CN108319889A - The method for carrying out EMG signal feature extraction using independent component analysis - Google Patents
The method for carrying out EMG signal feature extraction using independent component analysis Download PDFInfo
- Publication number
- CN108319889A CN108319889A CN201710046506.XA CN201710046506A CN108319889A CN 108319889 A CN108319889 A CN 108319889A CN 201710046506 A CN201710046506 A CN 201710046506A CN 108319889 A CN108319889 A CN 108319889A
- Authority
- CN
- China
- Prior art keywords
- emg signal
- signal
- carrying
- feature extraction
- formula
- 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
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Theoretical Computer Science (AREA)
- Artificial Intelligence (AREA)
- Signal Processing (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Investigating Or Analysing Biological Materials (AREA)
Abstract
The present invention provides a kind of method carrying out EMG signal feature extraction using independent component analysis, which is characterized in that includes the following steps：Step 1, it is assumed that EMG signal vector is x (t)=[x1(t), x2(t) ... ..., xN(t)]T, N is the dimension of EMG signal；EMG signal is modeled as by a series of built-up stochastic variable of isolated components, i.e. x (t)=As (t), wherein s (t)=[s by step 21(t), s2(t) ... ..., sM(t)]T, M is the number of wherein isolated component, M≤N；Step 3 finds a matrix W and makes y (t)=Wx (t), allows y (t) to be the estimation of isolated component s (t), y (t) is exactly the feature vector of the EMG signal extracted.The present invention is used as the feature of EMG signal by extracting the isolated component in EMG signal, this method is not while losing EMG signal feature, reduce the degree of correlation between EMG signal feature, reduce Characteristic Number, to simplify the process of EMG post-processings, the technologies such as muscular fatigue detection are can be widely applied to.
Description
Technical field
The present invention relates to a kind of methods carrying out EMG signal feature extraction using independent component analysis, belong to physiological signal
Process field.
Background technology
EMG is human muscle's current signal, is human body important biomolecule electric current index, our finger, the fortune of palm, wrist
It is dynamic, it is being macroscopically to combine what control generated by multiple extensors and flexors in forearm.This stimulation is by central nervous system
It is transmitted to the electricity order generation of movement meat fiber, this electric signal is known as muscle current signal on biology.Muscle electric current
Signal, can pass through muscle, subcutaneous fat, and reach skin surface, this, which decays and is doped with the electromyography signal of interference signal, claims
For epidermis muscle current signal.By to the main epidermis muscle current signal of polylith it is continuous pickup, amplification, filtering, rise always
A series of signal processing such as point lookup, frequency spectrum, energy, small echo, can extract the characteristic parameter of each EMG signal, and pass through
Characteristic parameter realizes the analysis of EMG signal.
Electromyography signal is briefly that nerve system of human body is controlling the faint electricity showed when respective muscle execution
Signal.Electromyography signal is having passed through meat fiber, and human body subcutaneous fat, when finally reaching skin surface, signal strength is by pole
Big reduction, and due to receive including electromyography signal crosstalk, pulse signal, static electricity on human body, around electromagnetic field etc. it is a variety of dry
Disturb the interference in source so that the signal-to-noise ratio of signal deteriorates significantly.How a kind of effective parser model is designed, and utmostly
Guarantee electromyography signal signal-to-noise ratio quality, it is vital that can characteristic parameter be correctly therefrom extracted for subsequent algorithm.
Invention content
The purpose of the present invention is to provide it is a kind of using independent component analysis carry out EMG signal feature extraction method, with
It solves the above problems.
Present invention employs following technical solutions：
A method of carrying out EMG signal feature extraction using independent component analysis, which is characterized in that including following step
Suddenly：
Step 1, it is assumed that EMG signal vector is x (t)=[x1(t), x2(t) ... ..., xN(t)]T, N is the dimension of EMG signal
Degree；
EMG signal is modeled as by a series of built-up stochastic variable of isolated components, i.e. x (t)=As by step 2
(t), wherein s (t)=[s1(t), s2(t) ... ..., sM(t)]T, M is the number of wherein isolated component, M≤N；
Step 3 finds a matrix W and makes y (t)=Wx (t), allows y (t) to be the estimation of isolated component s (t), y (t) is just
It is the feature vector of the EMG signal extracted.
Further, the method for the invention for carrying out EMG signal feature extraction using independent component analysis, can also have
There is such feature：In step 3, definition estimation signal y (t) and its each component yi(t) the KL distances between areKL distances describe p (y) andBetween similarity degree, whenWhen, D (W)=0, between each component of estimation signal y (t) independently of each other at this time.
Further, the method for the invention for carrying out EMG signal feature extraction using independent component analysis, can also have this
The feature of sample：Wherein, the computational methods of D (w) are as follows：
And
H (y)=H (x)+log | det (W) | (2)
Each component yiProbability can use Gram-Charlier sequences yiHigher Order Cumulants approximation obtain
Wherein μ (yi) be mean value it is 0, variance and yiThe probability density function of the careless variable of the consistent Gauss of variance,Hk(y) it is Chebyshev-Hermite multinomials, H3(y)=y3- 3y, H4(y)=
y4-6y2+ 3, (2) and (3) formula, which is substituted into (1) formula, to be had
Local derviation is asked to have (4) formula
The mathematic expectaion in (5) formula is replaced with instantaneous value, W is iterated to calculate out using gradient descent method, is had
It is further simplified, and removes the mathematic expectaion in Higher Order Cumulants, write as matrix form, have
Wn+1=Wn+μ[I-f(y)yT]Wn (7)
WhereinMatrix W is iterated to calculate out according to formula (7).
Further, the method for the invention for carrying out EMG signal feature extraction using independent component analysis, can also have this
The feature of sample：
It is calculated again after calculating matrix W：
Y (t)=Wx (t) (8),
Obtain the feature of EMG signal.
Advantageous effect of the invention
The present invention mainly proposes a kind of method using the independent characteristic for minimizing mutual information extraction EMG signal, should
The advantage of algorithm is, as carried out independent characteristic amount extraction unit in the algorithm of iterative calculation weight matrix and (8) in formula (7)
Point, follow-up signal analysis and judgement can effectively be carried out by extracting EMG signal characteristic parameter by this method.
The present invention is used as the feature of EMG signal by extracting the isolated component in EMG signal, and this method is not losing EMG
While signal characteristic, the degree of correlation between EMG signal feature is reduced, reduces Characteristic Number, to simplify EMG post-processings
Process can be widely applied to the technologies such as muscular fatigue detection.
Specific implementation mode
The following specifically describes the specific implementation modes of the present invention.
First, it is assumed that EMG signal vector is x (t)=[x1(t), x2(t) ... ..., XN(t)]T, N is the dimension of EMG signal
Degree, EMG signal can be modeled as by a series of built-up stochastic variable of isolated components, i.e. x (t)=As (t), wherein s (t)
=[s1(t), s2(t) ... ..., sM(t)]T, M is the number of wherein isolated component, M≤N.One matrix W of searching is needed to make now
Y (t)=Wx (t) is obtained, allows y (t) to be the estimation of isolated component s (t), y (t) is exactly the feature vector of the EMG signal extracted, can
For post-processing.
Using the method calculating matrix W for minimizing mutual information, since the prior information of isolated component can not be predicted, we
Only know between each component of s (t) independently of each other, thus our purpose is exactly so that phase between each components of the y (t) estimated
It is mutually independent, definition estimation signal y (t) and its each component yi(t) the KL distances between are
KL distances describe p (y) andBetween similarity degree, whenWhen, D (W)=0 is said at this time
Between each component of bright estimation signal y (t) independently of each other.
The computational methods of D (W) are given below, first
And
H (y)=H (x)+log | det (W) | (2)
Each component yiProbability can use Gram-Charlier sequences yiHigher Order Cumulants approximation obtain
Wherein μ (yi) be mean value it is 0, variance and yiThe probability density function of the careless variable of the consistent Gauss of variance,Hk(y) it is Chebyshev-Hermite multinomials, H3(y)=y3- 3y, H4(y)=
y4-6y2+ 3, (2) and (3) formula, which is substituted into (1) formula, to be had
Local derviation is asked to have (4) formula
The mathematic expectaion in (5) formula is replaced with instantaneous value, W is iterated to calculate out using gradient descent method, is had
It is further simplified, and removes the mathematic expectaion in Higher Order Cumulants, write as matrix form, have
Wn+1=Wn+μ[I-f(y)yT]Wn (7)
WhereinMatrix W is iterated to calculate out according to formula (7)
Afterwards, it then calculates
Y (t)=Wx (t) (8)
It can be obtained the feature of EMG signal.
Using the extracting method of the present invention, the feature extracted the feature used in tional identification that compares is many less, Gu
In the case of fixed 97% discrimination, the characteristic value number using independent component analysis is 2, and is used required for conventional method
Characteristic value number be 5, it can be seen that the algorithm be capable of it is prodigious reduce characteristic value number, to reduce identification and storage
Complexity.
Claims (4)
1. a kind of method carrying out EMG signal feature extraction using independent component analysis, which is characterized in that include the following steps：
Step 1, it is assumed that EMG signal vector is x (t)=[x1(t), x2(t) ... ..., xN(t)]T, N is the dimension of EMG signal；
EMG signal is modeled as by a series of built-up stochastic variable of isolated components, i.e. x (t)=As (t) by step 2,
Middle s (t)=[s1(t), s2(t) ... ..., sM(t)]T, M is the number of wherein isolated component, M≤N；
Step 3 finds a matrix W and makes y (t)=Wx (t), allows y (t) to be the estimation of isolated component s (t), y (t) is exactly to carry
The feature vector of the EMG signal of taking-up.
2. the method as described in claim 1 for carrying out EMG signal feature extraction using independent component analysis, it is characterised in that：
In step 3, definition estimation signal y (t) and its each component yi(t) the KL distances between are
KL distances describe p (y) andBetween similarity degree, whenWhen, D (W)=0 estimates at this time
Between each component of meter signal y (t) independently of each other.
3. the method as claimed in claim 2 for carrying out EMG signal feature extraction using independent component analysis, it is characterised in that：
Wherein, the computational methods of D (w) are as follows：
And
H (y)=H (x)+log | det (W) | (2)
Each component yiProbability can use Gram-Charlier sequences yiHigher Order Cumulants approximation obtain
Wherein μ (yi) be mean value it is 0, variance and yiThe probability density function of the careless variable of the consistent Gauss of variance,Hk(y) it is Chebyshev-Hermite multinomials, H3(y)=y3- 3y, H4(y)=y4-
6y2+ 3, (2) and (3) formula, which is substituted into (1) formula, to be had
Local derviation is asked to have (4) formula
The mathematic expectaion in (5) formula is replaced with instantaneous value, W is iterated to calculate out using gradient descent method, is had
It is further simplified, and removes the mathematic expectaion in Higher Order Cumulants, write as matrix form, have
Wn+1=Wn+μ[1-f(y)yT]Wn (7)
WhereinMatrix W is iterated to calculate out according to formula (7).
4. the method as claimed in claim 3 for carrying out EMG signal feature extraction using independent component analysis, it is characterised in that：
It is calculated again after calculating matrix W：
Y (t)=Wx (t) (8),
Obtain the feature of EMG signal.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710046506.XA CN108319889A (en) | 2017-01-18 | 2017-01-18 | The method for carrying out EMG signal feature extraction using independent component analysis |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710046506.XA CN108319889A (en) | 2017-01-18 | 2017-01-18 | The method for carrying out EMG signal feature extraction using independent component analysis |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108319889A true CN108319889A (en) | 2018-07-24 |
Family
ID=62892993
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710046506.XA Pending CN108319889A (en) | 2017-01-18 | 2017-01-18 | The method for carrying out EMG signal feature extraction using independent component analysis |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108319889A (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102697493A (en) * | 2012-05-03 | 2012-10-03 | 北京工业大学 | Method for rapidly and automatically identifying and removing ocular artifacts in electroencephalogram signal |
| CN103761424A (en) * | 2013-12-31 | 2014-04-30 | 杭州电子科技大学 | Electromyography signal noise reducing and aliasing removing method based on second-generation wavelets and ICA (independent component analysis) |
| CN105447475A (en) * | 2015-12-21 | 2016-03-30 | 安徽大学 | Independent component analysis based glancing signal sample optimization method |
| CN105447445A (en) * | 2015-11-09 | 2016-03-30 | 天津商业大学 | High-spectral image unmixing method based on differential search |
| CN105640500A (en) * | 2015-12-21 | 2016-06-08 | 安徽大学 | Scanning signal feature extraction method based on independent component analysis and recognition method |
-
2017
- 2017-01-18 CN CN201710046506.XA patent/CN108319889A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102697493A (en) * | 2012-05-03 | 2012-10-03 | 北京工业大学 | Method for rapidly and automatically identifying and removing ocular artifacts in electroencephalogram signal |
| CN103761424A (en) * | 2013-12-31 | 2014-04-30 | 杭州电子科技大学 | Electromyography signal noise reducing and aliasing removing method based on second-generation wavelets and ICA (independent component analysis) |
| CN105447445A (en) * | 2015-11-09 | 2016-03-30 | 天津商业大学 | High-spectral image unmixing method based on differential search |
| CN105447475A (en) * | 2015-12-21 | 2016-03-30 | 安徽大学 | Independent component analysis based glancing signal sample optimization method |
| CN105640500A (en) * | 2015-12-21 | 2016-06-08 | 安徽大学 | Scanning signal feature extraction method based on independent component analysis and recognition method |
Non-Patent Citations (3)
| Title |
|---|
| SHANG XIAOJING ET AL: "Feature Extraction and Classification of sEMG", 《IEEE》 * |
| 巫书航: "独立分量分析法", 《道客巴巴》 * |
| 谈春祥 等: "独立分量分析在表面肌电信号分解中的应用", 《生物医学工程研究》 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Li et al. | Deep convolutional neural network based ECG classification system using information fusion and one‐hot encoding techniques | |
| Kang et al. | ECG authentication system design based on signal analysis in mobile and wearable devices | |
| US10524734B2 (en) | Method and system for measuring beat parameters | |
| CN109700450B (en) | Heart rate detection method and electronic equipment | |
| CN109620210B (en) | A kind of electrocardiosignal classification method of the CNN based on from coding mode in conjunction with GRU | |
| CN108784693B (en) | P300 single extraction technology based on independent component analysis and Kalman smoothing | |
| Luz et al. | Evaluating the use of ECG signal in low frequencies as a biometry | |
| Matos et al. | Embedded system for individual recognition based on ECG Biometrics | |
| CN105686801A (en) | An atrial fibrillation spontaneous termination predication method based on inherent time scale decomposing | |
| CN107273827B (en) | Electrocardiosignal R wave detection method and device | |
| CN107361765B (en) | Brain wave analysis method and device | |
| Kshirsagar et al. | Classification of ECG-signals using artificial neural networks | |
| CN109124620A (en) | A kind of atrial fibrillation detection method, device and equipment | |
| Xiong et al. | Application of the chaos theory in the analysis of EMG on patients with facial paralysis | |
| Biswal | ECG signal analysis using modified S‐transform | |
| CN113128384B (en) | Brain-computer interface software key technical method of cerebral apoplexy rehabilitation system based on deep learning | |
| CN112215196B (en) | Electrocardiogram identity recognition method | |
| CN111652132B (en) | Non-line-of-sight identity recognition method and device based on deep learning and storage medium | |
| CN108319889A (en) | The method for carrying out EMG signal feature extraction using independent component analysis | |
| CN114861706A (en) | Electrocardio identity recognition method based on quality evaluation and deep transfer learning | |
| Salustri et al. | Fetal magnetocardiographic signals extracted by'signal subspace'blind source separation | |
| Orlandi et al. | Motor-unit ordering of blindly-separated surface-EMG signals for gesture recognition | |
| Labati et al. | HeartCode: A novel binary ECG-based template | |
| Zhang et al. | Motor imagery electroencephalogram de-noising method based on EEMD and improved wavelet threshold | |
| Guo et al. | EEG signals denoising using bayesian estimation |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180724 |