Abstract
Brain-Computer Interfaces (BCIs) allows a user to communicate or control external devices without using muscles. The P300 speller, which uses P300 as the feature quantity, is one of the most popular BCIs. P300 is different with each person and depends on the physical condition slightly. Thus, users are required to do machine learning every time before use. However, this pre-training is a burden to users. This paper proposes an incremental learning using unknown data to reduce the training time. Consequently, this paper shows that the proposed method gives not only the reduction of the training time but also directly use of P300 speller without machine learning by using the data of last time.
