1. INTRODUCTION
Brain-computer interface (BCI) realizes a direct communication between the human brain and the external environment by translating human intentions into control signals [1]. A BCI allows an individual with severe motor disabilities or aphasia to have effective control over devices such as computers, wheelchairs and music instruments. A BCI system detects the presence of specific patterns in a brain activity and translates these patterns into meaningful control commands. Recently, electroencephalogram (EEG)-based BCI is attracting much attention due to their noninvasiveness and high communication speed. The information transfer rate (ITR) is commonly used to evaluate the communication speed of a BCI. Current EEG-based BCIs fall into four main categories such as sensorimotor activities, P300, visual evoked potentials (VEP), and common spatial pattern (CSP) [1]. In particular, a VEP-based BCI has received increasing attention due to their advantages of little user training, ease of use, and a high ITR [1]–[3].