Optimization Strategy for SSVEP-Based BCI in Spelling Program Application

This paper describes an optimization strategy for steady state visual evoked potential (SSVEP)-based brain-computer interface (BCI), especially in a spelling program application. In this application, there are at least three components for implementing a complete BCI application: stimulator, signal processing, and application (spelling program). Ideally, those three components should run on different processing units in order to obtain optimum performance. But integrating those three components in one computer system also gives advantages: make it easier for the subject to concentrate and simplifies the system configuration. There are two main parts that need to be optimized: the flickering animation and the spelling system. When optimizing the flickering animation, we will focus on the display driver technology and programming aspects. The optimization of spelling system will be focused on the layout and representation of the letter matrix. We tested our program on several computers for the following parameters: frequency range, frequency resolution, and frequency stability. It can be concluded that using a computer monitor as the stimulator, the maximum synthesizable stimulator frequency is always half of its minimum refresh-rate, no matter which software technology is applied (DirectX or OpenGL). The maximum synthesizable frequency of up to 30 Hz with frequency resolution 0.11 Hz is achieved. We have tested our system on 106 subjects during CeBIT 2008 in Hannover, Germany. Mean accuracy for the spelling system is 92.5%. Therefore, the optimization strategy described here led to a stable and reliable system that performed effectively across most subjects without requiring extensive expert help or expensive hardware.