CN108960182B - P300 event related potential classification identification method based on deep learning - Google Patents
P300 event related potential classification identification method based on deep learning Download PDFInfo
- Publication number
- CN108960182B CN108960182B CN201810794963.1A CN201810794963A CN108960182B CN 108960182 B CN108960182 B CN 108960182B CN 201810794963 A CN201810794963 A CN 201810794963A CN 108960182 B CN108960182 B CN 108960182B
- Authority
- CN
- China
- Prior art keywords
- layer
- event
- network
- deep learning
- related potential
- 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.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/12—Classification; Matching
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/02—Preprocessing
- G06F2218/04—Denoising
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Biomedical Technology (AREA)
- Data Mining & Analysis (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Signal Processing (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)
Abstract
A P300 event related potential classification and identification method based on deep learning belongs to the technical field of medical and physiological signal detection and processing analysis. And carrying out high-pass and low-pass filtering on the original signal by adopting a Butterworth filter to remove artifacts and power frequency interference. And amplifying data by using a primary superposition average technology, carrying out normalization and time domain truncation processing on the electroencephalogram signals, and formulating corresponding supervision signals according to signal types. After the preprocessing of the data is completed, the electroencephalogram data are divided into a training set and a verification set, a deep learning network capable of classifying and identifying the P300 event-related potential is constructed, and the network feature extraction capability is improved. The trained network finally predicts the probability that the input signal contains the P300 event-related potential. And finally, predicting the target character according to the experimental paradigm and the probability value output by the network. Experiments show that the algorithm of the invention has good performance, and can obtain better character recognition accuracy under the condition of reducing the experiment times.
Description
Technical Field
The invention belongs to the technical field of medical and physiological signal detection, processing and analysis, and relates to a method for classifying and identifying P300 event-related potentials in an electroencephalogram signal, in particular to a method for classifying and identifying P300 event-related potentials by utilizing deep learning.
Background
In a character spelling brain-computer interface (BCI) system, a character spelling function is realized by detecting the potential related to a P300 event related to human cognition in an electroencephalogram signal. If the P300 event-related potential can be classified and identified efficiently and accurately, the practical application of the BCI system is facilitated. In early research, the traditional signal processing and analyzing method is applied, characteristics such as the maximum amplitude difference and the waveform area of a signal are extracted and sent to a classifier for classification and identification, but the method is too simple, and the problem of low classification and identification accuracy exists. Kaper et al use a Support Vector Machine (SVM) as a P300 signal classifier, and train the classifier to directly classify and identify the acquired EEG signals, so that the accuracy of P300 potential identification is improved, but the P300 potential identification is not very ideal. Li sends the data after artifact removal by Independent Component Analysis (ICA) to the SVM classifier to improve the classification effect, thereby improving the character recognition accuracy. Although the signal after ICA preprocessing is easier to be accurately recognized, a certain time is consumed in data preprocessing. At present, some researchers research P300 time-dependent potential recognition based on deep learning, and although the classification accuracy is better than that of the conventional machine learning method, the classification accuracy still needs to be improved under the condition of reducing the number of experiments.
Disclosure of Invention
The invention mainly aims to solve the problems in the prior art and provide a P300 event-related potential classification and identification method based on deep learning.
The technical scheme adopted by the invention is as follows:
a P300 event related potential classification and identification method based on deep learning is characterized in that a classifier of P300 event related potentials is constructed by deep learning on the basis of brain electrical signal preprocessing under visual stimulation, the probability that current signals contain P300 event related potentials is output by a network through a supervised training network, and then character identification is completed by combining an experimental paradigm. The method comprises the following steps:
A. preprocessing the EEG signal under the visual stimulation
A1. And acquiring electroencephalogram data under visual stimulation. The Butterworth filter is used for filtering the EEG signals under the visual stimulation, and noises such as power frequency interference, electromyographic interference and the like are filtered.
A2. And D, performing data enhancement on the data processed in the step A1 by using a superposition average method, and acquiring more electroencephalogram signals containing event-related evoked potentials.
A3. The signal obtained from A2 is subjected to zero-mean, unit variance normalization.
A4. And performing time domain truncation on the signal obtained by the A3, and manufacturing corresponding supervision signals for the P300 event-related potential and the non-P300 event-related potential.
B. Deep learning network capable of classifying and identifying P300 event-related potentials is constructed
B1. And dividing the preprocessed electroencephalogram signals and the corresponding supervisory signals into a training set and a testing set.
B2. The traditional convolutional neural network in deep learning is improved, a third layer of serially connected convolutional layers in the traditional convolutional neural network is changed into 3 different convolutional layers connected in parallel, the improved convolutional neural network is used as a classification decision device of P300 visual evoked potentials, and the probability that the electroencephalogram signals contain P300 event-related potentials is predicted by utilizing the network.
C. And (4) character classification and identification, and determining the class of the character by utilizing the network output probability and the experimental paradigm.
And B, obtaining the probability that the input electroencephalogram signal contains the P300 event-related potential, determining the row number and the column number of the event-related potential induced in the visual stimulator according to a formula (3), and finally determining the predicted target character.
And P (k, i) is the output probability of the network in n experiments, wherein k represents the experiment serial number, and i is the corresponding row number or column number. The column and row with the highest search probability are the predicted target character position (x, y).
The data enhancement method for the electroencephalogram data in the A2. Because of the limitation of an experimental paradigm, the acquired electroencephalogram data contain less electroencephalogram data of P300 event-related potentials, so that the event-related potentials stimulated at different moments are subjected to primary superposition averaging, and the electroencephalogram data containing the P300 event-related potentials detected in the simulated scalp are used for balancing different types of data.
The network CNN _ L suitable for electroencephalogram signal analysis and processing is invented in the B2 for the traditional convolutional neural network improvement method for deep learning in combination with the inclusion V1 network, the network structure is shown in figure 2, the network has seven layers, the first layer is an input layer, and input data are loaded; the second layer is a convolution layer with convolution kernel size equal to the number of signal channels, which is equivalent to performing spatial filtering on input data; in the third layer, the series connection convolutional layer is changed into three parallel connection convolutional layers with different convolutional kernel sizes, namely time domain filtering is performed on different time scales, and the model feature extraction capability is improved; the fourth layer is a connecting layer and is used for integrating the characteristic diagram obtained by the third layer; the fifth layer is a pooling layer used for reducing network parameters and preventing a small number of training samples from generating an overfitting problem; the sixth layer is a convolutional layer for extracting more abstract features; and the seventh layer is a full connection layer, and the probability that the current input electroencephalogram signal is judged to be the P300 event-related potential by the network is obtained after the characteristics obtained by the sixth layer are subjected to weighted calculation. The improved convolutional neural network in deep learning is used as a classification model, cross entropy is used as a cost function, parameters are updated by an Adam parameter optimization method, and the network is trained by using training skills such as dropout and regularization.
The invention has the beneficial effects that: the method provided by the invention can realize classification and identification of the P300 event related potential in the electroencephalogram signal under the condition of reducing the experiment times, thereby effectively improving the character identification accuracy and being beneficial to the practical application of a character spelling BCI system.
Drawings
Fig. 1 is a system block diagram of P300 event-related potential classification identification based on deep learning of the present invention.
Fig. 2 is a block diagram of an improved convolutional neural network of the present invention.
FIG. 3 is a graph of character recognition accuracy for different experimental times for test samples of the method of the present invention.
Detailed Description
In order to make the implementation objects, technical solutions and advantages of the present invention clearer, the following detailed description is further described with reference to the technical solutions of the present invention and the accompanying drawings:
a general block diagram of a deep learning-based P300 event-related potential classification identification method is shown in FIG. 1. The method can be divided into three links, which are respectively: and signal preprocessing, constructing a classification network and character recognition. The signal preprocessing part is used for filtering some artifacts and noise interference of the electroencephalogram signals obtained through detection by using a filter and expanding data samples by using a data amplification technology. The function of constructing the classification network is to construct a network capable of predicting the probability that the input electroencephalogram signal contains the P300 event-related potential. The role of character recognition is to utilize the output of the network in conjunction with an experimental paradigm to recognize characters. The method comprises the following specific steps:
and A, preprocessing an electroencephalogram sample signal acquired under visual stimulation. The method mainly comprises the following steps:
A1. a matrix visual stimulator with 6 rows and 6 columns formed by 36 characters is placed in front of a subject, the subject stares at a target character in the visual stimulator, in one experiment, the visual stimulator randomly traverses and flickers all rows and columns at a fixed frequency for 12 times in total, and each flicker can enable the brain to generate corresponding electroencephalogram signals. The brain electrical signals are obtained from 64 lead recordings conforming to the 10/20 international standard. Each sample data size is 240 x 64, where 240 represents the number of sample points in the time domain of a single sample data and 64 represents the number of channels of a single sample data. Firstly, a high-pass filter with 3-order cutoff frequency of 0.1Hz and a low-pass filter with 6-order cutoff frequency of 20Hz are designed, and an original signal sequentially passes through the two filters according to a channel to remove power frequency interference and some artifacts.
A2. And limited by the experimental paradigm, one experiment corresponds to one target character, and each target character is experimented 15 times, so that the total number of the experiments is 185 characters. Only 2 of the 12 sample data obtained in one experiment contain event-related potentials, which results in data imbalance. Data amplification is therefore performed on the data for event-related potentials. And (3) carrying out primary superposition averaging on the event-related potentials stimulated at different moments, and simulating the electroencephalogram data containing the P300 event-related potentials detected in the scalp.
A3. And carrying out normalization processing on the obtained signals. Because the pre-normalized data can accelerate the training speed of the deep learning network and improve the classification and identification capability of the network, the acquired electroencephalogram signal x needs to be subjected toi,jAnd carrying out normalization processing of zero mean and unit variance according to channels. The transformation formula is as follows:
where i represents the number of sample points on the time axis, j represents the number of channels of the EEG signal,
represents the average, σ, of all sample points of the jth channeljRepresenting the variance of the sample point on the jth channel.
A4. The signal obtained from A3 is truncated in time domain, and since the event-related potential appears about 300ms after stimulation, the data length of each channel is only 0-667ms, i.e. 160 sampling points. The electroencephalogram signals are supervised and trained by using a deep learning network, so that the electroencephalogram signals containing P300 event related potential and electroencephalogram signals not containing P300 event related potential need to be distinguished respectively, and corresponding supervision signals y are manufactured according to the formula (2).
And B, constructing a deep learning network capable of classifying and identifying the P300 event-related potential.
The step B specifically comprises the following steps:
B1. the preprocessed electroencephalogram signals and corresponding supervisory signals are divided into a training set and a testing set, the training set comprises 85 characters, 25500 sample data are counted, and the size of each sample data is a matrix of 160 multiplied by 64 and is used for training network parameters. Where 160 represents the number of sample points per sample and 64 represents the number of signal channels. The test set contains 100 characters, and 18000 sample data in total are used for testing the network performance.
B2. The traditional convolutional neural network is improved, a network CNN _ L suitable for electroencephalogram analysis and processing is invented by combining an inclusion V1 network, and the network structure is shown in figure 2. The network has seven layers, the first layer is the input layer, and input data is loaded. The second layer is a convolution layer with convolution kernel size equal to the number of signal channels, which is equivalent to performing spatial filtering on the input data. In the third layer, the series connection convolutional layer is changed into three parallel connection convolutional layers with different convolutional kernel sizes, which is equivalent to time-domain filtering on different time scales, and the extraction capability of model features is improved. The fourth layer is a connection layer for integrating the feature maps obtained by the third layer. The fifth layer is a pooling layer for reducing network parameters and preventing over-fitting problems with small numbers of training samples. The sixth layer is a convolutional layer for extracting more abstract features. And the seventh layer is a full connection layer, and the probability that the current input electroencephalogram signal is judged to be the P300 event-related potential by the network is obtained after the characteristics obtained by the sixth layer are subjected to weighted calculation. The improved convolutional neural network in deep learning is used as a classification model, cross entropy is used as a cost function, parameters are updated by an Adam parameter optimization method, and the network is trained by using training skills such as dropout and regularization.
And step C, in one experiment, the probability that the input electroencephalogram signals contain the P300 event-related potential can be obtained in the step B, then the row number and the column number of the event-related potential induced in the visual stimulator can be determined according to the formula (3), and finally the predicted target character is determined. However, one experiment cannot accurately determine the position of the target character, and the output probability P (k, i) of the network in n experiments needs to be accumulated for each time, where k represents the experiment serial number and i is the corresponding row number or column number. And the column and the row with the maximum search probability are the positions (x, y) of the predicted target characters, and further the recognition accuracy of the network to the characters is calculated.
And comparing the character recognition accuracy with the original target character to obtain the character recognition accuracy under different experiment times. FIG. 3 shows the accuracy of character recognition for different experimental times for test samples using the method of the present invention. It can be seen from the figure that when the number of experiments is greater than 11, the character recognition rate of the method of the present invention exceeds 95%, and when the number of experiments is greater than 9, the character recognition accuracy of the method of the present invention exceeds 90%. Under the condition of reducing the number of experiments, the method of the invention also improves the accuracy of character recognition.
The above description is only for the preferred embodiment of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art should be considered to be within the technical scope of the present invention, and the technical solutions and the inventive concepts thereof according to the present invention should be equivalent or changed within the scope of the present invention.
Claims (2)
1. A P300 event-related potential classification and identification method based on deep learning is characterized by comprising the following steps:
A. preprocessing the EEG signal under the visual stimulation
A1. Acquiring an electroencephalogram signal under visual stimulation; filtering the EEG signal under visual stimulation by using a Butterworth filter, and filtering power frequency interference and electromyographic interference noise;
A2. performing data enhancement on the data processed in the step A1 by using a superposition average method to obtain more electroencephalogram signals containing event-related evoked potentials;
A3. carrying out zero mean value and unit variance normalization processing on the signal obtained by the A2;
A4. performing time domain truncation on the signal obtained from the A3, and manufacturing corresponding supervision signals for the P300 event-related potential and the non-P300 event-related potential;
B. deep learning network capable of classifying and identifying P300 event-related potentials is constructed
B1. Dividing the preprocessed electroencephalogram signals and corresponding monitoring signals into a training set and a testing set;
B2. the method comprises the steps of improving a traditional convolutional neural network in deep learning, changing a series connection convolutional layer of a third layer in the traditional convolutional neural network into 3 different convolutional layers connected in parallel, using the improved convolutional neural network as a classification decision device of P300 visual evoked potentials, and predicting the probability of P300 event-related potentials contained in electroencephalogram signals by using the network;
C. character classification and recognition, and determining character category by using network output probability and experimental paradigm
B, determining the probability that the input electroencephalogram signal contains the P300 event-related potential according to a formula (3), determining the row number and the column number of the event-related potential induced in the visual stimulator according to the formula (3), and finally determining a predicted target character;
p (k, i) is the output probability of the network in n experiments, wherein k represents the experiment serial number, and i is the corresponding row number or column number; the column and the row with the maximum search probability are the positions (x, y) of the predicted target characters;
the network CNN _ L suitable for electroencephalogram signal analysis and processing is invented in the B2 by combining an inclusion V1 network with a traditional convolutional neural network improvement method for deep learning, the network structure has seven layers, the first layer is an input layer, and input data are loaded; the second layer is a convolution layer with convolution kernel size equal to the number of signal channels, which is equivalent to performing spatial filtering on input data; in the third layer, the series connection convolution layer is changed into three convolution layers which are connected in parallel and have different convolution volumes and sizes, namely time-domain filtering is carried out on different time scales, and the extraction capability of model features is improved; the fourth layer is a connecting layer and is used for integrating the characteristic diagram obtained by the third layer; the fifth layer is a pooling layer used for reducing network parameters and preventing a small number of training samples from generating an overfitting problem; the sixth layer is a convolutional layer for extracting more abstract features; the seventh layer is a full connection layer, and the probability that the current input electroencephalogram signal is judged to be the P300 event-related potential by the network is obtained after the characteristics obtained by the sixth layer are subjected to weighted calculation; the improved convolutional neural network in deep learning is used as a classification model, cross entropy is used as a cost function, parameters are updated by an Adam parameter optimization method, and a training skill training network is regularized by applying dropout.
2. The deep learning-based P300 event-related potential classification and identification method according to claim 1, wherein the data enhancement method for electroencephalogram data in a2 specifically comprises: and performing primary superposition averaging on the event-related potentials stimulated at different moments, and simulating the detected P300 event-related potentials to balance different classes of data.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810794963.1A CN108960182B (en) | 2018-07-19 | 2018-07-19 | P300 event related potential classification identification method based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810794963.1A CN108960182B (en) | 2018-07-19 | 2018-07-19 | P300 event related potential classification identification method based on deep learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108960182A CN108960182A (en) | 2018-12-07 |
| CN108960182B true CN108960182B (en) | 2021-11-05 |
Family
ID=64481844
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810794963.1A Active CN108960182B (en) | 2018-07-19 | 2018-07-19 | P300 event related potential classification identification method based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108960182B (en) |
Families Citing this family (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109684296A (en) * | 2018-12-14 | 2019-04-26 | 深圳大学 | Brain line identification database construction method, device, computer equipment and storage medium |
| CN109784027B (en) * | 2019-01-11 | 2022-03-22 | 西安电子科技大学 | Identity verification method based on dual electroencephalogram signal characteristics |
| CN111488964A (en) * | 2019-01-29 | 2020-08-04 | 北京市商汤科技开发有限公司 | Image processing method and device and neural network training method and device |
| CN110163180A (en) * | 2019-05-29 | 2019-08-23 | 长春思帕德科技有限公司 | Mental imagery eeg data classification method and system |
| CN110916672A (en) * | 2019-11-15 | 2020-03-27 | 中南民族大学 | Old people daily activity monitoring method based on one-dimensional convolutional neural network |
| CN111012341B (en) * | 2020-01-08 | 2022-04-22 | 东南大学 | Artifact removal and electroencephalogram signal quality evaluation method based on wearable electroencephalogram equipment |
| CN111428601B (en) * | 2020-03-18 | 2023-08-18 | 五邑大学 | P300 signal identification method, device and storage medium based on MS-CNN |
| CN111950366B (en) * | 2020-07-08 | 2023-10-24 | 北京理工大学 | Convolutional neural network motor imagery electroencephalogram classification method based on data enhancement |
| CN112446289B (en) * | 2020-09-25 | 2023-08-22 | 华南理工大学 | Method for improving P300 spelling device performance |
| CN112381124B (en) * | 2020-10-30 | 2023-11-14 | 华南理工大学 | Method for improving brain-computer interface performance based on dynamic inverse learning network |
| CN112650386B (en) * | 2020-12-07 | 2023-05-16 | 华南师范大学 | Brain-computer interface character output system based on PCA-PCN |
| CN113017645B (en) * | 2021-02-25 | 2022-03-29 | 华南理工大学 | P300 signal detection method based on void convolutional neural network |
| CN112884063B (en) * | 2021-03-11 | 2023-07-14 | 广东工业大学 | P300 signal detection and identification method based on multi-element space-time convolution neural network |
| CN113408444B (en) * | 2021-06-24 | 2022-12-09 | 西安交通大学 | Event-related potential signal classification method based on CNN-SVM |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102200833A (en) * | 2011-05-13 | 2011-09-28 | 天津大学 | Speller brain-computer interface (SCI) system and control method thereof |
| CN102609090A (en) * | 2012-01-16 | 2012-07-25 | 中国人民解放军国防科学技术大学 | Electrocerebral time-frequency component dual positioning normal form quick character input method |
| CN105068644A (en) * | 2015-07-24 | 2015-11-18 | 山东大学 | Method for detecting P300 electroencephalogram based on convolutional neural network |
| CN107463250A (en) * | 2017-07-11 | 2017-12-12 | 天津大学 | The method for improving P300 spellings device using effect under Mental Workload state |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9058473B2 (en) * | 2007-08-29 | 2015-06-16 | International Business Machines Corporation | User authentication via evoked potential in electroencephalographic signals |
-
2018
- 2018-07-19 CN CN201810794963.1A patent/CN108960182B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102200833A (en) * | 2011-05-13 | 2011-09-28 | 天津大学 | Speller brain-computer interface (SCI) system and control method thereof |
| CN102609090A (en) * | 2012-01-16 | 2012-07-25 | 中国人民解放军国防科学技术大学 | Electrocerebral time-frequency component dual positioning normal form quick character input method |
| CN105068644A (en) * | 2015-07-24 | 2015-11-18 | 山东大学 | Method for detecting P300 electroencephalogram based on convolutional neural network |
| CN107463250A (en) * | 2017-07-11 | 2017-12-12 | 天津大学 | The method for improving P300 spellings device using effect under Mental Workload state |
Non-Patent Citations (4)
| Title |
|---|
| Convolutional Neural Networks for P300 Detection with Application to Brain-Computer Interfaces;Hubert Cecotti et al;《IEEE TRANSACTIONS ON PATTERN ANAL丫SIS AND MACHINE INTELLIGENCE》;20110331;第433-445页 * |
| EEGNet:A Compact Convolutional Network for EEG一based Brain-Computer Interfaces;Vernon J et al;《arxiv》;20161123;1-20页 * |
| 基于人工智能机器视觉技术的古建筑表层损伤检测;赵鹏等;《物联网技术》;20170930;第7卷(第9期);第14-16页 * |
| 基于卷积神经网络的脑电信号检测与脑机接口实现;董贤光;《中国优秀硕士学位论文全文数据库 医药卫生科技辑》;20170215;全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108960182A (en) | 2018-12-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108960182B (en) | P300 event related potential classification identification method based on deep learning | |
| WO2020151075A1 (en) | Cnn-lstm deep learning model-based driver fatigue identification method | |
| CN110070105B (en) | Electroencephalogram emotion recognition method and system based on meta-learning example rapid screening | |
| CN108959895B (en) | Electroencephalogram EEG (electroencephalogram) identity recognition method based on convolutional neural network | |
| CN108304917A (en) | A kind of P300 signal detecting methods based on LSTM networks | |
| CN105894039A (en) | Emotion recognition modeling method, emotion recognition method and apparatus, and intelligent device | |
| CN106108894A (en) | A kind of emotion electroencephalogramrecognition recognition method improving Emotion identification model time robustness | |
| CN112884063B (en) | P300 signal detection and identification method based on multi-element space-time convolution neural network | |
| Arnau-González et al. | ES1D: A deep network for EEG-based subject identification | |
| CN111714118A (en) | Brain cognition model fusion method based on ensemble learning | |
| CN113662560B (en) | Method for detecting seizure-like discharge between attacks, storage medium and device | |
| CN113017645B (en) | P300 signal detection method based on void convolutional neural network | |
| Zhang et al. | Deep Learning Models for EEG-based Rapid Serial Visual Presentation Event Classification. | |
| CN109325410B (en) | Electroencephalogram EEG (electroencephalogram) feature extraction method based on convolutional neural network | |
| CN109009098B (en) | Electroencephalogram signal feature identification method under motor imagery state | |
| AU2013100576A4 (en) | Human Identification with Electroencephalogram (EEG) for the Future Network Security | |
| CN110338760B (en) | Schizophrenia three-classification method based on electroencephalogram frequency domain data | |
| Hadiyoso et al. | Signal Dynamics Analysis for Epileptic Seizure Classification on EEG Signals. | |
| Dhivya et al. | A review on machine learning algorithm for EEG signal analysis | |
| CN110458066B (en) | Age group classification method based on resting electroencephalogram data | |
| CN111772629A (en) | Brain cognitive skill transplantation method | |
| CN108962379A (en) | A kind of mobile phone assisted detection system of cerebral nervous system disease | |
| CN113408444B (en) | Event-related potential signal classification method based on CNN-SVM | |
| CN105787459A (en) | ERP signal classification method based on optimal score sparse determination | |
| CN113143275B (en) | Electroencephalogram fatigue detection method for quantitative evaluation of sample and characteristic quality in combined manner |
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 |