CN108113665A - A kind of automatic noise-reduction method of electrocardiosignal - Google Patents
A kind of automatic noise-reduction method of electrocardiosignal Download PDFInfo
- Publication number
- CN108113665A CN108113665A CN201711339403.9A CN201711339403A CN108113665A CN 108113665 A CN108113665 A CN 108113665A CN 201711339403 A CN201711339403 A CN 201711339403A CN 108113665 A CN108113665 A CN 108113665A
- Authority
- CN
- China
- Prior art keywords
- electrocardiosignal
- value
- reserve pool
- training
- network
- 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
-
- 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
- A61B5/24—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
- A61B5/72—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
Abstract
The invention discloses a kind of automatic noise reduction algorithm of electrocardiosignal, processing procedure is：A）It establishes echo state network and initializes；B）It obtains human ecg signal and builds training set on this basis；C）Using the on-line training Algorithm for Training echo state network based on recurrent least square method, trained echo state network is obtained；D）Test set is built, test set is inputted into trained echo state network, obtains clean electrocardiosignal.Processing through the method for the present invention, the clean electrocardiosignal after denoising not only effectively filter out noise, and have recovered electrocardiosignal characteristics of low-frequency ripple, remain the effective information of electrocardiosignal.
Description
Technical field
It is detected automatically the present invention relates to electrocardiosignal and analysis method, specifically a kind of automatic noise reduction side of electrocardiosignal
Method.
Background technology
Angiocardiopathy has become the number one killer for health of people, has high incidence, high disability rate, high mortality.
Wherein, the low-frequency component of electrocardiosignal is very critically important to the intelligent diagnostics of angiocardiopathy, such as the ST of one of its low-frequency component
Section, the potential change situation after represent sequences of ventricular depolarization, in a period of time that ventricular bipolar starts, is diagnosis of myocardial ischemia
Important indicator.But at the same time, electrocardiosignal has the general character of bioelectrical signals in itself：Amplitude is faint, low frequency, impedance are big,
Randomness etc., these features cause electrocardiosignal to be highly prone to the interference of various noises in gatherer process.It is especially remote nowadays
Under journey medical context, noise during ecg signal acquiring presents increasingly complex feature, so as to easily cause the low of electrocardiosignal
The phenomenon that frequency part is flooded by Complex Noise, and effective information is lost.
There are three types of currently used methods：The first, empirical mode decomposition (EMD) is extraction intrinsic mode function (IMF)
So as to carry out noise reduction to noisy ECG signal, 95% white Gaussian noise as a result can be removed.However, EMD can not be isolated
The signal similar to ECG signal frequency.Therefore, this noise-reduction method is possible to filter out from signal by P ripples and T ripples, so as to cause
Mistaken diagnosis.Second, principal component analysis (PCA) and independent component analysis (ICA) contribute to from non-Gaussian noise signal find
The blind source separate technology of independent source signal, for having apparent advantage with interior filtering.But its model is for noisy ECG signal
Slight change it is very sensitive, be not too much applicable in ECG signal contain more complicated noise situation.The third, small echo is gone
Method for de-noising is achieved in terms of high-frequency noise is removed good as a result, but the selection of its wavelet basis and the setting of threshold value are to need to lead to
Substantial amounts of experiment is crossed to determine, and has direct influence to final filter effect.It can be seen that has algorithm in noise reduction process
Effective low-frequency information of electrocardiosignal is easily lost, influences the judgement of doctor.
The content of the invention
The object of the present invention is to provide a kind of automatic noise reduction algorithm of electric signal, to be protected while electrocardiosignal noise reduction is realized
Characteristics of low-frequency ripple is stayed, overcomes the problems, such as that existing algorithm is easily lost effective low-frequency information of electrocardiosignal in noise reduction process.
The object of the present invention is achieved like this：
The automatic noise reduction algorithm of a kind of electrocardiosignal, including following steps：
A) establish echo state network and initialize；
B) obtain human ecg signal and build training set on this basis；
C the on-line training Algorithm for Training echo state network based on recurrent least square method) is utilized, obtains and preserves training
Good echo state network；
D) with adjacent electrocardiosignal structure test set after training set, test set is inputted into trained echo state net
Network obtains clean electrocardiosignal.
The automatic noise reduction algorithm of the electrocardiosignal, step A) include initialization weight matrix and network parameter initialization
Setting, wherein：
1. initialization weight matrix process is：
The weight matrix W that input unit is connected with reserve pool insideinWith reserve pool intrinsic nerve member connection weight matrix W
Random initializtion is allowed to all obey being uniformly distributed；
By the connection weight matrix W of output unit and reserve poolback, reserve pool and output unit connection weight matrix Wout
It is initialized as null matrix；
2. network parameter initial setting process is：
Input neuron number is set as 4, and output neuron number is set as 1, and neuron number is set as N in reserve pool
=1000 ± 300；
The sparse degree SD=m/N of reserve pool, wherein, m is the neuron number being connected with each other in reserve pool, SD value ranges
For 1~5%；
The value of reserve pool spectral radius SR is debugged with 0.1 for step-length according in the range of 0~1, with result it is optimal when
Corresponding SR values be final value, SR=max { abs (characteristic value of W) }；
It in the range of -1~1 with 0.01 is that step-length is debugged that the value of reserve pool input unit scale IS, which is, with result
Corresponding IS values are final value when optimal.
The automatic noise reduction algorithm of the electrocardiosignal, step B) detailed process is as follows：
1. several continuous hearts are intercepted from acquired human ecg signal data claps x (n), n=1,2 ..., M, M=55
± 2, then with the noisy electrocardiosignal x (n) at current time, the noisy electrocardiosignal x (n-1) of its last moment, current time
The first derivative x ' (n) of noisy electrocardiosignal, the second dervative x " (n) of the noisy electrocardiosignal at current time form echo shape
The input signal x of state networkr(n), it is expressed as
It is d (n) to define the clean heart beat of data corresponding to the noisy electrocardiosignal x (n) at current time；
By xr(n) training set (x is formed with d (n)r(n), d (n), n=1,2 ..., M)；
2. it definesWherein g () is output neuron activation primitive, is elected as identical
Function, i.e. output neuron are a linear neuron, and in the function, matrix v (n) is defined as Wherein r (n) is reserve pool state variable；
3. define error e (n)=d (n)-y (n).
The automatic noise reduction algorithm of the electrocardiosignal, step C) detailed process is：
1. the reserve pool state variable initial value of selected network is 0, i.e. r (0)=0, training set (xr(n), d (n), n=1,
2 ..., M) in sample xr(n) W is passed throughin, d (n) is by WbackIt is added separately in reserve pool, utilizes formula r (n)=(1- α) [f
(W·r(n-1)+Win·u(n)+WbackD (n))]+α r (n-1) carry out reserve pool state variable r (n) update；
Wherein f () is reserve pool neuron activation functions, elects nonlinear tanh () function as；
U (n)=ψ (n-1) v (n), ψ (n) are the correlation matrix of v (n) and pass through formula It is updated, wherein,λ is forgetting factor, and value is 0.95
~1 scope is debugged with 0.0001 for step-length, using result it is optimal when λ value as final value, ψ (0)=δ I, I is unit square
Battle array, δ be a minimum positive number, value in the range of 0~0.001 with 0.00001 for step-length debug, with result it is optimal when δ
It is worth for final value；
α is forgetting rate, and value is debugged in the range of 0~1 with 0.1 step-length, using α value of result when optimal as most
Whole value；
2. according to formula Wout(n)=Wout(n-1)+k (n) e (n) carry out output weight WoutUpdate；
3. by updating Wout(n) and r (n) makes y (n) approach clean ECG signal d (n), makes error e (n) that can be similar to
0, obtain the echo state network of training completion.
The automatic noise reduction algorithm of the electrocardiosignal, step D) detailed process is：
The M heart intercepted during building training set adjacent T heart bat structure test set, test set (X after clappingr
(n), n=1,2 ..., T, T=200 ± 10), wherein,I.e. it includes the electrocardiosignal X at current time
(n), an electrocardiosignal X (n-1), its first derivative X ' (n) and second dervative X " (n) thereon；
By test set (Xr(n), n=1,2 ..., T) in Xr(n) echo state network that input training is completed, network are defeated
Go out clean electrocardiosignal Y (n).
Processing through the method for the present invention, the clean electrocardiosignal after denoising not only effectively filters out noise, and has recovered electrocardio
Signal characteristics of low-frequency ripple remains the effective information of electrocardiosignal.
Description of the drawings
Fig. 1 is the method for the present invention implementation process flow chart.
Fig. 2 is echo state network structure diagram.
Fig. 3 is human ecg signal waveform configuration schematic diagram.
Fig. 4 is noisy electrocardiosignal figure before filtering.
Fig. 5 is clean electrocardiosignal figure after filtering.
Specific embodiment
The present embodiment inside saves as 128.00GB, Win7,64 bit manipulation in Intel Xeon CPU E5-2697 2.70GHz
It is realized in the computer of system, entire electrocardiosignal Algorithms for Automatic Classification is realized using Matlab language.
With reference to Fig. 1, implementation process of the invention is as follows：
A echo state network (Echo State Network, ESN)) is established：
1., initialization weight matrix：
Random initializtion input unit and the weight matrix W being connected inside reserve poolinIt is connected with reserve pool intrinsic nerve member
Weight matrix W, and it is made all to obey and is uniformly distributed；
The connection weight matrix W of output unit and reserve poolback, reserve pool and output unit connection weight matrix WoutJust
Beginning turns to null matrix；
2., parameter setting：
The input neuron number of ESN is set as 4, and output neuron number is set as 1, of the neuron in reserve pool
Number is reserve pool scale N=1000；
The sparse degree SD of reserve pool of ESN is calculated by formula S D=m/N, wherein, m is connected with each other in reserve pool
M is taken as m=20 in this example, then SD=2% by neuron number；
Reserve pool spectral radius SR is defined as SR=max { abs (characteristic value of W) }, since as SR ＜ 1, ESN just has back
Sound state behavior, SR values are taken as SR=according to being debugged with 0.1 for step-length in the range of 0~1 when as a result optimal in this example
0.4；
Define WSThe sparse matrix that the equally distributed sparse degree of obedience generated at random for one is SD, then in reserve pool
Portion's neuron connection weight matrix W=SRWS, to ensure the spectral radius of W as SR；
Reserve pool input unit scale IS is to need be multiplied scale factor before input signal is connected to reserve pool,
It is multiplied by input signal with it, so as to input signal is converted into the input range of reserve pool neuron activation functions, god
It is [- 1,1] through first activation primitive input range, the value of IS is debugged in the range of -1~1 with 0.01 for step-length in this example,
As a result IS=[0.01 when optimal；0.01；0.9；0.8]；
B human body electrocardio original signal) is gathered with the frequency acquisition of 250Hz, waveform configuration is as shown in figure 3, and be stored as
Then the electrocardio original signal data stored with TXT documents is read computer by the data mode of TXT documents with Matlab softwares
In, as original electro-cardiologic signals, as shown in figure 4, structure training set：
Define xr(n) input signal for being ESN is to intercept the continuous M=55 heart from original electro-cardiologic signals to clap x (n),
N=1,2 ..., 55, with the noisy electrocardiosignal x (n) at current time, the noisy electrocardiosignal x (n-1) of its last moment, current
The first derivative x ' (n) of the noisy electrocardiosignal at moment, the second dervative x " (n) of the noisy electrocardiosignal at current time are formed,
It is expressed asFor the matrix of N × 4；Definition clean heart beat of data corresponding with working as front center bat is d
(n)；By xr(n) training set (x is formed with d (n)r(n), d (n), n=1,2 ..., 55).
DefinitionWherein g () is output neuron activation primitive, elects identical letter as
Number, i.e. output neuron are a linear neuron, and in the function, matrix v (n) is defined as
Wherein r (n) is the state variable of reserve pool, and diag (IS) represents to generate a diagonal matrix by diagonal of IS；
Define error e (n)=d (n)-y (n)；
C the on-line training Algorithm for Training echo state network based on recurrent least square method) is utilized：
1. the state variable initial value of the reserve pool of selected network is 0, i.e. r (0)=0, training set (xr(n), d (n), n=
1,2 ..., 55) in sample xr(n) W is passed throughin, d (n) is by WbackIt is added separately in reserve pool, utilizes formula r (n)=(1-
α)[f(W·r(n-1)+Win·u(n)+WbackD (n))]+α r (n-1) carry out reserve pool state r (n) update：
Wherein f () is reserve pool neuron activation functions, and stronger non-linear, choosing is needed when being modeled due to network
For nonlinear tanh () function；
U (n)=ψ in the function-1(n-1) v (n), ψ (n) are the inverse matrix of the correlation matrix of v (n) and under passing through
Formula is updated：
Wherein,λ is forgetting factor, fixed
Justice for a scope be 0.95 ＜ λ ＜ 1 constant, value is step-length with 0.0001, is debugged, with result it is optimal when λ value
For final value, λ=0.9999, ψ (0)=δ are taken in this example-1I, I be unit matrix, δ be a minimum positive number, value
Scope is 0~0.001, is debugged with 0.00001 for step-length, with result, optimal (input signal i.e. in training set puts network into again
In, when the error of obtained y (n) and d (n) is similar to 0) when δ values be final value, take δ=0.00172 in this example；
α is forgetting rate in the function, is defined as a positive number less than 1, value range is 0~1, with 0.1 step-length
It is debugged, using α value of result when optimal as final value, value is α=0.8 in this example.
2. according to formula Wout(n)=Wout(n-1)+k (n) e (n) carry out output weight WoutUpdate.
3. by updating Wout(n) and r (n) makes y (n) approach clean ECG signal d (n), makes error e (n) that can be similar to
0, it obtains and preserves the intact echo state network of training.
When y (n) approaches d (n), i.e., error e (n) is similar to 0, and network is stable state, preserves trained time shape
State network can be used to electrocardiosignal denoising.
D) verify：
55 hearts intercepted during building training set adjacent T=200 heart bat structure test set, test set after clapping
(Xr(n), n=1,2 ..., 200), wherein,Electrocardiosignal X (n) i.e. it includes current time, its
A upper electrocardiosignal X (n-1), its first derivative X ' (n) and second dervative X " (n)；
By test set (Xr(n), n=1,2 ..., 200) in Xr(n) echo state network that input training is completed, network
Clean electrocardiosignal Y (n) is exported, as shown in Figure 5.
It is opposite with normal electro-cardiologic signal waveforms (Fig. 2) that noisy electrocardiosignal (such as Fig. 4) is can be seen that from more than handling result
Than can be seen that, electrocardiosignal characteristics of low-frequency ripple T ripples and P ripple serious distortions under the interference of noise, the processing through the method for the present invention,
Clean electrocardiosignal (such as Fig. 5) after denoising not only effectively filters out noise, and has recovered electrocardiosignal characteristics of low-frequency ripple, remains
The effective information of electrocardiosignal.
Claims (5)
1. a kind of automatic noise reduction algorithm of electrocardiosignal, it is characterized in that, including following steps：
A) establish echo state network and initialize；
B) obtain human ecg signal and build training set on this basis；
C the on-line training Algorithm for Training echo state network based on recurrent least square method) is utilized, obtains and preserves trained
Echo state network；
D) with adjacent electrocardiosignal structure test set after training set, test set is inputted into trained echo state network,
Obtain clean electrocardiosignal.
2. the automatic noise reduction algorithm of electrocardiosignal according to claim 1, it is characterized in that, step A) include initialization weights
Matrix and network parameter initializing set, wherein：
1. initialization weight matrix process is：
The weight matrix W that input unit is connected with reserve pool insideinIt is random with reserve pool intrinsic nerve member connection weight matrix W
Initialization is allowed to all obey being uniformly distributed；
By the connection weight matrix W of output unit and reserve poolback, reserve pool and output unit connection weight matrix WoutInitialization
For null matrix；
2. network parameter initial setting process is：
Input neuron number is set as 4, and output neuron number is set as 1, and neuron number is set as N=in reserve pool
1000±300；
The sparse degree SD=m/N of reserve pool, wherein, m is the neuron number being connected with each other in reserve pool, and SD value ranges are 1
~5%；
The value of reserve pool spectral radius SR is debugged according in the range of 0~1 with 0.1 for step-length, and with result, optimal when institute is right
The SR values answered be final value, SR=max { abs (characteristic value of W) }；
It in the range of -1~1 with 0.01 is that step-length is debugged that the value of reserve pool input unit scale IS, which is, optimal with result
When corresponding IS values be final value.
3. the automatic noise reduction algorithm of electrocardiosignal according to claim 2, it is characterized in that, step B) detailed process is as follows：
1. several continuous hearts bat x (n) are intercepted from acquired human ecg signal data, n=1,2 ..., M, M=55 ± 2,
Then contained with the noisy electrocardiosignal x (n) at current time, the noisy electrocardiosignal x (n-1) of its last moment, current time
The make an uproar first derivative x ' (n) of electrocardiosignal, the second dervative x " (n) of the noisy electrocardiosignal at current time forms echo state net
The input signal x of networkr(n), it is expressed as
It is d (n) to define the clean heart beat of data corresponding to the noisy electrocardiosignal x (n) at current time；
By xr(n) training set (x is formed with d (n)r(n), d (n), n=1,2 ..., M)；
2. it definesWherein g () is output neuron activation primitive, elects identity function as,
I.e. output neuron is a linear neuron, and in the function, matrix v (n) is defined as Its
Middle r (n) is reserve pool state variable；
3. define error e (n)=d (n)-y (n).
4. the automatic noise reduction algorithm of electrocardiosignal according to claim 3, it is characterized in that, step C) detailed process is：
1. the reserve pool state variable initial value of selected network is 0, i.e. r (0)=0, training set (xr(n), d (n), n=1,
2 ..., M) in sample xr(n) W is passed throughin, d (n) is by WbackIt is added separately in reserve pool, utilizes formula r (n)=(1- α)
[f(W·r(n-1)+Win·u(n)+WbackD (n))]+α r (n-1) carry out reserve pool state variable r (n) update；
Wherein f () is reserve pool neuron activation functions, elects nonlinear tanh () function as；
U (n)=ψ (n-1) v (n), ψ (n) are the correlation matrix of v (n) and pass through formula It is updated, wherein,λ is forgetting factor, and value is 0.95
~1 scope is debugged with 0.0001 for step-length, using result it is optimal when λ value as final value, ψ (0)=δ I, I is unit square
Battle array, δ be a minimum positive number, value in the range of 0~0.001 with 0.00001 for step-length debug, with result it is optimal when δ
It is worth for final value；
α is forgetting rate, and value is debugged in the range of 0~1 with 0.1 step-length, using α value of result when optimal finally to take
Value；
2. according to formula Wout(n)=Wout(n-1)+k (n) e (n) carry out output weight WoutUpdate；
3. by updating Wout(n) and r (n) makes y (n) approach clean ECG signal d (n), makes error e (n) that can be similar to 0, obtains
The echo state network completed to training.
5. the automatic noise reduction algorithm of electrocardiosignal according to claim 4, it is characterized in that, step D) detailed process is：
The M heart intercepted during building training set adjacent T heart bat structure test set, test set (X after clappingr(n), n=
1,2 ..., T, T=200 ± 10), wherein,Electrocardiosignal X (n) i.e. it includes current time, its
A upper electrocardiosignal X (n-1), its first derivative X ' (n) and second dervative X " (n)；
By test set (Xr(n), n=1,2 ..., T) in Xr(n) echo state network that input training is completed, network output are dry
Circumcise electric signal Y (n).
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711339403.9A CN108113665B (en) | 2017-12-14 | 2017-12-14 | Automatic noise reduction method for electrocardiosignal |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711339403.9A CN108113665B (en) | 2017-12-14 | 2017-12-14 | Automatic noise reduction method for electrocardiosignal |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108113665A true CN108113665A (en) | 2018-06-05 |
| CN108113665B CN108113665B (en) | 2020-10-30 |
Family
ID=62229138
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201711339403.9A Active CN108113665B (en) | 2017-12-14 | 2017-12-14 | Automatic noise reduction method for electrocardiosignal |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108113665B (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110169768A (en) * | 2019-07-08 | 2019-08-27 | 河北大学 | A kind of automatic noise-reduction method of electrocardiosignal |
| CN111209853A (en) * | 2020-01-05 | 2020-05-29 | 天津大学 | Optical fiber sensing vibration signal mode identification method based on AdaBoost-ESN algorithm |
| CN112515637A (en) * | 2020-12-02 | 2021-03-19 | 山东省人工智能研究院 | Electrocardiosignal noise reduction method based on group sparsity characteristic |
| CN113456043A (en) * | 2021-07-08 | 2021-10-01 | 军事科学院系统工程研究院卫勤保障技术研究所 | Continuous blood pressure detection method and device |
| CN113729649A (en) * | 2021-09-29 | 2021-12-03 | 九次元(重庆)智能技术有限公司 | Cardiovascular and cerebrovascular disease identification system based on echo state network |
| CN115105088A (en) * | 2022-06-20 | 2022-09-27 | 山东省人工智能研究院 | Improved electrocardiosignal denoising method based on wavelet domain sparse characteristic |
| CN115581465A (en) * | 2022-11-21 | 2023-01-10 | 毕胜普生物科技有限公司 | Coronary heart disease risk assessment method and device, and sudden cardiac death risk assessment method and system |
| CN116304777A (en) * | 2023-04-12 | 2023-06-23 | 中国科学院大学 | Self-adaptive electrocardiosignal denoising method and system based on reference signal during rest |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104157293A (en) * | 2014-08-28 | 2014-11-19 | 福建师范大学福清分校 | Signal processing method for enhancing target voice signal pickup in sound environment |
| CN104346517A (en) * | 2013-08-02 | 2015-02-11 | 杨凤琴 | Echo state network based prediction method and prediction device |
| CN104424507A (en) * | 2013-08-28 | 2015-03-18 | 杨凤琴 | Prediction method and prediction device of echo state network |
| CN107049303A (en) * | 2017-03-10 | 2017-08-18 | 哈尔滨工业大学 | It is a kind of that the adaptive Fetal ECG signal extraction system and method led with many chests is led based on many abdomens |
| CN107145937A (en) * | 2017-04-28 | 2017-09-08 | 河南科技大学 | Echo state network Time Series Forecasting Methods based on elastic SCAD penalty functions |
-
2017
- 2017-12-14 CN CN201711339403.9A patent/CN108113665B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104346517A (en) * | 2013-08-02 | 2015-02-11 | 杨凤琴 | Echo state network based prediction method and prediction device |
| CN104424507A (en) * | 2013-08-28 | 2015-03-18 | 杨凤琴 | Prediction method and prediction device of echo state network |
| CN104157293A (en) * | 2014-08-28 | 2014-11-19 | 福建师范大学福清分校 | Signal processing method for enhancing target voice signal pickup in sound environment |
| CN107049303A (en) * | 2017-03-10 | 2017-08-18 | 哈尔滨工业大学 | It is a kind of that the adaptive Fetal ECG signal extraction system and method led with many chests is led based on many abdomens |
| CN107145937A (en) * | 2017-04-28 | 2017-09-08 | 河南科技大学 | Echo state network Time Series Forecasting Methods based on elastic SCAD penalty functions |
Non-Patent Citations (4)
| Title |
|---|
| ANDRIUS PETRENAS ET AL: "Reservoir Computing for Extraction of Low Amplitude Atrial Activity in Atrial Fibrillation", 《COMPUTING IN CARDIOLOGY》 * |
| ANDRIUS PETRENAS ET AL: "Ventricular Activity Cancellation in ECG Using an Adaptive Echo State Network", 《THE 6TH IEEE INTERNATIONAL CONFERENCE ON INTELLIGENT DATA ACQUISITION AND ADVANCED COMPUTING SYSTEMS: TECHNOLOGY AND APPLICATIONS》 * |
| 李昳霏: "基于ESN的单导联动态心电信号的QRST消除", 《中国优秀硕士学位论文全文数据库 医药卫生科技辑》 * |
| 韩敏等: "基于回声状态网络的脑电信号特征提取", 《生物医学工程学杂志》 * |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110169768A (en) * | 2019-07-08 | 2019-08-27 | 河北大学 | A kind of automatic noise-reduction method of electrocardiosignal |
| CN111209853A (en) * | 2020-01-05 | 2020-05-29 | 天津大学 | Optical fiber sensing vibration signal mode identification method based on AdaBoost-ESN algorithm |
| CN112515637A (en) * | 2020-12-02 | 2021-03-19 | 山东省人工智能研究院 | Electrocardiosignal noise reduction method based on group sparsity characteristic |
| CN112515637B (en) * | 2020-12-02 | 2021-06-15 | 山东省人工智能研究院 | Electrocardiosignal noise reduction method based on group sparsity characteristic |
| CN113456043A (en) * | 2021-07-08 | 2021-10-01 | 军事科学院系统工程研究院卫勤保障技术研究所 | Continuous blood pressure detection method and device |
| CN113729649A (en) * | 2021-09-29 | 2021-12-03 | 九次元(重庆)智能技术有限公司 | Cardiovascular and cerebrovascular disease identification system based on echo state network |
| CN115105088A (en) * | 2022-06-20 | 2022-09-27 | 山东省人工智能研究院 | Improved electrocardiosignal denoising method based on wavelet domain sparse characteristic |
| CN115581465A (en) * | 2022-11-21 | 2023-01-10 | 毕胜普生物科技有限公司 | Coronary heart disease risk assessment method and device, and sudden cardiac death risk assessment method and system |
| CN116304777A (en) * | 2023-04-12 | 2023-06-23 | 中国科学院大学 | Self-adaptive electrocardiosignal denoising method and system based on reference signal during rest |
| CN116304777B (en) * | 2023-04-12 | 2023-11-03 | 中国科学院大学 | Self-adaptive electrocardiosignal denoising method and system based on reference signal during rest |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108113665B (en) | 2020-10-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108113665A (en) | A kind of automatic noise-reduction method of electrocardiosignal | |
| Obeid et al. | Evaluation of spike-detection algorithms fora brain-machine interface application | |
| EP1613209B1 (en) | Biomedical signal denoising techniques | |
| Javadi et al. | Classification of ECG arrhythmia by a modular neural network based on mixture of experts and negatively correlated learning | |
| JP4657202B2 (en) | Physiological event detector and method thereof | |
| Clancy et al. | Adaptive whitening of the electromyogram to improve amplitude estimation | |
| EP1614049B1 (en) | Biomedical signal analysis techniques using wavelets | |
| CN110226919B (en) | Electrocardiosignal type detection method and device, computer equipment and storage medium | |
| CN105726018A (en) | Automatic atrial fibrillation detection method irrelevant to RR interphase | |
| CN113397555A (en) | Arrhythmia classification algorithm of C-LSTM for physiological parameter monitoring | |
| CN107233093B (en) | R wave detection method and device and electronic equipment | |
| CN109948396B (en) | Heart beat classification method, heart beat classification device and electronic equipment | |
| CN109222964B (en) | Atrial fibrillation detection apparatus and storage medium | |
| CN110367968B (en) | Right bundle branch retardation detection method, device, equipment and storage medium | |
| Kumar et al. | Optimal selection of wavelet function and decomposition level for removal of ECG signal artifacts | |
| CN110367936A (en) | Electrocardiograph signal detection method and device | |
| CN109077721B (en) | Atrial fibrillation detection apparatus and storage medium | |
| Azzerboni et al. | A new approach to detection of muscle activation by independent component analysis and wavelet transform | |
| Qureshi et al. | Performance analysis of adaptive algorithms for removal of low frequency noise from ECG signal | |
| CN110226918B (en) | Electrocardiosignal type detection method and device, computer equipment and storage medium | |
| Rao et al. | Performance identification of different heart diseases based on neural network classification | |
| Zubair et al. | Removal of motion artifacts from ECG signals by combination of recurrent neural networks and deep neural networks | |
| Turnip et al. | Extraction of P and T waves from electrocardiogram signals with modified Hamilton algorithm | |
| Khandait et al. | ECG signal processing using classifier to analyses cardiovascular disease | |
| Suchetha et al. | Denoising and arrhythmia classification using EMD based features 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 |