CN113951898A - P300 electroencephalogram signal detection method and device for data migration, electronic device and medium - Google Patents
P300 electroencephalogram signal detection method and device for data migration, electronic device and medium Download PDFInfo
- Publication number
- CN113951898A CN113951898A CN202111202081.XA CN202111202081A CN113951898A CN 113951898 A CN113951898 A CN 113951898A CN 202111202081 A CN202111202081 A CN 202111202081A CN 113951898 A CN113951898 A CN 113951898A
- Authority
- CN
- China
- Prior art keywords
- domain
- data set
- training
- electroencephalogram signal
- target electroencephalogram
- 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
Images
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
- A61B5/369—Electroencephalography [EEG]
- A61B5/372—Analysis of electroencephalograms
Abstract
The invention discloses a method and a device for detecting a P300 electroencephalogram signal of data migration, electronic equipment and a medium, wherein the method comprises the following steps: performing multi-domain preprocessing on a target electroencephalogram signal containing a P300 component and a migration data set; extracting multi-domain characteristics from the target electroencephalogram signal subjected to multi-domain preprocessing and the migration data set; taking a migration data set subjected to multi-domain feature extraction as a training set, taking a target electroencephalogram signal training data set subjected to multi-domain feature extraction as a test set, and taking a target electroencephalogram signal training data set subjected to multi-domain feature extraction which is correctly classified as a pre-trained target electroencephalogram training set by using a multi-domain combination classifier based on a confidence coefficient; and (3) performing secondary training on the multi-domain combination classifier by using a pre-trained target electroencephalogram training set, and classifying a target electroencephalogram signal test data set subjected to multi-domain feature extraction, thereby obtaining a final target electroencephalogram signal P300 classification result.
Description
Technical Field
The application relates to the technical field of electroencephalogram signal processing, in particular to a method and a device for detecting a P300 electroencephalogram signal by data migration, an electronic device and a medium.
Background
The P300 electroencephalogram signal is a forward peak which can be generated about 300 milliseconds after a person receives 'target stimulation' which is mixed with conventional irrelevant stimulation and occurs with a small probability, is a very common event-related potential component, is often used as an electroencephalogram signal for controlling and communicating external equipment in a brain-computer interface, is also a main object of lie detection research such as hidden information test and the like, and is also often used for evaluating the cognitive function of a tested person. The P300 component in the electroencephalogram signal is detected, and the distinction of the P300 component is the key point of the P300 electroencephalogram signal processing research. Many researches show that brain diseases such as epilepsy can affect the form of P300 components, so that the P300 amplitude is reduced, the latency is increased, the induced electroencephalogram signal discrimination of target stimulation and irrelevant stimulation is low, the difficulty of identifying the P300 signal is increased, and adverse effects can be caused in the application of a P300-based brain-computer interface. Meanwhile, the P300 electroencephalogram data volume obtained by an epileptic is often small, and the detection effect is limited.
Disclosure of Invention
The embodiment of the application aims to provide a method and a device for detecting a P300 electroencephalogram signal by data migration, an electronic device and a medium, so as to solve the technical problems of abnormal P300 signal form, small data volume and high detection difficulty of an epileptic patient.
According to a first aspect of the embodiments of the present application, a method for detecting a P300 electroencephalogram signal of data migration is provided, including:
performing multi-domain preprocessing on a target electroencephalogram signal containing a P300 component and a migration data set;
extracting multi-domain features from the target electroencephalogram signal and the migration data set after the multi-domain preprocessing;
taking the migration data set subjected to multi-domain feature extraction as a training set, taking the target electroencephalogram signal training data set subjected to multi-domain feature extraction as a test set, and taking the target electroencephalogram signal training data set subjected to multi-domain feature extraction which is correctly classified as a pre-trained target electroencephalogram training set by using a multi-domain combination classifier based on a confidence coefficient;
and performing secondary training on the multi-domain combination classifier by using a pre-trained target electroencephalogram training set, and classifying the target electroencephalogram signal test data set subjected to multi-domain feature extraction, thereby obtaining a final target electroencephalogram signal P300 classification result.
Further, the multi-domain preprocessing is carried out on the target brain electrical signal containing the P300 component and the migration data set, and comprises the following steps:
respectively unifying sampling frequency of a target electroencephalogram signal containing a P300 component and a migration data set;
respectively carrying out spatial domain preprocessing on the target electroencephalogram signal and the migration data set with uniform sampling frequency by an electrode selection and common average reference method;
respectively carrying out time domain preprocessing on the target electroencephalogram signal and the migration data set after the uniform sampling frequency through an extreme value adjusting and normalizing method;
and respectively carrying out frequency domain preprocessing on the target electroencephalogram signal and the migration data set with unified sampling frequency by a wavelet packet decomposition method.
Further, extracting multi-domain features from the target electroencephalogram signal and the migration data set which are subjected to the multi-domain preprocessing comprises the following steps:
respectively extracting time domain energy entropy as time domain characteristics from the target electroencephalogram signal subjected to the multi-domain preprocessing and the migration data set;
extracting energy information of wavelet coefficients as time-frequency characteristics from the target electroencephalogram signal subjected to multi-domain preprocessing and the migration data set respectively;
and respectively extracting signal space domain characteristics of the target electroencephalogram signal subjected to the multi-domain preprocessing and the migration data set by using independent component analysis.
Further, taking the migration data set subjected to multi-domain feature extraction as a training set, taking the target electroencephalogram signal training data set subjected to multi-domain feature extraction as a test set, and taking the target electroencephalogram signal training data set subjected to multi-domain feature extraction which is correctly classified as a pre-trained target electroencephalogram training set by using a multi-domain combination classifier based on a confidence coefficient, the method comprises the following steps:
pre-training a multi-domain combination classifier based on a confidence coefficient by using a migration data set to obtain a multi-domain combination classifier which is trained once;
taking the migration data set subjected to multi-domain feature extraction as a training set of the multi-domain combined classifier which is trained for one time;
taking the target electroencephalogram signal training data set subjected to multi-domain feature extraction as a test set of the multi-domain combined classifier of the primary training;
and outputting a target electroencephalogram signal training data set which is correctly classified by the multi-domain combined classifier of the primary training and is subjected to multi-domain feature extraction as a pre-trained target electroencephalogram training set.
Further, the multi-domain combined classifier based on the confidence coefficient comprises two linear discriminant analysis classifiers with different threshold constants and a naive Bayes classifier, wherein the linear discriminant analysis classifier takes the distance between the feature vector X and the classification line as the confidence coefficient, and the naive Bayes classifier takes the difference of the two classification probabilities as the confidence coefficient.
Further, for the two linear discriminant analysis classifiers with different threshold constants, one of the linear discriminant analysis classifiers is used as a basic component classifier, a confidence coefficient threshold is selected, the classification result with the confidence coefficient higher than the threshold is directly accepted, and for the classification sample with the confidence coefficient lower than the threshold, a naive Bayes classifier and the other linear discriminant analysis classifier with different threshold constants are used as a reference component classifier.
Further, performing secondary training on the multi-domain combination classifier by using a pre-trained target electroencephalogram training set, and classifying the target electroencephalogram signal test data set subjected to multi-domain feature extraction, so as to obtain a final target electroencephalogram signal P300 classification result, wherein the classification result comprises:
performing secondary training on the multi-domain combined classifier subjected to the primary training by using a pre-trained target electroencephalogram training set to obtain a multi-domain combined classifier subjected to the secondary training;
taking a pre-trained target electroencephalogram training set as a training set of the secondarily-trained multi-domain combined classifier;
taking a target electroencephalogram signal test data set subjected to multi-domain feature extraction as a test set of the secondarily-trained multi-domain combined classifier;
and outputting the classification result of the multi-domain combined classifier of the secondary training as a final target electroencephalogram signal P300 classification result.
According to a second aspect of the embodiments of the present application, there is provided a P300 electroencephalogram signal detection apparatus for data migration, including:
the preprocessing module is used for performing multi-domain preprocessing on the target electroencephalogram signal containing the P300 component and the migration data set;
the characteristic extraction module is used for extracting multi-domain characteristics from the target electroencephalogram signal and the migration data set which are subjected to the multi-domain preprocessing;
the pre-training module is used for taking the migration data set subjected to multi-domain feature extraction as a training set, taking the target electroencephalogram signal training data set subjected to multi-domain feature extraction as a test set, and taking the target electroencephalogram signal training data set subjected to multi-domain feature extraction which is correctly classified as a pre-trained target electroencephalogram training set by using a multi-domain combination classifier based on a confidence coefficient;
and the secondary training classification module is used for carrying out secondary training on the multi-domain combination classifier by using a pre-trained target electroencephalogram training set and classifying the target electroencephalogram signal test data set subjected to multi-domain feature extraction so as to obtain a final target electroencephalogram signal P300 classification result.
According to a third aspect of embodiments of the present application, there is provided an electronic apparatus, including: one or more processors; a memory for storing one or more programs; when executed by the one or more processors, cause the one or more processors to implement a method as described in the first aspect.
According to a fourth aspect of embodiments herein, there is provided a computer-readable storage medium having stored thereon computer instructions, characterized in that the instructions, when executed by a processor, implement the steps of the method according to the first aspect.
The technical scheme provided by the embodiment of the application can have the following beneficial effects:
according to the technical scheme, the migration data set with large data volume is introduced, so that the problems that electroencephalogram signal samples of epileptics are few and learning is difficult are solved, the data distribution of the migration data set is learned and utilized, and the final classification accuracy is improved. Because a migration data set of a P300 signal form standard is introduced, the problem that the electroencephalogram signal form of an epileptic patient is abnormal is solved, the interference of samples with poor quality in a training data set on a classifier is reduced, and the final classification accuracy is improved. Because multi-domain feature extraction is used, multi-domain features of the P300 signal can be fully extracted, and the P300 signal is better characterized. Because the multi-domain combined classifier is used, the advantages of various classifiers can be combined, and the high total classification accuracy can be realized in less operation time.
It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the application.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and together with the description, serve to explain the principles of the application.
FIG. 1 is a flowchart illustrating a method for P300 brain electrical signal detection for data migration according to an exemplary embodiment.
FIG. 2 is a flowchart illustrating a confidence coefficient based multi-domain combined classifier in accordance with an exemplary embodiment.
FIG. 3 illustrates a block diagram of a data migrated P300 brain electrical signal detection apparatus, according to an example embodiment.
Detailed Description
Reference will now be made in detail to the exemplary embodiments, examples of which are illustrated in the accompanying drawings. When the following description refers to the accompanying drawings, like numbers in different drawings represent the same or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present application. Rather, they are merely examples of apparatus and methods consistent with certain aspects of the present application, as detailed in the appended claims.
The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. As used in this application and the appended claims, the singular forms "a", "an", and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It should also be understood that the term "and/or" as used herein refers to and encompasses any and all possible combinations of one or more of the associated listed items.
It is to be understood that although the terms first, second, third, etc. may be used herein to describe various information, such information should not be limited to these terms. These terms are only used to distinguish one type of information from another. For example, first information may also be referred to as second information, and similarly, second information may also be referred to as first information, without departing from the scope of the present application. The word "if" as used herein may be interpreted as "at … …" or "when … …" or "in response to a determination", depending on the context.
Fig. 1 is a flowchart illustrating a method for detecting a P300 electroencephalogram signal for data migration according to an exemplary embodiment, where the method is applied to a terminal, and may include the following steps:
step S11, performing multi-domain preprocessing on the target electroencephalogram signal containing the P300 component and the migration data set;
step S12, extracting multi-domain features from the target electroencephalogram signal and the migration data set after the multi-domain preprocessing;
step S13, taking the migration data set subjected to multi-domain feature extraction as a training set, taking the target electroencephalogram signal training data set subjected to multi-domain feature extraction as a test set, and taking the target electroencephalogram signal training data set subjected to multi-domain feature extraction which is correctly classified as a pre-trained target electroencephalogram training set by using a multi-domain combination classifier based on a confidence coefficient;
and step S14, performing secondary training on the multi-domain combination classifier by using a pre-trained target electroencephalogram training set, and classifying the target electroencephalogram signal test data set subjected to multi-domain feature extraction, so as to obtain a final target electroencephalogram signal P300 classification result.
According to the technical scheme, the migration data set with large data volume is introduced, so that the problems that electroencephalogram signal samples of epileptics are few and learning is difficult are solved, the data distribution of the migration data set is learned and utilized, and the final classification accuracy is improved. Because a migration data set of a P300 signal form standard is introduced, the problem that the electroencephalogram signal form of an epileptic patient is abnormal is solved, the interference of samples with poor quality in a training data set on a classifier is reduced, and the final classification accuracy is improved. Because multi-domain feature extraction is used, multi-domain features of the P300 signal can be fully extracted, and the P300 signal is better characterized. Because the multi-domain combined classifier is used, the advantages of various classifiers can be combined, and the high total classification accuracy can be realized in less operation time. The method can effectively classify the P300 electroencephalogram signals with nonstandard forms and poor data quality, such as the P300 signals of epileptics.
In a specific implementation of step S11, performing multi-domain preprocessing on the target electroencephalogram signal containing the P300 component and the migration data set, the step may include the following sub-steps:
step S111, respectively unifying sampling frequencies of a target electroencephalogram signal containing a P300 component and a migration data set;
specifically, the sampling frequency of the target electroencephalogram signal containing the P300 component and the electroencephalogram signal of the migration data set is unified to 500Hz by using an interpolation method and a downsampling technology, so that the time scales of the features are consistent during subsequent multi-domain preprocessing and multi-domain feature extraction, and the operation is convenient.
Step S112, respectively carrying out spatial domain preprocessing on the target electroencephalogram signal and the migration data set with unified sampling frequency through an electrode selection and common average reference method;
specifically, using common average reference to remove correlated noise in the electroencephalogram signals obtained in step S111, subtracting the average of all the electrode original electroencephalogram signals from the electrode original electroencephalogram signals after common average reference processing, and setting the original signals as X for the original electroencephalogram signals of n electrode channelsi,i∈[1,n]The signal after the common average reference is:
and selecting signals of Fz, Cz and Pz electrode channels as main analysis objects by using channel selection, so that the computation amount is reduced.
Step S113, respectively carrying out time domain preprocessing on the target electroencephalogram signal and the migration data set with unified sampling frequency by an extreme value adjusting and normalizing method;
specifically, extremum adjustment is performed in a time domain, and the maximum and minimum 5% amplitude abnormal signal samples in the electroencephalogram signal obtained in step S111 are respectively replaced by appropriate acceptable thresholds C1 and C2, where the thresholds satisfy that 5% of samples in the electroencephalogram signal are greater than C1 and 5% of samples are less than C2; normalization is carried out, the electroencephalogram signal obtained in S111 is set as S, and the normalized signal is set as S', so that
U and L are the maximum value and the minimum value of the normalized projection space, and are respectively set to be 1 and-1, so that the electroencephalogram signal is normalized to the range of the minimum value of-1 and the maximum value of 1, and subsequent calculation is facilitated.
And S114, respectively carrying out frequency domain preprocessing on the target electroencephalogram signal and the migration data set with the uniform sampling frequency by a wavelet packet decomposition method.
Specifically, the db4 mother wavelet is used for wavelet packet decomposition, the electroencephalogram signal obtained in S111 is filtered, and the 0.25-30Hz electroencephalogram signal is extracted.
And S115, performing effective signal extraction on the target electroencephalogram signal and the migration data set with the uniform sampling frequency through time domain cutting.
Specifically, an effective signal of 200-600ms after the stimulation is intercepted from the electroencephalogram signal obtained in S111 is used for detecting a P300 component and distinguishing target stimulation from irrelevant stimulation.
In a specific implementation of step S12, extracting multi-domain features from the target electroencephalogram signal and migration data set subjected to the multi-domain preprocessing may include the following sub-steps:
step S121, extracting time domain energy entropy as time domain features from the target electroencephalogram signal subjected to the multi-domain preprocessing and the migration data set respectively;
specifically, the time domain energy entropy is used as a time domain feature, the electroencephalogram signal is divided into 10 segments on the average in the time domain, the time domain energy entropy of each segment is based on 2, a negative value of the logarithm of the proportion of the time domain energy of the segment signal in the sum of the total energy of each segment is used as the time domain feature, and each segment signal is set to be xi，i∈[1,10]Time domain energy entropy E of each segmentiComprises the following steps:
step S122, extracting energy information of wavelet coefficients as time-frequency characteristics from the target electroencephalogram signal subjected to the multi-domain preprocessing and the migration data set respectively;
specifically, the energy information of the wavelet coefficients is used as time-frequency characteristics, db4 is used as a mother wavelet function, wavelet decomposition is carried out on the electroencephalogram signals, wavelet approximation coefficients corresponding to a 0-16Hz frequency band are extracted, and squaring is carried out to obtain the energy information of the wavelet approximation coefficients as the time-frequency characteristics.
And S123, respectively extracting signal space domain characteristics of the target electroencephalogram signal subjected to the multi-domain preprocessing and the migration data set by using independent component analysis.
Specifically, the spatial domain features are obtained by using independent component analysis, the signals of the Fz, Cz and Pz channels are subjected to independent component analysis by using an Infmax algorithm, a 3 x 3 mixing matrix is obtained, and the expanded 9 features are used as the spatial domain features.
In the specific implementation of step S13, taking the migration data set subjected to multi-domain feature extraction as a training set, taking the target electroencephalogram signal training data set subjected to multi-domain feature extraction as a test set, and using a multi-domain combination classifier based on a confidence coefficient, and taking the target electroencephalogram signal training data set subjected to multi-domain feature extraction with correct classification as a pre-trained target electroencephalogram training set, with reference to fig. 2, the step may include the following sub-steps:
step S131, pre-training the multi-domain combined classifier based on the confidence coefficient by using the migration data set to obtain a multi-domain combined classifier which is trained once;
specifically, the migration data set subjected to feature extraction is used as a training set, and a multi-domain combined classifier based on a confidence coefficient is pre-trained to obtain a multi-domain combined classifier which is trained once.
And S132, classifying the training data of the target electroencephalogram signal by using the multi-domain combined classifier for one-time training, and outputting the classified target electroencephalogram signal training data which is correctly classified and subjected to multi-domain feature extraction as a pre-trained target electroencephalogram training set.
Specifically, the target electroencephalogram signal training data set subjected to multi-domain feature extraction is used as a test set of the multi-domain combination classifier subjected to one-time training, the multi-domain combination classifier subjected to one-time training is used for classifying the test set, and the classified target electroencephalogram signal training data subjected to multi-domain feature extraction and correctly classified are selected according to the classification result and output to be used as a pre-trained target electroencephalogram training set.
Further, the multi-domain combined classifier based on the confidence coefficient comprises two linear discriminant analysis classifiers with different threshold constants and a naive Bayes classifier, wherein the linear discriminant analysis classifier takes the distance between the feature vector X and the classification line as the confidence coefficient, and the naive Bayes classifier takes the difference of the two classification probabilities as the confidence coefficient.
Specifically, in the linear discriminant analysis classifier, for the feature vector X, X ═ X (X)1,x2,…,xn)THaving a linear function
Wherein WL＝(w1,w2,…,wn)TIs an n-dimensional weight vector, and theta is a threshold constant. For the binary problem, there is a decision mechanism: d (X)>0，X∈C1,d(X)<0，X∈C2D (x) 0 is a classification boundary. The optimal W can be calculated using Fisher's criterionLAnd theta are substituted into | d (X) |, namely the distance between the feature vector X and the classification line, namely the confidence coefficient of the linear discriminant analysis classifier.
In a naive bayes classifier, the feature vector X, X ═ X1,x2,…,xn)TSolving for each class C under the condition that the feature vector appearsiProbability of occurrence P (C)i|X)，P(C1I X) and P (C)2| X), and a difference Δ P greater than 0 is classified as C1And if Δ P is less than 0, it is classified as C2. Therefore, | Δ P | is the confidence coefficient of the naive bayes classifier.
Further, for the two linear discriminant analysis classifiers with different threshold constants, one of the linear discriminant analysis classifiers is used as a basic component classifier, a confidence coefficient threshold is selected, the classification result with the confidence coefficient higher than the threshold is directly accepted, and for the classification sample with the confidence coefficient lower than the threshold, a naive Bayes classifier and the other linear discriminant analysis classifier with different threshold constants are used as a reference component classifier.
Specifically, the confidence coefficient is normalized to eliminate the effect of variability of the confidence coefficient, and the Logistic curve is adjusted to the confidence coefficient to improve the confidence coefficient discrimination.
In the construction of the combined classifier, the linear discriminant analysis classifier LDA1 with the highest speed but the general classification accuracy is used as a basic component classifier, and the absolute value | d of the distance between the feature vector X in the LDA1 and a classification line is used1(X) is used as a confidence coefficient, and a confidence coefficient threshold is set as the mean value of the absolute values of the distances between the characteristic vector X and the classification line in the training set
Classification results with confidence coefficients above a threshold are accepted directly.
For confidence systemThe classification samples with numbers below the threshold were used as reference component classifiers using a naive bayes classifier and a linear discriminant analysis classifier LDA2 with different threshold constants. Confidence coefficient of LDA2 is | d2(X) | takes the binary classification difference | Δ P | of the naive Bayes classifier as its confidence coefficient.
To eliminate the effect of confidence coefficient variability, the confidence coefficient | d is set1(X)|、|d2The (X) | and | Δ P | are normalized, and the confidence coefficient is adjusted by an S-shaped curve in order to improve the confidence coefficient discrimination. And taking the confidence coefficient after normalization and S-shaped curve adjustment as weight, carrying out weighted average on the classification results of the three component classifiers, and taking the classification of the weighted average as a final classification result.
In the specific implementation of step S14, the multi-domain combination classifier is secondarily trained using a pre-trained target electroencephalogram training set, and the target electroencephalogram signal test data set extracted by the multi-domain features is classified, so as to obtain a final target electroencephalogram signal P300 classification result, which may include the following sub-steps:
step S141, performing secondary training on the multi-domain combined classifier subjected to primary training by using a pre-trained target electroencephalogram training set to obtain a multi-domain combined classifier subjected to secondary training;
specifically, a pre-trained target electroencephalogram training set subjected to multi-domain feature extraction is used as a training set of the multi-domain combined classifier for the secondary training, and the multi-domain combined classifier for the primary training is subjected to the secondary training to obtain the multi-domain combined classifier for the secondary training.
And S142, classifying the test set of the target electroencephalogram signal by using the multi-domain combined classifier of the secondary training, and outputting the classification result as the final classification result of the target electroencephalogram signal P300.
Specifically, a target electroencephalogram signal test data set subjected to multi-domain feature extraction is used as a test set of the multi-domain combination classifier of the secondary training, the test set is classified by using the multi-domain combination classifier of the secondary training, and a classification result is output as a final target electroencephalogram signal P300 classification result.
Due to the introduction of the migration data set, such as the data II in the BCI competition III, the problems that electroencephalogram signal samples of epileptics are few and learning is difficult can be solved due to the large sample amount of the migration data set, and the classification accuracy of the test set in the final target data is improved by learning and utilizing the data distribution of the migration data set. Due to the fact that the migration data set P300 is standard in signal form, the problem that electroencephalogram signal form of epileptics is abnormal can be solved, samples with poor quality in a training data set in target data are screened out through migration pre-training, interference of the samples with poor quality in the training data set on a classifier is reduced, and the effect of improving the classification accuracy of a test set in the final target data is achieved.
Corresponding to the embodiment of the method for detecting the P300 electroencephalogram signals of data migration, the application also provides an embodiment of a device for detecting the P300 electroencephalogram signals of data migration.
FIG. 3 is a block diagram illustrating a data migrated P300 brain electrical signal detection apparatus, according to an example embodiment. Referring to fig. 3, the apparatus includes:
the preprocessing module 21 is configured to perform multi-domain preprocessing on the target electroencephalogram signal containing the P300 component and the migration data set;
the feature extraction module 22 is configured to extract multi-domain features from the target electroencephalogram signal and the migration data set after the multi-domain preprocessing;
the pre-training module 23 is configured to use the migration data set subjected to multi-domain feature extraction as a training set, use the target electroencephalogram signal training data set subjected to multi-domain feature extraction as a test set, and use a multi-domain combination classifier based on a confidence coefficient to use the target electroencephalogram signal training data set subjected to multi-domain feature extraction and correctly classified as a pre-trained target electroencephalogram training set;
and the secondary training classification module 24 is configured to perform secondary training on the multi-domain combination classifier by using a pre-trained target electroencephalogram training set, and classify the target electroencephalogram signal test data set subjected to multi-domain feature extraction, so as to obtain a final target electroencephalogram signal P300 classification result.
With regard to the apparatus in the above-described embodiment, the specific manner in which each module performs the operation has been described in detail in the embodiment related to the method, and will not be elaborated here.
For the device embodiments, since they substantially correspond to the method embodiments, reference may be made to the partial description of the method embodiments for relevant points. The above-described embodiments of the apparatus are merely illustrative, and the units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the scheme of the application. One of ordinary skill in the art can understand and implement it without inventive effort.
Correspondingly, the present application also provides an electronic device, comprising: one or more processors; a memory for storing one or more programs; when executed by the one or more processors, cause the one or more processors to implement a data migrated P300 brain electrical signal detection method as described above.
Accordingly, the present application also provides a computer readable storage medium, on which computer instructions are stored, wherein the instructions, when executed by a processor, implement a data migration P300 electroencephalogram signal detection method as described above.
Other embodiments of the present application will be apparent to those skilled in the art from consideration of the specification and practice of the disclosure disclosed herein. This application is intended to cover any variations, uses, or adaptations of the invention following, in general, the principles of the application and including such departures from the present disclosure as come within known or customary practice within the art to which the invention pertains. It is intended that the specification and examples be considered as exemplary only, with a true scope and spirit of the application being indicated by the following claims.
It will be understood that the present application is not limited to the precise arrangements described above and shown in the drawings and that various modifications and changes may be made without departing from the scope thereof. The scope of the application is limited only by the appended claims.
Claims (10)
1. A P300 electroencephalogram signal detection method for data migration is characterized by comprising the following steps:
performing multi-domain preprocessing on a target electroencephalogram signal containing a P300 component and a migration data set;
extracting multi-domain features from the target electroencephalogram signal and the migration data set after the multi-domain preprocessing;
taking the migration data set subjected to multi-domain feature extraction as a training set, taking the target electroencephalogram signal training data set subjected to multi-domain feature extraction as a test set, and taking the target electroencephalogram signal training data set subjected to multi-domain feature extraction which is correctly classified as a pre-trained target electroencephalogram training set by using a multi-domain combination classifier based on a confidence coefficient;
and performing secondary training on the multi-domain combination classifier by using a pre-trained target electroencephalogram training set, and classifying the target electroencephalogram signal test data set subjected to multi-domain feature extraction, thereby obtaining a final target electroencephalogram signal P300 classification result.
2. The method of claim 1, wherein performing multi-domain preprocessing of the target brain electrical signal and the migrated data set containing the P300 component comprises:
respectively unifying sampling frequency of a target electroencephalogram signal containing a P300 component and a migration data set;
respectively carrying out spatial domain preprocessing on the target electroencephalogram signal and the migration data set with uniform sampling frequency by an electrode selection and common average reference method;
respectively carrying out time domain preprocessing on the target electroencephalogram signal and the migration data set after the uniform sampling frequency through an extreme value adjusting and normalizing method;
and respectively carrying out frequency domain preprocessing on the target electroencephalogram signal and the migration data set with unified sampling frequency by a wavelet packet decomposition method.
3. The method of claim 1, wherein extracting multi-domain features from the target brain electrical signal and migration dataset after the multi-domain preprocessing comprises:
respectively extracting time domain energy entropy as time domain characteristics from the target electroencephalogram signal subjected to the multi-domain preprocessing and the migration data set;
extracting energy information of wavelet coefficients as time-frequency characteristics from the target electroencephalogram signal subjected to multi-domain preprocessing and the migration data set respectively;
and respectively extracting signal space domain characteristics of the target electroencephalogram signal subjected to the multi-domain preprocessing and the migration data set by using independent component analysis.
4. The method of claim 1, wherein the step of using the migration data set subjected to multi-domain feature extraction as a training set, using the target electroencephalogram signal training data set subjected to multi-domain feature extraction as a test set, and using a multi-domain combination classifier based on a confidence coefficient to use the target electroencephalogram signal training data set subjected to multi-domain feature extraction and correctly classified as a pre-trained target electroencephalogram training set comprises:
pre-training a multi-domain combination classifier based on a confidence coefficient by using a migration data set to obtain a multi-domain combination classifier which is trained once;
taking the migration data set subjected to multi-domain feature extraction as a training set of the multi-domain combined classifier which is trained for one time;
taking the target electroencephalogram signal training data set subjected to multi-domain feature extraction as a test set of the multi-domain combined classifier of the primary training;
and outputting a target electroencephalogram signal training data set which is correctly classified by the multi-domain combined classifier of the primary training and is subjected to multi-domain feature extraction as a pre-trained target electroencephalogram training set.
5. The method according to claim 1, wherein the confidence coefficient-based multi-domain combined classifier comprises two linear discriminant analysis classifiers with different threshold constants and a naive Bayes classifier, wherein the linear discriminant analysis classifier takes a distance of the feature vector X from the classification line as a confidence coefficient, and wherein the naive Bayes classifier takes a difference of two classification probabilities as its confidence coefficient.
6. The method according to claim 1, wherein for the two linear discriminant analysis classifiers with different threshold constants, one of the linear discriminant analysis classifiers is used as a base component classifier, a confidence coefficient threshold is selected, the classification results with confidence coefficients higher than the threshold are directly accepted, and for the classification samples with confidence coefficients lower than the threshold, a naive bayes classifier and another linear discriminant analysis classifier with a different threshold constant are used as a reference component classifier.
7. The method of claim 4, wherein the multi-domain combination classifier is secondarily trained using a pre-trained target EEG training set, and the multi-domain feature extracted target EEG test data set is classified, so as to obtain a final target EEG P300 classification result, comprising:
performing secondary training on the multi-domain combined classifier subjected to the primary training by using a pre-trained target electroencephalogram training set to obtain a multi-domain combined classifier subjected to the secondary training;
taking a pre-trained target electroencephalogram training set as a training set of the secondarily-trained multi-domain combined classifier;
taking a target electroencephalogram signal test data set subjected to multi-domain feature extraction as a test set of the secondarily-trained multi-domain combined classifier;
and outputting the classification result of the multi-domain combined classifier of the secondary training as a final target electroencephalogram signal P300 classification result.
8. A P300 electroencephalogram signal detection device for data migration is characterized by comprising:
the preprocessing module is used for performing multi-domain preprocessing on the target electroencephalogram signal containing the P300 component and the migration data set;
the characteristic extraction module is used for extracting multi-domain characteristics from the target electroencephalogram signal and the migration data set which are subjected to the multi-domain preprocessing;
the pre-training module is used for taking the migration data set subjected to multi-domain feature extraction as a training set, taking the target electroencephalogram signal training data set subjected to multi-domain feature extraction as a test set, and taking the target electroencephalogram signal training data set subjected to multi-domain feature extraction which is correctly classified as a pre-trained target electroencephalogram training set by using a multi-domain combination classifier based on a confidence coefficient;
and the secondary training classification module is used for carrying out secondary training on the multi-domain combination classifier by using a pre-trained target electroencephalogram training set and classifying the target electroencephalogram signal test data set subjected to multi-domain feature extraction so as to obtain a final target electroencephalogram signal P300 classification result.
9. An electronic device, comprising:
one or more processors;
a memory for storing one or more programs;
when executed by the one or more processors, cause the one or more processors to implement the method of any one of claims 1-7.
10. A computer-readable storage medium having stored thereon computer instructions, which when executed by a processor, perform the steps of the method according to any one of claims 1-7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111202081.XA CN113951898B (en) | 2021-10-15 | 2021-10-15 | P300 electroencephalogram signal detection method and device for data migration, electronic equipment and medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111202081.XA CN113951898B (en) | 2021-10-15 | 2021-10-15 | P300 electroencephalogram signal detection method and device for data migration, electronic equipment and medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113951898A true CN113951898A (en) | 2022-01-21 |
| CN113951898B CN113951898B (en) | 2023-03-10 |
Family
ID=79464694
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111202081.XA Active CN113951898B (en) | 2021-10-15 | 2021-10-15 | P300 electroencephalogram signal detection method and device for data migration, electronic equipment and medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113951898B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114403899A (en) * | 2022-02-08 | 2022-04-29 | 浙江浙大西投脑机智能科技有限公司 | Depression detection method combining cerebral neuron spike potential and local field potential |
Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20100191139A1 (en) * | 2009-01-28 | 2010-07-29 | Brainscope, Inc. | Method and Device for Probabilistic Objective Assessment of Brain Function |
| CN104523268A (en) * | 2015-01-15 | 2015-04-22 | 江南大学 | Electroencephalogram signal recognition fuzzy system and method with transfer learning ability |
| CN104799852A (en) * | 2015-05-19 | 2015-07-29 | 北京工业大学 | Method for extracting movement imagination electroencephalogram characteristics based on ultralimit learning machine self encoding |
| CN107049239A (en) * | 2016-12-28 | 2017-08-18 | 苏州国科康成医疗科技有限公司 | Epileptic electroencephalogram (eeg) feature extracting method based on wearable device |
| CN109964278A (en) * | 2017-03-30 | 2019-07-02 | 艾腾怀斯股份有限公司 | Pass through the system and method for the error in evaluated in parallel classifier the first classifier of output calibration |
| CN111134667A (en) * | 2020-01-19 | 2020-05-12 | 中国人民解放军战略支援部队信息工程大学 | Electroencephalogram signal-based time migration emotion recognition method and system |
| CN111380558A (en) * | 2018-12-29 | 2020-07-07 | 北京四维图新科技股份有限公司 | Method, device, server and storage medium for ordering points of interest |
| CN111444747A (en) * | 2019-01-17 | 2020-07-24 | 复旦大学 | Epileptic state identification method based on transfer learning and cavity convolution |
| CN111695500A (en) * | 2020-06-11 | 2020-09-22 | 齐鲁工业大学 | Method and system for recognizing motor imagery task of stroke patient based on transfer learning |
| CN111700592A (en) * | 2020-07-02 | 2020-09-25 | 河南科技大学 | Method and system for acquiring epilepsia electroencephalogram automatic classification model and classification system |
| CN112869711A (en) * | 2021-01-19 | 2021-06-01 | 华南理工大学 | Automatic sleep staging and migration method based on deep neural network |
| CN113080864A (en) * | 2021-04-07 | 2021-07-09 | 电子科技大学 | Common sleep disease detection method through automatic sleep staging results |
| US20210282701A1 (en) * | 2020-03-16 | 2021-09-16 | nCefalon Corporation | Method of early detection of epileptic seizures through scalp eeg monitoring |
-
2021
- 2021-10-15 CN CN202111202081.XA patent/CN113951898B/en active Active
Patent Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20100191139A1 (en) * | 2009-01-28 | 2010-07-29 | Brainscope, Inc. | Method and Device for Probabilistic Objective Assessment of Brain Function |
| CN104523268A (en) * | 2015-01-15 | 2015-04-22 | 江南大学 | Electroencephalogram signal recognition fuzzy system and method with transfer learning ability |
| CN104799852A (en) * | 2015-05-19 | 2015-07-29 | 北京工业大学 | Method for extracting movement imagination electroencephalogram characteristics based on ultralimit learning machine self encoding |
| CN107049239A (en) * | 2016-12-28 | 2017-08-18 | 苏州国科康成医疗科技有限公司 | Epileptic electroencephalogram (eeg) feature extracting method based on wearable device |
| CN109964278A (en) * | 2017-03-30 | 2019-07-02 | 艾腾怀斯股份有限公司 | Pass through the system and method for the error in evaluated in parallel classifier the first classifier of output calibration |
| CN111380558A (en) * | 2018-12-29 | 2020-07-07 | 北京四维图新科技股份有限公司 | Method, device, server and storage medium for ordering points of interest |
| CN111444747A (en) * | 2019-01-17 | 2020-07-24 | 复旦大学 | Epileptic state identification method based on transfer learning and cavity convolution |
| CN111134667A (en) * | 2020-01-19 | 2020-05-12 | 中国人民解放军战略支援部队信息工程大学 | Electroencephalogram signal-based time migration emotion recognition method and system |
| US20210282701A1 (en) * | 2020-03-16 | 2021-09-16 | nCefalon Corporation | Method of early detection of epileptic seizures through scalp eeg monitoring |
| CN111695500A (en) * | 2020-06-11 | 2020-09-22 | 齐鲁工业大学 | Method and system for recognizing motor imagery task of stroke patient based on transfer learning |
| CN111700592A (en) * | 2020-07-02 | 2020-09-25 | 河南科技大学 | Method and system for acquiring epilepsia electroencephalogram automatic classification model and classification system |
| CN112869711A (en) * | 2021-01-19 | 2021-06-01 | 华南理工大学 | Automatic sleep staging and migration method based on deep neural network |
| CN113080864A (en) * | 2021-04-07 | 2021-07-09 | 电子科技大学 | Common sleep disease detection method through automatic sleep staging results |
Non-Patent Citations (5)
| Title |
|---|
| GHANAHSHYAM B. KSHIRSAGAR;: "Performance Improvement for Devanagari Script Input Based P300 Speller", 《2018 5TH INTERNATIONAL CONFERENCE ON SIGNAL PROCESSING AND INTEGRATED NETWORKS (SPIN)》 * |
| XIAOCHEN LIU; JIZHONG SHEN; WUFENG ZHAO: "P300-based deception detection of mock network fraud with modified genetic algorithm and combined classification", 《2019 7TH INTERNATIONAL WINTER CONFERENCE ON BRAIN-COMPUTER INTERFACE (BCI)》 * |
| YANG LIU; CHENGUANG YANG; ZHIJUN LI: "The Application of Transfer Learning in P300 Detection", 《 2018 IEEE INTERNATIONAL CONFERENCE ON CYBORG AND BIONIC SYSTEMS 》 * |
| 梁健威: "基于多域融合与遗传算法的P300测谎研究", 《中国优秀博硕士学位论文全文数据库(硕士)信息科技辑》 * |
| 黄建: "基于迁移学习的脑电信号分类算法研究", 《中国优秀硕士学位论文全文数据库 基础科学辑》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114403899A (en) * | 2022-02-08 | 2022-04-29 | 浙江浙大西投脑机智能科技有限公司 | Depression detection method combining cerebral neuron spike potential and local field potential |
| CN114403899B (en) * | 2022-02-08 | 2023-07-25 | 浙江浙大西投脑机智能科技有限公司 | Depression detection device combining brain neuron spike potential and local field potential |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113951898B (en) | 2023-03-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Franke et al. | Bayes optimal template matching for spike sorting–combining fisher discriminant analysis with optimal filtering | |
| JP6359123B2 (en) | Inspection data processing apparatus and inspection data processing method | |
| CN105930663B (en) | Hand tremor signal and audio signal classification method based on evolution fuzzy rule | |
| CN113951898B (en) | P300 electroencephalogram signal detection method and device for data migration, electronic equipment and medium | |
| Siuly et al. | Classification of EEG signals using sampling techniques and least square support vector machines | |
| Fan et al. | Modeling voice pathology detection using imbalanced learning | |
| CN113378620B (en) | Cross-camera pedestrian re-identification method in surveillance video noise environment | |
| Boubchir et al. | Classification of EEG signals for detection of epileptic seizure activities based on LBP descriptor of time-frequency images | |
| Rahmawati et al. | Classify epilepsy and normal Electroencephalogram (EEG) signal using wavelet transform and K-nearest neighbor | |
| CN112378942B (en) | White spirit grade classification and identification method based on nuclear magnetic resonance fingerprint | |
| TWI775186B (en) | Rf fingerprint signal processing device and rf fingerprint signal processing method | |
| Yazdani et al. | Classification of EEG signals correlated with alcohol abusers | |
| Übeyli | Implementing eigenvector methods/probabilistic neural networks for analysis of EEG signals | |
| CN114705431A (en) | Rolling bearing fault diagnosis method based on multi-parameter screening criterion and GWO-PNN | |
| CN114117141A (en) | Self-adaptive density clustering method, storage medium and system | |
| CN109272020B (en) | Method and system for processing outliers in electromyographic data | |
| CN112200228A (en) | Epileptic seizure state identification method based on two-dimensional convolutional neural network | |
| CN107657278B (en) | Optimal sample number sampling method for multi-classification of electroencephalogram signal modes | |
| CN111079824A (en) | Method for classifying epilepsy by comparing vector distances | |
| KR102382538B1 (en) | Method for detecting noise channels of surface emg signals | |
| Haggag et al. | Neural spike representation using cepstrum | |
| Jantan et al. | A multi-model analysis for driving fatigue detection using EEG signals | |
| Goerg | A nonparametric frequency domain EM algorithm for time series classification with applications to spike sorting and macro‐economics | |
| CN115622589A (en) | Frequency hopping communication radiation source individual identification method based on feature space processing | |
| CN112201258A (en) | AMBP-based noise robustness camouflage voice detection method |
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 |