CN112107310A - ECG identity recognition method based on IWT and AGA-BP models - Google Patents
ECG identity recognition method based on IWT and AGA-BP models Download PDFInfo
- Publication number
- CN112107310A CN112107310A CN202011065153.6A CN202011065153A CN112107310A CN 112107310 A CN112107310 A CN 112107310A CN 202011065153 A CN202011065153 A CN 202011065153A CN 112107310 A CN112107310 A CN 112107310A
- Authority
- CN
- China
- Prior art keywords
- wave
- point
- ecg
- decomposition
- value
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/117—Identification of persons
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7203—Signal processing specially adapted for physiological signals or for diagnostic purposes for noise prevention, reduction or removal
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7235—Details of waveform analysis
- A61B5/725—Details of waveform analysis using specific filters therefor, e.g. Kalman or adaptive filters
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7235—Details of waveform analysis
- A61B5/7253—Details of waveform analysis characterised by using transforms
- A61B5/726—Details of waveform analysis characterised by using transforms using Wavelet transforms
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7235—Details of waveform analysis
- A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
- A61B5/7267—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems involving training the classification device
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
Abstract
The invention discloses an ECG identity recognition method based on IWT and AGA-BP models, which specifically comprises the following steps: step 1, collecting electrocardiosignals and denoising; step 2, positioning the R wave peak point of the denoised ECG signal by adopting a wavelet positioning method; step 3, determining the position of a QRS complex through the R wave peak point obtained in the step 2, and determining the peak point, the starting point and the end point of the P wave and the T wave; and 4, combining the QRS complex, the peak point, the starting point and the end point of the P wave and the T wave obtained in the steps 2 and 3 to obtain a feature vector, and then identifying the ECG signal by using an AGA-BP model. The ECG identity recognition method based on the IWT model and the AGA-BP model solves the problem of poor electrocardiosignal recognition accuracy in the prior art.
Description
Technical Field
The invention belongs to the technical field of biological feature identification methods, and relates to an ECG identity identification method based on IWT and AGA-BP models.
Background
With the continuous development of science and technology, biological characteristics increasingly show unique advantages, and identification by using biological characteristics is more and more interesting. Compared with the traditional biological characteristics, the electrocardiosignals have a plurality of great advantages, and are difficult to imitate firstly because the electrocardiosignals come from the inside of the body; secondly, any surviving individual can have electrocardiosignals, so that the individual cannot be forgotten or lost; in addition, the electrocardiosignal is used as a one-dimensional signal, the processing is simple, the calculated amount is small, and the identification speed is higher. In summary, the great advantages of Electrocardiograph (ECG) signals make them a significant component of the field of biometric identification in the 21 st century.
The ECG-based identity recognition faces three important problems, one is the preprocessing of the electrocardiogram signal, the electrocardiogram signal initially acquired from the human body has a large amount of harmonics, the waveform quality is not high, so the electrocardiogram signal needs to be preprocessed, the quality of the result of the preprocessing of the electrocardiogram signal directly determines the quality of the later feature extraction and recognition, and at present, the electrocardiogram signal is preprocessed by using Empirical Mode Decomposition (EMD) and Wavelet Transform (WT) in the prior research. Secondly, the ECG performs feature extraction, a typical electrocardiographic waveform can be composed of main parts such as P wave, QRS complex, T wave, and the like, and the current research mainly focuses on two aspects: on one hand, the characteristic extraction is carried out on different wave groups, and most documents extract the QRS wave groups because the QRS wave groups can reflect the changes of the left ventricular depolarization potential and the right ventricular depolarization potential and time. Some documents also perform feature extraction on P-waves and T-waves for other purposes; on the other hand, a new feature extraction method is provided, and the currently common extraction methods comprise multi-scale autoregressive model (MSARM) extraction, autocorrelation extraction and transformation identification extraction, wherein the transformation identification extraction is divided into Discrete Cosine Transform (DCT) extraction, Discrete Fourier Transform (DFT) extraction, Walsh Hadamard Transform (WHT) extraction, wavelet transform extraction and the like according to different transformation methods. Thirdly, the electrocardiogram signal classification method, and the common classification methods include a Support Vector Machine (SVM), a bp (back propagation) neural network, a deep learning, a convolutional neural network, and some other improved methods.
In the traditional ECG identification method, a basic filtering algorithm is used in a data preprocessing stage, the filtering effect is not good, so that an accurate feature point position cannot be obtained in a feature extraction stage, meanwhile, the electrocardiogram signal classification method basically uses a single machine learning or intelligent algorithm, but the identification precision and the convergence speed of the single algorithm in the ECG identification process are not good, and the problems can influence the final identification result. Therefore, based on the above problems, the present invention provides an ECG identity recognition method based on IWT and AGA-BP models, where IWT is a WT algorithm optimized based on an infinite Kalman Filter (UKF) algorithm. The method of the invention applies IWT to improve the effect of signal preprocessing and uses an AGA-BP model to ensure that the ECG identity recognition can quickly reach the optimal recognition precision.
Disclosure of Invention
The invention aims to provide an ECG identity recognition method based on IWT and AGA-BP models, which solves the problem of poor electrocardiosignal recognition accuracy in the prior art.
The technical scheme adopted by the invention is that the ECG identity recognition method based on IWT and AGA-BP models is implemented according to the following steps:
step 2, positioning R wave peak points of the denoised ECG signal by adopting a wavelet positioning method, performing R wave omission detection and error detection investigation, and simultaneously determining the sampling interval between the adjacent R wave peak points;
step 3, determining the position of a QRS complex through the R wave peak point obtained in the step 2, and determining the peak point, the starting point and the end point of the P wave and the T wave;
and 4, combining the QRS complex, the peak point, the starting point and the end point of the P wave and the T wave obtained in the steps 2 and 3 to obtain a feature vector, and then identifying the ECG signal by using an AGA-BP model.
The present invention is also characterized in that,
the step 1 specifically comprises the following steps:
step 1.1, obtaining original electrocardiographic data in a mode of equipment reading or database acquisition, and then drawing the obtained original electrocardiographic data by using an ECG algorithm to obtain a matrix in which the ECG data is stored, wherein the matrix is an ECG signal to be processed;
step 1.2, the ECG signal obtained in the step 1.1 is divided according to a standard electrocardio period T to obtain n groups of periodic signal sequences X (n), and then a recursive algorithm of a least square method is used for modeling the periodic signal sequences X (n) to obtain a mathematical model thereof;
1.2.1, segmenting electrocardiosignals:
the ECG signal obtained in step 1.1 is segmented according to a QRS wave period T, and the following sequence is obtained:
step 1.2.2, periodic sequence modeling:
carrying out self-adaptive modeling on X (n) obtained in 1.2.1, determining the order and each order coefficient of the model by using a fixed order criterion of F test and a recursive least square method, and finally determining the self-adaptive model as a first-order model, wherein the obtained mathematical model formula is as follows:
X(n)＝Φ*X(n-1) (2)；
wherein phi is a coefficient matrix obtained by least squares recursion.
Step 1.3, performing wavelet decomposition reconstruction on the original electrocardiogram data and the mathematical model obtained in the step 1.2 by adopting a Mallat algorithm to respectively obtain original ECG wavelet decomposition signals under multiple scales and wavelet decomposition signals of the mathematical model under multiple scales;
step 1.4, filtering the wavelet decomposition signals of the mathematical model under the multiple scales after wavelet decomposition by an unscented Kalman filtering algorithm by using original ECG wavelet decomposition signals under the multiple scales as a measurement equation and using wavelet decomposition signals of the mathematical model under the multiple scales as a state equation;
and 1.5, reconstructing the electrocardiosignals subjected to the unscented Kalman filtering by using a Mallat algorithm to obtain denoised ECG signals.
The step 2 specifically comprises the following steps:
step 2.1, carrying out four-layer discrete wavelet decomposition on the ECG signal denoised in the step 1 through a two-sample strip wavelet filter to obtain an ECG signal subjected to cubic scale decomposition;
step 2.2, based on the ECG signal after the cubic scale decomposition obtained in the step 2.1, searching an R wave maximum value and an R wave minimum value under the cubic scale, and determining a suspected R wave peak value point;
and 2.3, performing missing detection and error detection on the R wave aiming at the suspected R wave peak value point determined in the step 2.2, and finally determining the accurate position of the R wave peak value point.
The step 2.2 specifically comprises the following steps:
step 2.2.1, finding the maximum value of the R wave: based on the ECG signal after the cubic scale decomposition obtained in the step 2.1, searching the maximum value of the R wave under the cubic scale decomposition, namely finding out the point with the slope larger than 0, assigning the point to be 1, assigning the rest points to be 0, and positioning the maximum value point in the sequence of 1 and 0;
searching for R wave minimum value: based on the ECG signal obtained after the cubic scale decomposition in the step 2.1, searching the minimum value of the R wave under the cubic scale decomposition, namely finding out the point with the slope smaller than 0, assigning the point to be 1, assigning the rest points to be 0, and positioning the minimum value point in the 1 and 0 sequence;
and 2.2.2, removing a maximum value and a minimum value point of which the absolute value is smaller than the threshold value by setting the threshold value as an average value of one third of adjacent signal periods under three decomposition scales to obtain an existence interval of the suspected R wave peak value, wherein a zero crossing point of the adjacent maximum and minimum value pairs is the suspected R wave peak value point.
The step 2.3 is specifically as follows:
setting the determination conditions as follows, when the distance between adjacent suspected R wave peak points is less than 0.4mean (RR), indicating that false detection exists, removing the suspected R wave peak point with the minimum absolute value under the three-time decomposition scale, when the distance between the adjacent R wave peak points is more than 1.6mean (RR), indicating that false detection exists, searching an absolute value and a maximum minimum value pair in two suspected R wave intervals under the three-time decomposition scale, positioning a zero crossing point of the searched maximum minimum value pair as the position of the missed detection R wave peak value, finally determining the accurate R wave peak point position, comparing the determined R wave peak point position with the actual R wave peak point in the original electrocardiogram data, and if the determined R wave peak point position and the actual R wave peak point in the original electrocardiogram data exist, artificially selecting a displacement correction quantity to correct the determined R wave peak point position, and obtaining the position of the final R wave peak point.
And after the positioning is successful, determining a sampling interval between two RR intervals, wherein the sampling interval is the period T of the electrocardiosignal, and the T can optimize the signal segmentation in the step 1.2, so that the superiority of the method is further improved.
The step 3 specifically comprises the following steps:
step 3.1 determining the QRS complex position:
based on the position of the R wave peak point obtained in the step 2, corresponding the position of the R wave peak point to a primary decomposition scale, determining the positions of the first three extreme points of the R wave peak point in the primary decomposition scale as a Q wave starting point, and determining the last three extreme points as an S wave terminal point;
step 3.2, determining peak points of the P wave and the T wave:
under the four decomposition scales, the determined QRS complex position is utilized, the Q wave interval with the forward starting point of 2/3RR is used as a P wave searching interval, the S wave interval with the backward ending point of 2/3RR is used as a T wave searching interval, the maximum extreme value pair is searched in each interval range, the midpoint is found and determined as the peak point of the P wave and the T wave, and simultaneously the maximum extreme value point and the minimum extreme value point searched in each interval range respectively determine the starting point and the ending point of the P wave and the T wave.
The step 4 specifically comprises the following steps:
step 4.1, using the P wave starting point, the P wave peak point, the P wave end point, the QRS wave starting point, the QRS wave R wave peak point, the QRS wave end point, the T wave starting point, the T wave peak point, and the T wave end point as feature points, obtaining 22 different feature vectors in a pairwise combination manner by using the distance between the feature points and the amplitude of the feature points to represent a periodic signal, where the periodic signal is an obtained feature data set sample, where the 22 different feature vectors include 16 distance feature vectors: R-R, R-Q, R-S, R-P, R-PBegin, R-Pend, R-T, R-TBegin, R-Tend, PBegin-Pend, TBegin-Tend, Q-P, S-T, P-T, Q-PBegin, S-Tend, 6 amplitude feature vectors: Q-R, S-R, PBegin-P, P-Q, T-TBegin and T-S, wherein, Begin is added after the letter to represent the starting point, end is added to represent the ending point, and peak point is not added;
step 4.2, after the characteristic data group samples are extracted in the step 4.1, assuming that the original electrocardiosignals have N cycles, selecting the characteristic data group of the former N/2 cycles as a training set, and selecting the characteristic data group of the latter N/2 cycles as a test set;
and 4.3, firstly, normalizing the training sample and the test sample simultaneously to obtain a training sample matrix and a test sample matrix. Then, ECG identification is performed by the AGA-BP model. And finally, comparing the actual training result with the given output category matrix to obtain the final identification accuracy.
The invention has the beneficial effects that: firstly, the invention combines the advantages of wavelet transformation and unscented Kalman filtering in signal noise processing to carry out filtering processing on electrocardiosignals, thereby greatly reducing the influence of measurement noise on subsequent characteristic point extraction, then, based on a wavelet transformation extraction algorithm, the invention can effectively select and extract electrocardiosignal characteristic point data, and finally, the invention adopts a fast convergent AGA-BP model to carry out classification identification on the electrocardiosignals, and the three processes complement each other to improve the electrocardio identification precision.
Drawings
FIG. 1 is a flow chart of an ECG identification method based on IWT and AGA-BP models according to the present invention;
FIG. 2 is a diagram of a Mallat algorithm filter bank implementation of wavelet decomposition and reconstruction in the ECG identification method based on IWT and AGA-BP models according to the present invention;
FIG. 3 is a flowchart of IWT algorithm in the ECG identification method based on IWT and AGA-BP models according to the present invention;
FIG. 4 is a flowchart of an ECG characteristic point extraction general algorithm in the ECG identification method based on IWT and AGA-BP models according to the present invention;
FIG. 5 is a schematic diagram illustrating the offset between the wavelet positioning R wave peak point and the actual peak R wave value point in the ECG identification method based on IWT and AGA-BP models according to the present invention;
FIG. 6 is a structural diagram of an AGA-BP model algorithm in an ECG identification method based on IWT and AGA-BP models.
Detailed Description
The present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.
The invention relates to an ECG identity recognition method based on IWT and AGA-BP models, the flow of which is shown in figure 1 and is implemented according to the following steps:
step 1.1, obtaining original electrocardiographic data in a mode of equipment reading or database acquisition, and then drawing the obtained original electrocardiographic data by using an ECG algorithm to obtain a matrix in which the ECG data is stored, wherein the matrix is an ECG signal to be processed;
step 1.2, the ECG signal obtained in the step 1.1 is divided according to a standard electrocardio period T to obtain n groups of periodic signal sequences X (n), and then a recursive algorithm of a least square method is used for modeling the periodic signal sequences X (n) to obtain a mathematical model thereof; the method specifically comprises the following steps:
1.2.1, segmenting electrocardiosignals:
the ECG signal obtained in step 1.1 is segmented according to a QRS wave period T, and the following sequence is obtained:
step 1.2.2, periodic sequence modeling:
carrying out self-adaptive modeling on X (n) obtained in 1.2.1, determining the order and each order coefficient of the model by using a fixed order criterion of F test and a recursive least square method, and finally determining the self-adaptive model as a first-order model, wherein the obtained mathematical model formula is as follows:
X(n)＝Φ*X(n-1) (2)；
wherein phi is a coefficient matrix obtained by least square recursion;
step 1.3, performing wavelet decomposition reconstruction on the original electrocardiogram data and the mathematical model obtained in the step 1.2 by adopting a Mallat algorithm to respectively obtain original ECG wavelet decomposition signals under multiple scales and wavelet decomposition signals of the mathematical model under multiple scales;
step 1.4, using MATLAB software, taking original ECG wavelet decomposition signals under multiple scales as a measurement equation, taking wavelet decomposition signals of a mathematical model under multiple scales as a state equation, and performing filtering processing on the wavelet decomposition signals of the mathematical model under multiple scales after wavelet decomposition by writing an unscented Kalman filtering algorithm;
and 1.5, reconstructing the electrocardiosignals subjected to the unscented Kalman filtering by using a Mallat algorithm to obtain denoised ECG signals.
Step 2, positioning R wave peak points of the denoised ECG signal by adopting a wavelet positioning method, performing R wave omission detection and error detection investigation, and simultaneously determining the sampling interval between the adjacent R wave peak points; the method specifically comprises the following steps:
step 2.1, carrying out four-layer discrete wavelet decomposition on the ECG signal denoised in the step 1 through a two-sample strip wavelet filter to obtain an ECG signal subjected to cubic scale decomposition;
step 2.2, based on the ECG signal after the cubic scale decomposition obtained in the step 2.1, searching an R wave maximum value and an R wave minimum value under the cubic scale, and determining a suspected R wave peak value point; the method specifically comprises the following steps:
step 2.2.1, finding the maximum value of the R wave: based on the ECG signal after the cubic scale decomposition obtained in the step 2.1, searching the maximum value of the R wave under the cubic scale decomposition, namely finding out the point with the slope larger than 0, assigning the point to be 1, assigning the rest points to be 0, and positioning the maximum value point in the sequence of 1 and 0;
searching for R wave minimum value: based on the ECG signal obtained after the cubic scale decomposition in the step 2.1, searching the minimum value of the R wave under the cubic scale decomposition, namely finding out the point with the slope smaller than 0, assigning the point to be 1, assigning the rest points to be 0, and positioning the minimum value point in the 1 and 0 sequence;
step 2.2.2, removing a maximum value and a minimum value point of which the absolute value is smaller than the threshold value by setting the threshold value as an average value of one third of adjacent signal periods under three decomposition scales to obtain an existence interval of a suspected R wave peak value, wherein a zero crossing point of the adjacent maximum and minimum value pairs is the suspected R wave peak value point;
step 2.3, performing missing detection and error detection investigation on the R wave aiming at the suspected R wave peak point determined in step 2.2, and finally determining the accurate R wave peak point position, specifically:
setting the determination conditions as follows, when the distance between adjacent suspected R wave peak points is less than 0.4mean (RR), indicating that false detection exists, removing the suspected R wave peak point with the minimum absolute value under the three-time decomposition scale, when the distance between the adjacent R wave peak points is more than 1.6mean (RR), indicating that false detection exists, searching an absolute value and a maximum minimum value pair in two suspected R wave intervals under the three-time decomposition scale, positioning a zero crossing point of the searched maximum minimum value pair as the position of the missed detection R wave peak value, finally determining the accurate R wave peak point position, comparing the determined R wave peak point position with the actual R wave peak point in the original electrocardiogram data, and if the determined R wave peak point position and the actual R wave peak point in the original electrocardiogram data exist, artificially selecting a displacement correction quantity to correct the determined R wave peak point position, obtaining the position of the final R wave peak point;
and after the positioning is successful, determining a sampling interval between two RR intervals, wherein the sampling interval is the period T of the electrocardiosignal, and the T can optimize the signal segmentation in the step 1.2, so that the superiority of the method is further improved.
Step 3, determining the position of a QRS complex through the R wave peak point obtained in the step 2, and determining the peak point, the starting point and the end point of the P wave and the T wave;
and 4, combining the QRS complex, the peak point, the starting point and the end point of the P wave and the T wave obtained in the steps 2 and 3 to obtain a feature vector, and then identifying the ECG signal by using an AGA-BP model.
The step 3 specifically comprises the following steps:
step 3.1 determining the QRS complex position:
based on the position of the R wave peak point obtained in the step 2, corresponding the position of the R wave peak point to a primary decomposition scale, determining the positions of the first three extreme points of the R wave peak point in the primary decomposition scale as a Q wave starting point, and determining the last three extreme points as an S wave terminal point;
step 3.2, determining peak points of the P wave and the T wave:
under the four decomposition scales, the determined QRS complex position is utilized, the Q wave interval with the forward starting point of 2/3RR is used as a P wave searching interval, the S wave interval with the backward ending point of 2/3RR is used as a T wave searching interval, the maximum extreme value pair is searched in each interval range, the midpoint is found and determined as the peak point of the P wave and the T wave, and simultaneously the maximum extreme value point and the minimum extreme value point searched in each interval range respectively determine the starting point and the ending point of the P wave and the T wave.
The step 4 specifically comprises the following steps:
step 4.1, using the P wave starting point, the P wave peak point, the P wave end point, the QRS wave starting point, the QRS wave R wave peak point, the QRS wave end point, the T wave starting point, the T wave peak point, and the T wave end point as feature points, obtaining 22 different feature vectors in a pairwise combination manner by using the distance between the feature points and the amplitude of the feature points to represent a periodic signal, where the periodic signal is an obtained feature data set sample, where the 22 different feature vectors include 16 distance feature vectors: R-R, R-Q, R-S, R-P, R-PBegin, R-Pend, R-T, R-TBegin, R-Tend, PBegin-Pend, TBegin-Tend, Q-P, S-T, P-T, Q-PBegin, S-Tend, 6 amplitude feature vectors: Q-R, S-R, PBegin-P, P-Q, T-TBegin and T-S, wherein, Begin is added after the letter to represent the starting point, end is added to represent the ending point, and peak point is not added;
step 4.2, after the characteristic data group samples are extracted in the step 4.1, assuming that the original electrocardiosignals have N cycles, selecting the characteristic data group of the former N/2 cycles as a training set, and selecting the characteristic data group of the latter N/2 cycles as a test set;
and 4.3, firstly, normalizing the training sample and the test sample simultaneously to obtain a training sample matrix and a test sample matrix. Then, ECG identification is performed by the AGA-BP model. And finally, comparing the actual training result with the given output category matrix to obtain the final identification accuracy. (after discussion with teacher and brother, teacher and brother representation does not necessarily explain how the training sample matrix and the testing sample matrix are applied to the network, since this is the basic step of the algorithm and not the focus here)
Examples
The ECG identity recognition method based on the IWT model and the AGA-BP model is implemented by the following steps:
step 1.1, acquiring original electrocardio data through an arrhythmia electrocardio database (MIT-BIH), reading and drawing any group of sample data in the electrocardio sample database by using an ECG algorithm reading program written by a Vorarlberg University of Applied Sciences University Robert Trtnig, and obtaining a matrix storing ECG data, wherein the matrix is an ECG signal to be processed;
step 1.2, the ECG signal obtained in the step 1.1 is divided according to a standard electrocardio period T to obtain n groups of periodic signal sequences X (n), and then a recursive algorithm (RLS algorithm) of a least square method is used for modeling the periodic signal sequences X (n) to obtain a mathematical model thereof; the method specifically comprises the following steps:
1.2.1, segmenting electrocardiosignals:
the ECG signal obtained in step 1.1 is segmented according to a QRS wave period T, and the following sequence is obtained:
step 1.2.2, periodic sequence modeling:
carrying out self-adaptive modeling on X (n) obtained in 1.2.1, determining the order and each order coefficient of the model by using a fixed order criterion of F test and a recursive least square method, and finally determining the self-adaptive model as a first-order model, wherein the obtained mathematical model formula is as follows:
X(n)＝Φ*X(n-1) (2)；
wherein phi is a coefficient matrix obtained by least square recursion;
as shown in fig. 2, step 1.3, performing wavelet decomposition and reconstruction on the original electrocardiographic data and the mathematical model obtained in step 1.2 by using a Mallat algorithm to obtain original ECG wavelet decomposition signals under multiple scales and wavelet decomposition signals of the mathematical model under multiple scales respectively;
step 1.4, using MATLAB software, taking original ECG wavelet decomposition signals under multiple scales as a measurement equation, taking wavelet decomposition signals of a mathematical model under multiple scales as a state equation, and performing filtering processing on the wavelet decomposition signals of the mathematical model under multiple scales after wavelet decomposition by writing an unscented Kalman filtering algorithm;
and 1.5, reconstructing the electrocardiosignals subjected to the unscented Kalman filtering by using a Mallat algorithm to obtain denoised ECG signals.
Step 2, positioning R wave peak points of the denoised ECG signal by adopting a wavelet positioning method, performing R wave omission detection and error detection investigation, and simultaneously determining the sampling interval between the adjacent R wave peak points; as shown in fig. 4, specifically:
step 2.1, carrying out four-layer discrete wavelet decomposition on the ECG signal denoised in the step 1 through a two-sample strip wavelet filter to obtain an ECG signal subjected to cubic scale decomposition;
2.1.1 tuning filter parameters:
the filter parameters are adjusted after referring to relevant documents as follows:
low-pass filter coefficient: 1/4,3/4,3/4,1/4
High pass filter coefficients: -1/4, -3/4,3/4,1/4
2.1.2 obtaining the ECG signal after 4-time scale decomposition by a two-sample-injection strip wavelet filter, wherein theoretical analysis shows that the R wave peak value of the ECG signal under the three-time scale of wavelet decomposition is the largest and most prominent, so that the R wave peak value point is detected based on the waveform under the three-time scale;
step 2.2, based on the ECG signal after the cubic scale decomposition obtained in the step 2.1, searching an R wave maximum value and an R wave minimum value under the cubic scale, and determining a suspected R wave peak value point; the method specifically comprises the following steps:
step 2.2.1, finding the maximum value of the R wave: based on the ECG signal after the cubic scale decomposition obtained in the step 2.1, searching the maximum value of the R wave under the cubic scale decomposition, namely finding out the point with the slope larger than 0, assigning the point to be 1, assigning the rest points to be 0, and positioning the maximum value point in the sequence of 1 and 0;
searching for R wave minimum value: based on the ECG signal obtained after the cubic scale decomposition in the step 2.1, searching the minimum value of the R wave under the cubic scale decomposition, namely finding out the point with the slope smaller than 0, assigning the point to be 1, assigning the rest points to be 0, and positioning the minimum value point in the 1 and 0 sequence;
step 2.2.2, removing a maximum value and a minimum value point of which the absolute value is smaller than the threshold value by setting the threshold value as an average value of one third of adjacent signal periods under three decomposition scales to obtain an existence interval of a suspected R wave peak value, wherein a zero crossing point of the adjacent maximum and minimum value pairs is the suspected R wave peak value point;
step 2.3, performing missing detection and error detection investigation on the R wave aiming at the suspected R wave peak point determined in step 2.2, and finally determining the accurate R wave peak point position, specifically:
setting the determination conditions as follows, when the distance between adjacent suspected R wave peak points is less than 0.4mean (RR), indicating that false detection exists, removing the suspected R wave peak point with the minimum absolute value under the three-time decomposition scale, when the distance between the adjacent R wave peak points is more than 1.6mean (RR), indicating that false detection exists, searching an absolute value and a maximum minimum value pair in two suspected R wave intervals under the three-time decomposition scale, positioning a zero crossing point of the searched maximum minimum value pair as the position of the missed detection R wave peak value, finally determining the accurate R wave peak point position, comparing the determined R wave peak point position with the actual R wave peak point in the original electrocardiogram data, and if the determined R wave peak point position and the actual R wave peak point in the original electrocardiogram data exist, artificially selecting a displacement correction quantity to correct the determined R wave peak point position, obtaining the position of the final R wave peak point; as shown in fig. 5, a certain displacement is found between the two, and the displacement correction amount is artificially selected to be 20 sampling points to the left.
And after the positioning is successful, determining a sampling interval between two RR intervals, wherein the sampling interval is the period T of the electrocardiosignal, and the T can optimize the signal segmentation in the step 1.2, so that the superiority of the method is further improved.
Step 3, determining the position of a QRS complex through the R wave peak point obtained in the step 2, and determining the peak point, the starting point and the end point of the P wave and the T wave;
and 4, combining the QRS complex, the peak point, the starting point and the end point of the P wave and the T wave obtained in the steps 2 and 3 to obtain a feature vector, and then identifying the ECG signal by using an AGA-BP model.
The step 3 specifically comprises the following steps:
step 3.1 determining the QRS complex position:
based on the position of the R wave peak point obtained in the step 2, corresponding the position of the R wave peak point to a primary decomposition scale, determining the positions of the first three extreme points of the R wave peak point in the primary decomposition scale as a Q wave starting point, and determining the last three extreme points as an S wave terminal point;
step 3.2, determining peak points of the P wave and the T wave:
under the four decomposition scales, the determined QRS complex position is utilized, the Q wave interval with the forward starting point of 2/3RR is used as a P wave searching interval, the S wave interval with the backward ending point of 2/3RR is used as a T wave searching interval, the maximum extreme value pair is searched in each interval range, the midpoint is found and determined as the peak point of the P wave and the T wave, and simultaneously the maximum extreme value point and the minimum extreme value point searched in each interval range respectively determine the starting point and the ending point of the P wave and the T wave.
Step 4 is specifically, as shown in fig. 6:
step 4.1, using the P wave starting point, the P wave peak point, the P wave end point, the QRS wave starting point, the QRS wave R wave peak point, the QRS wave end point, the T wave starting point, the T wave peak point, and the T wave end point as feature points, obtaining 22 different feature vectors in a pairwise combination manner by using the distance between the feature points and the amplitude of the feature points to represent a periodic signal, where the periodic signal is an obtained feature data set sample, where the 22 different feature vectors include 16 distance feature vectors: R-R, R-Q, R-S, R-P, R-PBegin, R-Pend, R-T, R-TBegin, R-Tend, PBegin-Pend, TBegin-Tend, Q-P, S-T, P-T, Q-PBegin, S-Tend, 6 amplitude feature vectors: Q-R, S-R, PBegin-P, P-Q, T-TBegin and T-S, wherein, Begin is added after the letter to represent the starting point, end is added to represent the ending point, and peak point is not added;
step 4.2, after the characteristic data group samples are extracted in the step 4.1, assuming that the original electrocardiosignals have N cycles, selecting the characteristic data group of the former N/2 cycles as a training set, and selecting the characteristic data group of the latter N/2 cycles as a test set;
the step 4.3 is specifically as follows:
step 4.3.1, normalizing the training sample and the test sample to obtain a training sample matrix and a test sample matrix;
step 4.3.2, firstly, an initial BP neural network is constructed, then, the weight of the initial BP neural network is trained and optimized by utilizing an AGA algorithm, the evolution is stopped after a certain number of evolution steps is reached, an optimal weight is obtained, and finally, the optimal weight is used for carrying out local search by adopting the BP algorithm, so that the optimal weight is quickly converged to a final optimal value.
And 4.3.3, detecting the BP algorithm result, returning to the step 4.3.2 if the identification result does not reach the expected target, and stopping if the identification result reaches the expected target.
The invention provides a novel electrocardiogram signal identity recognition system, which comprises electrocardiogram signal preprocessing, electrocardiogram signal feature extraction and electrocardiogram signal identity recognition classification. Firstly, an Improved Wavelet Transform (IWT) algorithm is proposed, and the electrocardiogram signal is filtered by adopting a Wavelet Transform and Unscented Kalman Filter (Unscented Kalman Filter, UKF) combined method, so that the influence of most non-gaussian noise and gaussian noise on waveform feature extraction is eliminated simultaneously. Secondly, extracting the characteristic points of the electrocardiosignals by adopting a wavelet transform positioning method. Finally, Adaptive Genetic Algorithm (AGA) combined with BP neural network Algorithm for electrocardiogram signal identification is proposed.
Claims (9)
1. The ECG identity recognition method based on the IWT model and the AGA-BP model is characterized by comprising the following steps:
step 1, acquiring electrocardiosignals, and preprocessing the acquired electrocardiosignals by using an IWT algorithm to obtain denoised ECG signals;
step 2, positioning R wave peak points of the denoised ECG signal by adopting a wavelet positioning method, performing R wave omission detection and error detection investigation, and simultaneously determining the sampling interval between the adjacent R wave peak points;
step 3, determining the position of a QRS complex through the R wave peak point obtained in the step 2, and determining the peak point, the starting point and the end point of the P wave and the T wave;
and 4, combining the QRS complex, the peak point, the starting point and the end point of the P wave and the T wave obtained in the steps 2 and 3 to obtain a feature vector, and then identifying the ECG signal by using an AGA-BP algorithm.
2. The method for identifying an ECG based on IWT and AGA-BP model according to claim 1, wherein the step 1 specifically comprises:
step 1.1, obtaining original electrocardiographic data in a mode of equipment reading or database acquisition, and then drawing the obtained original electrocardiographic data by using an ECG algorithm to obtain a matrix in which the ECG data is stored, wherein the matrix is an ECG signal to be processed;
step 1.2, the ECG signal obtained in the step 1.1 is divided according to a standard electrocardio period T to obtain n groups of periodic signal sequences X (n), and then a recursive algorithm of a least square method is used for modeling the periodic signal sequences X (n) to obtain a mathematical model thereof;
step 1.3, performing wavelet decomposition reconstruction on the original electrocardiogram data and the mathematical model obtained in the step 1.2 by adopting a Mallat algorithm to respectively obtain original ECG wavelet decomposition signals under multiple scales and wavelet decomposition signals of the mathematical model under multiple scales;
step 1.4, using MATLAB software, taking original ECG wavelet decomposition signals under multiple scales as a measurement equation, taking wavelet decomposition signals of a mathematical model under multiple scales as a state equation, and performing filtering processing on the wavelet decomposition signals of the mathematical model under multiple scales after wavelet decomposition by writing an unscented Kalman filtering algorithm;
and 1.5, reconstructing the electrocardiosignals subjected to the unscented Kalman filtering by using a Mallat algorithm to obtain denoised ECG signals.
3. The method for ECG identification based on IWT and AGA-BP model according to claim 1, wherein the step 2 specifically comprises:
step 2.1, carrying out four-layer discrete wavelet decomposition on the ECG signal denoised in the step 1 through a two-sample strip wavelet filter to obtain an ECG signal subjected to cubic scale decomposition;
step 2.2, based on the ECG signal after the cubic scale decomposition obtained in the step 2.1, searching an R wave maximum value and an R wave minimum value under the cubic scale, and determining a suspected R wave peak value point;
and 2.3, performing missing detection and error detection on the R wave aiming at the suspected R wave peak value point determined in the step 2.2, and finally determining the accurate position of the R wave peak value point.
4. The method for ECG identification based on IWT and AGA-BP model according to claim 3, wherein the step 2.2 is specifically:
step 2.2.1, finding the maximum value of the R wave: based on the ECG signal after the cubic scale decomposition obtained in the step 2.1, searching the maximum value of the R wave under the cubic scale decomposition, namely finding out the point with the slope larger than 0, assigning the point to be 1, assigning the rest points to be 0, and positioning the maximum value point in the sequence of 1 and 0;
searching for R wave minimum value: based on the ECG signal obtained after the cubic scale decomposition in the step 2.1, searching the minimum value of the R wave under the cubic scale decomposition, namely finding out the point with the slope smaller than 0, assigning the point to be 1, assigning the rest points to be 0, and positioning the minimum value point in the 1 and 0 sequence;
and 2.2.2, removing a maximum value and a minimum value point of which the absolute value is smaller than the threshold value by setting the threshold value as an average value of one third of adjacent signal periods under three decomposition scales to obtain an existence interval of the suspected R wave peak value, wherein a zero crossing point of the adjacent maximum and minimum value pairs is the suspected R wave peak value point.
5. The method for ECG identification based on IWT and AGA-BP model according to claim 4, wherein the step 2.3 is specifically:
setting the determination conditions as follows, when the distance between adjacent suspected R wave peak points is less than 0.4mean (RR), indicating that false detection exists, removing the suspected R wave peak point with the minimum absolute value under the three-time decomposition scale, when the distance between the adjacent R wave peak points is more than 1.6mean (RR), indicating that false detection exists, searching an absolute value and a maximum minimum value pair in two suspected R wave intervals under the three-time decomposition scale, positioning a zero crossing point of the searched maximum minimum value pair as the position of the missed detection R wave peak value, finally determining the accurate R wave peak point position, comparing the determined R wave peak point position with the actual R wave peak point in the original electrocardiogram data, and if the determined R wave peak point position and the actual R wave peak point in the original electrocardiogram data exist, artificially selecting a displacement correction quantity to correct the determined R wave peak point position, and obtaining the position of the final R wave peak point.
And after the positioning is successful, determining a sampling interval between two RR intervals, wherein the sampling interval is the period T of the electrocardiosignal, and the T can optimize the signal segmentation in the step 1.2, so that the superiority of the method is further improved.
6. The ECG identification method based on IWT and AGA-BP model according to claim 5, wherein the step 3 specifically comprises:
step 3.1 determining the QRS complex position:
based on the position of the R wave peak point obtained in the step 2, corresponding the position of the R wave peak point to a primary decomposition scale, determining the positions of the first three extreme points of the R wave peak point in the primary decomposition scale as a Q wave starting point, and determining the last three extreme points as an S wave terminal point;
step 3.2, determining peak points of the P wave and the T wave:
under the four decomposition scales, the determined QRS complex position is utilized, the Q wave interval with the forward starting point of 2/3RR is used as a P wave searching interval, the S wave interval with the backward ending point of 2/3RR is used as a T wave searching interval, the maximum extreme value pair is searched in each interval range, the midpoint is found and determined as the peak point of the P wave and the T wave, and simultaneously the maximum extreme value point and the minimum extreme value point searched in each interval range respectively determine the starting point and the ending point of the P wave and the T wave.
7. The method for ECG identification based on IWT and AGA-BP model according to claim 6, wherein the step 4 specifically comprises:
step 4.1, using the P wave starting point, the P wave peak point, the P wave end point, the QRS wave starting point, the QRS wave R wave peak point, the QRS wave end point, the T wave starting point, the T wave peak point, and the T wave end point as feature points, obtaining 22 different feature vectors in a pairwise combination manner by using the distance between the feature points and the amplitude of the feature points to represent a periodic signal, where the periodic signal is an obtained feature data set sample, where the 22 different feature vectors include 16 distance feature vectors: R-R, R-Q, R-S, R-P, R-PBegin, R-Pend, R-T, R-TBegin, R-Tend, PBegin-Pend, TBegin-Tend, Q-P, S-T, P-T, Q-PBegin, S-Tend, 6 amplitude feature vectors: Q-R, S-R, PBegin-P, P-Q, T-TBegin and T-S, wherein, Begin is added after the letter to represent the starting point, end is added to represent the ending point, and peak point is not added;
step 4.2, after the characteristic data group samples are extracted in the step 4.1, assuming that the original electrocardiosignals have N cycles, selecting the characteristic data group of the former N/2 cycles as a training set, and selecting the characteristic data group of the latter N/2 cycles as a test set;
and 4.3, firstly, normalizing the training sample and the test sample simultaneously to obtain a training sample matrix and a test sample matrix. Then, ECG identification is performed by the AGA-BP model. And finally, comparing the actual training result with the given output category matrix to obtain the final identification accuracy. (after discussion with teacher and brother, teacher and brother representation does not necessarily explain how the training sample matrix and the testing sample matrix are applied to the network, since this is the basic step of the algorithm and not the focus here)
8. The method for identifying an ECG based on IWT and AGA-BP model according to claim 2, wherein the step 1.2 is specifically:
1.2.1, segmenting electrocardiosignals:
the ECG signal obtained in step 1.1 is segmented according to a QRS wave period T, and the following sequence is obtained:
step 1.2.2, periodic sequence modeling:
carrying out self-adaptive modeling on X (n) obtained in 1.2.1, determining the order and each order coefficient of the model by using a fixed order criterion of F test and a recursive least square method, and finally determining the self-adaptive model as a first-order model, wherein the obtained mathematical model formula is as follows:
X(n)＝Φ*X(n-1) (2)；
wherein phi is a coefficient matrix obtained by least squares recursion.
9. The method for ECG identification based on IWT and AGA-BP model according to claim 7, wherein the step 4.3 is specifically:
step 4.3.1, normalizing the training sample and the test sample to obtain a training sample matrix and a test sample matrix;
step 4.3.2, firstly, an initial BP neural network is constructed, then, the weight of the initial BP neural network is trained and optimized by utilizing an AGA algorithm, the evolution is stopped after a certain number of evolution steps is reached, an optimal weight is obtained, and finally, the optimal weight is used for carrying out local search by adopting the BP algorithm, so that the optimal weight is quickly converged to a final optimal value.
And 4.3.3, detecting the BP algorithm result, returning to the step 4.3.2 if the identification result does not reach the expected target, and stopping if the identification result reaches the expected target.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011065153.6A CN112107310A (en) | 2020-09-30 | 2020-09-30 | ECG identity recognition method based on IWT and AGA-BP models |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011065153.6A CN112107310A (en) | 2020-09-30 | 2020-09-30 | ECG identity recognition method based on IWT and AGA-BP models |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112107310A true CN112107310A (en) | 2020-12-22 |
Family
ID=73797274
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011065153.6A Pending CN112107310A (en) | 2020-09-30 | 2020-09-30 | ECG identity recognition method based on IWT and AGA-BP models |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112107310A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112971839A (en) * | 2021-02-05 | 2021-06-18 | 云南大学 | Heart sound classification method based on feedforward convolutional neural network |
| CN113192046A (en) * | 2021-05-14 | 2021-07-30 | 中北大学 | Automatic identification method for radial distribution function graph |
| CN113951891A (en) * | 2021-11-11 | 2022-01-21 | 西安博远恒达电气科技有限公司 | ECG (electrocardiogram) identity recognition method based on space-time combination feature vector |
| CN114224360A (en) * | 2021-12-27 | 2022-03-25 | 长春工程学院 | EEG signal processing method and equipment based on improved EMD-ICA and storage medium |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103083013A (en) * | 2013-01-18 | 2013-05-08 | 哈尔滨工业大学深圳研究生院 | Electrocardio signal QRS complex wave detection method based on morphology and wavelet transform |
| KR20140041327A (en) * | 2012-09-27 | 2014-04-04 | 삼성전자주식회사 | Method and system for determining qrs complexes in electrocardiogram signals |
| WO2017148452A1 (en) * | 2016-03-03 | 2017-09-08 | 深圳竹信科技有限公司 | Electrocardiography signal waveform feature point extraction method and device |
| CN107239684A (en) * | 2017-05-22 | 2017-10-10 | 吉林大学 | A kind of feature learning method and system for ECG identifications |
| CN108272451A (en) * | 2018-02-11 | 2018-07-13 | 上海交通大学 | A kind of QRS wave recognition methods based on improvement wavelet transformation |
| CN110236530A (en) * | 2019-06-20 | 2019-09-17 | 武汉中旗生物医疗电子有限公司 | A kind of electrocardiosignal QRS wave group localization method, device and computer storage medium |
-
2020
- 2020-09-30 CN CN202011065153.6A patent/CN112107310A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR20140041327A (en) * | 2012-09-27 | 2014-04-04 | 삼성전자주식회사 | Method and system for determining qrs complexes in electrocardiogram signals |
| CN103083013A (en) * | 2013-01-18 | 2013-05-08 | 哈尔滨工业大学深圳研究生院 | Electrocardio signal QRS complex wave detection method based on morphology and wavelet transform |
| WO2017148452A1 (en) * | 2016-03-03 | 2017-09-08 | 深圳竹信科技有限公司 | Electrocardiography signal waveform feature point extraction method and device |
| CN107239684A (en) * | 2017-05-22 | 2017-10-10 | 吉林大学 | A kind of feature learning method and system for ECG identifications |
| CN108272451A (en) * | 2018-02-11 | 2018-07-13 | 上海交通大学 | A kind of QRS wave recognition methods based on improvement wavelet transformation |
| CN110236530A (en) * | 2019-06-20 | 2019-09-17 | 武汉中旗生物医疗电子有限公司 | A kind of electrocardiosignal QRS wave group localization method, device and computer storage medium |
Non-Patent Citations (2)
| Title |
|---|
| SANWAN: "ECG信号读取，检测QRS，P，T 波（基于小波去噪与检测），基于BP神经网络的身份识别", HTTP://WWW.BUBUKO.COM/INFODETAIL-226147.HTML * |
| 王海军: "AGA-BP模型在遥感影像分类中的应用", 计算机测量与控制, vol. 25, no. 5 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112971839A (en) * | 2021-02-05 | 2021-06-18 | 云南大学 | Heart sound classification method based on feedforward convolutional neural network |
| CN113192046A (en) * | 2021-05-14 | 2021-07-30 | 中北大学 | Automatic identification method for radial distribution function graph |
| CN113951891A (en) * | 2021-11-11 | 2022-01-21 | 西安博远恒达电气科技有限公司 | ECG (electrocardiogram) identity recognition method based on space-time combination feature vector |
| CN114224360A (en) * | 2021-12-27 | 2022-03-25 | 长春工程学院 | EEG signal processing method and equipment based on improved EMD-ICA and storage medium |
| CN114224360B (en) * | 2021-12-27 | 2023-10-10 | 长春工程学院 | EEG signal processing method, equipment and storage medium based on improved EMD-ICA |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112107310A (en) | ECG identity recognition method based on IWT and AGA-BP models | |
| CN112257518A (en) | ECG identity recognition method based on WT and WOA-PNN algorithm | |
| CN107203692B (en) | Electrocardio data digital signal processing method based on deep convolutional neural network | |
| CN110840402B (en) | Atrial fibrillation signal identification method and system based on machine learning | |
| CN106951753B (en) | Electrocardiosignal authentication method and device | |
| CN110786850B (en) | Electrocardiosignal identity recognition method and system based on multi-feature sparse representation | |
| CN109758145B (en) | Automatic sleep staging method based on electroencephalogram causal relationship | |
| CN103714281A (en) | Identity recognition method based on electrocardiosignals | |
| CN108154085B (en) | Method and device for identity recognition based on electrocardiogram data and electronic equipment | |
| CN111345816B (en) | Multi-lead QRS complex detection method, device, equipment and storage medium | |
| CN108509823A (en) | The detection method and device of QRS complex | |
| CN111797674A (en) | MI electroencephalogram signal identification method based on feature fusion and particle swarm optimization algorithm | |
| CN112353397A (en) | Electrocardiogram signal identity recognition method | |
| CN111053552A (en) | QRS wave detection method based on deep learning | |
| CN113486752B (en) | Emotion recognition method and system based on electrocardiosignal | |
| CN110169767A (en) | A kind of search method of electrocardiosignal | |
| CN113069124A (en) | Electrocardio monitoring method based on CNN-ET model | |
| CN113343869A (en) | Electroencephalogram signal automatic classification and identification method based on NTFT and CNN | |
| CN115486849A (en) | Electrocardiosignal quality evaluation method, device and equipment | |
| CN109063652B (en) | Signal processing method, system and computer storage medium | |
| CN110916645A (en) | QRS wave identification method combining wavelet transformation and image segmentation network | |
| CN113066544B (en) | FVEP characteristic point detection method based on CAA-Net and LightGBM | |
| CN114861706A (en) | Electrocardio identity recognition method based on quality evaluation and deep transfer learning | |
| CN112818315B (en) | Electrocardiosignal identity recognition method and system fusing multi-feature information | |
| CN114118157A (en) | Illumination information diagnosis method and system based on plant electric signals |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |