Abstract
In this paper we present a method of shared control strategy which combine electroencephalography (EEG) signals and heuristic algorithm, to guide a virtual dot on the two dimensional maze to reach the goal. One of the most main problems of EEG-based brain computer interfaces (BCIs) is the low classify accuracy. It is hard to build a generalize classifier to identify all the people’s EEG signals, in order to using people’s intention as the control factor with the low information rate, recent works have explored shared-control strategies which the system does not only execute the decoded commands from signals’ owner, but also involved in executing the task have been set up beforehand. That is, the system’s execution result can more close to people’s intension. Our shared-control system use error-related potentials (ErrP) as feedback which only be detected when the subjects feel wrong or confused. ErrP can be evoked steady in 0.3~0.6 milliseconds after the stimulation happened, with the subjects’ assessment of the target moving in the maze, transfer them as feedback into our heuristic function, we can guide the target to reach the goal without knowing the goal’s location efficiently. Shared-control strategies in BCI systems such as we presented here may prove to be the foundation for complex BCIs capable of doing more than we ever imagined.