CN109998524A - A kind of electrocardiosignal classification method based on variation mode decomposition theory and K nearest neighbor algorithm - Google Patents
A kind of electrocardiosignal classification method based on variation mode decomposition theory and K nearest neighbor algorithm Download PDFInfo
- Publication number
- CN109998524A CN109998524A CN201910249396.6A CN201910249396A CN109998524A CN 109998524 A CN109998524 A CN 109998524A CN 201910249396 A CN201910249396 A CN 201910249396A CN 109998524 A CN109998524 A CN 109998524A
- Authority
- CN
- China
- Prior art keywords
- electrocardiosignal
- variation mode
- energy
- mode decomposition
- classification
- 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
Classifications
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
- A61B5/316—Modalities, i.e. specific diagnostic methods
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
- A61B5/316—Modalities, i.e. specific diagnostic methods
- A61B5/318—Heart-related electrical modalities, e.g. electrocardiography [ECG]
-
- 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
- A61B5/7203—Signal processing specially adapted for physiological signals or for diagnostic purposes for noise prevention, reduction or removal
- A61B5/7207—Signal processing specially adapted for physiological signals or for diagnostic purposes for noise prevention, reduction or removal of noise induced by motion artifacts
-
- 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
-
- 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
Landscapes
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Heart & Thoracic Surgery (AREA)
- Surgery (AREA)
- Signal Processing (AREA)
- Veterinary Medicine (AREA)
- Public Health (AREA)
- Biophysics (AREA)
- Pathology (AREA)
- Biomedical Technology (AREA)
- Artificial Intelligence (AREA)
- Medical Informatics (AREA)
- Molecular Biology (AREA)
- General Health & Medical Sciences (AREA)
- Animal Behavior & Ethology (AREA)
- Psychiatry (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physiology (AREA)
- Evolutionary Computation (AREA)
- Fuzzy Systems (AREA)
- Mathematical Physics (AREA)
- Cardiology (AREA)
- Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)
Abstract
The electrocardiosignal classification method based on variation mode decomposition theory and K nearest neighbor algorithm that the invention discloses a kind of, comprising the following steps: intercept existing electrocardiosignal as sample data；Electrocardiosignal is pre-processed；The variation mode decomposition that 8 layers are carried out to electrocardiosignal, extracts the energy feature of each mode as characteristic of division；By the energy feature of mode 1,2,4,5,6,7 as input, Classification and Identification is carried out to signal with K nearest neighbor algorithm.The experimental results showed that this method can realize the quick and precisely classification to electrocardiosignal, facilitate heart condition-inference.
Description
Technical field
The present invention relates to electrocardiosignals to identify field, by carrying out power feature extraction to electrocardiosignal, to realize the heart
Electric signal classification, and in particular to one kind is based on the electrocardiosignal of variation mode decomposition theory (VMD) and K nearest neighbor algorithm (KNN)
Classification method.
Background technique
Arrhythmia cordis is the important diseases in terms of cardiovascular disease, is often occurred together with other cardiovascular diseases, it is considered to be
The sign that major disease occurs.The diagnosis of arrhythmia cordis mainly passes through analysis electrocardiosignal, it has objectively responded the activity of heart
State, help doctor judge to play the role of in the state of an illness it is vital, so, study electrocardiosignal classification method, raising
Classification accuracy is significant.
The main research about electrocardiosignal classification has at present: Han Junze proposes a kind of based on Wavelet transformation and K mean value
The QRS complex detection algorithm of cluster；Jeen-Shing Wang et al. extracts R wave and nearby puts as feature, passes through probabilistic neural net
Network (PNN) classifies to electrocardiosignal；R Benali et al. is using the method for wavelet neural network (WNN) to five class electrocardios
Signal realizes classification；Yan Haolin et al. proposes a kind of electrocardio classification sampled based on depth conviction net (DBN) and ecg wave form
Algorithm etc..Most of the above machine classification process is both needed to put into numerous studies on positioning waveform position or extraction QRS wave, and neglects
Application slightly in terms of electrocardiosignal energy.
So in view of the above problems, devising a kind of electrocardiosignal combined based on variation mode decomposition and K arest neighbors
Sorting algorithm, does not need the process for carrying out complicated extraction waveform, and sample is carried out VMD points by direct intercepted samples data respectively
Solution is finally carried out Classification and Identification to signal with KNN algorithm, is realized and believed electrocardio using the energy of each mode as characteristic of division
Number classification.
Summary of the invention
For the classification accuracy for improving electrocardiosignal, identification difficulty is reduced, the invention proposes one kind to be based on variation mode
The quick and precisely classification to electrocardiosignal may be implemented, specifically in the electrocardiosignal classification method of resolution theory and K nearest neighbor algorithm
Steps are as follows:
(1) existing electrocardiosignal is intercepted as sample data.
(2) pretreatment about baseline drift and Hz noise is carried out to all electrocardio samples, removal contains baseline drift
Modal components, and delete the modal components for significantly having 50~60Hz Hz noise.
(3) because the primary frequency range of electrocardiosignal is 0.5~100Hz, and 90% spectrum energy concentrate on 0.5~
Between 45Hz, the VMD number of plies decomposed is set to 8 layers, category signal is treated and carries out VMD decomposition, obtain 8 modal components, wherein
Variation mode decomposition method is as follows: X is electrocardiosignal, obtains K modal components { μ by K layers of variation mode decompositionk}=
{μ1...μk}；The center frequency of each modal components is expressed as { ωk}={ ω1...{ωk}:, during iterative solution constantly
Modal components and its center frequency are updated, update step (a), (b), (c), (d) are as follows:
(a) it initializesIt is 0 with n；
(b) μ is updated according to following two formulakAnd ωk, ∝ is penalty factor (value 2000), and λ is Lagrange multiplier operator；
(c) λ is updated, wherein τ is noise factor, τ=0；
If (d)Then stop iteration, otherwise gains execution step
(b), ε indicates constraint IF condition, ε=0.01.
(4) each mode energy is calculated separately, modal components energy proportion feature is solved:
N is sampling number
Gross energy are as follows:
Obtain energy feature:
(5) it successively assigns sample to digital label, inputs in KNN and carry out Classification and Identification, to improve classification accuracy and mentioning
The performance of high-class device and the difficulty for reducing operation, the screening preferable feature of validity is inside 8 energy features to reduce spy
Dimension is levied, because dimension is in contrast lower thus, all combinations of 8 energy features are directly tested using traversal, it is real
Test the energy feature shown using mode 1,2,4,5,6,7, classification accuracy highest；Arest neighbors numerical value is set as in KNN algorithm
2, range formula uses manhatton distance.
The present invention realizes electrocardiosignal classification, with existing detection mode by carrying out power feature extraction to electrocardiosignal
Compared to there is following characteristics:
1. the present invention reduces interference from human factor using the method directly analyzed electrocardiosignal, what exclusion was manually misread
It influences, the interpretation to electrocardiosignal is directly realized by by machine；
2. the method that the present invention skips the reading wave character generallyd use, does not need the step of carrying out complicated extraction waveform,
The classification of electrocardiosignal is realized by analysing energy feature；
3. the KNN sorting algorithm that the present invention uses is a kind of classification method of simple maturation, aobvious in the performance of electrocardio classification method
It writes.
Detailed description of the invention
Fig. 1 is flow chart of the method for the present invention；
Fig. 2 is that the feature extracted summarizes；
Fig. 3 is typical ecg wave form figure；
Fig. 4 is baseline drift processing figure；
Fig. 5 is Hz noise processing figure；
Fig. 6 is the modal components figure after VMD is decomposed；
Fig. 7 is typical energy ratio chart；
Fig. 8 is VMD-KNN classification results figure.
Specific embodiment
With reference to the accompanying drawing and specific embodiment the present invention is further described.
Hardware environment for implementation is different computer, and software environment is: Win7 and Matlab R2014a.We
Use Matlab software realization method proposed by the present invention.The electrocardiogram (ECG) data for being used to classification loses from the MIT-BIH rhythm of the heart
Regular data library is current one of the recognized standard ecg database in the world, shares 48 double leads records, herein all samples
It is all from using MLII lead, each record time is about 30 minutes, sample frequency 360Hz.In this experiment, for 5 classes
Ecg wave form is analyzed, and is respectively: normal beats (N), right bundle branch block (RBBB), left bundle branch block
(LBBB), ventricular premature beat (PVC), atrial premature beats (APB), every class heart clap used record and quantity as shown in Fig. 2, total extraction
300 groups of electrocardio samples.
Fig. 1 is flow chart of the method for the present invention, and following steps are shown in specific implementation:
(1) the electrocardio type to be extracted of determination, such as normal beats read the complete of No. 112 records by taking No. 112 records as an example first
Whole electrocardiographic recording；Corresponding number in annotated code is searched, it is 1 that the normal heart, which claps corresponding number, searches all annotations in this record
It is clapped for 1 heart, centered on annotation point, front and back intercepts at 1000 points as sampled data altogether, guarantees in this data length all
All it is that heart bat with wholeheartedly clapping type, within the scope of this no longer intercepts, accomplishes not repeat to sample, all samples for the condition that meets are protected
It deposits；For the reliability for guaranteeing experiment, same electrocardio type is chosen from multiple records, so randomly choosing 20 in 112 records
Group sample is finally completed the extraction to sample in 112 records as experiment sample, and Fig. 3 is the 5 typical sample datas of class electrocardio
Figure.
(2) we are concentrated mainly on baseline drift, on Hz noise to the pretreatment of electrocardiosignal, and baseline drift is by exhaling
It inhales with caused by other body movements, frequency is very low, is not more than 1Hz；50~60Hz of frequency range of Hz noise, after tested,
Exist in the record 105,214,228 used.Electrocardiosignal is decomposed into multiple modal components using VMD, removal contains baseline
Drift modal components, and the modal components that Hz noise is had in 3 records are deleted, finally remaining mode is reconstructed and is denoised
Electrocardiosignal complete pretreatment to electrocardiosignal, Fig. 4 and Fig. 5 are respectively baseline drift processing figure and Hz noise processing
Figure.
(3) 8 layers of Decomposition order are determined, VMD decomposition is carried out to all signals to be sorted, obtains 8 modal components, Fig. 6 is
Modal components figure after VMD decomposition.
(4) each mode energy is calculated separately, modal components energy proportion feature is solved:
N is sampling number
Gross energy are as follows:
Obtain energy feature:
Fig. 7 is the typical energy ratio of 5 class electrocardiosignals.
(5) 5 class samples are successively assigned to digital label 0~4, respectively represents 5 class electrocardios in the test and training in KNN
Type, normal beats are label 0, and right bundle branch block is label 1, and left bundle branch block is label 2, and ventricular premature beat is mark
Label 3, atrial premature beats are label 4,20 groups are randomly selected in every class electrocardio sample as test set, remainder is as training
Collection, then share 100 groups of test samples, 200 groups of sample trainings.By the energy feature of mode 1,2,4,5,6,7 as input, utilize
KNN algorithm realizes classification, and arest neighbors numerical value is set as 2 in KNN algorithm, and range formula uses manhatton distance, Fig. 8 VMD-
KNN classification results figure.
Claims (4)
1. a kind of electrocardiosignal classification method based on variation mode decomposition theory and K nearest neighbor algorithm, it is characterised in that including
Following steps:
(1) existing electrocardiosignal is intercepted as sample data；
(2) baseline drift and the pretreatment of Hz noise are removed to all electrocardio samples；
(3) 8 layers of Decomposition order are determined, category signal is treated and carries out variation mode decomposition, is i.e. VMD obtains 8 modal components；
(4) each mode energy is calculated separately, modal components energy proportion feature is solved；
(5) screening shows good feature, then utilizes K nearest neighbor algorithm, i.e. KNN realizes electrocardiosignal classification.
2. a kind of electrocardiosignal classification side based on variation mode decomposition theory and K nearest neighbor algorithm according to claim 1
Method, it is characterised in that: in step (3), used variation mode decomposition in electrocardiosignal classification and optimal Decomposition has been determined
The number of plies is 8 layers；Variation mode decomposition method is as follows: X is electrocardiosignal, obtains K modal components by K layers of variation mode decomposition；The center frequency of each modal components is expressed as:, in iteration
Modal components and its center frequency are constantly updated in solution procedure, update step (a), (b), (c), (d) are as follows:
(a) it initializesIt is 0 with n；
(b) it is updated according to following two formulaWith,For penalty factor (value 2000), λ is Lagrange multiplier operator；
(c) λ is updated, wherein τ is noise factor, τ=0；
If (d), then stop iteration, otherwise gain execution step
(b), ε indicates constraint IF condition, ε=0.01.
3. a kind of electrocardiosignal classification side based on variation mode decomposition theory and K nearest neighbor algorithm according to claim 1
Method, it is characterised in that: in step (4), calculate separately the energy of each mode for classifying:
, N is sampling number
Gross energy are as follows:
；
Energy feature are as follows:
。
4. a kind of electrocardiosignal classification side based on variation mode decomposition theory and K nearest neighbor algorithm according to claim 1
Method, it is characterised in that: in step (5), from 8 energy features inside screen more effective feature further to reduce intrinsic dimensionality,
Final feature is determined as the energy feature of mode 1,2,4,5,6,7, successively assigns sample to digital label, inputs in KNN and carry out
Classification and Identification, arest neighbors number are set as 2, and range formula uses manhatton distance.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910249396.6A CN109998524A (en) | 2019-03-29 | 2019-03-29 | A kind of electrocardiosignal classification method based on variation mode decomposition theory and K nearest neighbor algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910249396.6A CN109998524A (en) | 2019-03-29 | 2019-03-29 | A kind of electrocardiosignal classification method based on variation mode decomposition theory and K nearest neighbor algorithm |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109998524A true CN109998524A (en) | 2019-07-12 |
Family
ID=67168894
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910249396.6A Pending CN109998524A (en) | 2019-03-29 | 2019-03-29 | A kind of electrocardiosignal classification method based on variation mode decomposition theory and K nearest neighbor algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109998524A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111557659A (en) * | 2020-05-22 | 2020-08-21 | 郑州大学 | Arrhythmia classification method based on multi-feature fusion and Stacking-DWKNN |
| CN114869294A (en) * | 2022-05-05 | 2022-08-09 | 电子科技大学 | Particle filter motion artifact suppression method based on VMD decomposition and LET model |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101496716A (en) * | 2009-02-26 | 2009-08-05 | 周洪建 | Measurement method for detecting sleep apnoea with ECG signal |
| CN108056773A (en) * | 2017-12-11 | 2018-05-22 | 重庆邮电大学 | Based on the algorithms of QRS complexes detection in electrocardiogram signal for improving variation mode decomposition |
-
2019
- 2019-03-29 CN CN201910249396.6A patent/CN109998524A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101496716A (en) * | 2009-02-26 | 2009-08-05 | 周洪建 | Measurement method for detecting sleep apnoea with ECG signal |
| CN108056773A (en) * | 2017-12-11 | 2018-05-22 | 重庆邮电大学 | Based on the algorithms of QRS complexes detection in electrocardiogram signal for improving variation mode decomposition |
Non-Patent Citations (1)
| Title |
|---|
| UDAY MAJI等: "Estimation of Arrhythmia Episode using Variational Mode Decomposition Technique", 《2015 IEEE INTERNATIONAL INSTRUMENTATION AND MEASUREMENT TECHNOLOGY CONFERENCE (I2MTC) PROCEEDINGS》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111557659A (en) * | 2020-05-22 | 2020-08-21 | 郑州大学 | Arrhythmia classification method based on multi-feature fusion and Stacking-DWKNN |
| CN114869294A (en) * | 2022-05-05 | 2022-08-09 | 电子科技大学 | Particle filter motion artifact suppression method based on VMD decomposition and LET model |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Marinho et al. | A novel electrocardiogram feature extraction approach for cardiac arrhythmia classification | |
| Houssein et al. | An efficient ECG arrhythmia classification method based on Manta ray foraging optimization | |
| CN107951485B (en) | Ambulatory ECG analysis method and apparatus based on artificial intelligence self study | |
| CN111046731B (en) | Transfer learning method and recognition method for gesture recognition based on surface electromyographic signals | |
| US11783924B2 (en) | ECG information processing method and ECG workstation | |
| Ahmed et al. | An investigative study on motifs extracted features on real time big-data signals | |
| de Chazal et al. | Automatic classification of ECG beats using waveform shape and heart beat interval features | |
| Ivaturi et al. | A comprehensive explanation framework for biomedical time series classification | |
| Feng et al. | Unsupervised semantic-aware adaptive feature fusion network for arrhythmia detection | |
| Wang et al. | Imbalanced heartbeat classification using EasyEnsemble technique and global heartbeat information | |
| Lu et al. | An arrhythmia classification algorithm using C-LSTM in physiological parameters monitoring system under internet of health things environment | |
| Park et al. | Preprocessing method for performance enhancement in CNN-based STEMI detection from 12-lead ECG | |
| CN109998524A (en) | A kind of electrocardiosignal classification method based on variation mode decomposition theory and K nearest neighbor algorithm | |
| Kuila et al. | Feature extraction and classification of mit-bih arrhythmia database | |
| Gan et al. | Parallel classification model of arrhythmia based on DenseNet-BiLSTM | |
| Tong et al. | Locating abnormal heartbeats in ECG segments based on deep weakly supervised learning | |
| Wang et al. | Myocardial infarction detection based on multi-lead ensemble neural network | |
| Pucer et al. | A topological approach to delineation and arrhythmic beats detection in unprocessed long-term ECG signals | |
| Chandrashekhar et al. | Pulse ID: the case for robustness of ECG as a biometric identifier | |
| Zou et al. | A generalizable and robust deep learning method for atrial fibrillation detection from long-term electrocardiogram | |
| Ji et al. | Heartbeat classification based on multifeature combination and stacking-DWKNN algorithm | |
| Khalaf et al. | A QRS-Detection Algorithm for Real-Time Applications. | |
| Kumari et al. | Classification of cardiac arrhythmia using hybrid genetic algorithm optimisation for multi-layer perceptron neural network | |
| Batal et al. | Temporal Data Mining for Healthcare Data. | |
| Ge et al. | Automated recognition of arrhythmia using deep neural networks for 12-lead electrocardiograms with fractional time–frequency domain extension |
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 | ||
| WD01 | Invention patent application deemed withdrawn after publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20190712 |