基于卷积神经网络的P300电位检测及在脑机接口系统中的应用

P300 potential detection based on convolutional neural network and its application in brain-computer interface system

为提取更深层、更原始的脑电信号特征,提高基于P300电位的脑机接口系统的性能,提出将卷积神经网络(CNN)应用到脑机接口系统的P300电位检测.首先,根据脑电信号的时间和空间特征,构建CNN的网络结构.然后,对脑电信号进行预处理,采用卷积层和下采样层进行特征提取.最后,通过全连接层实现P300电位的检测.结果显示,卷积神经网络对P300电位具有很好的特征学习能力,取得了较好的分类结果,为进一步提高脑机接口系统的性能提供了有效手段.

In order to extract deeper,rawer features of EEG signals and improve the performance of braincomputer interface system based on P300 potential,the convolutional neural network（ CNN） was used to detect P300 potential in a brain-computer interface system. First,the network structure of CNN was constructed according to the temporal and spatial feature of EEG signals. Then,the EEG signals were preprocessed,and feature was extracted using convolutional layers and downsampling layers,and the detection of the P300 was realized through a fully connected layer. The results showed that the CNN had a good ability for feature learning of P300,and had achieved better classification results,which provided an effective means for further improving the performance of brain-computer interface systems.