ITSEF: Inception-based two-stage ensemble framework for P300 detection

To address the problems of low signal-to-noise ratio, significant individual differences between subjects, and class imbalance in P300-based brain-computer interface (BCI), this paper proposes a novel Inception-based two-stage ensemble framework (ITSEF) to improve detection accuracy. Firstly, an Inception-based convolutional neural network (ICNN) is designed to extract multi-scale features and conduct cross-channel learning. In addition, a two-stage ensemble framework (TSEF) combined with a pre-training and fine-tuning strategy is developed, aiming to enhance the classification performance of the minority class and improve the generalization ability of the model. The framework comprises a conventional learning branch and a re-balancing branch, each based on an ICNN pre-trained with a different loss function. The prediction results of both branches are dynamically weighted by a cumulative learning strategy, so that the model gradually shifts its learning focus from the majority class to the minority class, comprehensively improving the identification ability for both classes. Experimental results on two datasets, Dataset II of BCI Competition III and BCIAUT-P300, demonstrate that the proposed ITSEF achieves state-of-the-art performance in the P300 classification task, with average classification accuracies of 86.16 % and 92.13 %, respectively. Compared with the existing state-of-the-art methods, the ITSEF achieves improvements of 4.61 % and 1.01 % on the two datasets, respectively. Furthermore, it exhibits significant improvements compared to baseline models and widely used class re-balancing strategies. The proposed ITSEF method provides an innovative deep learning framework for P300 signal analysis and has application potential in the field of P300-BCI.