Method and system for constructing CNN-GB model, data feature classification method
A technology of model and EEG data, applied in the field of building CNN-GB model, to overcome insufficient training and improve accuracy
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Method used
Image
Examples
Embodiment 1
[0068] Attached figure 1 As shown, the method for constructing a CNN-GB model of the present invention includes the following steps:
[0069] S100. Collect EEG data and preprocess the EEG data;
[0070] S200. Build a CNN network model based on the Caffe deep learning framework. The CNN network model includes but is not limited to convolutional layer, pooling layer, fully connected layer and loss layer;
[0071] S300. Select a part of the EEG data as the training set, train the CNN network model through the training set and update the network parameters of the CNN network model to obtain the trained CNN network model, and output the characteristics of the EEG data;
[0072] S400: Train the GB network model based on the characteristics of the EEG data and update the network parameters of the GB network model to obtain the trained GB network model;
[0073] The trained CNN network model is used to extract features of EEG data, and the trained GB network model is used to classify the extrac...
Embodiment 2
[0120] The present invention provides a system for constructing a CNN-GB model, including:
[0121] EEG data acquisition module, used to collect and store EEG data, and preprocess EEG data;
[0122] The CNN network configuration module is used to construct the CNN network model, and to update the network parameters of the CNN network model through error back propagation and stochastic gradient descent to obtain the trained CNN network model;
[0123] The GB network configuration module is used to construct a GB network model, and is used to train the GB network model and update the network parameters of the GB network model through the characteristics of the EEG data to obtain the trained GB network model.
[0124] Among them, the EEG data acquisition module is a module with the following functions:
[0125] Based on Dataset I in the international BCI Competition III database, the EEG signal acquisition equipment is used to sample data according to the preset frequency, and the collecte...
Embodiment 3
[0140] The present invention provides a data feature extraction and classification method, including the following steps:
[0141] L100. Collect EEG data and preprocess the EEG data. The EEG data is multi-channel data, and the EEG data has a training set and a test set;
[0142] L200, taking the training set as input, constructing a CNN-GB model according to the method of constructing a CNN-GB model disclosed in Example 1, to obtain a trained CNN network model and a trained GB network model;
[0143] L300, taking the test set as input, extracting features of EEG data through the trained CNN network model, and classifying the extracted features through the trained GB network model.
[0144] As a further improvement of this embodiment, it also includes evaluating the classification of the obtained test set, including the following steps:
[0145] The label of the test set is predicted by the trained GB network model, and the predicted test set and label are compared with the label of the ...
PUM
Login to View More Abstract
Description
Claims
Application Information
Login to View More