagarkoti et al.2012|An algorithm for fetal heart rate detection using wavelet transform

---

KR101420345B1|2014-07-16|System and Method for Detecting Ｒ Wave of ECG Signal based on ＳＯＭ

---

Tan et al.2013|Real-time baseline wander removal in ECG signal based on weighted local linear regression smoothing

---

Gaamouri et al.2018|Denoising ECG signals by using extended kalman filter to train multi-layer perceptron neural network

---

Illanes-Manriquez2010|An automatic multi-lead electrocardiogram segmentation algorithm based on abrupt change detection

---

Mishra et al.2021|Denoising of Electrocardiogram Signal Using S-Transform Based Time–Frequency Filtering Approach

---

KR20190019668A|2019-02-27|An apparatus for detecting r-peak in electrocardiogram signal by using wavelet transform and modified shannon energy envelope and the method thereof

同族专利:

---

公开号 | 公开日

---

NL2025147B1|2021-04-21|

---

CN110755069A|2020-02-07|

---

CN110755069B|2020-10-16|

引用文献:

---

公开号 | 申请日 | 公开日 | 申请人 | 专利标题

---

---

US5259387A|1991-09-09|1993-11-09|Quinton Instrument Company|ECG muscle artifact filter system|

---

WO2010103542A2|2009-03-09|2010-09-16|Council Of Scientific & Industrial Research|Ecg device with impulse and channel switching adc noise filter and error corrector for derived leads|

---

EP3052007A1|2013-10-01|2016-08-10|Koninklijke Philips N.V.|Device, method and system for processing a physiological signal|

---

CN108937919A|2018-06-04|2018-12-07|湖北工业大学|A method of eliminating ECG baseline drift|CN112006679B|2020-08-28|2021-08-03|山东省人工智能研究院|Wearable electrocardiosignal R wave detection method based on window variance transformation|

---

CN112487880B|2020-11-13|2021-09-24|河北省科学院应用数学研究所|Median filter, device and system based on maximum error criterion|

法律状态:

优先权:

---

申请号 | 申请日 | 专利标题

---

CN201911027439.2A|CN110755069B|2019-10-25|2019-10-25|Dynamic electrocardiosignal baseline drift correction method for jump mutation noise|

---