On P300 Signal Recognition Algorithms Based on Convolutional Neural Network

As we know, the P300 signal has the characteristics of low signal-to-noise ratio, strong randomness and great individual difference, it is hard to identify in a common way. A novel method based on convolutional neural network (CNN) was presented in this paper. Firstly, the original signal was preprocessed. Then, a suitable CNN structure was designed to extract the feature, according to the temporal and spatial characteristics of the P300 signal. Finally, the Softmax classifier was used to classify the P300 signals. In the construction of CNN model, the traditional CNN structure was improved and the dropout in the network training process was introduced to prevent the model from over-fitting due to the lack of training samples. The presented algorithm was tested on the dataset II of the BCI competition III and achieved 96.69% average accuracy. The result shows that this method exhibits significant performance improvement in terms of accuracy and efficiency. Furthermore, the method provides a new and effective way to improve the performance of brain-computer interface system based on P300 signal.