BCI adaptation using incremental-SVM learning

Abstract

Brain-computer interface (BCI) systems allow the user to interact with a computer by merely thinking. Successful BCI operation depends on the continuous adaptation of the system to the user. This paper presents an implementation of this adaptation using incremental support vector machines (SVM). This approach is tested on three subjects and three types of mental activities across ten sessions. The results show that the continuous adaptation of the BCI to the user's brain activity brings clear advantages over a non-adapting approach.