CN108596142A - A kind of cardioelectric characteristic extracting process based on PCANet - Google Patents
A kind of cardioelectric characteristic extracting process based on PCANet Download PDFInfo
- Publication number
- CN108596142A CN108596142A CN201810434968.3A CN201810434968A CN108596142A CN 108596142 A CN108596142 A CN 108596142A CN 201810434968 A CN201810434968 A CN 201810434968A CN 108596142 A CN108596142 A CN 108596142A
- Authority
- CN
- China
- Prior art keywords
- heart
- matrix
- clapped
- sorted
- pcanet
- 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.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/08—Feature extraction
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/02—Preprocessing
- G06F2218/04—Denoising
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/12—Classification; Matching
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Data Mining & Analysis (AREA)
- Artificial Intelligence (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Signal Processing (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)
Abstract
Technical scheme of the present invention includes a kind of cardioelectric characteristic extracting process based on PCANet, which is characterized in that is included the following steps:S10, pretreatment acquisition training set and collection to be sorted are carried out to electrocardiogram, S20, the feature extraction for being carried out heart bat to training set and collection to be sorted respectively using PCANet, S30, the heart extracted using training set are clapped feature training grader and use it for the classification that the collection heart to be sorted claps feature;Beneficial effects of the present invention are:The step of having robustness to the noise of ECG signal, simplify noise remove, clapping the unbalanced heart has preferable classifying quality, improves the efficiency and accuracy rate of ecg characteristics extraction, alleviates the pressure that doctor identifies electrocardiogram, reduces the probability of doctor's mistaken diagnosis.
Description
Technical field
The present invention relates to a kind of cardioelectric characteristic extracting process based on PCANet, belong to medical signals process field.
Background technology
Currently, with the development of computer technology, data mining, deep learning isotype identification technology are gradually applied
To in medical signals processing.Currently known area of pattern recognition includes the skills such as electrocardiogram, electroencephalogram, Medical Image Processing
Art field.In electrocardiogram field, electrocardiogram assisting in diagnosis and treatment equipment has had significant progress, it can excavate electrocardio
Profound information in figure simultaneously carries out efficient automatic identification.
Electrocardiogram automatic identification technology includes three steps, is pretreatment, feature extraction and classifying respectively, i.e., will pre-process
The heart afterwards is clapped excavates further feature by characteristic extraction step, and these features are identified using grader.Wherein feature
Extraction step is particularly important, while being also the key point of the present invention.
However, before carrying out feature extraction, generally require to carry out ECG signal processing, the denoising during being somebody's turn to do is particularly heavy
It wants.First, when we carry out heart bat sampling by instrument and equipment, due to being caused there are the influence of body other organs of the body
The heart that we obtain claps waveform, and there are noise pollutions, and the noise of electrocardiogram (ECG) signal includes that baseline drift is made an uproar under normal circumstances
Sound, industrial frequency noise etc. need to clap progress noise remove, mesh to the heart using necessary measure in pretreatment stage in the prior art
The numerous kinds of noise remove methods such as preceding existing wavelet analysis, median filter, but it is inevitable while the removal of noise
Cause the loss of useful information in signal.Secondly, although existing numerous sorting algorithms are to a variety of hearts for being balanced in quantity now
Bat is trained can obtain preferable evaluation criterion with classification, but when being trained to the bat of the number of samples unbalanced heart, past
The heart of advantage, which is clapped, toward on to quantity has significant classification advantage, and it is relatively low to clap recognition accuracy to the less heart of training.Therefore
The present invention proposes a kind of heart bat feature extracting method for noise robustness and the unbalanced robustness of sample size..
Invention content
To solve the above problems, the purpose of the present invention is to provide a kind of cardioelectric characteristic extracting process based on PCANet,
It has robustness to noise, need not clap noise to the heart in preprocessing process and be removed, also need not be to heart umber of beats amount
Equalization processing is carried out, preferable classifying quality can be obtained, mitigates pressure of the doctor by the ecg analysis state of an illness.
Technical solution is used by the present invention solves the problems, such as it:
A kind of cardioelectric characteristic extracting process based on PCANet, includes the following steps:S10, electrocardiogram is pre-processed
Acquisition training set and collection to be sorted, S20, the feature extraction for carrying out heart bat to training set and collection to be sorted respectively using PCANet,
S30, the heart extracted using training set are clapped feature training grader and use it for the classification that the collection heart to be sorted claps feature;
The step S10 includes:
S11, the R waves wave crest point for detecting ECG signal simultaneously put each interception certain amount sampled point forwards, backwards on the basis of it
It is clapped as a holocentric;
S12, it is divided into several holocentrics to clap entire electrocardiogram;
S13, the holocentric amplitude of beat value is normalized;
S14, the holocentric bat Jing Guo normalized is divided into training set and collection two parts to be sorted;
The step S20 includes:
S21, training set and/or collection to be sorted are made into the output that Second Order Convolution processing obtains corresponding heart bat using PCA algorithms
Matrix;
S22, the output matrix clapped training set and/or the collection heart to be sorted carry out binary system Hash coding and blocked histogram
Processing obtains the feature vector that training set and/or the collection heart to be sorted are clapped;
The step S30 includes:
S31, grader is trained using the feature vector that the training set heart is clapped;
S32, by the trained grader of feature vector input that the collection heart to be sorted is clapped classify simultaneously output category result.
Further, in the step S11, the quantity of sampled point is intercepted depending on sample frequency.
Further, in the step S21, making Second Order Convolution processing using PCA algorithms includes:
Extract L1A first layer PCA filters;
By L1A first layer PCA filters clap matrix with each heart and are first time process of convolution acquisition L1A local feature
Matrix;
Extract L2A second layer PCA filters;
By L2A second layer PCA filters do second of process of convolution with the local feature matrix and obtain L2A part
Eigenmatrix.
Further, when extracting the feature of sample to be sorted, directly application is filtered by the first layer PCA that training set extracts
Device and second layer PCA filters.
Further, the method for the extraction PCA filters includes:
It is that the heart claps matrix that the heart, which is clapped vector reconstruction,;
Centralization handles the heart and claps matrix;
Matrix, which is clapped, using the heart of centralization sets up pending matrix;
It treats processing array and carries out principal component analysis acquisition PCA filters.
Further, the expression of the PCA filters is as follows:
Wherein XXTFor the covariance matrix of X, ql() extracts the feature vector of matrix in bracket, matk1,k2() will be in bracket
Vector be reconstructed into matrix respectively, these matrixes are PCA filters.
Further, the binary system Hash coding includes:Two are carried out to all matrixes in a local feature matrix group
Value carries out Hash coding by hash function, and the decimal system is combined as a matrix.
Further, the blocked histogram, which is handled, includes:Block size is chosen, and overlap coefficient is chosen, conversion vector, histogram
Statistics is connect with vector, obtains the feature vector that the heart is clapped.
Further, the grader includes but not limited to Linear SVM, KNN graders, BP neural network grader and random
Forest.
The beneficial effects of the invention are as follows:A kind of cardioelectric characteristic extracting process based on PCANet that the present invention uses, to making an uproar
Sound has stronger robustness, the heart to not removing noise clap carry out feature extraction and being input in grader can also obtain compared with
Good classifying quality;The unbalanced heart of quantity can be clapped and carry out accurate classification;Dimension is carried out to heart bat feature to carry
It rises, by intrinsic dimensionality that the heart is clapped into improving two orders of magnitude, then feature is clapped to this high-dimensional heart using grader and is carried out
Classification can obtain considerable effect.
Description of the drawings
Fig. 1 is the overview flow chart of the present invention;
Fig. 2 is the pretreatment schematic diagram of the present invention;
Fig. 2-A are the electrocardio schematic diagrames of the 2-A in Fig. 2 of the present invention;
Fig. 2-B are the electrocardio schematic diagrames of the 2-B in Fig. 2 of the present invention;
Fig. 2-C are the electrocardio schematic diagrames of the 2-C in Fig. 2 of the present invention;
Fig. 2-D are the electrocardio schematic diagrames of the 2-D in Fig. 2 of the present invention;
Fig. 3 is that the first layer PCA filters of the present invention establish schematic diagram;
Fig. 4 is that the second layer PCA filters of the present invention establish schematic diagram;
Fig. 5 is the process of convolution schematic diagram twice of the present invention;
Fig. 6 is that the heart of the present invention claps the schematic diagram of feature output processing;
Fig. 7 is the classification results figure of a certain embodiment of the present invention.
Specific implementation mode
To make the objectives, technical solutions, and advantages of the present invention clearer, right in the following with reference to the drawings and specific embodiments
The present invention is described in detail.The game graph demand of the present invention and the system and method for design are suitable for game movie image
Exploitation.
Referring to Fig.1, a kind of cardioelectric characteristic extracting process based on PCANet of the invention, includes the following steps:It is S10, right
Electrocardiogram carries out pretreatment and obtains training set and collection to be sorted, S20, using PCANet respectively to training set and it is to be sorted collect into
The feature extraction that the row heart is clapped, S30, the heart extracted using training set are clapped feature training grader and use it for the collection heart to be sorted
Clap the classification of feature;
The step S10 includes:
S11, the R waves wave crest point for detecting ECG signal simultaneously put each interception certain amount sampled point forwards, backwards on the basis of it
It is clapped as a holocentric;
S12, it is divided into several holocentrics to clap entire electrocardiogram;
S13, the holocentric amplitude of beat value is normalized;
S14, the holocentric bat Jing Guo normalized is divided into training set and collection two parts to be sorted;
The step S20 includes:
S21, training set and/or collection to be sorted are made into the output that Second Order Convolution processing obtains corresponding heart bat using PCA algorithms
Matrix;
S22, the output matrix clapped training set and/or the collection heart to be sorted carry out binary system Hash coding and blocked histogram
Processing obtains the feature vector that training set and/or the collection heart to be sorted are clapped;
The step S30 includes:
S31, grader is trained using the feature vector that the training set heart is clapped;
S32, by the trained grader of feature vector input that the collection heart to be sorted is clapped classify simultaneously output category result.
In order to keep explanation more specific, the present embodiment enumerates some specific data and formula is illustrated the invention.
With reference to Fig. 2 and Fig. 2-A to Fig. 2-D, for pretreatment stage, including:
Detection electrocardiogram is simultaneously divided into several holocentrics bats:
With MLII, V5 two in the Relational databases such as a large amount of patient ECG either MIT-BIH in the present embodiment
For the ECG signal of lead as experimental subjects, sample frequency is 360 to 250Hz.It is calculated using Pan-Tompkins in the present embodiment
Method carries out R wave detections to a large amount of ECG signal, and R wave crest locations are labeled;By all R waves waves in ECG signal
The midpoint that peak dot is clapped as each heart is arranged the 149th sample point on the left of the R waves wave crest point that each heart is clapped and is clapped as the corresponding heart
Starting point, the 150th, right side of the R waves wave crest point heart is clapped claps terminal as the corresponding heart, and 300 points that have altogether of origin-to-destination is taken to make
Vectorial sample is clapped for a heart.
The heart claps vectorial normalized:
For normalized, it is intended to which the amplitude for the curve for clapping all hearts is reduced between 0 to 1, and this method can
Final classification results are promoted to a certain extent, and the method for normalizing that the present embodiment uses standardizes for Min-Max.
Wherein y is the original value of certain point during the heart is clapped, and min is the minimum value in a heart is clapped, and max is in the heart is clapped
Maximum value, x are the amplitude after point normalization.
Set up training set and collection to be sorted:
The heart after normalization is clapped and carries out category division according to label, the five classes heart therein is taken out according to the standard of AAMI
Clap, a certain distance quantitatively had per one kind heart bat, take out in being clapped per a kind of heart sub-fraction (about ten/
One), and by the ten of this five class/ wholeheartedly clap and be constructed as the training set heart and clap, the training set heart is clapped for being carried in next feature
It takes and generates filter and the training to grader in step.The bat of the remaining heart is constructed as collection to be sorted, the collection heart to be sorted is clapped
By the classification for grader.
With reference to Fig. 3-Fig. 5, for the feature extraction phases that the training set heart is clapped, including:
With reference to Fig. 3, it is that the heart claps matrix that being needed before extraction first layer PCA filters, which first will intentionally clap vector reconstruction, this
Assume there is N number of heart to clap vector in embodiment, each heart claps vector and contains 300 sampled points, and it is that m × n is big that the heart, which is clapped vector reconstruction,
The small heart claps matrix, and reconstruct mode is as follows:
Vector is clapped to all hearts and carries out above-mentioned processing method, N number of heart is finally obtained and claps matrix.
First layer PCA filters are extracted using Principal Component Analysis (PCA algorithms), clap each heart the extraction process of matrix
It is as follows:
First stage block samples:
It uses the window p1 of k1xk2 sizes to clap matrix to the heart and carries out step-length as 1 sliding, obtained matrix will be slided every time
It extracts, i.e., it is final to obtain 300 sampling block matrix, it is cascaded, the sampling block matrix of matrix is clapped as i-th of heart
It indicates as follows:
First stage centralization:
300 sampling block matrix that i-th of heart is clapped to matrix carry out zero averaging, that is, make each sampling after going averagely
The average value of all elements in block matrix is 0, claps i-th of heart the zero averaging of matrix, obtains 300 pending samplings
Block, it is specific as follows:
First stage reconstruct combination:
To intentionally clapping matrix into row block sampling and centralization processing, 300 pending sampling blocks of N groups are obtained, by every group
Pending sampling block is reconstructed into vector form again, and is combined into the pending heart containing 300N column vectors and claps matrix:
Obtain first layer PCA filters:
It is assumed that being L in first layer PCA filter quantity1, the purpose of PCA algorithms is exactly by finding a series of standard just
Matrix is handed over to minimize reconstructed error:
The solution of this problem is exactly the principal component analysis of classics, the i.e. preceding L of the covariance matrix of matrix X1A feature to
Then amount is reconstructed processing to each feature vector and obtains L1A first layer PCA filters indicate as follows:
With reference to Fig. 5, L is obtained1After a first layer PCA filters, it is clapped into matrix with each heart and is done at first time convolution
Reason, it is specific as follows:
Wherein ΥiMatrix is clapped for the heart,For local feature matrix.After the step, each original heart is clapped matrix and is reflected
It penetrates as L1New matrix is named as local feature matrix, by L by a new matrix1The collection of a new matrix composition is combined into local feature square
Battle array group, therefore the bat of N number of heart will finally be mapped as N number of local feature matrix group.
With reference to Fig. 4, second layer PCA filters are extracted, including:
Second stage block samples:
The process is similar to first stage block sampling process, sample objects for intentionally clap pass through first time process of convolution after
The L of acquisition1A local feature matrix, sampling window p2 sizes are r1xr2, finally obtain several sampling block matrix.
Second stage centralization:
All sampling block matrix obtained to previous step do centralization processing, processing method and first stage centralization
Method is consistent, and which is not described herein again;A certain amount of pending sampling block is finally obtained,
Wherein m and n is respectively the line number and columns of sampling block.
Second stage reconstruct recombination:
Reconstruction processing is done to each pending sampling block and obtains vector of samples, obtained all vector of samples are combined as waiting for
Processing array,
The above processing is done to the local feature matrix group intentionally clapped, is obtained
Obtain second layer PCA filters:
It is consistent with first layer PCA filtered methods are obtained, it is formed by choosing the corresponding feature vector of covariance matrix,
Wherein YYTFor the covariance matrix of Y, qλ() extracts the feature vector of matrix in bracket, matk1,k2() will be in bracket
Vector be reconstructed into matrix respectively, these matrixes are second layer PCA filters, thus obtain L2Second layer PCA filters.
With reference to Fig. 5, L is obtained2After a second layer PCA filters, itself and each local feature matrix are done into the second secondary volume
Product processing, it is specific as follows:
WhereinFor i-th local feature matrix pass through withThe obtained secondary local feature matrix group of convolution.
With reference to Fig. 6, PCANet output layers are handled, including:
Binary system Hash encodes:
Binary conversion treatment is done to all secondary local feature matrixes, one time local feature group will be obtained by binary conversion treatment
Obtain L2A binaryzation time local feature matrix, is then encoded by Hash by L2A binaryzation time local feature matrix is converted into ten
System matrix, formula are as follows:
Wherein H () function is similar to a unit-step function, and element value is mapped as 0 and 1,For decimal system matrix,
The function is by L2Each element chemical conversion 0 to 2 of a secondary local feature matrixλ-1Between to get to element in only comprising whole
Number and zero, and by L2A matrix corresponding element is added to obtain a decimal system matrix.
Blocked histogram processing:
By L1A secondary local feature group all encodes to obtain L by binary system Hash1A decimal system matrix, first a certain
Block sampling is done to each decimal system matrix under overlap coefficient, and converts the sampling block of acquisition to vector of samples, one will be passed through
All vector of samples that decimal system matrix obtains are combined as pending matrix, for each pending matrix, are classified as B
Block handles the matrix using the method for statistics with histogram, converts eigenmatrix to vector form.Finally by L1It is a
The L that decimal system matrix maps1A vector is cascaded into a feature long vector, this feature long vector is to extract
The heart claps feature;Formula is:
The feature extracting method and training set clapped for the collection heart to be sorted are almost the same, and specific method is not repeating, still
The first layer PCA filters and the second layer extracted by training set can be directly applied when extracting the feature of sample to be sorted
PCA filters, therefore PCA filter extraction steps need not be executed again when treating category set and carrying out feature extraction.
The present embodiment using Linear SVM, KNN graders, BP neural network grader and random forest Various Classifiers on Regional come
Verify the feature extraction effect of the present invention.
Category criteria is clapped according to the heart that American Medical instrument promotes association (AAMI) specified, the bat of the 15 class hearts is integrated into 5 classes,
It is N, S, V, F, Q respectively, the content that this five classes heart is clapped is as shown in table 1.
1 AAMI of table specifies the heart to clap category division table
The present embodiment tests to the classifying quality of the present invention using the noisy ECG signal of MIT-BIH databases, table 2
Number is clapped for the pretreated all kinds of hearts, it is 107168 that the heart, which claps sum,.By table 2 this it appears that the quantity of all kinds of hearts bat is tight
Weight is uneven, wherein the minimum F class hearts clap 1 percent clapped less than the N class hearts.The present embodiment has carried out the experiment of 10 subseries,
Every time experiment from it is all kinds of it is middle randomly select about ten respectively/ wholeheartedly clap composition training set train the grader, the heart to clap sum and be
10715, the remaining heart is clapped carrys out testing classification effect as collection to be sorted.
Classification | The heart claps sum | Training set heart umber of beats amount | Collection heart umber of beats amount to be sorted |
N | 90411 | 9041 | 89370 |
S | 2778 | 277 | 2501 |
V | 7227 | 722 | 6505 |
F | 802 | 80 | 722 |
Q | 5950 | 595 | 5355 |
It amounts to | 107168 | 10715 | 96453 |
2 heart of table claps sample number scale
When being classified using various graders, parameter configuration is identical, as shown in table 3.
Project | Parameter |
The heart claps matrix size | 15x20 |
First stage PCA filter quantity | 9 |
Second stage PCA filter quantity | 9 |
Block sample size in PCA filter extraction process | 7x7 |
Blocked histogram processing stage block sample size | 7x7 |
Blocked histogram processing stage block samples Duplication | 0.5 |
The classifier parameters allocation list of 3 the present embodiment of table
The classification results of acquisition are as shown in Figure 7, it can be seen that the ecg characteristics extraction side based on PCANet of the invention
Method has robustness to noise, carries out to prevent removal from making an uproar without the removal for carrying out noise when the heart claps feature extraction classification
The harmful effect that sound process extracts ecg characteristics, this method has significant advantage when to there is the bat of the noise heart to classify;And
And it is also preferable to unbalanced beat classification effect.
The above, only presently preferred embodiments of the present invention, the invention is not limited in the above embodiments, as long as
It reaches the technique effect of the present invention with identical means, should all belong to the scope of protection of the present invention.In the protection model of the present invention
Its technical solution and/or embodiment can have a variety of different modifications and variations in enclosing.
Claims (9)
1. a kind of cardioelectric characteristic extracting process based on PCANet, which is characterized in that include the following steps:S10, to electrocardiogram into
Row pretreatment obtains training set and collection to be sorted, S20, the spy for carrying out heart bat to training set and collection to be sorted respectively using PCANet
Sign extraction, S30, the heart extracted using training set are clapped feature training grader and use it for point that the collection heart to be sorted claps feature
Class;
The step S10 includes:
S11, detect ECG signal R waves wave crest point and put using on the basis of it forwards, backwards respectively interception certain amount sampled point as
One holocentric is clapped;
S12, it is divided into several holocentrics to clap entire electrocardiogram;
S13, the holocentric amplitude of beat value is normalized;
S14, the holocentric bat Jing Guo normalized is divided into training set and collection two parts to be sorted;
The step S20 includes:
S21, training set and/or collection to be sorted are made into the output matrix that Second Order Convolution processing obtains corresponding heart bat using PCA algorithms;
S22, the output matrix clapped training set and/or the collection heart to be sorted carry out binary system Hash coding and blocked histogram processing
Obtain the feature vector that training set and/or the collection heart to be sorted are clapped;
The step S30 includes:
S31, grader is trained using the feature vector that the training set heart is clapped;
S32, by the trained grader of feature vector input that the collection heart to be sorted is clapped classify simultaneously output category result.
2. the cardioelectric characteristic extracting process according to claim 1 based on PCANet, it is characterised in that:The step S11
In, the quantity of sampled point is intercepted depending on sample frequency.
3. the cardioelectric characteristic extracting process according to claim 1 based on PCANet, it is characterised in that:The step S21
In, making Second Order Convolution processing using PCA algorithms includes:
Extract L1A first layer PCA filters;
By L1A first layer PCA filters clap matrix with each heart and are first time process of convolution acquisition L1A local feature matrix;
Extract L2A second layer PCA filters;
By L2A second layer PCA filters do second of process of convolution with the local feature matrix and obtain L2A secondary local feature
Matrix.
4. the cardioelectric characteristic extracting process according to claim 3 based on PCANet, it is characterised in that:It is to be sorted extracting
When the feature of sample, directly using the first layer PCA filters and second layer PCA filters extracted by training set.
5. the cardioelectric characteristic extracting process according to claim 3 based on PCANet, it is characterised in that:The extraction PCA
The method of filter includes:
It is that the heart claps matrix that the heart, which is clapped vector reconstruction,;
Centralization handles the heart and claps matrix;
Matrix, which is clapped, using the heart of centralization sets up pending matrix;
It treats processing array and carries out principal component analysis acquisition PCA filters.
6. the cardioelectric characteristic extracting process according to claim 3 based on PCANet, it is characterised in that:The PCA filtering
The expression of device is as follows:
Wherein XXTFor the covariance matrix of X, ql() extracts the feature vector of matrix in bracket, matk1,k2() by bracket to
Amount is reconstructed into matrix respectively, these matrixes are PCA filters.
7. the cardioelectric characteristic extracting process according to claim 1 based on PCANet, it is characterised in that:The binary system is breathed out
Uncommon coding includes:Binaryzation is carried out to all matrixes in a local feature matrix group, Hash volume is carried out by hash function
Code, the decimal system are combined as a matrix.
8. the cardioelectric characteristic extracting process according to claim 1 based on PCANet, it is characterised in that:The piecemeal histogram
Figure is handled:Block size is chosen, and overlap coefficient is chosen, and conversion vector, statistics with histogram is connect with vector, obtains the spy that the heart is clapped
Sign vector.
9. the cardioelectric characteristic extracting process according to claim 1 based on PCANet, it is characterised in that:The grader packet
Include but be not limited to Linear SVM, KNN graders, BP neural network grader and random forest.
Priority Applications (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810434968.3A CN108596142B (en) | 2018-05-09 | 2018-05-09 | PCANet-based electrocardiogram feature extraction method |
PCT/CN2018/093662 WO2019214026A1 (en) | 2018-05-09 | 2018-06-29 | Ecg feature extraction method employing pcanet |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810434968.3A CN108596142B (en) | 2018-05-09 | 2018-05-09 | PCANet-based electrocardiogram feature extraction method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN108596142A true CN108596142A (en) | 2018-09-28 |
CN108596142B CN108596142B (en) | 2022-01-11 |
Family
ID=63635914
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810434968.3A Active CN108596142B (en) | 2018-05-09 | 2018-05-09 | PCANet-based electrocardiogram feature extraction method |
Country Status (2)
Country | Link |
---|---|
CN (1) | CN108596142B (en) |
WO (1) | WO2019214026A1 (en) |
Cited By (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109602414A (en) * | 2018-11-12 | 2019-04-12 | 安徽心之声医疗科技有限公司 | A kind of ecg signal data Enhancement Method of multi-angle of view conversion |
CN109620211A (en) * | 2018-11-01 | 2019-04-16 | 吉林大学珠海学院 | A kind of intelligent abnormal electrocardiogram aided diagnosis method based on deep learning |
CN110025308A (en) * | 2019-04-09 | 2019-07-19 | 澳门大学 | A kind of cardioelectric characteristic extracting process, the heart clap recognition methods and device |
CN110141214A (en) * | 2019-04-23 | 2019-08-20 | 首都师范大学 | A kind of mask method of electrocardiogram identification and its application |
CN110738166A (en) * | 2019-10-14 | 2020-01-31 | 西南大学 | Fishing administration monitoring system infrared target identification method based on PCNN and PCANet and storage medium |
CN112257573A (en) * | 2020-10-21 | 2021-01-22 | 吉林大学 | ECG identity recognition method based on t-SNE and Adaboost |
CN112883803A (en) * | 2021-01-20 | 2021-06-01 | 武汉中旗生物医疗电子有限公司 | Deep learning-based electrocardiosignal classification method and device and storage medium |
CN113486752A (en) * | 2021-06-29 | 2021-10-08 | 吉林大学 | Emotion identification method and system based on electrocardiosignals |
CN113647954A (en) * | 2021-07-07 | 2021-11-16 | 吉林大学 | Cardiovascular disease identification method, device and medium of two-channel hybrid network model |
Families Citing this family (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112200763A (en) * | 2020-08-24 | 2021-01-08 | 江苏科技大学 | Liver function grading method based on liver CT image |
CN112883812B (en) * | 2021-01-22 | 2024-05-03 | 广东白云学院 | Deep learning-based lung sound classification method, system and storage medium |
CN114707546B (en) * | 2022-03-31 | 2024-03-26 | 山东科技大学 | 12-lead ECG signal classification method based on combined two-dimensional features |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20070208265A1 (en) * | 2006-03-02 | 2007-09-06 | Jean-Philippe Couderc | Method and system for assessing repolarization abnormalities |
CN103815897A (en) * | 2014-02-28 | 2014-05-28 | 吉林大学 | Electrocardiogram characteristic extraction method |
CN104102915A (en) * | 2014-07-01 | 2014-10-15 | 清华大学深圳研究生院 | Multiple-template matching identity recognition method based on ECG (Electrocardiogram) under electrocardiogram abnormality state |
Family Cites Families (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US8696585B2 (en) * | 2008-09-30 | 2014-04-15 | Nellcor Puritan Bennett Ireland | Detecting a probe-off event in a measurement system |
CN106725428B (en) * | 2016-12-19 | 2020-10-27 | 中国科学院深圳先进技术研究院 | Electrocardiosignal classification method and device |
CN107122788B (en) * | 2017-03-09 | 2020-08-25 | 吉林大学 | Identity recognition method and device based on electrocardiosignals |
CN107239684A (en) * | 2017-05-22 | 2017-10-10 | 吉林大学 | A kind of feature learning method and system for ECG identifications |
CN107456226A (en) * | 2017-08-15 | 2017-12-12 | 东南大学 | Wearable ECG monitoring and diagnostic system based on cloud |
-
2018
- 2018-05-09 CN CN201810434968.3A patent/CN108596142B/en active Active
- 2018-06-29 WO PCT/CN2018/093662 patent/WO2019214026A1/en active Application Filing
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20070208265A1 (en) * | 2006-03-02 | 2007-09-06 | Jean-Philippe Couderc | Method and system for assessing repolarization abnormalities |
CN103815897A (en) * | 2014-02-28 | 2014-05-28 | 吉林大学 | Electrocardiogram characteristic extraction method |
CN103815897B (en) * | 2014-02-28 | 2015-07-15 | 吉林大学 | Electrocardiogram characteristic extraction method |
CN104102915A (en) * | 2014-07-01 | 2014-10-15 | 清华大学深圳研究生院 | Multiple-template matching identity recognition method based on ECG (Electrocardiogram) under electrocardiogram abnormality state |
Non-Patent Citations (2)
Title |
---|
陈义: "心电信号的异常心律分类算法研究", 《中国优秀硕士学位论文全文数据库.医药卫生科技辑》 * |
顾凌云: "基于PCANet和SVM的谎言测试研究", 《电子学报》 * |
Cited By (14)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109620211A (en) * | 2018-11-01 | 2019-04-16 | 吉林大学珠海学院 | A kind of intelligent abnormal electrocardiogram aided diagnosis method based on deep learning |
CN109602414A (en) * | 2018-11-12 | 2019-04-12 | 安徽心之声医疗科技有限公司 | A kind of ecg signal data Enhancement Method of multi-angle of view conversion |
CN110025308A (en) * | 2019-04-09 | 2019-07-19 | 澳门大学 | A kind of cardioelectric characteristic extracting process, the heart clap recognition methods and device |
CN110025308B (en) * | 2019-04-09 | 2021-09-10 | 澳门大学 | Electrocardio feature extraction method, heart beat identification method and device |
CN110141214A (en) * | 2019-04-23 | 2019-08-20 | 首都师范大学 | A kind of mask method of electrocardiogram identification and its application |
CN110738166A (en) * | 2019-10-14 | 2020-01-31 | 西南大学 | Fishing administration monitoring system infrared target identification method based on PCNN and PCANet and storage medium |
CN110738166B (en) * | 2019-10-14 | 2023-04-18 | 西南大学 | Fishing administration monitoring system infrared target identification method based on PCNN and PCANet and storage medium |
CN112257573B (en) * | 2020-10-21 | 2022-06-24 | 吉林大学 | ECG identity recognition method based on t-SNE and Adaboost |
CN112257573A (en) * | 2020-10-21 | 2021-01-22 | 吉林大学 | ECG identity recognition method based on t-SNE and Adaboost |
CN112883803A (en) * | 2021-01-20 | 2021-06-01 | 武汉中旗生物医疗电子有限公司 | Deep learning-based electrocardiosignal classification method and device and storage medium |
CN112883803B (en) * | 2021-01-20 | 2023-09-01 | 武汉中旗生物医疗电子有限公司 | Electrocardiogram signal classification method, device and storage medium based on deep learning |
CN113486752A (en) * | 2021-06-29 | 2021-10-08 | 吉林大学 | Emotion identification method and system based on electrocardiosignals |
CN113486752B (en) * | 2021-06-29 | 2023-06-16 | 吉林大学 | Emotion recognition method and system based on electrocardiosignal |
CN113647954A (en) * | 2021-07-07 | 2021-11-16 | 吉林大学 | Cardiovascular disease identification method, device and medium of two-channel hybrid network model |
Also Published As
Publication number | Publication date |
---|---|
CN108596142B (en) | 2022-01-11 |
WO2019214026A1 (en) | 2019-11-14 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108596142A (en) | A kind of cardioelectric characteristic extracting process based on PCANet | |
CN106473750B (en) | Personal identification method based on photoplethysmographic optimal period waveform | |
CN104523266B (en) | A kind of electrocardiosignal automatic classification method | |
CN110786850B (en) | Electrocardiosignal identity recognition method and system based on multi-feature sparse representation | |
CN104102915B (en) | Personal identification method based on ECG multi-template matching under a kind of anomalous ecg state | |
Arif | Robust electrocardiogram (ECG) beat classification using discrete wavelet transform | |
CN106951753B (en) | Electrocardiosignal authentication method and device | |
CN108537100A (en) | A kind of electrocardiosignal personal identification method and system based on PCA and LDA analyses | |
CN108511055B (en) | Ventricular premature beat recognition system and method based on classifier fusion and diagnosis rules | |
Zhang et al. | A framework for automatic time-domain characteristic parameters extraction of human pulse signals | |
CN106377247A (en) | Feature selection-based arrhythmia classification method | |
CN108030494A (en) | Electrocardiosignal error flag training sample recognition methods based on cross validation | |
CN103083013A (en) | Electrocardio signal QRS complex wave detection method based on morphology and wavelet transform | |
CN101002682A (en) | Method for retrieval and matching of hand back vein characteristic used for identification of status | |
CN109948396B (en) | Heart beat classification method, heart beat classification device and electronic equipment | |
CN105320969A (en) | A heart rate variability feature classification method based on multi-scale Renyi entropy | |
CN113486752B (en) | Emotion recognition method and system based on electrocardiosignal | |
CN104887263A (en) | Identity recognition algorithm based on heart sound multi-dimension feature extraction and system thereof | |
CN113057648A (en) | ECG signal classification method based on composite LSTM structure | |
CN113116300A (en) | Physiological signal classification method based on model fusion | |
CN115530788A (en) | Arrhythmia classification method based on self-attention mechanism | |
Wei et al. | Effective extraction of Gabor features for adaptive mammogram retrieval | |
CN1540568A (en) | Identification and authenticaton method | |
Li et al. | A novel abnormal ECG beats detection method | |
Muthuvel et al. | ECG signal feature extraction and classification using harr wavelet transform and neural network |
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 | ||
GR01 | Patent grant | ||
GR01 | Patent grant |