CN108875580A - A kind of multiclass Mental imagery EEG signal identification method based on because imitating network - Google Patents
A kind of multiclass Mental imagery EEG signal identification method based on because imitating network Download PDFInfo
- Publication number
- CN108875580A CN108875580A CN201810463020.0A CN201810463020A CN108875580A CN 108875580 A CN108875580 A CN 108875580A CN 201810463020 A CN201810463020 A CN 201810463020A CN 108875580 A CN108875580 A CN 108875580A
- Authority
- CN
- China
- Prior art keywords
- motor imagery
- network
- cause
- causal
- matrix
- 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
- 238000000034 method Methods 0.000 title claims abstract description 19
- 230000003340 mental effect Effects 0.000 title 1
- 239000011159 matrix material Substances 0.000 claims abstract description 52
- 230000001364 causal effect Effects 0.000 claims abstract description 40
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 19
- 238000012706 support-vector machine Methods 0.000 claims abstract description 10
- 238000007781 pre-processing Methods 0.000 claims abstract description 4
- 238000005259 measurement Methods 0.000 claims description 18
- 238000012549 training Methods 0.000 claims description 7
- 230000009467 reduction Effects 0.000 claims description 6
- 239000013598 vector Substances 0.000 claims description 5
- 238000012417 linear regression Methods 0.000 claims description 4
- 238000012360 testing method Methods 0.000 claims description 3
- 238000011946 reduction process Methods 0.000 claims 1
- 230000008569 process Effects 0.000 description 4
- 206010033799 Paralysis Diseases 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 230000003993 interaction Effects 0.000 description 2
- 230000009471 action Effects 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 210000003710 cerebral cortex Anatomy 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000007659 motor function Effects 0.000 description 1
- 210000003205 muscle Anatomy 0.000 description 1
- 238000011160 research Methods 0.000 description 1
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
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- 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
-
- 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)
- Data Mining & Analysis (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Signal Processing (AREA)
- Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)
Abstract
本发明公开了一种基于因效网络的多类运动想象脑电信号识别方法；具体为1、使用脑电采集设备采集驾驶脑电信号；2、对采集到的脑电信号进行预处理，包括降频、降噪；3、构建运动想象因效网络；4、给运动想象因效网路矩阵进行编码；5、使用线性支持向量机对运动想象因效网络的特征进行分类；本发明只需使用运动想象因效网络矩阵训练一次卷积神经网络，和使用8个特征值训练一次线性支持向量机分类器，我们就可以进行在线识别多类运动想象脑电信号。
The invention discloses a multi-type motor imagery EEG signal recognition method based on a causal network; specifically, 1. Using an EEG acquisition device to collect driving EEG signals; 2. Preprocessing the collected EEG signals, including 3. Construct a motor imagery cause-effect network; 4. Encode the motor imagery cause-effect network matrix; 5. Use a linear support vector machine to classify the characteristics of the motor imagery cause-effect network; the present invention only needs Using the motor imagery causal network matrix to train a convolutional neural network, and using 8 eigenvalues to train a linear support vector machine classifier, we can recognize multiple types of motor imagery EEG signals online.
Description
技术领域technical field
本发明属于生物电信号识别领域，涉及多类运动想象脑电信号识别方法，具体是一种基于因效网络的多类运动想象脑电信号识别方法。The invention belongs to the field of bioelectrical signal recognition, and relates to a multi-type motor imagery EEG signal recognition method, in particular to a multi-type motor imagery EEG signal recognition method based on a cause-effect network.
背景技术Background technique
运动想象是一种特殊的运动功能状态，可表述为人们就执行某个动作所进行的内心排演，由于并不引发肢体、肌肉的活动，它与实际运动存在着明显的不同。运动想象作为近年来最具有临床应用前景的康复治疗新方法之一，引起了越来越多研究者的关注，已有研究表明，它能够与实际运动一样有效激活运动相关的大脑皮层区域。Motor imagery is a special state of motor function, which can be expressed as an inner rehearsal for people to perform a certain action. Because it does not trigger the activities of limbs and muscles, it is obviously different from the actual movement. As one of the most promising new rehabilitation methods in recent years, motor imagery has attracted more and more researchers' attention. Studies have shown that it can activate motor-related cerebral cortex regions as effectively as actual exercise.
随着人工智能技术的不断发展和进步，康复机器人在国际上已经逐步成为临床康复治疗的重要技术手段之一。结合运动想象的康复机器人可以辅助脑损伤肢体瘫痪患者进行康复训练，这一技术给脑损伤肢体瘫痪患者带来了福音。但是目前在线识别患者的运动想象任务以及多类运动想象脑电信号的识别已经成为康复机器人设计中亟待解决的问题，需要进一步研究和方法上的拓展。With the continuous development and progress of artificial intelligence technology, rehabilitation robots have gradually become one of the important technical means of clinical rehabilitation in the world. Rehabilitation robots combined with motor imagination can assist patients with brain-injured limb paralysis to perform rehabilitation training. This technology has brought good news to patients with brain-injured limb paralysis. However, at present, the online identification of patients' motor imagery tasks and the identification of multiple types of motor imagery EEG signals have become urgent problems in the design of rehabilitation robots, which require further research and methodological expansion.
发明内容Contents of the invention
针对以上问题，提出一种基于因效网络矩阵的多类运动想象脑电信号识别方法。该方法具有离线训练、在线识别的特点。Aiming at the above problems, a multi-type motor imagery EEG signal recognition method based on the cause-effect network matrix is proposed. This method has the characteristics of offline training and online recognition.
按照本发明提供的技术方案，提出了一种基于因效网络的多类运动想象脑电信号识别方法，包括如下5个步骤：According to the technical solution provided by the present invention, a method for recognizing multi-type motor imagery EEG signals based on a causal network is proposed, including the following five steps:
步骤1、使用脑电采集设备采集64个导联的多类运动想象脑电信号；Step 1. Use EEG acquisition equipment to collect 64 leads of multi-type motor imagery EEG signals;
步骤2、对采集到的脑电信号进行预处理，包括降频、降噪。Step 2. Perform preprocessing on the collected EEG signals, including frequency reduction and noise reduction.
提取每个信道中两个频带，分别是α(8-15HZ)、β(16-30HZ)。因为在运动想象期间这两个频带起着重要的作用。Extract two frequency bands in each channel, namely α(8-15HZ) and β(16-30HZ). Because these two frequency bands play an important role during motor imagery.
步骤3、构建运动想象因效网络；Step 3, construct motor imagery cause-effect network;
使用多变量格兰杰因果关系构建运动想象因效网络。具体步骤为：Construction of a motor imagery causal network using multivariate Granger causality. The specific steps are:
3-1.确定运动想象因效网络节点；3-1. Determine the motor imagery causal network nodes;
把每个EEG导联对应的电极覆盖的区域定义为一个节点。The area covered by the electrodes corresponding to each EEG lead is defined as a node.
3-2.使用多变量格兰杰因果关系度量节点之间的因果关系；3-2. Use multivariate Granger causality to measure the causal relationship between nodes;
使用多变量格兰杰因果关系度量各节点之间的因果关系，多变量格兰杰因果关系(MVGC)是基于格兰杰因果关系的概念，然后推广到相互关联的X、Y集合和条件多变量情况下Z的相互作用。Use multivariate Granger causality to measure the causal relationship between nodes. Multivariate Granger causality (MVGC) is based on the concept of Granger causality, and then extended to interrelated X, Y sets and conditional multivariate Interaction of Z in variable case.
首先定义表示向量的垂直级联，X-表示X的滞后变量，Y-表示Y的滞后变量，Z-表示Z的滞后变量，∑(X)表示X的n×n的协方差矩阵，∑(X，Y)表示X，Y的协方差矩阵，tr[]表示矩阵的迹。first define Represents the vertical concatenation of vectors, X - represents the lagged variable of X, Y - represents the lagged variable of Y, Z - represents the lagged variable of Z, ∑(X) represents the n×n covariance matrix of X, ∑(X, Y) represents the covariance matrix of X and Y, and tr[] represents the trace of the matrix.
在运动想象任务期间节点被预测节点X在另一个节点Y上的多元线性回归：Multiple linear regression of nodes being predicted node X on another node Y during a motor imagery task:
X＝A·Y+εX＝A·Y+ε
其中A是n×m的矩阵，包含着回归系数，随机向量ε包含残差。该模型的系数通过残差和回归预测因子Y之间施加零相关来唯一确定。由Yule-Walkers公式我们可以得到：Where A is an n×m matrix containing regression coefficients, and the random vector ε contains residuals. The coefficients of this model are uniquely determined by imposing a zero correlation between the residuals and the regression predictor Y. From the Yule-Walkers formula we can get:
A＝∑(X，Y)∑(Y)-1 A=∑(X,Y)∑(Y) -1
残差的协方差矩阵：The covariance matrix of the residuals:
∑(ε)＝∑(X|Y)＝∑(X)-∑(X，Y)∑(Y)-1∑(X，Y)T ∑(ε)=∑(X|Y)=∑(X)-∑(X, Y)∑(Y) -1 ∑(X, Y) T
假设现在我们有三个节点Xt，Yt，Zt。然后为了测量给定Z从Y到X的格兰杰因果关系，我们想要比较以下两个多变量自回归MVAR模型：Suppose now we have three nodes X t , Y t , Z t . Then to measure the Granger causality from Y to X given Z, we want to compare the following two multivariate autoregressive MVAR models:
因此，被预测节点X首先在先前的自身延迟上加上条件变量Z的滞后，然后再加上预测变量Y的滞后。根据标准的格兰杰因果关系度量，即由回归残差方差比率给出。那么根据我们的公式可以得出：Therefore, the predicted node X first adds the lag of the condition variable Z to the previous own lag, and then adds the lag of the predictor variable Y. According to the standard measure of Granger causality, i.e. given by the ratio of variances of the regression residuals. Then according to our formula we can get:
使用格兰杰因果关系的标准度量仅限于预测变量和预测变量Y和X进行定义，因为第一个回归残差方差总是大于或等于第二个人回归残差方差，所以上式总是大于等于0。我们现在考虑预测变量和预测变量不在被约束为单变量情况下，即多变量G因果关系，对于多变量的预测情况，G因果关系还没有一个标准的定义，一种可能性是简单的使用多变量的均方误差，即多变量残差的总方差或预期多元残差的平方长度。因此我们可以得出：Using standard measures of Granger causality is limited to predictors and predictors Y and X for definition, since the first regression residual variance is always greater than or equal to the second individual regression residual variance, so the above formula is always greater than or equal to 0. We now consider the case where predictors and predictors are not constrained to be univariate, that is, multivariate G causality. For multivariate prediction, there is no standard definition of G causality. One possibility is to simply use multiple The mean square error of the variable, which is the total variance of the multivariate residuals or the squared length of the expected multivariate residuals. Therefore we can derive:
上式就可以计算出多变量格兰杰因果关系。The above formula can calculate the multivariate Granger causality.
3-3.根据因果关系构建因效网络；3-3. Construct a causal-effect network according to the causal relationship;
利用多变量格兰杰因果关系量化节点之间的关系，经过显著性检验，确定各节点之间的因果流，即多变量格兰杰因果关系的大小为节点之间边的大小，多变量格兰杰因果关系的方向为节点间边的方向。The multivariate Granger causality is used to quantify the relationship between nodes, and after the significance test, the causal flow between nodes is determined, that is, the size of the multivariate Granger causality is the size of the edge between nodes, and the multivariate lattice The direction of the Ranger causality is the direction of the edges between the nodes.
步骤4、给运动想象因效网路矩阵进行编码；Step 4, encoding the motor imagery cause-effect network matrix;
因效网络矩阵就是因效网络图的矩阵表示。我们使用卷积神经网络给因效网络矩阵进行编码，通过训练卷积神经网络让其生成8个因效网络的测量值。虽然训练过程比较复杂，但是训练只需一次就好，训练一旦完成，我们就可以通过这个卷积神经网络来处理我们的因效网络矩阵，让其为我们的因效网络矩阵生成8个测量值，这8个测量值就是因效网络的特征。具体步骤如下：The cause-effect network matrix is the matrix representation of the cause-effect network diagram. We use a convolutional neural network to encode the causal network matrix, and train the convolutional neural network to generate 8 measurements of the causal network. Although the training process is more complicated, it only needs to be trained once. Once the training is completed, we can process our cause-effect network matrix through this convolutional neural network and let it generate 8 measurements for our cause-effect network matrix. , these eight measurements are the characteristics of the cause-effect network. Specific steps are as follows:
4-1.标注因效网络矩阵；4-1. Mark the cause-effect network matrix;
4-2.用已经标注好的因效网络矩阵训练卷积神经网络；4-2. Use the marked cause-effect network matrix to train the convolutional neural network;
训练卷积神经网络，确保相同运动想象任务下生成的测量值之间的欧式距离小于阈值T，不同运动想象任务下生成的测量值之间的欧式距离大于阈值T；Train the convolutional neural network to ensure that the Euclidean distance between measurements generated under the same motor imagery task is less than a threshold T, and that the Euclidean distance between measurements generated under different motor imagery tasks is greater than a threshold T;
4-3让卷积神经网络分别为每个运动想象任务因效网络矩阵生成8个测量值。4-3 Let the convolutional neural network generate 8 measurements separately for each motor imagery task cause-effect network matrix.
步骤5、使用线性支持向量机对运动想象因效网络的特征进行分类；Step 5, using a linear support vector machine to classify the features of the motor imagery causal network;
最后这一步，就是在找到数据库中，与我们运动想象因效网络矩阵的测量值比较接近的那个运动想象因效网络矩阵。我们需要做的是，训练一个线性支持向量机分类器，它可以从新的因效网络矩阵中获取测量结果，并找到最匹配的那个因效网络矩阵，分类器的结果就是最匹配的运动想象任务。The last step is to find the motor imagery cause-effect network matrix in the database that is relatively close to the measured value of our motor imagery cause-effect network matrix. What we need to do is to train a linear support vector machine classifier, which can take the measurement results from the new causal network matrix and find the best matching causal network matrix. The result of the classifier is the best matching motor imagery task .
本发明有益效果如下：The beneficial effects of the present invention are as follows:
只需使用运动想象因效网络矩阵训练一次卷积神经网络，和使用8个特征值训练一次线性支持向量机分类器，我们就可以进行在线识别多类运动想象脑电信号。Just use the motor imagery causal network matrix to train a convolutional neural network, and use 8 eigenvalues to train a linear support vector machine classifier, and we can recognize multiple types of motor imagery EEG signals online.
附图说明Description of drawings
图1多类运动想象任务识别流程图；Figure 1 Flowchart of multi-category motor imagery task recognition;
具体实施方式：Detailed ways:
下面结合具体实施例对本发明作进一步说明。以下描述仅作为示范和解释，并不对本发明作任何形式上的限制。The present invention will be further described below in conjunction with specific examples. The following description is only for demonstration and explanation, and does not limit the present invention in any form.
如图1所示，一种基于因效网络的多类运动想象脑电信号识别方法，该方法具体包括以下步骤：As shown in Figure 1, a method for recognizing multi-type motor imagery EEG signals based on a causal network, the method specifically includes the following steps:
步骤1、使用脑电采集设备采集64个导联的多类运动想象脑电信号；Step 1. Use EEG acquisition equipment to collect 64 leads of multi-type motor imagery EEG signals;
步骤2、对采集到的脑电信号进行预处理，包括降频、降噪。Step 2. Perform preprocessing on the collected EEG signals, including frequency reduction and noise reduction.
步骤3、构建运动想象因效网络；Step 3, construct motor imagery cause-effect network;
步骤4、给运动想象因效网路矩阵进行编码；Step 4, encoding the motor imagery cause-effect network matrix;
步骤5、使用线性支持向量机对运动想象因效网络的特征进行分类；Step 5, using a linear support vector machine to classify the features of the motor imagery causal network;
所述的步骤2中，提取每个信道中两个频带，分别是α(8-15HZ)、β(16-30HZ)。因为在运动想象期间这两个频带起着重要的作用。In the step 2, two frequency bands in each channel are extracted, namely α (8-15HZ) and β (16-30HZ). Because these two frequency bands play an important role during motor imagery.
所述的步骤3中，使用多变量格兰杰因果关系构建运动想象因效网络。具体步骤为：In Step 3, the motor imagery causality network was constructed using multivariate Granger causality. The specific steps are:
3-1.确定运动想象因效网络节点；3-1. Determine the motor imagery causal network nodes;
把每个EEG导联对应的电极覆盖的区域定义为一个节点。The area covered by the electrodes corresponding to each EEG lead is defined as a node.
3-2.使用多变量格兰杰因果关系度量节点之间的因果关系；3-2. Use multivariate Granger causality to measure the causal relationship between nodes;
使用多变量格兰杰因果关系度量各节点之间的因果关系，多变量格兰杰因果关系(MVGC)是基于格兰杰因果关系的概念，然后推广到相互关联的X、Y集合和条件多变量情况下Z的相互作用。Use multivariate Granger causality to measure the causal relationship between nodes. Multivariate Granger causality (MVGC) is based on the concept of Granger causality, and then extended to interrelated X, Y sets and conditional multivariate Interaction of Z in variable case.
首先定义表示向量的垂直级联，X-表示X的滞后变量，Y-表示Y的滞后变量，Z-表示Z的滞后变量，∑(X)表示X的nxn的协方差矩阵，∑(X，Y)表示X，Y的协方差矩阵，tr[]表示矩阵的迹。first define Represents the vertical concatenation of vectors, X - represents the lagged variable of X, Y - represents the lagged variable of Y, Z - represents the lagged variable of Z, ∑(X) represents the nxn covariance matrix of X, ∑(X, Y) Represents the covariance matrix of X, Y, and tr[] represents the trace of the matrix.
在运动想象任务期间被预测节点X在另一个节点Y上的多元线性回归：Multiple linear regression of predicted node X on another node Y during a motor imagery task:
X＝A·Y+εX＝A·Y+ε
其中A是n×m的矩阵，包含着回归系数，随机向量ε包含残差。该模型的系数通过残差和回归预测因子Y之间施加零相关来唯一确定。由Yule-Walkers公式我们可以得到：Where A is an n×m matrix containing regression coefficients, and the random vector ε contains residuals. The coefficients of this model are uniquely determined by imposing a zero correlation between the residuals and the regression predictor Y. From the Yule-Walkers formula we can get:
A＝∑(X，Y)∑(Y)-1 A=∑(X,Y)∑(Y) -1
残差的防方差矩阵：Antivariance matrix for the residuals:
∑(ε)＝∑(X|Y)＝∑(X)-∑(X，Y)∑(Y)-1∑(X，Y)T ∑(ε)=∑(X|Y)=∑(X)-∑(X, Y)∑(Y) -1 ∑(X, Y) T
假设现在我们有三个节点Xt，Yt，Zt。然后为了测量给定Z从Y到X的格兰杰因果关系，我们想要比较以下两个多变量自回归MVAR模型：Suppose now we have three nodes X t , Y t , Z t . Then to measure the Granger causality from Y to X given Z, we want to compare the following two multivariate autoregressive MVAR models:
因此，被预测节点X首先在先前的自身延迟上加上条件变量Z的滞后，然后再加上预测变量Y的滞后。根据标准的格兰杰因果关系度量，即由回归残差方差比率给出。那么根据我们的公式可以得出：Therefore, the predicted node X first adds the lag of the condition variable Z to the previous own lag, and then adds the lag of the predictor variable Y. According to the standard measure of Granger causality, i.e. given by the ratio of variances of the regression residuals. Then according to our formula we can get:
使用格兰杰因果关系的标准度量仅限于预测变量和预测变量Y和X进行定义，因为第一个回归残差方差总是大于或等于第二个人回归残差方差，所以上式总是大于等于0。我们现在考虑预测变量和预测变量不在被约束为单变量情况下，即多变量G因果关系，对于多变量的预测情况，G因果关系还没有一个标准的定义，一种可能性是简单的使用多变量的均方误差，即多变量残差的总方差或预期多元残差的平方长度。因此我们可以得出：Using standard measures of Granger causality is limited to predictors and predictors Y and X for definition, since the first regression residual variance is always greater than or equal to the second individual regression residual variance, so the above formula is always greater than or equal to 0. We now consider the case where predictors and predictors are not constrained to be univariate, that is, multivariate G causality. For multivariate prediction, there is no standard definition of G causality. One possibility is to simply use multiple The mean square error of the variable, which is the total variance of the multivariate residuals or the squared length of the expected multivariate residuals. Therefore we can derive:
上式就可以计算出多变量格兰杰因果关系。The above formula can calculate the multivariate Granger causality.
3-3.根据因果关系构建因效网络；3-3. Construct a causal-effect network according to the causal relationship;
利用多变量格兰杰因果关系量化节点之间的关系，经过显著性检验，确定各节点之间的因果流，即多变量格兰杰因果关系的大小为节点之间边的大小，多变量格兰杰因果关系的方向为节点间边的方向。The multivariate Granger causality is used to quantify the relationship between nodes, and after the significance test, the causal flow between nodes is determined, that is, the size of the multivariate Granger causality is the size of the edge between nodes, and the multivariate lattice The direction of the Ranger causality is the direction of the edges between the nodes.
所述的步骤4中，给运动想象因效网路矩阵进行编码；In the described step 4, encode the motor imagery cause-effect network matrix;
因效网络矩阵就是因效网络图的矩阵表示。我们使用卷积神经网络给因效网络矩阵进行编码，通过训练卷积神经网络让其生成8个因效网络的测量值。虽然训练过程比较复杂，但是训练只需一次就好，训练一旦完成，我们就可以通过这个卷积神经网络来处理我们的因效网络矩阵，让其为我们的因效网络矩阵生成8个测量值，这8个测量值就是因效网络的特征。具体步骤如下：The cause-effect network matrix is the matrix representation of the cause-effect network diagram. We use a convolutional neural network to encode the causal network matrix, and train the convolutional neural network to generate 8 measurements of the causal network. Although the training process is more complicated, it only needs to be trained once. Once the training is completed, we can process our cause-effect network matrix through this convolutional neural network and let it generate 8 measurements for our cause-effect network matrix. , these eight measurements are the characteristics of the cause-effect network. Specific steps are as follows:
4-1.标注因效网络矩阵；4-1. Mark the cause-effect network matrix;
4-2.用已经标注好的因效网络矩阵训练卷积神经网络；4-2. Use the marked cause-effect network matrix to train the convolutional neural network;
训练卷积神经网络，确保相同运动想象任务下生成的测量值之间的欧式距离小于阈值T，不同运动想象任务下生成的测量值之间的欧式距离大于阈值T；Train the convolutional neural network to ensure that the Euclidean distance between measurements generated under the same motor imagery task is less than a threshold T, and that the Euclidean distance between measurements generated under different motor imagery tasks is greater than a threshold T;
4-3让卷积神经网络分别为每个运动想象任务因效网络矩阵生成8个测量值。4-3 Let the convolutional neural network generate 8 measurements separately for each motor imagery task cause-effect network matrix.
步骤5、使用线性支持向量机对运动想象因效网络的特征进行分类；Step 5, using a linear support vector machine to classify the features of the motor imagery causal network;
最后这一步，就是在找到数据库中，与我们运动想象因效网络矩阵的测量值比较接近的那个运动想象因效网络矩阵。我们需要做的是，训练一个分类器，它可以从新的因效网络矩阵中获取测量结果，并找到最匹配的那个因效网络矩阵，分类器的结果就是最匹配的运动想象任务。The last step is to find the motor imagery cause-effect network matrix in the database that is relatively close to the measured value of our motor imagery cause-effect network matrix. What we need to do is to train a classifier that can take the measurements from the new causal network matrix and find the one that best matches the causal network matrix, and the result of the classifier is the best matched motor imagery task.
Claims (4)
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810463020.0A CN108875580A (en) | 2018-05-15 | 2018-05-15 | A kind of multiclass Mental imagery EEG signal identification method based on because imitating network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810463020.0A CN108875580A (en) | 2018-05-15 | 2018-05-15 | A kind of multiclass Mental imagery EEG signal identification method based on because imitating network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108875580A true CN108875580A (en) | 2018-11-23 |
Family
ID=64333920
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810463020.0A Pending CN108875580A (en) | 2018-05-15 | 2018-05-15 | A kind of multiclass Mental imagery EEG signal identification method based on because imitating network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108875580A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109711383A (en) * | 2019-01-07 | 2019-05-03 | 重庆邮电大学 | Time-frequency domain-based convolutional neural network motor imagery EEG signal recognition method |
| CN110969108A (en) * | 2019-11-25 | 2020-04-07 | 杭州电子科技大学 | Limb action recognition method based on autonomic motor imagery electroencephalogram |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102521505A (en) * | 2011-12-08 | 2012-06-27 | 杭州电子科技大学 | Brain electric and eye electric signal decision fusion method for identifying control intention |
| CN104305993A (en) * | 2014-11-12 | 2015-01-28 | 中国医学科学院生物医学工程研究所 | Electroencephalogram source localization method based on granger causality |
| CN104473636A (en) * | 2014-12-30 | 2015-04-01 | 天津大学 | Brain fatigue network analysis method based on partial orientation coherence |
| CN106691378A (en) * | 2016-12-16 | 2017-05-24 | 深圳市唯特视科技有限公司 | Deep learning vision classifying method based on electroencephalogram data |
-
2018
- 2018-05-15 CN CN201810463020.0A patent/CN108875580A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102521505A (en) * | 2011-12-08 | 2012-06-27 | 杭州电子科技大学 | Brain electric and eye electric signal decision fusion method for identifying control intention |
| CN104305993A (en) * | 2014-11-12 | 2015-01-28 | 中国医学科学院生物医学工程研究所 | Electroencephalogram source localization method based on granger causality |
| CN104473636A (en) * | 2014-12-30 | 2015-04-01 | 天津大学 | Brain fatigue network analysis method based on partial orientation coherence |
| CN106691378A (en) * | 2016-12-16 | 2017-05-24 | 深圳市唯特视科技有限公司 | Deep learning vision classifying method based on electroencephalogram data |
Non-Patent Citations (5)
| Title |
|---|
| ADAM B. BARRETT ET AL.: "Multivariate Granger Causality and Generalized Variance", 《ARXIV》 * |
| 佘青山 等: "基于感兴趣脑区 LASSO-Granger因果关系的脑电特征提取算法", 《电子与信息学报》 * |
| 曾庆山 等: "基于 CSP 与卷积神经网络算法的多类运动想象脑电信号分类", 《科学技术与工程》 * |
| 王卫星 等: "基于卷积神经网络的脑电信号上肢运动意图识别", 《浙江大学学报》 * |
| 黄登凤: "基于PDC的注意相关脑电因效网络分析", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109711383A (en) * | 2019-01-07 | 2019-05-03 | 重庆邮电大学 | Time-frequency domain-based convolutional neural network motor imagery EEG signal recognition method |
| CN109711383B (en) * | 2019-01-07 | 2023-03-31 | 重庆邮电大学 | Convolutional neural network motor imagery electroencephalogram signal identification method based on time-frequency domain |
| CN110969108A (en) * | 2019-11-25 | 2020-04-07 | 杭州电子科技大学 | Limb action recognition method based on autonomic motor imagery electroencephalogram |
| CN110969108B (en) * | 2019-11-25 | 2023-04-07 | 杭州电子科技大学 | Limb action recognition method based on autonomic motor imagery electroencephalogram |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110537922B (en) | A method and system for lower limb motion recognition during human walking based on deep learning | |
| Cheng et al. | A rapid spiking neural network approach with an application on hand gesture recognition | |
| Yu et al. | Hand medical monitoring system based on machine learning and optimal EMG feature set | |
| Shi et al. | Feature extraction and classification of lower limb motion based on sEMG signals | |
| CN104107042A (en) | Electromyographic signal gait recognition method based on particle swarm optimization and support vector machine | |
| CN105426842B (en) | Multiclass hand motion recognition method based on support vector machines and surface electromyogram signal | |
| CN113128552A (en) | Electroencephalogram emotion recognition method based on depth separable causal graph convolution network | |
| CN102930284B (en) | Surface electromyogram signal pattern recognition method based on empirical mode decomposition and fractal | |
| CN104537382A (en) | Electromyographic signal gait recognition method for optimizing support vector machine based on genetic algorithm | |
| Shao et al. | Single-channel SEMG using wavelet deep belief networks for upper limb motion recognition | |
| CN101587546A (en) | Identification method of lower limb action pattern based on supporting vector multivariate classification | |
| Zhao et al. | Deep CNN model based on serial-parallel structure optimization for four-class motor imagery EEG classification | |
| CN111178288B (en) | Human body posture recognition method and device based on local error layer-by-layer training | |
| CN101732110A (en) | Method for fusing brain electricity and muscle electricity signal chaos characteristics for hand motion identification | |
| Gao et al. | Human gait recognition based on multiple feature combination and parameter optimization algorithms | |
| CN101859377A (en) | EMG signal classification method based on multi-core support vector machine | |
| CN107618018B (en) | Manipulator action speed proportional control method based on myoelectricity | |
| Oweis et al. | ANN-based EMG classification for myoelectric control | |
| Chada et al. | An efficient approach for physical actions classification using surface EMG signals | |
| CN107808166B (en) | Electromyographic feature extraction method for linear Laplace discrimination of MEMD tensor | |
| CN112036357B (en) | Upper limb action recognition method and system based on surface electromyographic signals | |
| CN108875580A (en) | A kind of multiclass Mental imagery EEG signal identification method based on because imitating network | |
| Zhang et al. | Deep learning analysis based on multi-sensor fusion data for hemiplegia rehabilitation training system for stoke patients | |
| CN115281676B (en) | Fatigue detection method based on GRU neural network and ECG signal | |
| CN116502066A (en) | Exoskeleton swing period prediction system and method based on BP neural network |
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: 20181123 |