Recent Advances in the Design of a 3-State Self-Paced (Asynchronous) Brain Computer Interface

Unlike synchronous brain computer interfaces (BCI), self-paced (asynchronous) BCIs have the advantage of being operational at all times. A 3-state self-paced BCI is capable of detecting two different brain states (e.g. two movements) from the ongoing EEG, while a 2-state one can only detect one brain state. This study improves the performance of a 3-state self-paced BCI designed to detect right and left hand extension movements. Instead of using the values of features at each instant of time, the improved BCI uses all past features' values to detect the presence of a movement at any specific time. After detecting the presence of a movement, the system uses spectral features to determine whether the detected movement is a right or a left hand extension. Using data from two able-bodied individuals, it is shown that the correct detection of a right or a left hand movement, on average, increases from 44.3% to 55.9%, for a fixed false positive rate of 1%. In differentiating between right and left hand movements the average performance increases from 64% to 68.5%. At the false positive rate of 0.5%, the average true positive rate increases from 20.2% to 27.6% and the differentiation rate between right and left hand extensions increase from 71% to is 72.5%.