WO2017148452A1

WO2017148452A1 - Electrocardiography signal waveform feature point extraction method and device

Info

Publication number: WO2017148452A1
Application number: PCT/CN2017/079425
Authority: WO
Inventors: 郑慧敏
Original assignee: 深圳竹信科技有限公司
Priority date: 2016-03-03
Filing date: 2017-04-05
Publication date: 2017-09-08
Also published as: CN105748066A; CN105748066B

Abstract

An electrocardiography signal waveform feature point extraction method, comprising QRS wave point, P wave point and T wave point position determination: processing an electrocardiography signal via a smooth wavelet transform (S2), determining a QRS wave embodiment optimal target layer and a T and P wave embodiment optimal target layer (S3), finding maximum value and minimum value pairs of the corresponding target layers (S4), removing maximum value and minimum value pairs which do not meet requirements (S5), performing error detection and miss detection on an R wave point position, obtaining a final R wave point position (S7), determining Q wave point and S wave point positions (S8), and determining P wave point and T wave point positions (S9). The QRS wave point, T wave point and P wave point position determination all use the smooth wavelet transform, and compared to a discrete wavelet transform, the smooth wavelet transform effectively avoids resolution damage during scale increasing without movement deformation, thereby effectively solving the problem in the prior art.

Description

Method and device for extracting characteristic points of ECG signal waveform

Technical field

The invention relates to the technical field of medical technology, in particular to a method and a device for extracting characteristic points of an electrocardiographic signal waveform.

Background technique

There are many kinds of wearable devices for monitoring human health in the prior art, generally collecting physiological information data of the human body through sensors, and then analyzing these physiological data to achieve the purpose of detecting the human body state. Among them, how to accurately extract the feature points (PQRST) in the ECG signal during the processing of ECG signals is especially important for the automatic determination of the realization state, where P wave represents the fluctuation of the left and right new house, QRS wave represents the contraction of the ventricle, T wave Represents ventricular repolarization.

The traditional ECG signal feature point extraction method mainly adopts the linear filter method. The processing speed of this method is fast and easy to implement, but the accuracy of the method is limited due to the frequency variability of the ECG signal. The combination of filtering and differential operation methods is also used to identify QRS waves, but the operation of this method is relatively complicated. With the development of wavelet theory, the wavelet method is used to extract the characteristic points of ECG signals. The methods commonly used in the prior art to process ECG signals are continuous wavelet transform (CWT) and discrete wavelet transform (DWT). The CWT processing method generates redundancy, and the calculation amount of the CWT processing method is twice the calculation amount of the DWT processing method. Therefore, the DWT processing method is a common method in the prior art.

The DWT processing method, in the JPEG image, the discrete cosine transform is to compress the image into 8*8 small blocks. This algorithm considers discarding the frequency information to achieve compression, because the higher the compression rate, the more the frequency information is discarded. In extreme cases, the JPEG image retains only the basic information of the appearance of the reflected image, and the fine image detail image is lost, resulting in loss of resolution of the DWT processing method when the scale becomes large, and there is no movement and no deformation.

Therefore, how to avoid the loss of resolution in the process of ECG signal processing has become a technical problem to be solved by those skilled in the art.

Summary of the invention

In view of this, the present invention provides a method for extracting characteristic points of an electrocardiographic signal waveform, which is a resolution loss in the process of ECG signal processing. The invention also provides an electrocardiogram signal waveform feature point extracting device

To achieve the above object, the present invention provides the following technical solutions:

A method for extracting characteristic points of an electrocardiogram signal waveform, comprising:

Step 1): reading the original ECG signal after noise reduction;

Step 2): processing the ECG signal in the step 1) by the stationary wavelet transform, selecting an optimal wavelet base, and layering the ECG signal according to the formula N=2 ⁿ , wherein N is selected The total number of signal points, n is the total number of layers, and n is a positive integer;

Step 3): The energy distribution, frequency analysis and cross-correlations average QRS wave reflects most of the details of the target layer coefficients swdn _i, and the detail coefficients swdn _j T wave and P wave reflected optimal target layer, which is swd The detail coefficient, n _i is the target layer in which the QRS wave is optimal, n _j is the target layer in which the T wave and the P wave are optimal, and i and j are the positions of the target layer;

Step 4): determining a minimum value pair of swdn _i and swdn _j ;

Step 5): removing, by the first search window, a maximum value minimum value of swdn _i and swdn _j that exceeds a range of heights of the first search window, when removing a pole of swdn _i that exceeds a height range of the first search window When the large value is small, the height T1 of the first search window is 0.25 rms (swdn _i ^2), and when the maximum value of the swdn _j exceeding the height range of the first search window is removed, the first The height of the search window is T1=0.25 rms (swdn _j ^2), the width of the first search window is d1, and the d1 is greater than the waveform time from the maximum value to the adjacent minimum value in the original signal. ;

Removing a negative maxima minimum value of swdn _i and swdn _j beyond the second search window height range by the second search window, when the swdn _i is removed, the negative maximum value exceeding the second search window height range is extremely small When the value is correct, the height of the second search window is T2=0.4 rms (swdn _i (a)^2), when the negative maximum value of swdn _j exceeding the height range of the second search window is removed, the second The height of the search window is T2=0.4 rms (swdn _j (a)^2), the width of the second search window is d2, and the d2 is greater than the maximum value in the original signal to the adjacent minimum value Waveform time, where a∈ the detail coefficient in the second search window;

The second search window is located after the first search window is 150ms-200ms;

Step 6): Determine the position of the R wave point in swdn _i and swdn _j , and the position of the zero point of each group of the minimum value of the maximum value is the position of the R wave point:

Step 7): error detection and missed detection of swd _i and swdn _j in the R wave point position;

Step 8): determining the position of the Q wave point and the position of the S wave point in swdn _i and swdn _j according to the position of the R wave point;

Step 9): Determine the P wave and the T wave according to the position of the Q wave point and the position of the S wave point in swdn _j .

Preferably, in the method for extracting characteristic points of the electrocardiographic signal waveform, the determining step of determining the minimum value of the maximum value in swdn _i and swdn _j in the step 4) is:

When determining the maximum value point swdn _i of the image is greater than the slope swdn _i 0 is designated as 1, the slope of swdn _i less than or equal named 0 0 image, wherein the change in slope is the maximum value point;

When it is determined that the minimum value of swdn _i, the slope is less than the image swdn _i 0 designated as 1, the swdn _i slope greater than or equal to the image 0 0 named, is the point where the minimum slope change point;

When determining the maximum value of swdn _j, is greater than the slope of the image swdn _j 0 designated as 1, the slope is less than or equal to the swdnj image named 0 0, where the slope change point is the maxima ;

When the minimum value point swdn _j is determined, the image is smaller than the slope swdn _j 0 designated as 1, the swdn _j slope is greater than or equal to the image 0 0 named, which is the maximum value of the change in slope point.

Preferably, in the method for extracting characteristic points of the electrocardiographic signal waveform, the step 8) determining the position of the Q wave point and the position of the S wave point in swdn _i and swdn _j are:

Q swdn _i determined in the preceding wave position of the R-wave swdn _i point position in the third position of the extreme points determined swdn _j S wave in the foregoing for the position of the R-wave swdn _j in a third position Extreme point position;

Using the fifth search window to determine the position of the S wave point in swdn _i and swdn _j , when determining the position of the S wave point in swdn _i , the height of the third search window is T4=0.9rms (swdn _i ^2) When determining the position of the S wave point in swdn _i , the height of the third search window is T4=0.9 rms (swdn _j ^2), and the width is d4, and the d4 is greater than the maximum value in the original signal. The waveform time between adjacent minimum values, the third extreme point located in the third search frame after the R wave point position is the S wave point position.

Preferably, in the method for extracting the feature points of the electrocardiographic signal waveform, the step 7) is specifically:

When the distance between adjacent two R-wave point positions RR<0.4mean is a misdetection, the position of the R-wave point with a small extremum is removed, and when the distance between adjacent two R-wave points is RR>1.6mean For the missed detection, a pair of maximum value minimum value pairs with the largest absolute value is found between the positions of the adjacent two R wave points that are missed.

Preferably, in the method for extracting the feature points of the electrocardiographic signal waveform, the step 9) is specifically:

Taking a Q wave in swdn _j as a reference, a fourth search window of 100/512s is set forward to determine a first maximum value minimum pair in the fourth search window, the first maximum value 0 is the position of the P wave point, and a fifth search window of 100/512 s is set backward according to the S wave in swdn _j to determine a second maximum value minimum pair in the fifth search window, the first The zero point of the second maximum value pair is the T wave point position.

An apparatus for extracting characteristic points of an electrocardiogram signal waveform, comprising:

a first reading unit, configured to read the original ECG signal after noise reduction;

a first processing unit, configured to process the ECG signal in the step 1) by the stationary wavelet transform, select an optimal wavelet base, and layer the ECG signal according to the formula N=2 ⁿ , wherein, N For the total number of selected signal points, n is the total number of layers, and n is a positive integer;

a second processing unit, configured to determine, according to the energy distribution, the frequency analysis, and the average cross-correlation, the detail coefficient swdn _i of the target layer in which the QRS wave is optimal, and the detail coefficient swdn _j of the target layer in which the T wave and the P wave are optimal, Where swd is the detail coefficient, n _i is the target layer in which the QRS wave is optimal, n _j is the target layer in which the T wave and the P wave are optimal, and i and j are the positions of the target layer;

a third processing unit, configured to determine a maximum value minimum value pair of swdn _i and swdn _j ;

a fourth processing unit, configured to remove, by the first search window, a maximum value minimum value of swdn _i and swdn _j that exceeds a range of the first search window height, when the swdn _i is removed beyond the first search window height When the maximum value of the range is a minimum value, the height T1 of the first search window is 0.25 rms (swdn _i ^2), when the maximum value of the swdn _j exceeding the height range of the first search window is removed The height of the first search window is T1=0.25 rms (swdn _j ^2), the width of the first search window is d1, and the d1 is greater than the maximum value from the maximum value in the original signal. Waveform time

The fifth processing unit is configured to determine the position of the R wave point in the swdn _i and the swdn _j , and the position of the zero point of each pair of the minimum value of the maximum value is the position of the R wave point:

a sixth processing unit for error detection and miss detection of the R wave point positions in swdn _i and swdn _j ;

a seventh processing unit, configured to determine a Q wave point position and an S wave point position in swdn _i and swdn _j according to the R wave point position;

The eighth processing unit is configured to determine the P wave and the T wave according to the Q wave point position and the S wave point position in swdn _j .

The image is preferably, in the above-described extraction device ECG waveform characteristic points, the third processing unit for, when it is determined that the maximum value of swdn _i, the slope is greater than 0 swdn _i are designated as 1, the The image whose slope is less than or equal to 0 in swdn _i is named 0, and the point where the slope changes is the maximum point;

When the minimum value of swdn _j is determined, the image whose slope is less than 0 in swdn _j is named 1, and the image whose slope is greater than or equal to 0 in swdn _j is named 0, and the point where the slope changes is the maximum value. point.

Preferably, in the extracting device for the characteristic points of the electrocardiographic signal waveform, the sixth processing unit is configured to detect a difference when the distance between the adjacent two R wave point positions is RR<0.4mean, and remove the minimum value. The position of the R wave point is a miss detection when the distance between adjacent two R wave points is RR>1.6mean, and a maximum absolute value is found between the positions of two adjacent R wave points that are missed. The value is a minimum value pair.

Preferably, in the device of the ECG waveform extracting feature points, the processing unit for determining a seventh swdn _i Q wave in the third position in front of the extreme points of the R-wave swdn _i point position position determining swdn _j is the S-wave position swdn _j preceding R-wave in the third extremum point position position;

Preferably, in the extracting device of the electrocardiographic signal waveform feature point, the eighth processing unit is configured to determine a fourth search window of 100/512s forward by using a Q wave in swdn _j as a reference. a first maximum value minimum pair in the four search window, wherein the zero point of the first maximum value pair is the P wave point position, and the S wave in the swdn _j is used as a reference, and a 100/512 s is set backward. The fifth search window determines a second maximum value minimum pair in the fifth search window, and the zero point of the second maximum value minimum pair is a T wave point position.

It can be seen from the above technical solution that the method for extracting the characteristic points of the ECG signal waveform provided by the present invention includes determining the position of the QRS wave point, the position of the P wave point, and the position of the T wave point. Determination process of QRS wave position: After reading the original ECG signal after noise reduction, the ECG signal is processed by stationary wavelet transform, the optimal wavelet base is selected and the ECG signal is layered, according to energy distribution, frequency analysis and average The cross-correlation determines that the QRS wave reflects the optimal target layer's detail coefficient swdn _i , and the T and P waves represent the optimal target layer's detail coefficient swdn _j , and then determines the maximum value of swdn _i and swdn _j , screened by the first and second search window of the search window is removed and swdn _j swdn _i do not meet the requirements for the maximum value minimum value, the R-wave is determined and the position swdn _i swdn _j, and for the swdn _i and swdn _j The position of the R wave point is misdetected and missed, the final R wave position is determined, and the positions of the Q wave point and the S wave point of swdn _i and swdn _j are determined according to the R wave position, thereby realizing swdn _i and swdn _j . The determination of the position of the QRS wave point. swdn _j and P-wave position T wave point position determination: swdn _j by determining the P wave of the QRS o'clock position and T waves position. The determination of the position of the QRS wave point, the position of the T wave point and the position of the P wave point are all determined by the stationary wavelet transform. The stationary wavelet transform can effectively avoid the damage of the resolution when the scale becomes large, and the movement does not deform. , effectively solve the problems existing in the prior art.

DRAWINGS

In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the embodiments or the description of the prior art will be briefly described below. Obviously, the drawings in the following description are only It is a certain embodiment of the present invention, and other drawings can be obtained from those skilled in the art without any creative work.

FIG. 1 is a flowchart of a method for extracting feature points of an electrocardiogram signal according to an embodiment of the present invention.

detailed description

The invention discloses a method for extracting characteristic points of an electrocardiographic signal waveform, which has a resolution loss in the process of ECG signal processing.

The technical solutions in the embodiments of the present invention are clearly and completely described in the following with reference to the accompanying drawings in the embodiments of the present invention. It is obvious that the described embodiments are only a part of the embodiments of the present invention, but not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative efforts are within the scope of the present invention.

Please refer to FIG. 1. FIG. 1 is a flowchart of a method for extracting feature points of an electrocardiogram signal according to an embodiment of the present invention.

The invention discloses a method for extracting characteristic points of an electrocardiogram signal waveform, comprising:

Step 1): reading the original ECG signal after noise reduction;

Step 2): processing the ECG signal in step 1) by stationary wavelet transform, selecting an optimal wavelet base, and layering the ECG signal according to the formula N=2 ⁿ , where N is the total number of selected signal points , n is the total number of layers, and n is a positive integer;

Step 3): determining, according to the energy distribution, the frequency analysis and the average cross-correlation, the detail coefficient swdn _i of the target layer in which the QRS wave is optimal, and the detail coefficient swdn _j of the target layer in which the T wave and the P wave are optimal, wherein swd is The detail coefficient, n _i is the target layer in which the QRS wave is optimal, n _j is the target layer in which the T wave and the P wave are optimal, and i and j are the positions of the target layer, assuming i=5, and the corresponding target layer is the first 5 layers, assuming j=6, the corresponding target layer is the 7th layer, i and j represent the position of the layer, not the total number of layers;

Step 4): determining a minimum value pair of swdn _i and swdn _j ;

Step 5): removing the maximum value of the swdn _i and swdn _j beyond the first search window height range by using the first search window, and removing the maximum value of the swdn _i beyond the first search window height range When the value is correct, the height of the first search window is T1=0.25 rms (swdn _i ^2). When the maximum value of the swdn _j exceeding the maximum value of the first search window is removed, the height of the first search window is T1= 0.25 rms (swdn _j ^ 2), the width of the first search window is d1, and d1 is greater than the waveform time from the maximum value to the adjacent minimum value in the original signal;

Removing a negative maxima minimum value of swdn _i and swdn _j beyond the second search window height range by the second search window, when the swdn _i is removed, the negative maximum value exceeding the second search window height range is extremely small When the value is correct, the height of the second search window is T2=0.4 rms (swdn _i (a)^2), when the negative maximum value of swdn _j exceeding the height range of the second search window is removed, the second The height of the search window is T2=0.4rms (swdn _j (a)^2), and the width of the second search window is d2, which is greater than the waveform time from the maximum value to the adjacent minimum value in the original signal, where a细节 the detail coefficient in the second search window;

Step 7): False detection and miss detection swd _i and swdn _j R position of the wave point, when the distance between adjacent two R wave points is RR<0.4mean, it is wrong detection, and the R wave point with small extremum is removed. Position, when the distance between adjacent two R wave points is RR>1.6mean, it is missed detection, and between the two adjacent R wave points of the missed detection, a pair of absolute maximum values are found. Correct;

In this scheme, the stationary wavelet transform is used to process the ECG signal instead of the process of processing the ECG signal by discrete wavelet transform in the prior art, which can effectively solve the problem that the existing ECG signal is processed by discrete wavelet transform without movement and deformation. The problem of loss of resolution caused by the increase in size has improved the processing effect of the ECG signal to some extent.

The determination of the maxima minimum value pair needs to be performed at the corresponding layer. It should be noted that the maxima minimum value pair includes a maximum value and a minimum value.

The criterion for removing the minimum value of the minimum value exceeding the height range of the first search window in step 5) is to remove the maximum value when the absolute value greater than T1 or the minimum value is greater than one of the conditions of T1. The value is a minimum value pair.

The first search window is used to remove the maximum value minimum value pair whose extreme value exceeds the threshold T1 range, and the second search window is used to remove the negative maximum value minimum value pair, and cooperate with the second search window through the first search window. Screening can ensure the location of the determined R wave points is more accurate and improve the accuracy of ECG signal detection.

The height calculation formula of the first search window uses all the detail coefficients of the target layer, and the height calculation formula of the second search window uses the detail coefficient located in the second search window.

The determining step of determining the minimum value of the maximum value in swdn _i and swdn _j in step 4) is:

When determining the maximum value swdn _i, the slope is greater than the image swdn _i 0 designated as 1, swdn _i is less than or equal to the slope of the image 0 0 named, is the point where the change in slope maxima;

When it is determined that the minimum value swdn _i, the slope is less than the image swdn _i 0 is designated as 1, swdn _i is greater than or equal to the slope of the image is named 0 0, the point where the slope change point is the minimum value;

When determining the maximum value swdn _j, larger than the image slope swdn _j 0 designated as 1, swdnj image slope less than or equal to 0 is designated 0, is the point where the change in slope maxima;

When it is determined that the minimum value swdn _j, the slope is less than the image swdn _j 0 designated as 1, swdn _j greater than or equal to the slope of the image is named 0 0, which is the point of change of slope maxima.

The existence of a maximum value minimum value pair is determined by the change point of the slope.

Step 8) Determine the specific steps of the Q wave point position and the S wave point position in swdn _i and swdn _j as follows:

Q swdn _i determined in the preceding wave position of the R-wave swdn _i extrema point position of the third point position determined swdn _j S wave in the foregoing for the position swdn _j R wave in the third position extrema position;

Using the fifth search window to determine the position of the S wave point in swdn _i and swdn _j , when determining the position of the S wave point in swdn _i , the height of the third search window is T4=0.9rms (swdn _i ^2), when determining swdn When the position of the S wave point in _i is, the height of the third search window is T4=0.9rms (swdn _j ^2), and the width is d4, and d4 is larger than the waveform time from the maximum value to the adjacent minimum value in the original signal. The third extreme point located in the third search box after the R wave position is the S wave point position.

Step 7) is specifically:

The accuracy of the determination of the position of the R wave point is further improved by the misdetection and the missed detection operation.

Step 9) is specifically:

Based on the Q wave in swdn _j , a fourth search window of 100/512s is set forward to determine a first maximum value pair in the fourth search window, and the zero point of the first maximum value pair is P wave position, with the S wave in swdn _j as the reference, a fifth search window of 100/512s is set backward to determine the second maximum value pair in the fifth search window, and the second maximum value is The 0 point is the T wave point position.

The fourth search window and the fifth search window have only the width and no height, and the maximum value minimum value pair here is not necessarily the maximum value minimum value pair, and the purpose is to find the nearest P wave point position and the distance S wave nearest to the Q wave. The position of the T wave point. What needs to be explained here is that the maximum value minimum pair includes a maximum value and a minimum value.

a first processing unit, configured to process the ECG signal in the step 1) by the stationary wavelet transform, select an optimal wavelet base, and layer the ECG signal according to the formula N=2 ⁿ , where N is the selected signal The total number of points, n is the total number of layers, and n is a positive integer;

a fourth processing unit, configured to remove, by the first search window, a maximum value minimum value of swdn _i and swdn _j that exceeds a first search window height range, when removing swdn _i exceeds a maximum of a first search window height range When the value is a minimum value, the height of the first search window is T1=0.25 rms (swdn _i ^2), when the maximum value of the swdn _j exceeding the height range of the first search window is removed, the first search window The height T1=0.25 rms (swdn _j ^2), the width of the first search window is d1, and d1 is greater than the waveform time from the maximum value to the adjacent minimum value in the original signal;

In this scheme, the stationary wavelet transform is used to process the ECG signal instead of the process of processing the ECG signal by discrete wavelet transform in the prior art, which can effectively solve the problem that the existing ECG signal is processed by discrete wavelet transform without movement and deformation. The problem of loss of resolution caused by getting bigger To a certain extent, the processing effect of the ECG signal is improved.

The criterion for removing the maximum value minimum value pair exceeding the first search window height range in the fourth processing unit is to remove the set pole when the absolute value greater than T1 or the minimum value of the minimum value is greater than one of the conditions of T1. Large value minimum value pair.

To further optimize the above technical solutions, the feature point extracting means ECG waveform provided by the invention, the third processing unit for, when it is determined that the maximum value swdn _i, the slope is greater than the image swdn _i 0 1 named The image whose slope is less than or equal to 0 in swdn _i is named 0, and the point where the slope changes is the maximum point;

The sixth processing unit is configured to perform a misdetection when the distance between the two adjacent R wave point positions is RR<0.4mean, and remove the R wave point position with a small extremum, when the position between the adjacent two R wave points is A miss detection is performed when the distance is RR>1.6mean, and a pair of maximum value minimum values having the largest absolute value is found between the positions of the adjacent two R wave points that are missed.

Seventh processing means for determining a point Q wave swdn _i in position in front of the R-wave swdn _i extrema point position of the third point position determined swdn _j S wave in front of the point point position position swdn _j R-wave The third extreme point position;

The eighth processing unit is configured to set a first search window of 100/512s forward by using a Q wave in swdn _j as a reference, and determine a first maximum value minimum pair in the fourth search window, the first maximum value minimum pair The 0 point is the position of the P wave point, and the S wave of swdn _j is used as a reference, and a fifth search window of 100/512s is set backward to determine the second maximum value pair in the fifth search window, the second largest The 0 point of the value minimum pair is the T wave point position.

The invention is illustrated by a specific embodiment:

The total number of selected signal points is N=4096, and the number of decomposition layers obtained according to the formula N=2 ⁿ is 12 layers;

If the length of the wavelet base is too large, it will be detrimental to real-time performance. The poor similarity of the signal to be analyzed will cause the original signal to have distortion after reconstruction, and consider the symmetry of the wavelet base. If the symmetry is not good, it is easy to cause The original signal has a phase shift after reconstruction. In summary, the wavelet base db4 is selected, and its symmetry is good, and it has certain similarity with the P-QRS-T wave of the ECG signal.

According to the energy distribution, frequency analysis and average mutuality, the fifth layer swd5 is selected for QRS wave extraction, and the sixth layer swd6 is used for T wave and P wave extraction.

The reason why the swd5 detail coefficient is selected to obtain the QRS wave:

1) Energy distribution, since the energy of ECG (electrocardiogram signal) is mainly concentrated on the QRS complex, correspondingly, after signal decomposition, QRS wave will be most obvious in the layer with the highest energy, and QRS can be obtained by MATLAB software. The wave group is most obvious in swd5;

2) Frequency analysis, according to the frequency range of QRS waveform in ECG, its energy is mainly concentrated in the range of 0~38Hz, the peak is concentrated at 10~20Hz, the center frequency is about 17Hz, the bandwidth is about 10Hz, and the detail coefficient of swd5 is Fourier. Transform, find that the center frequency of swd5 is also close to the frequency range of the QRS complex;

3) The average cross-correlation, according to the calculation can also get the best correlation between the detail coefficient of swd5 and the original ECG signal.

It can be seen from the above that swd5 has the greatest correlation with QRS waves in the energy distribution, frequency domain and time domain.

Reasons for choosing the swd6 detail factor:

1) Energy distribution, after signal decomposition, T wave and P wave are most obvious in swd6;

2) Frequency analysis, Fourier transform is performed on the detail coefficient of swd6, and it is found that the center frequency of swd6 is also close to the frequency range of T wave and P wave;

3) The average cross-correlation, according to the calculation can also get the best correlation between the detail coefficient of swd6 and the original ECG signal.

It can be seen from the above that swd6 has the highest correlation with T wave and P wave in energy distribution, frequency domain and time domain.

Find the swd5 maxima minimum value pair, which contains a maximal value and a minimum value. There are many ways to obtain the maximum value of the minimum value. In this scheme, the maximum value of the minimum value is obtained by the slope of the image. Specifically, the image with the slope greater than 0 in swd5 is named 1, and the slope in swd5 is less than or equal to 0. The image is named 0, and the point where the slope changes is the point where the slope changes from 1 to 0 or from 0 to 1 is the maximum point; the image with the slope less than 0 in swd5 is named 1, and the slope in swd5 is greater than or equal to The image of 0 is named 0, and the point at which the slope changes is the point at which the slope changes from 1 to 0 or from 0 to 1 is a minimum point.

Find the swd6 maxima minimum value. The image with a slope greater than 0 in swd6 is named 1. The image with a slope less than or equal to 0 in swd6 is named 0. The slope change point is the slope from 1 to 0 or The point where 0 becomes 1 is the maximum point; the image with the slope less than 0 in swd6 is named 1, and the image with the slope greater than or equal to 0 in swd6 is named 0, and the point where the slope changes is the slope changes from 1 to 0 or The point from 0 to 1 is the minimum point.

A part of the maximum value minimum value of swd5 and swd5 is removed by the first search window, and the removal principle is that the maximum value is larger than the threshold T1, and the absolute value of the minimum value is larger than T1, and the maximum value is extremely small. Value pairs can be removed as long as one of the conditions is met.

The inverse of the maximum value of swd5 and swd5 is removed through the second search window.

The second search window appears after the first search window is 150ms-200ms. Preferably, the second search window appears after the first search window 200ms, and the width of the first search window and the second search window is 40mm.

After step 4), the maximum value of swd5 and swd5 is determined. The maximum value and the minimum value of each group are 0 points, that is, the R wave position of the corresponding layer.

In order to avoid losing the R wave position of swd5 and swd5, and avoiding redundant R wave position of swd5 and swd5, it is necessary to perform error detection and miss detection on the R wave position in step 6), among which the error detection and leakage The principle of detection is that when the distance between adjacent two R wave points is RR<0.4mean, the error is detected, and the position of the R wave point with a small extremum is removed, and the distance between the positions of two adjacent R wave points is RR. >1.6mean is the missed detection, and a pair of absolute maximum value pairs with the largest absolute value is found between the positions of the adjacent two R wave points that are missed. The missed detection and the misdetection of the R-wave point position are not limited to the above methods, and may be other implementations that can be realized by those skilled in the art, and are not specifically limited herein. Where mean is the average of the RR.

Determining the Q wave point position and the S wave point position of swd5 and swd5, since the extreme point position of the R wave front has been recorded during the sliding process of the first search window and the second search window, the third of the R wave front The position of the extreme point is the position of the Q wave point; the determination of the position of the S wave point needs to be obtained through the third search window, and the third extreme point in the third search window is the position of the S wave point. In the process of determining the position of the Q wave point and the position of the S wave point, the third extreme point is either a maximum point or a minimum point.

It should be noted here that the determination of the QRS wave on swd6 is not a true QRS wave, and the position of the QRS wave detected by the layer is used to locate the T wave and the P wave position.

The swd6 hung T wave point position and the P wave point position are determined by the fourth search window and the fifth search window, and only one set of the maximum value minimum pair exists in the sliding range of the fourth search window and the fifth search window, wherein each group The maximum value minimum pair includes a maximum value and a minimum value, and the 0 point between the adjacent maximum value and the minimum value in the fourth search window is the T wave point position, and the adjacent maximum value and the minimum value in the fifth search window The 0 point between the values is the P wave point position.

The ECG waveform waveform feature point extraction process ends.

According to the general rule of the human body ECG signal, the widths of the first search window, the second search window and the third search window are 40 ms, which can meet the requirements.

Preferably, the second search window is located 200 ms after the first search window.

The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments are obvious to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but the scope of the invention is to be accorded

Claims

A method for extracting characteristic points of an electrocardiographic signal waveform, comprising:

Step 1): reading the original ECG signal after noise reduction;

Step 2): processing the ECG signal in the step 1) by the stationary wavelet transform, selecting an optimal wavelet base, and layering the ECG signal according to the formula N=2 n , wherein N is selected The total number of signal points, n is the total number of layers, and n is a positive integer;

Step 3): determining, according to the energy distribution, the frequency analysis and the average cross-correlation, the detail coefficient swdn i of the target layer in which the QRS wave is optimal, and the detail coefficient swdn j of the target layer in which the T wave and the P wave are optimal, wherein swd is The detail coefficient, n i is the target layer in which the QRS wave is optimal, n j is the target layer in which the T wave and the P wave are optimal, and i and j are the positions of the target layer;

Step 4): determining a minimum value pair of swdn i and swdn j ;

Step 5): removing, by the first search window, a maximum value minimum value of swdn i and swdn j that exceeds a range of heights of the first search window, when removing a pole of swdn i that exceeds a height range of the first search window When the large value is small, the height T1 of the first search window is 0.25 rms (swdn i ^2), and when the maximum value of the swdn j exceeding the height range of the first search window is removed, the first The height of the search window is T1=0.25 rms (swdn j ^2), the width of the first search window is d1, and the d1 is greater than the waveform time from the maximum value to the adjacent minimum value in the original signal. ;

Removing a negative maxima minimum value of swdn i and swdn j beyond the second search window height range by the second search window, when the swdn i is removed, the negative maximum value exceeding the second search window height range is extremely small When the value is correct, the height of the second search window is T2=0.4 rms (swdn i (a)^2), when the negative maximum value of swdn j exceeding the height range of the second search window is removed, the second The height of the search window is T2=0.4 rms (swdn j (a)^2), the width of the second search window is d2, and the d2 is greater than the maximum value in the original signal to the adjacent minimum value Waveform time, where a∈ the detail coefficient in the second search window;

The second search window is located after the first search window is 150ms-200ms;

Step 6): Determine the position of the R wave point in swdn i and swdn j , and the position of the zero point of each group of the minimum value of the maximum value is the position of the R wave point:

Step 7): error detection and missed detection of swd i and swdn j in the R wave point position;

Step 8): determining the position of the Q wave point and the position of the S wave point in swdn i and swdn j according to the position of the R wave point;

Step 9): Determine the P wave and the T wave according to the position of the Q wave point and the position of the S wave point in swdn j .
The method for extracting characteristic points of an electrocardiographic signal according to claim 1, wherein the determining step of determining a minimum value pair of swdn i and swdn j in the step 4) is:

When determining the maximum value point swdn i of the image is greater than the slope swdn i 0 is designated as 1, the slope of swdn i less than or equal named 0 0 image, wherein the change in slope is the maximum value point;

When it is determined that the minimum value of swdn i, the slope is less than the image swdn i 0 designated as 1, the swdn i slope greater than or equal to the image 0 0 named, is the point where the minimum slope change point;

When determining the maximum value of swdn j, is greater than the slope of the image swdn j 0 designated as 1, the slope is less than or equal to the swdnj image named 0 0, where the slope change point is the maxima ;

When the minimum value point swdn j is determined, the image is smaller than the slope swdn j 0 designated as 1, the swdn j slope is greater than or equal to the image 0 0 named, which is the maximum value of the change in slope point.
The method for extracting characteristic points of an electrocardiographic signal according to claim 1, wherein

The specific steps of the step 8) determining the position of the Q wave point and the position of the S wave point in swdn i and swdn j are:

Q swdn i determined in the preceding wave position of the R-wave swdn i point position in the third position of the extreme points determined swdn j S wave in the foregoing for the position of the R-wave swdn j in a third position Extreme point position;

Using the fifth search window to determine the position of the S wave point in swdn i and swdn j , when determining the position of the S wave point in swdn i , the height of the third search window is T4=0.9rms (swdn i ^2) When determining the position of the S wave point in swdn i , the height of the third search window is T4=0.9 rms (swdn j ^2), and the width is d4, and the d4 is greater than the maximum value in the original signal. The waveform time between adjacent minimum values, the third extreme point located in the third search frame after the R wave point position is the S wave point position.
The method for extracting characteristic points of an electrocardiogram signal according to claim 1, wherein the step 7) is specifically:

When the distance between adjacent two R-wave point positions RR<0.4mean is a misdetection, the position of the R-wave point with a small extremum is removed, and when the distance between adjacent two R-wave points is RR>1.6mean For the missed detection, a pair of maximum value minimum value pairs with the largest absolute value is found between the positions of the adjacent two R wave points that are missed.
The method for extracting characteristic points of an electrocardiogram signal according to claim 1, wherein the step 9) is specifically:

Taking a Q wave in swdn j as a reference, a fourth search window of 100/512s is set forward to determine a first maximum value minimum pair in the fourth search window, the first maximum value 0 is the position of the P wave point, and a fifth search window of 100/512 s is set backward according to the S wave in swdn j to determine a second maximum value minimum pair in the fifth search window, the first The zero point of the second maximum value pair is the T wave point position.
An apparatus for extracting characteristic points of an electrocardiographic signal waveform, comprising:

a first reading unit, configured to read the original ECG signal after noise reduction;

a first processing unit, configured to process the ECG signal in the step 1) by the stationary wavelet transform, select an optimal wavelet base, and layer the ECG signal according to the formula N=2 n , wherein, N For the total number of selected signal points, n is the total number of layers, and n is a positive integer;

a second processing unit, configured to determine, according to the energy distribution, the frequency analysis, and the average cross-correlation, the detail coefficient swdn i of the target layer in which the QRS wave is optimal, and the detail coefficient swdn j of the target layer in which the T wave and the P wave are optimal, Where swd is the detail coefficient, n i is the target layer in which the QRS wave is optimal, n j is the target layer in which the T wave and the P wave are optimal, and i and j are the positions of the target layer;

a third processing unit, configured to determine a maximum value minimum value pair of swdn i and swdn j ;

a fourth processing unit, configured to remove, by the first search window, a maximum value minimum value of swdn i and swdn j that exceeds a range of the first search window height, when the swdn i is removed beyond the first search window height When the maximum value of the range is a minimum value, the height T1 of the first search window is 0.25 rms (swdn i ^2), when the maximum value of the swdn j exceeding the height range of the first search window is removed The height of the first search window is T1=0.25 rms (swdn j ^2), the width of the first search window is d1, and the d1 is greater than the maximum value from the maximum value in the original signal. Waveform time

Removing a negative maxima minimum value of swdn i and swdn j beyond the second search window height range by the second search window, when the swdn i is removed, the negative maximum value exceeding the second search window height range is extremely small When the value is correct, the height of the second search window is T2=0.4 rms (swdn i (a)^2), when the negative maximum value of swdn j exceeding the height range of the second search window is removed, the second The height of the search window is T2=0.4 rms (swdn j (a)^2), the width of the second search window is d2, and the d2 is greater than the maximum value in the original signal to the adjacent minimum value Waveform time, where a∈ the detail coefficient in the second search window;

The second search window is located after the first search window is 150ms-200ms;

The fifth processing unit is configured to determine the position of the R wave point in the swdn i and the swdn j , and the position of the zero point of each pair of the minimum value of the maximum value is the position of the R wave point:

a sixth processing unit for error detection and miss detection of the R wave point positions in swdn i and swdn j ;

a seventh processing unit, configured to determine a Q wave point position and an S wave point position in swdn i and swdn j according to the R wave point position;

The eighth processing unit is configured to determine the P wave and the T wave according to the Q wave point position and the S wave point position in swdn j .
The ECG waveform feature extracting means the point of claim 6, wherein the third processing means for, when it is determined that the maximum value of swdn i, is greater than the slope of the image swdn i 0 named Is 1, the image in the swdn i with a slope less than or equal to 0 is named 0, wherein the point where the slope changes is a maximum point;

When it is determined that the minimum value of swdn i, the slope is less than the image swdn i 0 designated as 1, the swdn i slope greater than or equal to the image 0 0 named, is the point where the minimum slope change point;

When determining the maximum value of swdn j, is greater than the slope of the image swdn j 0 designated as 1, the slope is less than or equal to the swdnj image named 0 0, where the slope change point is the maxima ;

When the minimum value point swdn j is determined, the image is smaller than the slope swdn j 0 designated as 1, the swdn j slope is greater than or equal to the image 0 0 named, which is the maximum value of the change in slope point.
The device for extracting characteristic points of an electrocardiographic signal according to claim 6, wherein the sixth processing unit is configured to detect a mismatch when the distance between adjacent two R-wave point positions is RR<0.4mean. , removing the R wave point position with a small extremum, when the distance between the adjacent two R wave point positions RR>1.6mean is a missed detection, and finding a pair of absolute between the two adjacent R wave point positions of the missed detection The largest value of the largest value of the minimum value pair.
The ECG waveform feature extracting means the point of claim 6, wherein said processing means for determining a seventh swdn i of the Q-wave position swdn i preceding the R-wave in the position of the point extremum position three, the determination swdn j S wave point position of said preceding R-wave swdn j third extremum point position position;

Using the fifth search window to determine the position of the S wave point in swdn i and swdn j , when determining the position of the S wave point in swdn i , the height of the third search window is T4=0.9rms (swdn i ^2) When determining the position of the S wave point in swdn i , the height of the third search window is T4=0.9 rms (swdn j ^2), and the width is d4, and the d4 is greater than the maximum value in the original signal. The waveform time between adjacent minimum values, the third extreme point located in the third search frame after the R wave point position is the S wave point position.
The apparatus for extracting characteristic points of an electrocardiogram signal according to claim 6, wherein the eighth processing unit is configured to set a fourth search of 100/512s forward based on the Q wave in swdn j . The window determines a first maximum value minimum value pair in the fourth search window, wherein the zero point of the first maximum value minimum value is a P wave point position, and is set backward according to the S wave in swdn j A fifth search window of 100/512s determines a second maximum value minimum pair in the fifth search window, and the zero point of the second maximum value minimum pair is the T wave point position.