CN107212882B - A kind of real-time detection method and system of EEG signals state change - Google Patents
A kind of real-time detection method and system of EEG signals state change Download PDFInfo
- Publication number
- CN107212882B CN107212882B CN201710349043.4A CN201710349043A CN107212882B CN 107212882 B CN107212882 B CN 107212882B CN 201710349043 A CN201710349043 A CN 201710349043A CN 107212882 B CN107212882 B CN 107212882B
- Authority
- CN
- China
- Prior art keywords
- eeg signals
- module
- time
- abnormality degree
- real
- 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.)
- Expired - Fee Related
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
- A61B5/369—Electroencephalography [EEG]
-
- 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
- A61B5/7235—Details of waveform analysis
- A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
Abstract
The invention discloses the real-time detection methods and system of a kind of EEG signals state change, wherein this method comprises: extracting the characteristic value of EEG signals；It is modeled using characteristic value of the autoregression model to the EEG signals of extraction, and calculates the timing abnormality degree of EEG signals by residual analysis；At the time of carrying out statistical check based on timing abnormality degree of the random power halter strap to EEG signals, and then determine that EEG signals state changes.
Description
Technical field
The invention belongs to EEG signals detection field more particularly to a kind of real-time detection methods of EEG signals state change
And system.
Background technique
EEG signals can provide a large amount of accurate effective physiology and psychographic information for clinical diagnosis, monitor brain telecommunications in real time
The brain function activity state that number can obtain observation object online, has been widely used in clinical diagnosis and business application.
During monitoring, the variation of real-time detection EEG signals state being capable of real-time finder's brain function activity state
Variation, can the mental disorders such as auxiliary diagnosis epilepsy breaking-out.However, in current clinical diagnosis, for EEG signals shape
The detection of state variation, mostly relies on eye-observation electroencephalogram dependent on experienced doctor.Since EEG data amount is huge, human eye
It checks that situation of breaking out extremely in EEG signals is often a uninteresting and time-consuming job for a long time, thus be easy to cause detection essence
The limitations such as low, detection time is long are spent, the requirement of real-time detection during monitoring is unable to reach.
Summary of the invention
In order to solve the deficiencies in the prior art, the present invention provides a kind of real-time detection sides of EEG signals state change
Method, this method make a policy to the change point of signal during EEG signals monitor, and are not necessarily to priori knowledge, can online in real time
It is detected, can be directly applied in EEG signals condition monitoring and variation detection related application.
A kind of real-time detection method of EEG signals state change of the invention, comprising:
Extract the characteristic value of EEG signals；
It is modeled using characteristic value of the autoregression model to the EEG signals of extraction, and brain electricity is calculated by residual analysis
The timing abnormality degree of signal；
Statistical check is carried out based on timing abnormality degree of the random power halter strap to EEG signals, and then determines EEG signals state hair
At the time of changing.
Further, the detailed process of the amplitude Characteristics value of EEG signals is extracted are as follows:
With the down-sampled rate of 1:n EEG signals are carried out with down-sampled, formation EEG signals voltage magnitude sequence；Wherein, n is
Positive integer greater than 1；
The sliding window for introducing a regular length extracts EEG signals voltage magnitude sequence in window in the sliding window
Time-domain Statistics feature；
By obtained Time-domain Statistics Fusion Features at one-dimensional characteristic vector.
The Time-domain Statistics Fusion Features of EEG signals are one-dimensional characteristic by down-sampled and sliding window setting technique by the present invention
Vector is unitized EEG signals, improves the accuracy of brain electric information state change detection.
Further, after being modeled using characteristic value of the autoregression model to the EEG signals of extraction, when certain a period of time
When the EEG signals at quarter no longer meet the model of above-mentioned foundation, then current time is change point.
Further, the detailed process of the timing abnormality degree of EEG signals is calculated by residual analysis are as follows:
Respectively by the characteristic value in one-dimensional characteristic vector and using autoregression model to the feature of the EEG signals of extraction
Value is modeled to obtain characteristic value and makees poor, obtains predicting error；
Prediction error is standardized；
By seeking the Euclidean distance between the prediction error mean after the prediction error after any standard and standardization,
Obtain the timing abnormality degree of EEG signals.
The present invention calculates the timing abnormality degree of EEG signals by introducing prediction error, can obtain standard
True calculated result, and then improve the accuracy and efficiency of EEG signals state change real-time detection.
Further, statistical check is carried out based on timing abnormality degree of the random power halter strap to EEG signals, and then determines brain electricity
Detailed process at the time of signal condition changes includes:
Firstly, building halter strap function；
Then, decision is carried out by Doob maximum inequality to judge whether current time is that EEG signals state occurs
At the time of variation.
Wherein, decision is carried out by Doob maximum inequality (Doob ' s Maximal Inequality), it may be assumed that
H0: 1 < M (c) < λ
HA:M(c)≥λ
If the martingale value of current time c is less than preset λ value, receive the hypothesis testing in above-mentioned inequality
H0, then current time c is not change point；Otherwise receive hypothesis testing HA, that is, think that current time c is change point.Threshold in inequality
Value λ is obtained by cross validation, can realize that the dynamic of testing result adjusts by changing the value of λ in inequality.
The present invention also provides a kind of real-time detection method systems of EEG signals state change.
A kind of real-time detection method system of EEG signals state change of the invention, comprising:
Characteristics extraction module is used to extract the characteristic value of EEG signals；
Abnormality degree computing module is used to model using characteristic value of the autoregression model to the EEG signals of extraction,
And the timing abnormality degree of EEG signals is calculated by residual analysis；
Change moment determining module, is used to carry out statistics inspection based on timing abnormality degree of the random power halter strap to EEG signals
At the time of testing, and then determine that EEG signals state changes.
Further, the characteristics extraction module includes:
Down-sampled module is used to carry out EEG signals with the down-sampled rate of 1:n down-sampled, formation EEG signals voltage
Amplitude sequence；Wherein, n is the positive integer greater than 1；
Sliding window characteristic extracting module is used to introduce the sliding window of a regular length, mentions in the sliding window
Take the Time-domain Statistics feature of EEG signals voltage magnitude sequence in window；
Fusion Features module, the Time-domain Statistics Fusion Features for being used to obtain are at one-dimensional characteristic vector.
Further, in the abnormality degree computing module, when the EEG signals at a certain moment no longer meet above-mentioned foundation
When model, then current time is change point.
Further, the abnormality degree computing module includes:
Predict error calculating module, be used for respectively by characteristic value in one-dimensional characteristic vector with utilize autoregression model
It is modeled to obtain characteristic value to the characteristic value of the EEG signals of extraction and makees poor, obtain predicting error；
It predicts error criterion module, is used to be standardized prediction error；
Euclidean distance computing module is used for by seeking the prediction error after any standard and the prediction after standardization
Euclidean distance between error mean obtains the timing abnormality degree of EEG signals.
Further, the variation moment determining module includes:
Halter strap function constructs module, is used to construct halter strap function；
Judgment module is used to carry out decision by Doob maximum inequality to judge whether current time is brain telecommunications
At the time of number state changes.
Compared with prior art, the beneficial effects of the present invention are:
The present invention carries out feature extraction to the EEG signals of acquisition first；On the basis of the EEG signals feature of extraction,
Modeling analysis is carried out to it using autoregression model, and the timing abnormality degree of the signal is calculated by residual analysis；According to obtaining
Timing abnormality degree, at the time of carrying out statistical check to it using random power halter strap, and then determine that EEG signals state changes.
This method EEG signals monitor during make a policy to the change point of signal, be not necessarily to priori knowledge, can online in real time into
Row detection can directly apply in EEG signals condition monitoring and variation detection related application.
Detailed description of the invention
The accompanying drawings constituting a part of this application is used to provide further understanding of the present application, and the application's shows
Meaning property embodiment and its explanation are not constituted an undue limitation on the present application for explaining the application.
Fig. 1 is a kind of real-time detection method flow chart of EEG signals state change of the invention；
Fig. 2 is the EEG signals of observation；
Testing result when Fig. 3 is λ=2；
Testing result when Fig. 4 is λ=3；
Testing result when Fig. 5 is λ=4；
Fig. 6 is a kind of structural schematic diagram of the real-time detection method system of EEG signals state change of the invention.
Specific embodiment
It is noted that following detailed description is all illustrative, it is intended to provide further instruction to the application.Unless another
It indicates, all technical and scientific terms used herein has usual with the application person of an ordinary skill in the technical field
The identical meanings of understanding.
It should be noted that term used herein above is merely to describe specific embodiment, and be not intended to restricted root
According to the illustrative embodiments of the application.As used herein, unless the context clearly indicates otherwise, otherwise singular
Also it is intended to include plural form, additionally, it should be understood that, when in the present specification using term "comprising" and/or " packet
Include " when, indicate existing characteristics, step, operation, device, component and/or their combination.
Fig. 1 is a kind of real-time detection method flow chart of EEG signals state change of the invention.
As shown in Figure 1, a kind of real-time detection method of EEG signals state change of the invention, comprising:
S101: the characteristic value of EEG signals is extracted.
For the present embodiment EEG signals shown in Fig. 2:
Specifically, the detailed process of the amplitude Characteristics value of EEG signals is extracted are as follows:
S1011: with the down-sampled rate of 1:n EEG signals are carried out with down-sampled, formation EEG signals voltage magnitude sequence；Its
In, n is the positive integer greater than 1.
In specific implementation, n is for 50:
For given EEG signals, it is carried out with the down-sampled rate of 1:50 first it is down-sampled, will be down-sampled after brain
Electric signal is expressed as { y1,y2,...,yN(indicating that the voltage magnitude of i-th of signaling point, N represent the sequence length of EEG signals).
S1012: the sliding window of a regular length is introduced, EEG signals voltage magnitude in window is extracted in the sliding window
The Time-domain Statistics feature of sequence.
It is introduced into the sliding window that a regular length is L (being set as 5 in application), sequence in window is extracted in the sliding window
Time-domain Statistics feature.For the signal at k moment, sliding window is set as { yk-L+1,yk-L+2,…,yk, with 5 in the sliding window
Time-domain Statistics feature fj(j=1,2,3,4,5) characterizes the feature at the moment, and used temporal signatures calculation is as follows:
f2=max { yk-L+1,yk-L+2,...,yk}
f3=min { yk-L+1,yk-L+2,...,yk}
S1013: by obtained Time-domain Statistics Fusion Features at one-dimensional characteristic vector.
By fj(j=1,2,3,4,5) is merged, and calculation is as follows:
Therefore, length is the EEG signals { y of N1,y2,…,yN{ q can be characterized as with one-dimensional vector1,q2,…,qN}.It is right
In signal { y1,y2,…,yL-1, feature is initialized as 0.
The Time-domain Statistics Fusion Features of EEG signals are one-dimensional characteristic by down-sampled and sliding window setting technique by the present invention
Vector is unitized EEG signals, improves the accuracy of brain electric information state change detection.
S102: being modeled using characteristic value of the autoregression model to the EEG signals of extraction, and passes through residual analysis meter
Calculate the timing abnormality degree of EEG signals.
Specifically, the detailed process of S102 includes:
S1021: assuming that given EEG signals sequence has been represented as { q1,q2,...,qN, with an autoregression line
Property model modeling analysis is carried out to the signal, modeling pattern is as follows:
qt=μ+β t+ εt
Wherein, μ is the mean value of EEG signals sequence, and β is variation tendency, εtFor an independent identically distributed error, and meet
It is desired for zero.Calculating can be passed throughObtain the value of μ and β, it may be assumed that
When the EEG signals at a certain moment no longer meet above-mentioned model, that is, think that the moment is change point.Obviously, once
Data distribution changes, two in above-mentioned model parameter μ and β will with it is significantly different in model built before.Assuming that in c
It carves signal distributions to change, the parameter in model is by original (μ1,β1) become (μ2,β2), a piecewise linearity mould can be passed through
Type describes this variation, it may be assumed that
Provide following construction testing model, it may be assumed that
H0:μ1=μ2andβ1=β2
HA:μ1≠μ2and/orβ1≠β2
When significant change does not occur for EEG signals state, μ and β are remained unchanged；If apparent variation occurs, μ and β
It will change.That is, working as H0It is rejected, i.e. HAWhen being true, the signal at current time is confirmed as change point.
S1022: prediction error { e is introducedtQuantify { q1,q2,...,qNData distribution timing fluctuate situation.
Assuming that (μ1,β1) by calculatingIt obtains, then qtLinear predictor can be by establishing in step (2-1) from returning
Model is returned to be calculatedThe then prediction error e at the momenttIt may be expressed as:
Wherein, | | | | it is Euclidean distance.It is calculated to simplify,It can be calculated asIn conjunction with being built in step (2-1)
Vertical piecewise linear model predicts error etValue can be divided into two kinds of situations:
1) as t < c, current demand signal obeys preset data distribution, predicts error etVery little is approximately equal to 0；
2) as t=c, current demand signal is not obeying original data distribution, the autoregression model established in step (2-1)
Significant change occurs for parameter, so that prediction error etBecome larger.
S1023: for the prediction error sequence { e of acquisitiont, it is standardized, it may be assumed that
Wherein,For sample average, σ is standard deviation.Finally, in the { z being calculated1,z2,...,zt-1On the basis of, number
It can be calculated according to abnormality degree are as follows:
st=s ({ z1,z2,…,zt-1},zt)=| | zt-Ht-1||
Wherein,
The present invention calculates the timing abnormality degree of EEG signals by introducing prediction error, can obtain standard
True calculated result, and then improve the accuracy and efficiency of EEG signals state change real-time detection.
S103: statistical check is carried out based on timing abnormality degree of the random power halter strap to EEG signals, and then determines EEG signals
At the time of state changes.
The detailed process of S103 includes:
S1031: abnormality degree { s1,s2,…,stOn the basis of, using based on random power halter strap (Random Power
Martingale, RPM) statistical check.It is a printenv data statistics inspection can be examined under unsupervised environment
Operation is surveyed, and can be executed with real-time online.
Firstly, building halter strap (Martingale) is as follows:
Wherein, (0,1) ξ ∈, is set as 0.8 in,Calculation formula it is as follows:
In formula: # { } is a counting function, θiThe random number being generally evenly distributed between [0,1].Due toM (t) can be obtained by iterative calculation.
S1032: decision is carried out by Doob maximum inequality (Doob ' s Maximal Inequality), it may be assumed that
H0: 1 < M (c) < λ
HA:M(c)≥λ
If the martingale value of current time c is less than preset λ value, receive the hypothesis testing in above-mentioned inequality
H0, then current time c is not change point；Otherwise receive hypothesis testing HA, that is, think that current time c is change point.Threshold in inequality
Value λ is obtained by cross validation, can realize that the dynamic of testing result adjusts by changing the value of λ in inequality, such as Fig. 3-Fig. 5
Respectively λ=2, λ=3 and λ=4 when testing result.
For realize EEG signals state change on-line real-time measuremen, for the EEG signals { y observed1,y2,...,
yt, if the t moment is not determined to change point, continues observation signal and detect t=t+1, continue to execute step S101-
S103；If t moment is confirmed as change point, which is reset into initial point, re-execute the steps S101-S103.
The present invention carries out feature extraction to the EEG signals of acquisition first；On the basis of the EEG signals feature of extraction,
Modeling analysis is carried out to it using autoregression model, and the timing abnormality degree of the signal is calculated by residual analysis；According to obtaining
Timing abnormality degree, at the time of carrying out statistical check to it using random power halter strap, and then determine that EEG signals state changes.
This method EEG signals monitor during make a policy to the change point of signal, be not necessarily to priori knowledge, can online in real time into
Row detection can directly apply in EEG signals condition monitoring and variation detection related application.
Fig. 6 is a kind of structural schematic diagram of the real-time detection method system of EEG signals state change of the invention.
As shown in fig. 6, a kind of real-time detection method system of EEG signals state change of the invention, comprising:
(1) characteristics extraction module is used to extract the characteristic value of EEG signals.
Wherein, the characteristics extraction module includes:
(1.1) down-sampled module is used to carry out EEG signals with the down-sampled rate of 1:n down-sampled, formation brain telecommunications
Number voltage magnitude sequence；Wherein, n is the positive integer greater than 1；
In specific implementation, n is for 50:
For given EEG signals, it is carried out with the down-sampled rate of 1:50 first it is down-sampled, will be down-sampled after brain
Electric signal is expressed as { y1,y2,...,yN(indicating that the voltage magnitude of i-th of signaling point, N represent the sequence length of EEG signals).
(1.2) sliding window characteristic extracting module is used to introduce the sliding window of a regular length, in the sliding window
To extract the Time-domain Statistics feature of EEG signals voltage magnitude sequence in window；
It is introduced into the sliding window that a regular length is L (being set as 5 in application), sequence in window is extracted in the sliding window
Time-domain Statistics feature.For the signal at k moment, sliding window is set as { yk-L+1,yk-L+2,...,yk, with 5 in the sliding window
Time-domain Statistics feature fj(j=1,2,3,4,5) characterizes the feature at the moment, and used temporal signatures calculation is as follows:
f2=max { yk-L+1,yk-L+2,…,yk}
f3=min { yk-L+1,yk-L+2,...,yk}
(1.3) Fusion Features module, the Time-domain Statistics Fusion Features for being used to obtain are at one-dimensional characteristic vector.
By fj(j=1,2,3,4,5) is merged, and calculation is as follows:
Therefore, length is the EEG signals { y of N1,y2,...,yN{ q can be characterized as with one-dimensional vector1,q2,...,qN}。
For signal { y1,y2,…,yL-1, feature is initialized as 0.
The Time-domain Statistics Fusion Features of EEG signals are one-dimensional characteristic by down-sampled and sliding window setting technique by the present invention
Vector is unitized EEG signals, improves the accuracy of brain electric information state change detection.
(2) abnormality degree computing module is used to build using characteristic value of the autoregression model to the EEG signals of extraction
Mould, and pass through the timing abnormality degree of residual analysis calculating EEG signals.
In the abnormality degree computing module, when the EEG signals at a certain moment no longer meet the model of above-mentioned foundation, then
Current time is change point.
Wherein, abnormality degree computing module includes:
(2.1) it predicts error calculating module, is used for respectively by characteristic value in one-dimensional characteristic vector and using from returning
Return model to be modeled to obtain characteristic value to the characteristic value of the EEG signals of extraction and make poor, obtains predicting error；
Specifically, it is assumed that given EEG signals sequence has been represented as { q1,q2,...,qN, with an autoregression line
Property model modeling analysis is carried out to the signal, modeling pattern is as follows:
qt=μ+β t+ εt
Wherein, μ is the mean value of EEG signals sequence, and β is variation tendency, εtFor an independent identically distributed error, and meet
It is desired for zero.Calculating can be passed throughObtain the value of μ and β, it may be assumed that
When the EEG signals at a certain moment no longer meet above-mentioned model, that is, think that the moment is change point.Obviously, once
Data distribution changes, two in above-mentioned model parameter μ and β will with it is significantly different in model built before.Assuming that in c
It carves signal distributions to change, the parameter in model is by original (μ1,β1) become (μ2,β2), a piecewise linearity mould can be passed through
Type describes this variation, it may be assumed that
Provide following construction testing model, it may be assumed that
H0:μ1=μ2andβ1=β2
HA:μ1≠μ2and/orβ1≠β2
When significant change does not occur for EEG signals state, μ and β are remained unchanged；If apparent variation occurs, μ and β
It will change.That is, working as H0It is rejected, i.e. HAWhen being true, the signal at current time is confirmed as change point.
Introduce prediction error { etQuantify { q1,q2,…,qNData distribution timing fluctuate situation.
Assuming that (μ1,β1) by calculatingIt obtains, then qtLinear predictor can be by establishing in step (2-1) from returning
Model is returned to be calculatedThe then prediction error e at the momenttIt may be expressed as:
Wherein, | | | | it is Euclidean distance.It is calculated to simplify,It can be calculated asIn conjunction with being built in step (2-1)
Vertical piecewise linear model predicts error etValue can be divided into two kinds of situations:
1) as t < c, current demand signal obeys preset data distribution, predicts error etVery little is approximately equal to 0；
2) as t=c, current demand signal is not obeying original data distribution, the autoregression model established in step (2-1)
Significant change occurs for parameter, so that prediction error etBecome larger.
(2.2) it predicts error criterion module, is used to be standardized prediction error；
For the prediction error sequence { e of acquisitiont, it is standardized, it may be assumed that
Wherein,For sample average, σ is standard deviation.
(2.3) Euclidean distance computing module, after being used for by seeking the prediction error after any standard and standardization
Prediction error mean between Euclidean distance, obtain the timing abnormality degree of EEG signals.
Finally, in the { z being calculated1,z2,...,zt-1On the basis of, data exception degree can calculate are as follows:
st=s ({ z1,z2,...,zt-1},zt)=| | zt-Ht-1||
Wherein,
(3) change moment determining module, be used to count based on timing abnormality degree of the random power halter strap to EEG signals
At the time of examining, and then determine that EEG signals state changes.
Wherein, variation moment determining module includes:
(3.1) halter strap function constructs module, is used to construct halter strap function；
Abnormality degree { s1,s2,...,stOn the basis of, using based on random power halter strap (Random Power Martingale,
RPM) statistical check.It is that the data statistics inspection an of printenv can carry out detection operation under unsupervised environment, and can
It is executed with real-time online.It is as follows to construct halter strap (Martingale):
Wherein, (0,1) ξ ∈, is set as 0.8 in,Calculation formula it is as follows:
In formula: # { } is a counting function, θiThe random number being generally evenly distributed between [0,1].Due toM (t) can be obtained by iterative calculation.
(3.2) judgment module, be used for by Doob maximum inequality carry out decision come judge current time whether be
At the time of EEG signals state changes.
Wherein, decision is carried out by Doob maximum inequality (Doob ' s Maximal Inequality), it may be assumed that
H0: 1 < M (c) < λ
HA:M(c)≥λ
If the martingale value of current time c is less than preset λ value, receive the hypothesis testing in above-mentioned inequality
H0, then current time c is not change point；Otherwise receive hypothesis testing HA, that is, think that current time c is change point.Threshold in inequality
Value λ is obtained by cross validation, can realize that the dynamic of testing result adjusts by changing the value of λ in inequality.
Above-mentioned, although the foregoing specific embodiments of the present invention is described with reference to the accompanying drawings, not protects model to the present invention
The limitation enclosed, those skilled in the art should understand that, based on the technical solutions of the present invention, those skilled in the art are not
Need to make the creative labor the various modifications or changes that can be made still within protection scope of the present invention.
Claims (3)
1. a kind of real-time detecting system of EEG signals state change characterized by comprising
Characteristics extraction module is used to extract the characteristic value of EEG signals；
Abnormality degree computing module is used to be modeled using characteristic value of the autoregression model to the EEG signals of extraction, and led to
Cross the timing abnormality degree that residual analysis calculates EEG signals；
Change moment determining module, is used to carry out statistical check based on timing abnormality degree of the random power halter strap to EEG signals, into
And at the time of determining that EEG signals state changes；
The characteristics extraction module includes:
Down-sampled module is used to carry out EEG signals with the down-sampled rate of 1:n down-sampled, formation EEG signals voltage magnitude
Sequence；Wherein, n is the positive integer greater than 1；
Sliding window characteristic extracting module is used to introduce the sliding window of a regular length, extracts window in the sliding window
The Time-domain Statistics feature of interior EEG signals voltage magnitude sequence；
Fusion Features module, the Time-domain Statistics Fusion Features for being used to obtain are at one-dimensional characteristic vector；
The abnormality degree computing module includes:
Predict error calculating module, be used for respectively by characteristic value in one-dimensional characteristic vector with using autoregression model to mentioning
The characteristic value of the EEG signals taken is modeled to obtain characteristic value and makees poor, obtains predicting error；
It predicts error criterion module, is used to be standardized prediction error；
Euclidean distance computing module is used for by seeking the prediction error after any standard and the prediction error after standardization
Euclidean distance between mean value obtains the timing abnormality degree of EEG signals.
2. the real-time detecting system of EEG signals state change as described in claim 1, which is characterized in that the abnormality degree meter
It calculates in module, when the EEG signals at a certain moment no longer meet the model of above-mentioned foundation, then current time is change point.
3. the real-time detecting system of EEG signals state change as described in claim 1, which is characterized in that the variation moment
Determining module includes:
Halter strap function constructs module, is used to construct halter strap function；
Judgment module is used to carry out decision by Doob maximum inequality to judge whether current time is EEG signals shape
At the time of state changes.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710349043.4A CN107212882B (en) | 2017-05-17 | 2017-05-17 | A kind of real-time detection method and system of EEG signals state change |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710349043.4A CN107212882B (en) | 2017-05-17 | 2017-05-17 | A kind of real-time detection method and system of EEG signals state change |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107212882A CN107212882A (en) | 2017-09-29 |
| CN107212882B true CN107212882B (en) | 2019-05-21 |
Family
ID=59944188
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710349043.4A Expired - Fee Related CN107212882B (en) | 2017-05-17 | 2017-05-17 | A kind of real-time detection method and system of EEG signals state change |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107212882B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109247936B (en) * | 2018-10-31 | 2023-06-13 | 山东大学 | Abnormal electroencephalogram behavior monitoring system and method for whole night sleep monitoring |
| CN110477910A (en) * | 2019-08-14 | 2019-11-22 | 深圳先进技术研究院 | Epileptic seizure prediction device, terminal device and computer readable storage medium |
| CN111671419B (en) * | 2020-06-12 | 2022-02-11 | 山东大学 | Electroencephalogram signal-based epilepsy early detection and identification method and system |
| CN116489236B (en) * | 2023-06-25 | 2023-09-12 | 中国科学院自动化研究所 | Electroencephalogram signal real-time lossless compression method and device, electronic equipment and storage medium |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20050119534A1 (en) * | 2003-10-23 | 2005-06-02 | Pfizer, Inc. | Method for predicting the onset or change of a medical condition |
| US8073963B1 (en) * | 2006-11-13 | 2011-12-06 | George Mason Intellectual Properties, Inc. | Data stream change detector |
| CN103815912B (en) * | 2014-01-26 | 2015-08-19 | 大连大学 | Old solitary people based on thermal infrared sensor array falls down behavior method of real-time |
| CN105675320B (en) * | 2016-01-06 | 2018-02-16 | 山东大学 | A kind of mechanical system running status method for real-time monitoring based on acoustic signal analysis |
-
2017
- 2017-05-17 CN CN201710349043.4A patent/CN107212882B/en not_active Expired - Fee Related
Also Published As
| Publication number | Publication date |
|---|---|
| CN107212882A (en) | 2017-09-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107212882B (en) | A kind of real-time detection method and system of EEG signals state change | |
| WO2019174142A1 (en) | Multi-mode degradation process modelling and remaining service life prediction method | |
| CN110572297B (en) | Network performance evaluation method, server and storage medium | |
| CN103761748A (en) | Method and device for detecting abnormal behaviors | |
| CN110309136B (en) | Method and system for filling missing data of database abnormal event | |
| WO2023165006A1 (en) | Predictive maintenance method and apparatus for industrial equipment based on health status index, and electronic device | |
| US11381583B1 (en) | Systems and methods for detecting anomalous system or network behavior | |
| CN115526515B (en) | Safety monitoring system of gate for water conservancy and hydropower | |
| CN105974223B (en) | A kind of method and system for on-line checking electrical equipment working condition | |
| CN103353883A (en) | Big data stream type cluster processing system and method for on-demand clustering | |
| Pastore et al. | S pi C o D yn: A Toolbox for the Analysis of Neuronal Network Dynamics and Connectivity from Multi-Site Spike Signal Recordings | |
| CN114325405A (en) | Battery pack consistency analysis method, modeling method, device, equipment and medium | |
| CN111161747B (en) | Prediction method and device based on Tensorflow awakening model and computer equipment | |
| CN116364203A (en) | Water quality prediction method, system and device based on deep learning | |
| CN116248674A (en) | Data transmission method, device and system | |
| CN114970635A (en) | Dynamic load leading model parameter identification method and system based on noise-like signal | |
| CN112526945B (en) | Full-process monitoring and early warning system with feedback and closed-loop control functions | |
| CN112228042B (en) | Method for judging working condition similarity of pumping well based on cloud edge cooperative computing | |
| CN111563078B (en) | Data quality detection method and device based on time sequence data and storage device | |
| CN111931798A (en) | Method for carrying out classification detection and service life prediction of cold head state | |
| CN106059093A (en) | Smart substation health condition monitoring system | |
| CN105843671B (en) | Resources of virtual machine security monitoring and risk pretreatment system based on cloud platform | |
| CN110263811A (en) | A kind of equipment running status monitoring method and system based on data fusion | |
| CN113349797B (en) | System for predicting epileptic seizure based on electroencephalogram critical nuclear parameters | |
| Chen et al. | A multimode anomaly detection method based on oc-elm for aircraft engine system |
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 | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20190521 Termination date: 20200517 |