TW202341932A - Method for separating high-frequency noise from an electrocardiogram - Google Patents

Abstract

A method for separating high-frequency noise from an electrocardiogram is adapted to prevent the electrocardiogram from being affected by the high-frequency noise. The method is to receive the heartbeat signal of the electrocardiogram and capture a QRS wave of the heartbeat signal. A high-frequency auxiliary signal is added in the captured QRS wave additionally to thereby increase the high-frequency signal of the QRS wave. The empirical mode decomposition is repeated to proceed a mode of a filtering procedure to thereby decompose the signal by the empirical mode decomposition and obtain a plurality of intrinsic mode functions. The first intrinsic mode function which represents the high-frequency component is adapted to minus the high-frequency auxiliary signal after being decomposed to thereby separate the high-frequency noise from the QRS wave. Finally, the high-frequency noise is removed from the QRS wave to thereby improve the accuracy of analysis and discrimination effectively.

Description

A method for separating high-frequency noise from electrocardiogram

The present invention relates to a design for detecting heartbeat signals, and in particular, to a method for separating high-frequency noise from electrocardiograms.

According to the investigation, in view of the advancement of science and technology and the improvement of living standards, modern people have gradually paid attention to the prevention and detection of major diseases. However, due to different social patterns and life paces, disease types may also change from acute to chronic, such as heart disease. diseases, cerebrovascular diseases, etc., therefore, thanks to the advancement of technology, through regular measurement and analysis of related physiological signals, the early signs of many major diseases can be detected, and for one of the major diseases As for heart disease, sudden cardiac death is a common cause, and sudden cardiac death often occurs when acute myocardial infarction is complicated by malignant arrhythmia. People with ventricular arrhythmia, especially those with ventricular arrhythmia, are in a high-risk group for myocardial infarction. Therefore, if Signs of ventricular arrhythmia can be correctly diagnosed before the event occurs, so that patients can receive appropriate treatment in advance, which helps to receive early treatment to improve the cure rate.

Therefore, electrocardiogram is now widely used to track heart disease-related conditions, and electrocardiogram is an important non-invasive tool that can detect cardiac abnormalities early. When abnormalities appear in the electrocardiogram, it indicates that there may be some kind of disorder in the heart function. Therefore, electrocardiogram is often used to diagnose heart rate or nerve conduction problems, including arrhythmia, conduction block, accelerated conduction, etc., as well as myocardial current changes, including myocardial infarction, atrial or ventricular hypertrophy, pericarditis, etc., while electrocardiogram It can also be used for general heartbeat monitoring, or with exercise electrocardiography to analyze whether there are abnormal changes in coronary artery hypoxia, or using 24-hour electrocardiography to check for occasional arrhythmias. However, electrocardiography is easy to measure. It is interfered by various noises, such as power interference, baseline drift, poor electrode contact, human interference, etc. In view of the aforementioned interference noise mixed in the ECG, the ECG will be affected during analysis. It is impossible to correctly detect and judge whether the heartbeat signal of the person being tested is an abnormal heartbeat signal, and whether the person being tested has signs of arrhythmia; therefore, how to reduce the impact of noise on the electrocardiogram analysis, so as to effectively and quickly determine whether the person being tested is The location of the patient's disease is the goal of concerted efforts by those in the technical field.

Therefore, the purpose of the present invention is to provide a method for separating high-frequency noise in electrocardiograms, which can effectively separate high-frequency noise in electrocardiograms, so as to reduce the impact of high-frequency noises on electrocardiogram analysis and effectively improve analysis. Accuracy of discrimination.

Therefore, the present invention is a method for separating electrocardiogram high-frequency noise, including the steps of receiving, retrieving, adding, separating and removing. Therefore, the heartbeat signal from an electrocardiograph is received, the QRS wave in the heartbeat signal is captured, and the QRS wave in the heartbeat signal is adjusted. The sampling frequency of the QRS wave is to add an additional high-frequency auxiliary signal to the QRS wave, thereby increasing the high-frequency signal component in the QRS wave. At the same time, the empirical mode is used for the QRS wave with the high-frequency auxiliary signal added. The decomposition method is used to decompose the signal and adopt the pattern of repeated screening procedures to decompose the signal into an empirical mode decomposition and obtain multiple essential mode functions, and then decompose the first one representing the high-frequency component. By subtracting the high-frequency auxiliary signal from an intrinsic modal function, the high-frequency noise of the QRS wave can be separated. Finally, the high-frequency noise is removed from the QRS wave, which can effectively greatly improve the accuracy of electrocardiogram analysis and discrimination.

The aforementioned and other technical contents, features and effects of the present invention will be clearly understood in the following detailed description of the preferred embodiments with reference to the drawings.

Referring to Figure 1, a preferred embodiment of a method for separating electrocardiogram high-frequency noise according to the present invention includes a receiving step, an acquisition step, an adjusting step, an adding step, a separating step and a removing step, etc.; Among them, in the receiving step, the heartbeat signal will first be received from an electrocardiograph (receiving step); and in the acquisition step, the QRS waveform will be found in the heartbeat signal and acquired. In view of the fact that a complete heartbeat signal usually It consists of dividing the waveform into P, Q, R, S and T parts, which is helpful to judge the heart rhythm condition. At the same time, its position has different representative meanings. The most common potential change is the QRS wave group, and the QRS waveform is not It is in a fixed order, and the abnormal potential frequency of the QRS waveform is often mixed with various interfering noise components. Especially the high-frequency noise mixed in it will affect the analysis and judgment, so the QRS waveform is specially captured. .

Continuing from the above, in the adding step, an additional high-frequency auxiliary signal is added to the QRS wave captured in the capturing step, so that the high-frequency signal component in the QRS wave is increased, especially in this implementation. In this example, the added high-frequency auxiliary signal can be a single frequency or a combination of multiple frequencies, which will not be described in detail here. At the same time, before adding the high-frequency auxiliary signal to the QRS wave, an additional adjustment step is added. , in order to adjust the sampling frequency of the QRS wave first, so that the adjusted sampling frequency of the QRS wave is an even multiple of the frequency of the high-frequency auxiliary signal, so that through the adjustment of the sampling frequency and the increase of the high-frequency auxiliary signal, it can The high-frequency component of the QRS wave with small amplitude that is not easily detected is effectively increased; please refer to Figure 2. As for the separation step, empirical mode decomposition will be used to signal the QRS wave with the aforementioned high-frequency auxiliary signal. Decomposition, and the empirical mode decomposition (EMD) uses a repeated screening process to find the intrinsic mode function (hereinafter referred to as IMF) in the signal, and in this screening The process of the program includes a finding process, a connecting process, a calculating process, a subtracting process, a checking process, a screening process and a disassembling process, that is, taking the original signal x ( n ) as an example, in the finding In the process, all the local maxima and local minima are found in the original signal of the QRS wave that has been adjusted and added, and in the connection process, all the local maxima and local minima are divided into cubic splines. (cubic spline) curve is connected into an upper envelope and a lower envelope, and then the calculation process will calculate the average of the aforementioned upper envelope and lower envelope to obtain the average envelope m ₁ ( n ), and then subtract the In the process, the average envelope is subtracted from the original signal to obtain the first component h ₁ ( n ) parameter, that is, the relationship is:

h ₁ ( n ) = x ( n ) - m ₁ ( n )

Continuing with the above, and at the same time, through the inspection process, the first component h ₁ ( n ) parameter is checked to see if it meets the conditions of the IMF. If it does not meet the conditions, the first component will be regarded as the original signal, and the screening will be repeated multiple times. Multiple component parameters with IMF conditions can be obtained. Therefore, if they are non-compliant during the inspection process after the subtraction process, the aforementioned four screening processes of finding, connecting, calculating, and subtracting are repeated, and The first component parameter is regarded as the original signal and is filtered for the second time to further obtain the second component h ₂ ( n ) parameter. The relationship formula is:

h ₁ ( n ) = x ( n ) - m ₁ ( n )

Continuing from the above, after repeated screening k times, the k-th component parameter that satisfies the IMF will be obtained, that is, the relationship is:

h _k ( n ) = h _{k -1} ( n ) - m _k ( n )

Then it is defined as the first IMF, c ₁ ( n ), that is, the relationship is:

c ₁ ( n ) = h ₁ ( n )

Therefore, the IMF condition used in the present invention is that the standard deviation of the k-th component and the two components before and after must be less than the threshold value 0.001. The standard deviation (SD) is defined as follows:

SD= ；where N is the length of the original signal.

In this screening process, the first IMF is subtracted from the original signal to obtain the first remaining signal r ₁ ( n ), that is, the relationship is:

r ₁ ( n ) = x ( n ) - c ₁ ( n )

And treat r ₁ ( n ) as the original signal, repeat the aforementioned screening process of finding, connecting, calculating, subtracting, checking to the screening, etc. 1 to 6, to obtain the second remaining signal r ₂ ( n ), That is, the relationship is:

r ₂ ( n ) = r ₁ ( n ) - c ₂ ( n )

After repeated screening n times, the I -th signal r _I ( n ) remains, that is, the relationship is:

r _I ( n ) = r _{I - 1} ( n ) - c _I ( n )

It can be known from the aforementioned relationship that after screening, it has become a monotonic function (monotonic function), that is, the number of extreme values is less than 3, and other IMFs cannot be decomposed and screened out, that is, the entire empirical mode decomposition is completed; and finally in this disassembly The original signal in the process can be disassembled into I IMF and 1 trend function, and its relationship can be expressed as:

x ( n )=

Then the high-frequency auxiliary signal is subtracted from the first intrinsic mode function representing the high-frequency component to separate the high-frequency noise of the QRS wave. Finally, the high-frequency noise is removed from the QRS wave to effectively improve the analysis. The accuracy of the judgment; therefore, the present invention first adds an additional high-frequency signal as an auxiliary signal to the captured heartbeat signal, and before adding the high-frequency auxiliary signal, the captured heartbeat signal must first be adjusted. The sampling frequency of the heartbeat signal, so that the sampling frequency of the heartbeat signal is an even multiple of the high-frequency auxiliary signal, that is, as shown in Figure 3, the input QRS wave x ( n ) must first go through upsampling ↑L and downsampling ↓M processing, where the selection of L and M satisfies two conditions, that is, condition 1 is the adjusted sampling frequency is the high frequency auxiliary signal frequency is an even multiple of , and condition 2 is the adjusted sampling frequency It should be close to and larger than the original sampling frequency ƒ _s , as shown in Table 1 below. Therefore, assuming that the sampling frequency ƒ _s of the input QRS wave is 2000 Hz, different high-frequency auxiliary signal frequencies are used, and the corresponding L and M value, and the adjusted sampling frequency will be as shown in Table 1; at the same time, when the high-frequency auxiliary signal has multiple frequencies, the adjusted sampling frequency must be an even multiple of each frequency, and as shown in Table 1 below, If the high-frequency signal package contains 150 Hz, 300 Hz, and 400 Hz at the same time, the adjusted sampling frequency of 2400 Hz can meet the conditions of even multiples of 150 Hz, 300 Hz, and 400 Hz.

Table 1

Continuing with the above, please refer to Figure 4(a)-(d), which shows the shape of the high-frequency auxiliary signal through empirical mode decomposition when there is no QRS wave, and compares the adjusted sampling frequency with empirical mode decomposition according to the present invention. The results of adding the QRS wave of the high-frequency auxiliary signal are shown in Figure 5(a)-(b). Therefore, taking the sampling frequency of 2000 Hz used in Figure 4(a)-(d) as an example, in Figure 4( a) and Figure 4(b) are the waveforms of high-frequency auxiliary signals of 350Hz and 250Hz respectively. It can be clearly seen that the sampling frequency 2000Hz presented is not an even multiple of the high-frequency auxiliary signal 350Hz, and in Figure The 350Hz waveform shown in Figure 4(a) is not symmetrical up and down, and the sampling frequency 2000Hz is an even multiple of the high-frequency auxiliary signal 250Hz, while the 250Hz waveform shown in Figure 4(b) is symmetrical up and down, and at the same time, with The upper image in Figure 4(c) and Figure 4(d) is the first IMF after empirical mode decomposition. The middle image in Figure 4(c) is the first IMF minus the high-frequency auxiliary signal. The lower image is is the spectrum diagram of the first IMF minus the high-frequency auxiliary signal. Therefore, as shown in the lower figure of Figure 4(c), it can be clearly seen that the harmonic signal occurs at 50Hz, 150Hz, 250Hz, and 350Hz. For the figure 4(d) shows that the empirical mode decomposition does not produce any additional signals for the 250Hz sine wave. Therefore, from the results shown in Figure 4(a)-(d), it can be found that the sampling frequency is not the frequency of the high-frequency auxiliary signal. At an even multiple, the signal generated is not completely symmetrical up and down, and will cause the high-frequency auxiliary signal to produce additional harmonic components after going through the filtering process of empirical mode decomposition.

On the other hand, in Figure 5(a), the s(n) + a(n) QRS wave s(n) plus the 150Hz high-frequency auxiliary signal a(n), and c1(n) in Figure 5(a) is the empirical From the first IMF after modal decomposition, it can be clearly seen that the additional 150Hz high-frequency auxiliary signal appears in the first IMF. From the spectrum of c1(n) in Figure 5(b), it is also It shows that the main energy of the first IMF is concentrated at 150Hz, and c1(n)-a(n) in Figure 5(a) is the result of removing the 150Hz high-frequency auxiliary information from the first IMF, and it can be clearly seen that It is known that only the small amplitude high-frequency component of the QRS wave is left. At the same time, in the spectrum of c1(n)-a(n) in Figure 5(b), it shows that the QRS wave s(n) is in the first IMF The spectrum in is mainly distributed around 150Hz to 350Hz, and corresponding to c2(n), c3(n), c4(n), and r4(n) in Figure 5(a) and Figure 5(b), it is relatively low frequency. The QRS wave component; therefore, because the sampling frequency is adjusted and the high-frequency auxiliary signal is added before decomposing the QRS wave, this can make the small-amplitude high-frequency band component in the heartbeat signal difficult to detect. The effective increase allows the empirical mode decomposition to increase the probability of separation of high-frequency noise in the original signal, thereby effectively and significantly improving the accuracy of electrocardiogram analysis and discrimination.

To summarize the above, the present invention is a method for separating electrocardiogram high-frequency noise. Through the steps of receiving, capturing, adjusting, adding, separating and removing, the QRS wave of the received heartbeat signal can be captured, and then the QRS wave can be captured. Before adding an additional high-frequency auxiliary signal to the QRS wave, the sampling frequency of the QRS wave can be adjusted first, so that the adjusted sampling frequency of the QRS wave is an even multiple of the high-frequency auxiliary signal, so that the samples can be passed through respectively. The frequency is adjusted and the high-frequency auxiliary signal is increased to enhance the high-frequency component, so that in this decomposition step, the empirical mode decomposition can be used to decompose it into an empirical mode decomposition using the repeated screening process mode. And obtain multiple intrinsic mode functions, and then subtract the high-frequency auxiliary signal from the first intrinsic mode function representing the high-frequency component to separate the high-frequency noise of the QRS wave. Finally, the high-frequency noise is removed from the QRS wave, thereby effectively improving the accuracy of analysis and discrimination.

However, the above descriptions are only for illustrating the preferred embodiments of the present invention, and should not be used to limit the scope of the present invention, that is, simple equivalent changes and modifications may be made based on the patent scope of the present invention and the contents of the description of the invention. , should still fall within the scope covered by the patent of this invention.

(this invention) None

Figure 1 is a flow chart of a preferred embodiment of the present invention. Figure 2 is a flow chart of the screening process of the decomposition step of the preferred embodiment. Figure 3 is a block diagram of using empirical mode decomposition to separate the high-frequency noise of the heartbeat signal. Figure 4(a) to (d) are schematic diagrams of the empirical mode decomposition results using high-frequency auxiliary signals when there is no QRS wave. Figures 5(a) to (b) are schematic diagrams of the results of QRS waves using empirical mode decomposition with adjusted sampling frequency and addition of high-frequency auxiliary signals.

Claims (4)

A method for separating electrocardiogram high-frequency noise, which includes: a receiving step, which receives a heartbeat signal from an electrocardiograph; An acquisition step is to acquire the QRS wave in the heartbeat signal; An additional step is to add an additional high-frequency auxiliary signal to the captured QRS wave, thereby increasing the high-frequency signal component in the QRS wave; A decomposition step, which uses empirical mode decomposition to decompose the QRS wave with the aforementioned high-frequency auxiliary signal, and the empirical mode decomposition adopts a repeated screening process to decompose the signal into an empirical mode. Decompose the state and obtain multiple essential mode functions; In a separation step, the high-frequency auxiliary signal is subtracted from the decomposed first essential mode function to separate the high-frequency noise of the QRS wave; and In a removal step, the high-frequency noise of the QRS wave can be removed by subtracting the separated high-frequency noise from the QRS wave. According to a method for separating electrocardiogram high-frequency noise described in claim 1, an adjustment step is added before the adding step to adjust the frequency of the QRS wave before adding the additional high-frequency auxiliary signal, so that the adjusted The sampling frequency of the QRS wave is an even multiple of the frequency of the high-frequency auxiliary signal. A method for separating electrocardiogram high-frequency noise according to claim 1, wherein the high-frequency auxiliary signal can be a single frequency signal or a combination of multiple frequency signals. A method for separating electrocardiogram high-frequency noise according to claim 1 or 2, wherein the filtering process includes: finding all local maxima and local minima in the original signal; The upper envelope and the lower envelope are connected to the local minimum by cubic spline curves respectively; calculate the average of the upper envelope and the lower envelope to obtain an average envelope; subtract the original signal from the Average the envelope to obtain the first component parameter; check whether the first component parameter meets the conditions of the intrinsic mode function. If it does not meet the conditions, the component will be regarded as the original signal, and the essential mode can be obtained by repeating the screening multiple times. Multiple component parameters of the function condition; subtract the first essential mode function from the original signal to obtain the first residual signal, and repeat the screening multiple times so that the first residual signal cannot be decomposed into other essences The modal function becomes a monotonic parameter with a small number of extreme values; and the original signal is decomposed into an essential modal function and a trend function.

Images