CN104644160A - Electrocardiogram pseudo-difference signal identification method and device - Google Patents

Abstract

The invention relates to the field of electrocardio signal processing, and provides an electrocardiogram artifact signal identification method and a device, which are used for solving the problems of complex calculation, high robustness and poor adaptability of the existing ECG artifact identification method, and mainly comprise the following steps: initializing an ECG signal; calculating a first information parameter R of the selected segmented ECG signal by correlation analysis, principal component analysis and frequency domain analysis, respectively_maxA second information parameter P_λ1And a third information parameter P_sn(ii) a The ECG segmented signal is identified based on the three information parameter values of the segment. Experimental results show that the accuracy, the sensitivity and the positive prediction rate of the ECG artifact signal recognition of the invention respectively reach 97.42%, 69.21% and 92.06%, so that the recognition accuracy is high.

Description

Electrocardiogram artifact signal identification method and device

technical field

The invention relates to the field of electrocardiogram signal processing, in particular to a method and device for identifying an electrocardiogram artifact signal.

Background technique

Electrocardiogram (ECG) is the potential difference generated on the body surface when the heart is excited, and it can reflect the electrophysiological activity of the heart. The methods of collecting ECG mainly include 3-lead ECG, 5-lead ECG and 12-lead ECG. Existing studies have shown that the ECG signal collected by any method may contain various interferences, which are not changes on the ECG due to cardiac excitation, so they are called "artifacts". There are many reasons for artifacts, for example, the body movement of the person being collected will cause motion artifacts in the ECG; a person's breathing will cause respiratory artifacts in the ECG; if the electrodes fall off during the ECG acquisition process, the recorded ECG Only distraction. From the severity of the interference, the artifact signal can be roughly divided into two categories: one is the signal that only causes minor interference to the original ECG, and this type of artifact signal can be filtered out by certain methods; The ECG produces a heavily disturbed or completely recorded noise signal, such artifacts can render the ECG devoid of any useful information. The present invention is mainly directed to the second type of artifact signal.

Such false signals in ECG are not only useless, but also increase the misdiagnosis rate of heart disease diagnosis based on ECG and reduce the accuracy of ECG monitoring. Manual screening of artifact signal fragments from ECG signals recorded for a long time (such as 48 hours, 7 days or even 1 month) is time-consuming and laborious, and the accuracy is low. Therefore, we urgently need a real-time and automatic calculation method to identify and mark the severely disturbed fragments (ie, ECG artifacts) in the ECG signal, so as to provide a solid guarantee for the further utilization of ECG.

Since the identification of ECG artifacts has very important practical significance, in recent years some scholars have begun to shift their research horizons to how to find ECG components that are submerged by interference and cannot be continued to be used by subsequent algorithms (ie, ECG artifacts) from ECG data. . The ECG cycle method is a recently proposed method that can automatically detect ECG artifacts. This method is designed based on fuzzy analysis, repolarization analysis and multi-parameter decision-making theory. Its advantage is that it does not require the setting of empirical thresholds in the judgment process. However, it must rely on ECG prior knowledge to complete artifact detection, and the length of input data is limited, so it cannot achieve good artifact location results for shorter ECG records. The dynamic ECG signal artifact recognition algorithm based on the principle of least squares can quickly and effectively identify the artifact components in the ECG signal caused by various reasons, but this method is easy to misjudge the ECG of arrhythmia patients as artifacts, which may Causes a high rate of missed diagnoses in subsequent disease diagnoses. The automatic artefact identification method based on wavelet transform can effectively and accurately identify sudden change artefacts at scales 5 and 6 by performing wavelet transform on the input signal. The effect is poor, so the robustness and adaptability of the method are poor.

It can be seen that the existing ECG artifact recognition methods either require prior knowledge, have a limit on the length of the data to be detected, or cannot be adapted to the ECG artifact recognition of arrhythmia patients, or have high computational complexity, poor robustness, and poor adaptability.

Contents of the invention

【Technical problems to be solved】

The object of the present invention is to provide a method and a device for identifying an electrocardiogram artifact signal, so as to at least solve one of the above technical problems.

【Technical solutions】

The present invention is achieved through the following technical solutions.

The present invention at first relates to a kind of electrocardiogram false difference signal recognition method, and this method comprises:

Step A: Segment the ECG signal to obtain a plurality of segmented ECG signals, select a segmented ECG signal, and the length of each segmented ECG signal obtained by the segment is greater than 2s;

Step B: respectively calculate the first information parameter R _max , the second information parameter P _λ1 and the third information parameter P _sn of the selected segmental ECG signal, the first information parameter R _max is the time window of the segmental ECG signal The maximum value of the internal shift correlation coefficient, the second information parameter P _λ1 is the contribution rate of the first principal component obtained after performing PCA operation on the segmented ECG signal, and the third information parameter P _sn is the segmental The ratio of the total signal amplitude of the spectrum of the ECG signal between 0.5 and 40 Hz to the total signal amplitude of the full frequency band;

Step C: Judging whether the second information parameter P _λ1 satisfies: P _λ1 >0.9, if it is satisfied, the selected segmented ECG signal is an ECG non-artifact signal and transfers to step G, otherwise, executes step D;

Step D: Judging whether the first information parameter R _max satisfies: R _max >0.61, judging whether the second information parameter P _λ1 satisfies: P _sn >0.85, judging whether the third information parameter P _sn satisfies: P _sn <0.44, judging the score Whether the maximum amplitude value POW of the spectrum of the segment ECG signal satisfies: POW>70, if the first information parameter R _max , the second information parameter P _λ1 , the third information parameter P _sn and the maximum amplitude value POW do not satisfy the corresponding inequality then Execute step F, otherwise, execute step E;

Step E: Judging whether the first information parameter R _max , the second information parameter P _λ1 and the third information parameter P _sn meet: 0.54*R _max +0.42*P _λ1 +0.04*P _sn >0.5, if satisfied, the selected The segmented ECG signal is an ECG non-artifact signal, otherwise the selected segmented ECG signal is an ECG artefact signal, and step E is transferred to step G after execution;

Step F: Determine whether the third information parameter P _sn satisfies: 0.77<P _sn <0.85, if satisfied, the selected segmented ECG signal is an ECG artifact signal, otherwise, the selected segmented ECG signal is an ECG non-artifacted signal Signal;

Step G: reselecting a segmented ECG signal, and repeating steps B to F until the identification of each segmented ECG signal in the ECG signals is completed.

As a preferred embodiment, the calculation method of the first information parameter R _max in the step B is:

Extract the ECG signal in the first 2s time window of the segmented ECG signal as a template; the time window of the template is shifted point by point in the segmented ECG signal with a sampling point as a step, and the shifted time window Correlation operation is performed between the ECG signal and the template, and the shift correlation coefficient of the segmented ECG signal in each time window is obtained by solving; after excluding the shift correlation coefficient with a value of 1, the maximum value of the remaining shift correlation coefficient is obtained to obtain An information parameter R _max .

As another preferred implementation manner, the length of the time window in the step B is 2s.

As another preferred implementation manner, the calculation method of the second information parameter P _λ1 in the step B is:

Segment the segmented ECG signal again according to the cardiac cycle to obtain the second segmented ECG signal; perform PCA operation on the second segmented ECG signal and extract the contribution rate of the first principal component as the second information parameter P _λ1 .

As another preferred implementation manner, the calculation method of the third information parameter _Psn in the step B is:

Calculate the frequency spectrum of the segmented ECG signal; calculate the ratio of the total signal amplitude of the segmented ECG signal spectrum between 0.5 and 40 Hz to the total amplitude of the full-frequency signal, and use the ratio as the third information parameter P _sn .

As another preferred implementation manner, in the step B, a fast FFT algorithm is used to calculate the frequency spectrum of the segmented ECG signal.

The present invention also relates to an electrocardiogram artifact signal identification device, which comprises:

The initialization module is used to segment the ECG signal to obtain a plurality of segmented ECG signals, and the length of each segmented ECG signal obtained by the segment is greater than 2s;

The first information parameter R _max calculation module is used to calculate the maximum value of the shift correlation coefficient in the time window of the segmented ECG signal;

The second information parameter P _λ1 calculation module is used to obtain the contribution rate of the first principal component after performing PCA operation on the segmented ECG signal;

The third information parameter P _sn calculation module is used to calculate the ratio of the total amplitude of the signal in the frequency spectrum of the segmented ECG signal between 0.5 and 40 Hz to the total amplitude of the signal in all frequency bands;

The first judging module is used to judge whether the second information parameter P _λ1 satisfies: P _λ1 >0.9, if satisfied, the segmented ECG signal selected is an ECG non-artifact signal, otherwise, the second judging module is notified;

The second judging module is used to judge whether the first information parameter R _max satisfies: R _max >0.61, judges whether the second information parameter P _λ1 satisfies: P _sn >0.85, and judges whether the third information parameter P _sn satisfies: P _sn < 0.44, judging whether the maximum amplitude value POW of the spectrum of the segmented ECG signal is satisfied: POW>70, if the first information parameter R _max , the second information parameter P _λ1 , the third information parameter P _sn and the maximum amplitude value POW are not satisfied The corresponding inequality then notifies the fourth judging module, otherwise notifies the third judging module;

The third judging module is used to judge whether the first information parameter R _max , the second information parameter P _λ1 and the third information parameter P _sn satisfy: 0.54*R _max +0.42*P _λ1 +0.04*P _sn >0.5, if satisfied Then the selected segmented ECG signal is an ECG non-aliased signal, otherwise the selected segmented ECG signal is an ECG artifacted signal;

The fourth judging module is used to judge whether the third information parameter P _sn satisfies: 0.77<P _sn <0.85, if satisfied, the selected segment ECG signal is an ECG artifact signal, otherwise the selected segment ECG signal is ECG non-artifact signal.

As a preferred implementation manner, the first information parameter R _max calculation module specifically includes:

A template extraction unit is used to extract the ECG signal in the 2s time window before each segmented ECG signal as a template;

A correlation operation unit is used to shift the time window of the template point by point in the segmented ECG signal with a sampling point as a step, and perform a correlation operation on the ECG signal in the shifted time window with the template to solve Obtain the shift correlation coefficient of the segmented ECG signal in the time window;

The maximum value unit is configured to obtain the maximum value of the remaining shifted correlation coefficients after excluding shifted correlation coefficients with a value of 1 to obtain the first information parameter R _max .

As another preferred implementation manner, the second information parameter P _λ1 calculation module specifically includes:

The segmentation unit is used to segment the segmented ECG signal again according to the cardiac cycle to obtain a second segmented ECG signal;

The PCA operation unit is used to perform PCA operation on the secondary segmented ECG signal processed by the segmentation unit and extract the contribution rate of the first principal component as the second information parameter P _λ1 .

As another preferred implementation manner, the third information parameter P _sn calculation module specifically includes:

The fast FFT unit is used to perform fast FFT to the segmented ECG signal to obtain the frequency spectrum of the segmented ECG signal;

The dividing unit is used to calculate the ratio of the total amplitude of the signal in the frequency spectrum of the segmented ECG signal between 0.5 and 40 Hz to the total amplitude of the signal in all frequency segments.

【Beneficial effect】

The technical scheme proposed by the present invention has the following beneficial effects:

(1) The present invention uses correlation analysis to reveal the rhythmicity of the ECG signal, and uses PCA technology to characterize the contribution of the ECG principal component in the input signal to the overall signal, and also distinguishes the ECG from the frequency through the frequency domain analysis. Difference signal and ECG pseudo-difference signal, the comprehensive use of different technologies enhances the robustness and applicability of the method of the present invention;

(2) The present invention uses three characteristic parameters of the local maximum of the correlation coefficient, the contribution rate of the first principal component and the spectral signal-to-noise ratio to characterize the ECG signal from different dimensions, and the three characteristic parameters are very good in the detection of artifacts played a complementary role, avoiding the risk of misidentification brought by the use of a single characteristic parameter, and improving the accuracy of the method of the present invention;

(3) The present invention can complete the identification of ECG artifact signals in a very short time window and without prior information, and has good real-time performance, robustness and adaptability.

Description of drawings

FIG. 1 is a schematic structural diagram of an electrocardiogram artifact signal identification device provided by Embodiment 1 of the present invention.

FIG. 2 is a flow chart of a method for identifying an electrocardiogram artifact signal provided by Embodiment 2 of the present invention.

Detailed ways

In order to make the purpose, technical solutions and advantages of the present invention clearer, the specific implementation manners of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Apparently, the described embodiments are part of the embodiments of the present invention, rather than All examples are not limitations of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without making creative efforts belong to the protection scope of the present invention.

FIG. 1 is a schematic structural diagram of an electrocardiogram artifact signal identification device provided by Embodiment 1 of the present invention. As shown in Figure 1, the device includes an initialization module 11, a first information parameter R _max calculation module 12, a second information parameter P _λ1 calculation module 13, a third information parameter P _sn calculation module 14, a first judgment module 15, a first information parameter The second judging module 16 , the third judging module 17 and the fourth judging module 18 .

The initialization module 11 is used to segment the ECG signal to obtain a plurality of segmented ECG signals, and select a segmented ECG signal as the segmented ECG signal to be identified, wherein the length of each segmented ECG signal obtained by segmenting is greater than 2s.

The first information parameter R _max calculation module 12 is used to calculate the maximum value of the shifted correlation coefficient within the time window of the segmented ECG signal. Specifically, the first information parameter R _max calculation module includes a template extraction unit, a correlation operation unit, and a maximum value unit, wherein the template extraction unit is used to extract the ECG signal within the 2s time window before each segmented ECG signal as a template, and the correlation The operation unit is used to shift the template point by point in the segmented ECG signal with a sampling point as the step size, and perform a correlation operation on the ECG signal in the shifted time window and the template to obtain the segmented ECG signal The unit for obtaining the maximum value of the shifted correlation coefficients within the time window is configured to obtain the maximum value of the remaining shifted correlation coefficients after excluding the shifted correlation coefficients with a value of 1 to obtain the first information parameter R _max .

The second information parameter P _λ1 calculation module 13 is used to obtain the contribution rate of the first principal component after performing PCA operation on the segmented ECG signal. Specifically, the second information parameter P _λ1 calculation module includes a segmentation unit and a PCA operation unit, wherein the segmentation unit is used to segment the segmented ECG signal again according to the cardiac cycle; the PCA operation unit is used to process the segmented unit. The secondary segmented ECG signal is subjected to PCA operation and the contribution rate of the first principal component is extracted as the second information parameter P _λ1 .

The third information parameter P _sn calculation module 14 is used to calculate the ratio of the total signal amplitude of the spectrum of the segmented ECG signal between 0.5 and 40 Hz to the total signal amplitude of all frequency bands. Specifically, the third information parameter P _sn calculation module includes a fast FFT unit and a division unit, wherein the fast FFT unit is used to perform fast FFT on the segmented ECG signal to obtain the spectrum of the segmented ECG signal, and the division unit is used to solve the segmented ECG The ratio of the total amplitude of the signal whose frequency spectrum is between 0.5 and 40 Hz to the total amplitude of the signal in all frequency bands.

The first judging module 15 is used to judge whether the second information parameter P _λ1 satisfies: P _λ1 >0.9, if so, the selected segment ECG signal is an ECG non-aliased signal, otherwise, inform the second judging module.

The second judging module 16 is used to judge whether the first information parameter R _max satisfies: R _max > 0.61, judges whether the second information parameter P _λ1 satisfies: P _sn > 0.85, and judges whether the third information parameter P _sn satisfies: P _sn < 0.44, judging whether the maximum amplitude value POW of the spectrum of the segmented ECG signal is satisfied: POW>70, if the first information parameter R _max , the second information parameter P _λ1 , the third information parameter P _sn and the maximum amplitude value POW are not satisfied The corresponding inequality is notified to the fourth judging module, otherwise, the third judging module is notified.

The third judging module 17 is used to judge whether the first information parameter R _max , the second information parameter P _λ1 and the third information parameter P _sn satisfy: 0.54*R _max +0.42*P _λ1 +0.04*P _sn >0.5, if satisfied Then the selected segment ECG signal is an ECG non-artifact signal, otherwise, the selected segment ECG signal is an ECG artifact signal.

The fourth judging module 18 is used to judge whether the third information parameter P _sn satisfies: 0.77<P _sn <0.85, if satisfied, the selected segment ECG signal is an ECG artifact signal, otherwise the selected segment ECG signal is ECG non-artifact signal.

For the method of using the device provided in Embodiment 1 to identify the electrocardiogram artifact signal, reference may be made to the following specific method embodiments.

FIG. 2 is a flowchart of a method for identifying an electrocardiogram artifact signal provided by Embodiment 2 of the present invention. As shown in Fig. 2, the method includes step S1 to step S7, each step will be described in detail below.

Step S1: Initialize the ECG signal.

Step S1 mainly includes segmenting the ECG signal to obtain a plurality of segmented ECG signals. Specifically, the ECG signal is uniformly segmented first, and the length of each segmented ECG signal is 3s. It should be noted that in step S1, the length of each segmented ECG signal should be greater than 2s.

Step S2: Select a segmented ECG signal, and calculate the first information parameter R _max , the second information parameter P _λ1 and the third information parameter P _sn .

Step S2 first selects a segmented ECG signal as the segmented ECG signal to be identified, and then respectively calculates the first information parameter R _max , the second information parameter P _λ1 and the third information parameter P _{sn of} the selected segmented ECG signal, Among them, the first information parameter R _max is the maximum value of the shifted correlation coefficient in the time window of the segmented ECG signal, and the second information parameter P _λ1 is the contribution of the first principal component obtained after PCA operation on the segmented ECG signal rate, and the third information parameter P _sn is the ratio of the total amplitude of the signal in the frequency spectrum of the segmented ECG signal between 0.5 and 40 Hz to the total amplitude of the signal in all frequency bands.

Specifically, the calculation method of the first information parameter R _max is:

Firstly, extract the ECG signal within the 2s time window before the segmented ECG signal as a template. The reason why the template length is 2s is that one ECG cycle of the ECG non-artifact signal is about 1s, so it can be guaranteed that in most cases the template contains The ECG signal of at least one electrocardiographic cycle, and the ECG artifact signal does not have this feature due to its randomness. It should be noted that the type of time window in this embodiment is a rectangular window, and the length of the time window is the same as the length of the template. is 2s;

Then the template is shifted point by point in the segmented ECG signal with a sampling point as the step length, and the ECG signal in the shifted time window is correlated with the template to solve the problem that the segmented ECG signal is in the time window If the ECG signal to be identified is an ECG non-artifact signal, its correlation coefficient has a local maximum value R _max , which shows a beat-by-beat morphological similarity. However, due to the randomness of the ECG artifact signal, the correlation coefficient will always maintain a low value during the solving process of the correlation operation. Therefore, the value of the first information parameter R _max of the ECG non-artifact signal is relatively large, while the value of the first information parameter R _max of the ECG artifact signal is very small. There is a large difference between the two, and the first information parameter R _max can be used to distinguish two types of signals;

Finally, after excluding the shifted correlation coefficients with a value of 1, the maximum value of the remaining shifted correlation coefficients is obtained to obtain the first information parameter R _max . It should be explained that because the template and the ECG signal in the time window before the shift are exactly the same, their autocorrelation value is always 1, so when solving the first information parameter R _max , it is necessary to use the local autocorrelation generated in the time window The maximum value is excluded because it does not truly reflect the beat-by-beat similarity of the ECG non-artifacted signal.

The calculation method of the second information parameter P _λ1 is:

First, segment the segmented ECG signal again according to the cardiac cycle. The specific re-segmentation method is as follows: firstly, locate the R peak in each cardiac cycle of the segmented ECG signal, and each segment of the second segmented ECG signal starts from the R peak The first 0.3 times the RR interval starts, and the end of the 0.7 times the RR interval after the R peak. After re-segmentation, the second segmented ECG signal can be obtained. It should be noted that in practical application of PCA, the input matrix should be measured multiple times. The data or the data of a single measurement of multiple individuals requires that the input data be multi-dimensional. However, for real-time collected ECG signals, this is difficult to achieve. In this embodiment, the PCA calculation can be obtained by segmenting the segmented ECG signals again. input data;

Then PCA operation is performed on the secondary segmented ECG signal and the contribution rate of the first principal component is extracted as the second information parameter P _λ1 . , while for the ECG artifact signal, the distribution of each principal component is very scattered, and the overall contribution rate of the principal components is low, so the second information parameter P _λ1 can be used to distinguish the two types of signals.

The calculation method of the third information parameter P _sn is:

First, fast FFT is performed on the segmented ECG signal to be identified to obtain the spectrum of the segmented ECG signal;

Then solve the ratio of the total signal amplitude of the spectrum of the segmented ECG signal between 0.5 and 40 Hz to the total signal amplitude of the full frequency band. This ratio is the third information parameter P _sn . In the power spectrum of the difference signal, the frequency range of the QRS complex is concentrated in the range of 3-40 Hz, and the frequency range of the P wave and the T wave is mainly concentrated in the range of 0.7-10 Hz, so this embodiment selects the frequency range of 0.5-40 Hz to calculate the third information Parameter P _sn , which can measure the extent to which the useful ECG information in the segmented ECG signal to be identified is submerged by noise. If the noise intensity is greater, the third information parameter P _sn will be lower. Conversely, if the noise intensity is smaller , the third information parameter P _sn is higher. Therefore, there is a large difference between the third information parameter P _sn of the ECG artifact signal and the ECG non-artifact signal.

Step S3: Identify by PCA, and judge whether the second information parameter P _λ1 satisfies: P _λ1 >0.9.

Step S3 mainly judges whether the second information parameter P _λ1 satisfies: P _λ1 >0.9, if the inequality is satisfied, the selected segmented ECG signal is an ECG non-artifact signal and then proceeds to step S7, if the inequality is not satisfied, Then step S4 is executed.

Step S4: Through the joint identification of correlation analysis, PCA and spectrum analysis, it is judged whether each information parameter does not satisfy the corresponding inequality.

Step S4 includes judging whether the first information parameter R _max satisfies: R _max >0.61, judging whether the second information parameter P _λ1 satisfies: P _sn >0.85, judging whether the third information parameter P _sn satisfies: P _sn <0.44, judging the score Whether the maximum amplitude value POW of the spectrum of the segment ECG signal satisfies: POW>70, if the first information parameter R _max , the second information parameter P _λ1 , the third information parameter P _sn and the maximum amplitude value POW do not satisfy the corresponding inequality then Execute step S6, otherwise, if one of the first information parameter R _max , the second information parameter P _λ1 , the third information parameter P _sn and the maximum amplitude value POW satisfies the corresponding inequality, then execute step S5. In step S4, only relying on the principal components of the input signal cannot fully judge its type, and it is necessary to consider finding its rhythm and frequency characteristics from the correlation coefficient and frequency domain to jointly judge. If the information parameters meet the corresponding conditions, it means This segmental ECG signal either has a relatively strong rhythm (R _max >0.61), or is similar in frequency to a noisy ECG signal (P _sn >0.85 or P _sn <0.44 or POW>70).

Step S5: identify by empirical formula, execute step S7 after step S5 is executed.

Step S5 mainly judges whether the first information parameter R _max , the second information parameter P _λ1 and the third information parameter P _sn satisfy the empirical formula: 0.54*R _max +0.42*P _λ1 +0.04*P _sn >0.5, if the inequality is satisfied , the selected segmented ECG signal is an ECG non-artifact signal, if the inequality is not satisfied, the selected segmented ECG signal is an ECG artifact signal, and step S7 is executed after step S5 is executed. The empirical formula is obtained by applying different weights to the first information parameter R _max , the second information parameter P _λ1 and the third information parameter P _sn respectively.

Step S6: identification by spectrum analysis.

Step S6 mainly judges whether the third information parameter P _sn satisfies: 0.77<P _sn <0.85, if the inequality is satisfied, the selected segment ECG signal is an ECG artifact signal, if the inequality is not satisfied, the selected segment ECG signal is The segment ECG signal is an ECG non-artifact signal. Since the frequency range of the ECG artifact signal mostly covers the frequency range of the normal ECG signal, its frequency distribution is highly random and relatively scattered, and is not concentrated in the frequency range of 0.5-40 Hz, so step S6 can further improve this Example accuracy.

Step S7: Judging whether the identification of each segmented ECG signal has been completed.

Specifically, step S7 first judges whether the identification of each segmented ECG signal has been completed, if the identification of each segmented ECG signal has been completed, the identification of the ECG signal ends, otherwise return to step S2 and continue to select unidentified segmental ECG The signal is identified.

In order to better describe Embodiment 2, the method provided in Embodiment 2 is used to conduct an artifact recognition experiment. The experiment includes artifact recognition for training data and artifact recognition for independent data.

The experimental data is described as follows:

Training data: The training data is generated by adding Gaussian white noise to normal human sinus ECG signals. Specifically, 40 groups of normal human sinus ECG signals are used, and the ECG signals are downloaded from the public MIT-BIH normal sinus rhythm database. The recording time of each group is 5 minutes, and the sampling rate is 128Hz. The annotation files of the MIT-BIH database ensure that each cardiac cycle is normally available. The position of the ECG artifact signal with noise interference was randomly generated, and the duration of the noise ranged from 30 to 150 s. When adding it, the start position and end point of the generated artifact were recorded and marked as the ECG artifact signal.

Independent data: This data consists of 25 sets of measured ECG record data containing real artifacts. Each set of recorded data has a different time length, the sampling frequency is 250Hz, and the total data length is 2395s. Among them, the signal marked as ECG has There are 11 segments, a total of 2227s, and the signal marked as ECG artifact has 6 segments, a total of 168s.

The method provided in Embodiment 2 is used to carry out the false signal recognition experiment, and the experimental results of the false signal recognition on the training data are shown in Table 1. The experimental results show that, for the artifact signals formed under different signal-to-noise ratios, the recognition accuracy, sensitivity and positive predictive rate of Embodiment 2 reached (0.9864±0.0017), (0.9899±0.0027) and (0.9925±0.0008) respectively. ), shows strong robustness, and has a strong ability to identify artifact signals and good performance, which shows that the selection of the threshold parameter in the embodiment of the second embodiment of the present invention has reached the optimum.

Table 1 Experimental results of identifying artifact signals in Embodiment 2 under different signal-to-noise ratios

Table 2 shows the experimental results of identifying artifact signals on independent data in Embodiment 2 of the present invention.

Table 2 Example 2 The experimental results of identifying real artifact signals

The experimental results in Table 2 show that the accuracy, sensitivity and positive predictive rate of embodiment two identifying ECG artifact signals are 97.42%, 69.21% and 92.06% respectively, so the accuracy of embodiment two identifying real artifact signals is very high , although the sensitivity of identifying ECG artifact signals is 69.21%, that is, a small amount of ECG artifact signals may be misjudged as ECG non-artifact signals, but this parameter is still better than other methods of the same kind, and it is not easy to cause missed diagnosis. In addition, the second embodiment has a sensitivity of 99.55% for identifying non-artifacted ECG signals, coupled with good robustness and real-time performance, therefore, the second embodiment of the present invention can meet the needs of practical applications.

It can be seen from the above embodiments that the embodiment of the present invention uses correlation analysis to reveal the rhythmicity of the ECG signal, and uses PCA technology to characterize the contribution of the ECG principal component in the input signal to the overall signal. From the frequency to distinguish two types of signals, ECG and ECG artifacts, the comprehensive use of different technologies enhances the robustness and applicability of the embodiment of the present invention; in addition, the embodiment of the present invention adopts the local maximum value of the correlation coefficient, the first principal component The three characteristic parameters of the contribution rate and the spectral signal-to-noise ratio characterize the ECG signal from different dimensions. The three characteristic parameters play a complementary role in the detection of artifacts, avoiding the misidentification caused by the use of a single characteristic parameter risk, improves the accuracy of this embodiment; finally, this embodiment can complete the identification of ECG artifact signals in a very short time window and without prior information, and has good real-time performance and robustness. and adaptability.

Claims (10)

1. an electrocardiogram artifact signal recognition method, is characterized in that comprising:

Steps A: carry out segmentation to ECG signal and obtain multiple segmentation ECG signal, selects a segmentation ECG signal, and the length of each segmentation ECG signal that described segmentation obtains all is greater than 2s;

Step B: the first information parameter R calculating selected segmentation ECG signal respectively _max, the second information parameter P _{λ 1}with the 3rd information parameter P _sn, described first information parameter R _maxfor the maximum of the time window internal shift correlation coefficient of segmentation ECG signal, described second information parameter P _{λ 1}for the contribution rate of first main constituent obtained after carrying out PCA computing to segmentation ECG signal, described 3rd information parameter P _snfor the signal total amplitude of frequency spectrum between 0.5 ~ 40Hz and the ratio of full rate segment signal total amplitude of segmentation ECG signal;

Step C: judge the second information parameter P _{λ 1}whether meet: P _{λ 1}>0.9, if met, selected segmentation ECG signal is the non-pseudo-difference signal of ECG and proceeds to perform step G, otherwise then performs step D;

Step D: judge first information parameter R _maxwhether meet: R _max>0.61, judges the second information parameter P _{λ 1}whether meet: P _sn>0.85, judges the 3rd information parameter P _snwhether meet: P _sn<0.44, judges whether the maximum amplitude value POW of the frequency spectrum of segmentation ECG signal meets: POW>70, if first information parameter R _max, the second information parameter P _{λ 1}, the 3rd information parameter P _snall do not meet corresponding inequality with maximum amplitude value POW and then perform step F, otherwise then perform step e;

Step e: judge first information parameter R _max, the second information parameter P _{λ 1}with the 3rd information parameter P _snwhether meet: 0.54*R _max+ 0.42*P _{λ 1}+ 0.04*P _sn>0.5, if met, selected segmentation ECG signal is the non-pseudo-difference signal of ECG, otherwise then selected segmentation ECG signal is ECG puppet difference signal, proceeds to and perform step G after step e executes;

Step F: judge the 3rd information parameter P _snwhether meet: 0.77<P _sn<0.85, if met, selected segmentation ECG signal is the pseudo-difference signal of ECG, otherwise then selected segmentation ECG signal is the non-pseudo-difference signal of ECG;

Step G: reselect a segmentation ECG signal, repeats step B to step F until complete the identification of each segmentation ECG signal in described ECG signal.

2. electrocardiogram artifact signal recognition method according to claim 1, is characterized in that the first information parameter R in described step B _maxcomputational methods be:

ECG signal in the front 2s time window of extraction segmentation ECG signal is as template; By the time window of described template in segmentation ECG signal with a sampled point for step-length pointwise displacement, by displacement after time window in ECG signal and template carry out related operation, solve and obtain the displacement correlation coefficient of segmentation ECG signal in each time window; After eliminating numerical value is the displacement correlation coefficient of 1, maximum is got to remaining displacement correlation coefficient and obtains first information parameter R _max.

3. electrocardiogram artifact signal recognition method according to claim 2, is characterized in that the length of the time window in described step B is 2s.

4. electrocardiogram artifact signal recognition method according to claim 1, is characterized in that the second information parameter P in described step B _{λ 1}computational methods be:

To segmentation ECG signal according to cardiac cycle again segmentation obtain secondary segmenting ECG signal; PCA computing is carried out to secondary segmenting ECG signal and the contribution rate extracting first main constituent as the second information parameter P _{λ 1}.

5. electrocardiogram artifact signal recognition method according to claim 1, is characterized in that the 3rd information parameter P in described step B _sncomputational methods be:

Calculate the frequency spectrum of segmentation ECG signal; Solve the signal total amplitude of frequency spectrum between 0.5 ~ 40Hz and the ratio of full rate segment signal total amplitude of segmentation ECG signal, using this ratio as the 3rd information parameter P _sn.

6. electrocardiogram artifact signal recognition method according to claim 5, is characterized in that adopting FFT to calculate the frequency spectrum of described segmentation ECG signal in described step B.

7. an electrocardiogram artifact signal identification device, is characterized in that comprising:

Initialization module, obtains multiple segmentation ECG signal for carrying out segmentation to ECG signal, and the length of each segmentation ECG signal that described segmentation obtains all is greater than 2s;

First information parameter R _maxcomputing module, for calculating the maximum of the time window internal shift correlation coefficient of segmentation ECG signal;

Second information parameter P _{λ 1}computing module, for obtaining the contribution rate of first main constituent after carrying out PCA computing to segmentation ECG signal;

3rd information parameter P _sncomputing module, for calculating the signal total amplitude of frequency spectrum between 0.5 ~ 40Hz and the ratio of full rate segment signal total amplitude of segmentation ECG signal;

First judge module, for judging the second information parameter P _{λ 1}whether meet: P _{λ 1}>0.9, if met, selected segmentation ECG signal is the non-pseudo-difference signal of ECG, otherwise then notifies the second judge module;

Second judge module, for judging first information parameter R _maxwhether meet: R _max>0.61, judges the second information parameter P _{λ 1}whether meet: P _sn>0.85, judges the 3rd information parameter P _snwhether meet: P _sn<0.44, judges whether the maximum amplitude value POW of the frequency spectrum of segmentation ECG signal meets: POW>70, if first information parameter R _max, the second information parameter P _{λ 1}, the 3rd information parameter P _snall do not meet corresponding inequality with maximum amplitude value POW and then notify the 4th judge module, otherwise then notify the 3rd judge module;

3rd judge module, for judging first information parameter R _max, the second information parameter P _{λ 1}with the 3rd information parameter P _snwhether meet: 0.54*R _max+ 0.42*P _{λ 1}+ 0.04*P _sn>0.5, if met, selected segmentation ECG signal is the non-pseudo-difference signal of ECG, otherwise then selected segmentation ECG signal is ECG puppet difference signal;

4th judge module, for judging the 3rd information parameter P _snwhether meet: 0.77<P _sn<0.85, if met, selected segmentation ECG signal is the pseudo-difference signal of ECG, otherwise then selected segmentation ECG signal is the non-pseudo-difference signal of ECG.

8. electrocardiogram artifact signal identification device according to claim 7, is characterized in that described first information parameter R _maxcomputing module specifically comprises:

Template extraction unit, for extracting ECG signal before each segmentation ECG signal in 2s time window as template;

Related operation unit, for by the time window of described template in segmentation ECG signal with a sampled point for step-length pointwise displacement, and the ECG signal in the time window after displacement and template are carried out related operation, solve and obtain the displacement correlation coefficient of segmentation ECG signal in time window;

Getting maximum unit, is get maximum to remaining displacement correlation coefficient after the displacement correlation coefficient of 1 to obtain first information parameter R for getting rid of numerical value _max.

9. electrocardiogram artifact signal identification device according to claim 7, is characterized in that described second information parameter P _{λ 1}computing module specifically comprises:

Segmenting unit, for segmentation ECG signal according to cardiac cycle again segmentation obtain secondary segmenting ECG signal;

PCA arithmetic element, the secondary segmenting ECG signal for obtaining segmenting unit process carry out PCA computing and the contribution rate extracting first main constituent as the second information parameter P _{λ 1}.

10. electrocardiogram artifact signal identification device according to claim 7, is characterized in that described 3rd information parameter P _sncomputing module specifically comprises:

Quick FFT unit, for carrying out quick FFT to segmentation ECG signal, obtains the frequency spectrum of segmentation ECG signal;

Divider, for solving the signal total amplitude of frequency spectrum between 0.5 ~ 40Hz and the ratio of full rate segment signal total amplitude of segmentation ECG signal.