The Application of Transfer Learning in P300 Detection

Abstract

The main task of a Brain-Computer Interface (BCI) is to use the EEG signals to realize communication of the human brain and computers or other external devices. This paper mainly studies the classification of P300 signals with the method of transfer learning. Convolutional neural network (CNN) and support vector machine (SVM) are used as the basic learning algorithm for transfer learning. Fine-tuning (FT) with which we built the CNN-FT model is introduced for CNN. Tradaboost algorithm, a novel transfer learning framework, is applied in SVM method to construct the SVM-TL model. These two models are trained, tested, compared and analyzed on the Data set II of BCI Competition III. The results show that CNN-FT model is better than CNN model and SVM-TL model is better than SVM model.