A dynamic stopping algorithm for P300 based brain computer interface systems

Abstract

P300 potentials are involuntary responses which were elicited when a subject recognizes a target item among a group of irrelevant items. In order to determine the target item that the subjects intend to select, multiple ERP responses for each stimulus must be evaluated. The number of intensification sequences required to select a target item may vary among the subjects as well as sessions which a subject participated in. Therefore, instead of using a predetermined number of intensification sequences, it should be determined automatically at the moment of selection. This paper proposes a dynamic stopping algorithm to determine required number of intensifications sequences. The algorithm uses the optimal operating point of the ROC (Receiver Operating Characteristics) curve to determine the threshold values. The proposed algorithm was tested on two different datasets which use row/column (RC) paradigm and region based (RB) paradigm. Dynamic stopping algorithm significantly improved SPM (symbol per minute) on both datasets by reducing number of intensification sequences and ratio of erroneous selections. Because it does not require to select two of the selections at same number of intensification sequences, RB paradigm provides more flexible, rapid and accurate BCI systems.