Hierarchical character selection for a brain computer interface spelling system

Abstract

Typing using brain computer interface provides a physically disabled person an important mode of communication with their external environment. The goal of this study is to develop a real-time brain computer interface (BCI) spelling interface using a single channel electroencephalograph (EEG) and steady state evoked potential (SSVEP). The study employs a hierarchical method to include all the alphabets and several special characters on a graphical user interface (GUI) that includes 5 flickering boxes in every level with frequencies set from 6 Hz to10 Hz. Subsequently, the real-time EEG signal was acquired from a single EEG channel specifically the Oz channel. Signals are processed and a command output is triggered to enter a sub-level or select the desired character that will be displayed on a Word file. The system was tested by typing a pangram with a total of 31 letters and 5 SPACE commands and results exhibited an average accuracy of 92.80%. Overall, the average information transfer rate (ITR) and characters per minute (CPM) with transitional delays were calculated to be 11.24 bits/min and 2.04 chars/min respectively. However, when the transitional delays were excluded from the computation, results showed an average ITR of 21.20 bits/min and a CPM of 3.76 characters per minute. It is noteworthy that one of the subjects was able to achieve a peak ITR of 54.73 bits/min and a maximum of 8.42 chars/minute when transitional delays are excluded. The current method was able to demonstrate fast response and practical application of the BCI spelling system using hierarchical method.