Research On P300 Brain Computer Interface System Based On Deep Learning

The BCI(Brian-computer interface)system is when people break away from the connection between the brain and muscle tissue,it can provide a non-neural conduction channel to help people establish a communication system to exchange information with the outside world.The most common EEG signal(Electroencephalogram)in the brain-computer interface is the P300 signal.P300 brain wave is an endogenous event-related potential,which can be captured as the subject responds to stimuli during the event decision-making process.The researchers can know the subject's decision by capturing the P300 brain electrical information result.P300 detection has many applications in the field of BCI.However,in actual applications,there are still problems such as poor system information transmission rate and low information transmission accuracy.In response to the above problems,this paper mainly studies the P300 brain-computer interface system based on deep learning network to do the following research work:In the pre-processing part of P300 EEG signals,this paper proposes to use the PC A(Principal Component Analysis)algorithm to reduce the dimension of P300 EEG signals.The principal component analysis algorithm performs linear transformation on the data in a statistically optimal way by extracting the correlation between the variables in the data.The principal component analysis algorithm calculates the feature vectors of different dimensions of the P300 EEG data,and assigns all the feature vectors to the weights.The weights correspond to the feature values of each feature vector.The researchers sort the feature vectors corresponding to the weights in descending order,and delete the feature vectors with smaller weights,so as to achieve the purpose of dimensionality reduction of multidimensional EEG data.From the comparison of experimental results,the principal component analysis algorithm not only ensures the high accuracy of P300 EEG signal recognition,but also improves the speed of character input.In the feature classification and recognition part of the P300 EEG signal,this paper proposes to use a new deep learning network algorithm:(1)Expand the single-core convolutional layer into a multi-core convolutional layer,deepen the depth of the convolutional neural network,and preprocess After the P300 EEG signal characteristics are optimized.The deep convolutional neural network algorithm is used to classify and recognize P300 EEG signals.This type of neural network is a multi-layer perceptron with a special topology and contains multiple hidden layers.The neural network is used for P300 character recognition,which allows automatic extraction of EEG features in each layer,and in addition to scaling and centering the input.vector,the original information can be retained as input without specific normalization.Compared with other classification algorithms,the deep convolutional neural network algorithm has many advantages in the classification of multi-dimensional EEG data(2)In view of the shortcomings of the traditional single classifier,this paper proposes an integrated classification algorithm to extract and classify the P300 EEG feature signal.The feature extraction capability of the integrated classification algorithm comes from multiple classifiers extracting signal features together.The parallel use of multiple classifiers improves the algorithm's computing power and more fully extracts the features of the P300 EEG signal.Through experimental verification,the multi-core convolutional neural network and system integrated classifier have improved the classification accuracy of the P300 brain-computer interface system,surpassing other classification algorithms that use the same kind of EEG data.In this paper,the above-mentioned EEG signal preprocessing,EEG signal feature classification and recognition and experimental comparison results are introduced and analyzed in detail.Finally,the development of the brain computer interface system and the next stage of work are prospected.