CN110059565A - A kind of P300 EEG signal identification method based on improvement convolutional neural networks - Google Patents
A kind of P300 EEG signal identification method based on improvement convolutional neural networks Download PDFInfo
- Publication number
- CN110059565A CN110059565A CN201910212733.4A CN201910212733A CN110059565A CN 110059565 A CN110059565 A CN 110059565A CN 201910212733 A CN201910212733 A CN 201910212733A CN 110059565 A CN110059565 A CN 110059565A
- Authority
- CN
- China
- Prior art keywords
- layer
- neural networks
- convolutional neural
- signal
- eeg
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- 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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/08—Feature extraction
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of based on the P300 EEG signal identification method for improving convolutional neural networks.The present invention comprises the steps of: 1, acquires P300 EEG signals using brain wave acquisition equipment；2, the EEG signals for choosing 16 channels carry out frequency reducing, noise reduction and resampling to collected EEG signals；3, five original dimension samples are reconstructed into two-dimensional matrix, carry out 15 superposed averages to increase signal-to-noise ratio, each P300 sample label is set as 1, and noise sample label is set as 0；4, new convolutional neural networks structure is constructed, pretreated data are sent into convolutional neural networks, determine network parameter, obtain the improved convolutional neural networks model for P300 EEG's Recognition by 5, training network；The present invention can effectively improve the discrimination of P300 signal using convolutional neural networks are improved to the progress feature extraction of P300 EEG signals and classification.
Description
Technical field
It is specifically a kind of based on the P300 brain telecommunications for improving convolutional neural networks the present invention relates to EEG signal identification method
Number recognition methods.
Background technique
Convolutional neural networks are a kind of deep learning algorithm models, and classical LeNet-5 model is as shown in Figure 1, exist
The fields such as figure identification, speech processes achieve satisfactory effect.Compared to traditional machine learning algorithm, convolutional Neural
Network can be automatically performed the extraction of feature and be abstracted in training without manual extraction feature, and carry out pattern classification simultaneously.
There is invariance to distortion such as scaling, translation, rotations with the model of convolutional neural networks training, there is very strong generalization.It is solely
Special weight is shared and local receptor field thought, and the parameter amount of neural network can be greatly decreased, while preventing over-fitting again
Reduce the complexity of neural network model.
As a kind of feedforward depth network, convolutional neural networks are not necessarily to too many data handling procedure.That is, collecting
Data only need to simply carry out pretreatment can be sent into network, a preferable model is obtained by the study of network.It is close
Year, as a kind of efficient deep neural network, convolutional neural networks already become the focus of all circles researcher and obtain
Deep discussion, and excellent result is shown in numerous research fields.In March, 2011 Cecotti et al. a paper
Error, which is used, for example-convolutional neural networks that we open deep learning combination electroencephalogramsignal signal analyzing returns algorithm work
For its training algorithm, using local receptor field and weight be shared etc., concepts combine specific space-time in EEG signals well
Information characteristics learn the inherent mode in data out；And in convolutional neural networks structure down-sampling operation to input when
Insensitivity is moved, even if occurring situations such as a small amount of offset or acquisition delay occur for electrode, convolutional Neural in signal acquisition process
Network can have an amendment well.
Being closely connected between convolutional neural networks are middle-level makes it particularly suitable for the processing of image with spatial information
And understanding, but since EEG signals are a kind of signals for combining time and space characteristics, in order not to make the spy after convolution algorithm
Miscellaneous space and temporal information, the convolution kernel in convolutional layer need pointedly to be set as vector rather than general pattern simultaneously in sign
Matrix in identification makes it only extract space characteristics or temporal characteristics.
Summary of the invention
The purpose of the present invention is being operated on the basis of classical convolutional neural networks structure in conjunction with dropout, one is proposed
Kind is based on the P300 EEG signal identification method for improving convolutional neural networks.This method can effectively to P300 EEG signals into
Row identification, highest recognition accuracy can reach 96.69%, and convolutional Neural can be effectively prevented in the dropout operation being added
Network over-fitting.
Technical solution provided by the invention: a kind of P300 EEG signal identification method based on improvement convolutional neural networks,
Include the following steps:
Step 1 acquires P300 EEG signals using brain wave acquisition equipment；
Step 2 chooses number respectively 32,34,36,41,9,11,13,42,47,49,51,53,55,56,60,62
The EEG signals in 16 channels pre-process collected EEG signals, including frequency reducing, noise reduction and resampling etc.；
Five original dimension samples are reconstructed into common two-dimensional matrix by step 3, are carried out 15 superposed averages and are increased noise
Than each P300 sample label being set as 1, noise sample label is set as 0；
The new convolutional neural networks structure of step 4, building, chooses an input layer, a convolutional layer, a down-sampling
Layer, a full articulamentum and an output layer, the size of convolution kernel are 1X16, and the size of down-sampling layer is 2X1, step-length 2；
Pretreated data are sent into convolutional neural networks, rule of thumb and many experiments are true by step 5, training network
Determine network parameter, obtains the improved convolutional neural networks model for P300 EEG's Recognition；
In the step 2, original signal is subjected to z-score standardization first, then uses the Butterworth of six ranks
Filter carries out the bandpass filtering of 0.5~30Hz to signal.Finally, data after choosing stimulation every time in 625ms, that is, 150 first
Sampled point carries out down-sampling, decimation factor 3.
In the step 5, training process is similar with traditional back-propagation algorithm, is broadly divided into two stages: first is that
The propagated forward stage；Second is that back-propagation phase.Specific training process is as follows:
In the propagated forward stage, input data successively passes through convolutional layer, down-sampling layer and full articulamentum, calculates separately each section
Spot net output.
Convolution algorithm is carried out using following formula in convolutional layer:
Wherein,Indicate j-th of characteristic pattern in l layers, MjIndicate the feature set of graphs of input,What l layers of expression used
Convolution kernel,Indicate the biasing of l layers of setting, f indicates activation primitive, and used herein is ReLU function, operational formula are as follows:
F (x)=max (0, x) (2)
Down-sampling layer uses mean value pond method, operational formula are as follows:
Wherein,Indicate l layers of weight parameter, down indicates down-sampling operation function.
Full articulamentum uses Softmax function to classify, and formula is as follows:
P indicates the probability that sample x is classified as to jth class,Product representation jth dimensional vector value, denominator indicate
The summation of all vector values.
It is calculated between output valve and actual value after obtaining classification output valve by forward calculation in back-propagation phase
Error.Then, error is subjected to backpropagation along output layer, optimizes network weight parameter.Use formula (5) as loss letter
Number:
Wherein, N indicates input sample number, and C indicates sample class number,Indicate n-th of sample for the kth of label
Dimension,Indicate k-th of the output that n-th of sample exports network.
Two processes of propagated forward and backpropagation are constantly repeated in CNN network, until the loss function of network reaches
It is minimum.In the training process, make the smallest weight parameter of loss function using gradient descent method searching.Calculation formula are as follows:
Wherein, a indicates that the learning rate in training process, λ indicate weight attenuation coefficient, wlAnd blRespectively indicate weight and partially
It sets.
The present invention has the beneficial effect that:
The shared unique thought with local receptor field of convolutional neural networks weight is used for reference, P300 EEG signals are known
Not, the results showed that by improved convolutional neural networks be used to detect P300 signal be it is feasible, effectively raise EEG signals
Discrimination, be the successful experiment that deep learning model is applied in brain machine interface system, at the same also be EEG signals feature
It extracts and classification provides new thinking.
Detailed description of the invention
The LeNet-5 structure chart of Fig. 1 classics；
The improved convolutional neural networks structure chart of Fig. 2；
Fig. 3 experimental result picture
Specific embodiment:
The present invention is further explained in the light of specific embodiments.It is described below only as demonstration and explanation, not
It is intended that the present invention is limited in any way.
As shown in Fig. 2, the present invention is as follows the step of realization:
Step 1 acquires P300 EEG signals using brain wave acquisition equipment.Wherein collected EEG signals be originated from 2 by
The international 10-20 lead system electrode cap that examination person wears, port number are 64 channels, sample frequency 240Hz；
Step 2 chooses number respectively 32,34,36,41,9,11,13,42,47,49,51,53,55,56,60,62
The EEG signals in 16 channels pre-process collected EEG signals, including frequency reducing, noise reduction and resampling.Pretreatment
Signal sampling frequencies afterwards are reduced to 80Hz, and primary frequency range is 0.5Hz~30Hz, and the sampled point of each sample becomes from 240
50；
Step 3, the five-tensor that pretreated EEG signals are 12x15x50x16x85, are reconstructed into 15300
The bivector sample of 50X16, and the sample label wherein comprising P300 signal is set as 1, the sample not comprising P300 signal
Label is set as 0；
Step 4, new convolutional neural networks structure are constituted by 5 layers.It is described in detail below:
First layer is input layer, and input data is to pass through pretreated P300 signal, and the dimension of each input sample is
50x16, wherein 50 be the sampling number in each channel, 16 be the port number of selection；The second layer is convolutional layer, right in this layer
Input sample carries out space filtering and convolution algorithm.Known by 4.1 section parameter selections, the convolution kernel number of this layer of setting is 20, greatly
Small is 1x16.Neuron is activated using ReLU function, has obtained 20 Feature Mapping figures, size 50x1；Third layer is adopted under being
Sample layer, main function are to reduce characteristic dimension.Use step-length for 2 mean value pond method, each neuron and phase in convolutional layer
It answers the region 2x1 of characteristic pattern to be connected, obtains the characteristic pattern that 20 sizes are 25x1；4th layer is full articulamentum, which introduces
Dropout operation, in the training process with 0.6 probability it is random turn off partial nerve member, to prevent CNN model excessively quasi-
It closes.The characteristic signal of each neuron node is converted into vector signal by full articulamentum, is differentiated using softmax classifier and is inputted
The classification of sample；Layer 5 is output layer, because the purpose of this paper is identification, whether input sample is P300 signal, finally
There are two types of classification results to export: ' 1 ' indicates P300 signal, and ' 0 ' indicates noise signal.
Step 5, training process are similar with traditional back-propagation algorithm, are broadly divided into two stages: first is that propagated forward
Stage；Second is that back-propagation phase.Specific training process is as follows:
In the propagated forward stage, input data successively passes through convolutional layer, down-sampling layer and full articulamentum, calculates separately each section
Spot net output.
Convolution algorithm is carried out using following formula in convolutional layer:
Wherein,Indicate j-th of characteristic pattern in l layers, MjIndicate the feature set of graphs of input,What l layers of expression used
Convolution kernel,Indicate the biasing of l layers of setting, f indicates activation primitive, and used herein is ReLU function, operational formula are as follows:
F (x)=max (0, x) (2)
Down-sampling layer uses mean value pond method, operational formula are as follows:
Wherein,Indicate l layers of weight parameter, down indicates down-sampling operation function.
Full articulamentum uses softmax function to classify, and formula is as follows:
P indicates the probability that sample x is classified as to jth class.
It is calculated between output valve and actual value after obtaining classification output valve by forward calculation in back-propagation phase
Error.Then, error is subjected to backpropagation along output layer, optimizes network weight parameter.Use formula (5) as loss letter
Number:
Wherein, N indicates input sample number, and C indicates sample class number,Indicate n-th of sample for the kth of label
Dimension,Indicate k-th of the output that n-th of sample exports network.
Two processes of propagated forward and backpropagation are constantly repeated in CNN network, until the loss function of network reaches
It is minimum.In the training process, make the smallest weight parameter of loss function using gradient descent method searching.Calculation formula are as follows:
Wherein, a indicates the learning rate in training process.
The data source of experiment P300 spelling device experiment, two subjects in third time world BCI contest in 2004 are equal
The experiment for carrying out 85 characters, is respectively adopted sample that preceding 55 characters of every subject generate as training set, and latter 30
The sample that character generates is as test set.P300 EEG's Recognition effect of the improved convolutional neural networks to two subjects
Fig. 3, it is shown that superiority of the improved CNN model in P300 signal identification, also illustrate different subjects it
Between P300 signal be implicitly present in individual difference.Further, it is also possible to find out that improved CNN model believes P300 in single experiment
Number classification achieve good effect.
Claims (2)
1. a kind of based on the P300 EEG signal identification method for improving convolutional neural networks, it is characterised in that include the following steps:
Step 1 acquires P300 EEG signals using brain wave acquisition equipment；
Step 2 chooses 16 that number is respectively 32,34,36,41,9,11,13,42,47,49,51,53,55,56,60,62
The EEG signals in channel pre-process the EEG signals in 16 channels, including frequency reducing, noise reduction and resampling；
Original five dimensions samples are reconstructed into common two-dimensional matrix by step 3, carry out 15 superposed averages to increase signal-to-noise ratio,
Each P300 sample label is set as 1, noise sample label is set as 0；
The new convolutional neural networks structure of step 4, building, chooses an input layer, a convolutional layer, down-sampling layer, one
A full articulamentum and an output layer, the size of convolution kernel are 1X16, and the size of down-sampling layer is 2X1, step-length 2；
It is described in detail below:
First layer is input layer, and input data is to pass through pretreated P300 signal, and the dimension of each input sample is
50x16, wherein 50 be the sampling number in each channel, 16 be the port number of selection；
The second layer is convolutional layer, carries out space filtering and convolution algorithm to input sample in this layer；The convolution kernel of this layer of setting
Number is 20, size 1x16；Neuron is activated using ReLU function, has obtained 20 Feature Mapping figures, size 50x1；
Third layer is down-sampling layer, use step-length for 2 mean value pond method, individual features figure in each neuron and convolutional layer
The region 2x1 be connected, obtain 20 sizes be 25x1 characteristic pattern；
4th layer is full articulamentum, which introduces dropout operation, is turned off in the training process with 0.6 probability is random
Partial nerve member；The characteristic signal of each neuron node is converted into vector signal by full articulamentum, uses softmax classifier
Differentiate the classification of input sample；
Layer 5 is output layer, and there are two types of classification results to export: ' 1 ' indicates P300 signal, and ' 0 ' indicates noise signal；
Pretreated data are sent into convolutional neural networks, determine network parameter, obtain improved use by step 5, training network
In the convolutional neural networks model of P300 EEG's Recognition.
2. as described in claim 1 a kind of based on the P300 EEG signal identification method for improving convolutional neural networks, feature
It is, is pre-processed in step 2 specifically: original signal is subjected to z-score standardization first, it is then fertile using the Bart of six ranks
This filter carries out the bandpass filtering of 0.5~30Hz to signal；Finally, choosing the data after stimulation every time in 625ms is preceding 150
A sampled point carries out down-sampling, decimation factor 3.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910212733.4A CN110059565A (en) | 2019-03-20 | 2019-03-20 | A kind of P300 EEG signal identification method based on improvement convolutional neural networks |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910212733.4A CN110059565A (en) | 2019-03-20 | 2019-03-20 | A kind of P300 EEG signal identification method based on improvement convolutional neural networks |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110059565A true CN110059565A (en) | 2019-07-26 |
Family
ID=67317401
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910212733.4A Pending CN110059565A (en) | 2019-03-20 | 2019-03-20 | A kind of P300 EEG signal identification method based on improvement convolutional neural networks |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110059565A (en) |
Cited By (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110399857A (en) * | 2019-08-01 | 2019-11-01 | 西安邮电大学 | A kind of brain electricity emotion identification method based on figure convolutional neural networks |
| CN110781751A (en) * | 2019-09-27 | 2020-02-11 | 杭州电子科技大学 | Emotional electroencephalogram signal classification method based on cross-connection convolutional neural network |
| CN110811558A (en) * | 2019-11-18 | 2020-02-21 | 郑州大学 | Sleep arousal analysis method based on deep learning |
| CN111428648A (en) * | 2020-03-26 | 2020-07-17 | 五邑大学 | Electroencephalogram signal generation network, method and storage medium |
| CN111428601A (en) * | 2020-03-18 | 2020-07-17 | 五邑大学 | Method, device and storage medium for identifying P300 signal based on MS-CNN |
| CN111461204A (en) * | 2020-03-30 | 2020-07-28 | 华南理工大学 | Emotion identification method based on electroencephalogram signals and used for game evaluation |
| CN111882036A (en) * | 2020-07-22 | 2020-11-03 | 广州大学 | Convolutional neural network training method, electroencephalogram signal identification method, device and medium |
| CN112241724A (en) * | 2020-10-30 | 2021-01-19 | 南京信息工程大学滨江学院 | Automatic identification method and system based on double-path convolution long-term and short-term neural network |
| CN112381008A (en) * | 2020-11-17 | 2021-02-19 | 天津大学 | Electroencephalogram emotion recognition method based on parallel sequence channel mapping network |
| CN112650386A (en) * | 2020-12-07 | 2021-04-13 | 华南师范大学 | PCA-PCN-based brain-computer interface character output system |
| CN112884063A (en) * | 2021-03-11 | 2021-06-01 | 广东工业大学 | P300 signal detection and identification method based on multivariate space-time convolution neural network |
| CN113180692A (en) * | 2021-02-11 | 2021-07-30 | 北京工业大学 | Electroencephalogram signal classification and identification method based on feature fusion and attention mechanism |
| CN113343869A (en) * | 2021-06-17 | 2021-09-03 | 中国科学院精密测量科学与技术创新研究院 | Electroencephalogram signal automatic classification and identification method based on NTFT and CNN |
| CN113598794A (en) * | 2021-08-12 | 2021-11-05 | 中南民族大学 | Training method and system for detection model of ice drug addict |
| CN113655884A (en) * | 2021-08-17 | 2021-11-16 | 河北师范大学 | Equipment control method, terminal and system |
| CN114997315A (en) * | 2022-06-13 | 2022-09-02 | 西安交通大学 | Error correlation potential classification method based on multichannel electroencephalogram integration |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20040199482A1 (en) * | 2002-04-15 | 2004-10-07 | Wilson Scott B. | Systems and methods for automatic and incremental learning of patient states from biomedical signals |
| CN105068644A (en) * | 2015-07-24 | 2015-11-18 | 山东大学 | Method for detecting P300 electroencephalogram based on convolutional neural network |
| CN108009635A (en) * | 2017-12-25 | 2018-05-08 | 大连理工大学 | A kind of depth convolutional calculation model for supporting incremental update |
| CN108959895A (en) * | 2018-08-16 | 2018-12-07 | 广东工业大学 | A kind of EEG signals EEG personal identification method based on convolutional neural networks |
| CN109271964A (en) * | 2018-10-11 | 2019-01-25 | 刘仕琪 | Emotion identification method and system based on deep learning model Yu length memory network |
| CN109325410A (en) * | 2018-08-16 | 2019-02-12 | 广东工业大学 | A kind of EEG signals EEG feature extracting method based on convolutional neural networks |
-
2019
- 2019-03-20 CN CN201910212733.4A patent/CN110059565A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20040199482A1 (en) * | 2002-04-15 | 2004-10-07 | Wilson Scott B. | Systems and methods for automatic and incremental learning of patient states from biomedical signals |
| CN105068644A (en) * | 2015-07-24 | 2015-11-18 | 山东大学 | Method for detecting P300 electroencephalogram based on convolutional neural network |
| CN108009635A (en) * | 2017-12-25 | 2018-05-08 | 大连理工大学 | A kind of depth convolutional calculation model for supporting incremental update |
| CN108959895A (en) * | 2018-08-16 | 2018-12-07 | 广东工业大学 | A kind of EEG signals EEG personal identification method based on convolutional neural networks |
| CN109325410A (en) * | 2018-08-16 | 2019-02-12 | 广东工业大学 | A kind of EEG signals EEG feature extracting method based on convolutional neural networks |
| CN109271964A (en) * | 2018-10-11 | 2019-01-25 | 刘仕琪 | Emotion identification method and system based on deep learning model Yu length memory network |
Non-Patent Citations (5)
| Title |
|---|
| YANG LI等: "Boosted Convolutional Neural Networks for Motor Imagery EEG Decoding with Multiwavelet-based Time-Frequency Conditional Granger Causality Analysis", 《ARXIV》 * |
| 曾向阳著: "《智能水中目标识别》", 31 March 2016, 国防工业出版社 * |
| 柳杨著: "《数字图像物体识别理论详解与实战》", 31 March 2018, 北京邮电大学出版社 * |
| 王建责任编辑: "《基于图像分析的植物及其病虫害识别方法研究》", 31 October 2018, 中国经济出版社 * |
| 许庆勇著: "《基于深度学习理论的纹身图像识别与检测研究》", 31 December 2018, 华中科技大学出版社 * |
Cited By (24)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110399857A (en) * | 2019-08-01 | 2019-11-01 | 西安邮电大学 | A kind of brain electricity emotion identification method based on figure convolutional neural networks |
| CN110781751A (en) * | 2019-09-27 | 2020-02-11 | 杭州电子科技大学 | Emotional electroencephalogram signal classification method based on cross-connection convolutional neural network |
| CN110811558A (en) * | 2019-11-18 | 2020-02-21 | 郑州大学 | Sleep arousal analysis method based on deep learning |
| CN111428601B (en) * | 2020-03-18 | 2023-08-18 | 五邑大学 | P300 signal identification method, device and storage medium based on MS-CNN |
| CN111428601A (en) * | 2020-03-18 | 2020-07-17 | 五邑大学 | Method, device and storage medium for identifying P300 signal based on MS-CNN |
| CN111428648A (en) * | 2020-03-26 | 2020-07-17 | 五邑大学 | Electroencephalogram signal generation network, method and storage medium |
| CN111428648B (en) * | 2020-03-26 | 2023-03-28 | 五邑大学 | Electroencephalogram signal generation network, method and storage medium |
| CN111461204A (en) * | 2020-03-30 | 2020-07-28 | 华南理工大学 | Emotion identification method based on electroencephalogram signals and used for game evaluation |
| CN111461204B (en) * | 2020-03-30 | 2023-05-26 | 华南理工大学 | Emotion recognition method based on electroencephalogram signals for game evaluation |
| CN111882036A (en) * | 2020-07-22 | 2020-11-03 | 广州大学 | Convolutional neural network training method, electroencephalogram signal identification method, device and medium |
| CN111882036B (en) * | 2020-07-22 | 2023-10-31 | 广州大学 | Convolutional neural network training method, electroencephalogram signal identification method, device and medium |
| CN112241724B (en) * | 2020-10-30 | 2023-12-15 | 南京信息工程大学滨江学院 | Automatic identification method and system based on double-path convolution long-term neural network |
| CN112241724A (en) * | 2020-10-30 | 2021-01-19 | 南京信息工程大学滨江学院 | Automatic identification method and system based on double-path convolution long-term and short-term neural network |
| CN112381008A (en) * | 2020-11-17 | 2021-02-19 | 天津大学 | Electroencephalogram emotion recognition method based on parallel sequence channel mapping network |
| CN112381008B (en) * | 2020-11-17 | 2022-04-29 | 天津大学 | Electroencephalogram emotion recognition method based on parallel sequence channel mapping 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 |
| CN112650386A (en) * | 2020-12-07 | 2021-04-13 | 华南师范大学 | PCA-PCN-based brain-computer interface character output system |
| CN113180692A (en) * | 2021-02-11 | 2021-07-30 | 北京工业大学 | Electroencephalogram signal classification and identification method based on feature fusion and attention mechanism |
| 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 |
| CN112884063A (en) * | 2021-03-11 | 2021-06-01 | 广东工业大学 | P300 signal detection and identification method based on multivariate space-time convolution neural network |
| CN113343869A (en) * | 2021-06-17 | 2021-09-03 | 中国科学院精密测量科学与技术创新研究院 | Electroencephalogram signal automatic classification and identification method based on NTFT and CNN |
| CN113598794A (en) * | 2021-08-12 | 2021-11-05 | 中南民族大学 | Training method and system for detection model of ice drug addict |
| CN113655884A (en) * | 2021-08-17 | 2021-11-16 | 河北师范大学 | Equipment control method, terminal and system |
| CN114997315A (en) * | 2022-06-13 | 2022-09-02 | 西安交通大学 | Error correlation potential classification method based on multichannel electroencephalogram integration |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110059565A (en) | A kind of P300 EEG signal identification method based on improvement convolutional neural networks | |
| CN110353702A (en) | A kind of emotion identification method and system based on shallow-layer convolutional neural networks | |
| CN110069958B (en) | Electroencephalogram signal rapid identification method of dense deep convolutional neural network | |
| CN106909784A (en) | Epileptic electroencephalogram (eeg) recognition methods based on two-dimentional time-frequency image depth convolutional neural networks | |
| CN109598222B (en) | EEMD data enhancement-based wavelet neural network motor imagery electroencephalogram classification method | |
| CN112381008B (en) | Electroencephalogram emotion recognition method based on parallel sequence channel mapping network | |
| CN105068644A (en) | Method for detecting P300 electroencephalogram based on convolutional neural network | |
| CN113627401A (en) | Myoelectric gesture recognition method of feature pyramid network fused with double-attention machine system | |
| CN108567418A (en) | A kind of pulse signal inferior health detection method and detecting system based on PCANet | |
| CN113705398B (en) | Music electroencephalogram space-time characteristic classification method based on convolution-long and short term memory network | |
| CN108256579A (en) | A kind of multi-modal sense of national identity quantization measuring method based on priori | |
| Lu et al. | Combined CNN and LSTM for motor imagery classification | |
| CN115381466A (en) | Motor imagery electroencephalogram signal classification method based on AE and Transformer | |
| CN113076878A (en) | Physique identification method based on attention mechanism convolution network structure | |
| CN113180659A (en) | Electroencephalogram emotion recognition system based on three-dimensional features and cavity full convolution network | |
| CN117520891A (en) | Motor imagery electroencephalogram signal classification method and system | |
| CN113128353B (en) | Emotion perception method and system oriented to natural man-machine interaction | |
| Kuang et al. | LSTM model with self-attention mechanism for EEG based cross-subject fatigue detection | |
| CN112163486A (en) | Electroencephalogram channel optimization method based on sparse learning and domain confrontation network | |
| Yu et al. | Encoding physiological signals as images for affective state recognition using convolutional neural networks | |
| CN113887374B (en) | Brain control water drinking system based on dynamic convergence differential neural network | |
| CN114469137B (en) | Cross-domain electroencephalogram emotion recognition method and system based on space-time feature fusion model | |
| CN112381124B (en) | Method for improving brain-computer interface performance based on dynamic inverse learning network | |
| CN114492560A (en) | Electroencephalogram emotion classification method based on transfer learning | |
| CN112668424B (en) | RBSAGAN-based data augmentation 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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190726 |