Multi-Layer Virtual Keyboard Interface Based on Character Frequency for Vietnamese BCI Speller System

Abstract

Abstract: The BCI (brain-computer interface) speller system comprises essential components,
including hardware for recording brain signals and software for processing these signals. EEG
(Electroencephalography) sensors record brain activity, which is then analyzed to classify into
specific commands for character selection by the software. The software’s GUI (Graphical User
Interface) facilitates user interaction by displaying the necessary symbols.
This article focuses on the design of a virtual keyboard interface aimed at efficiently selecting
Vietnamese characters for text composition in the BCI spleller system. Our contributions include a
multi-layered keyboard design, where keys are organized based on functional groups and the
frequency of character usage in Vietnamese. The multiple layered keyboard features keys that
represent individual Vietnamese character, groups of related characters, or functional keys for
specific editing tasks.
In our Vietnamese BCI speller system, EEG signals are captured using the EPOC Flex. During
EEG signal processing, we employ a Linear Discriminant Analysis (LDA) classifier with motor
imagery tasks involving right-hand and foot movements. Experimental results with our virtual
keyboard interface indicate that without suggestions, the typing rate achieved was 2.37
characters/min, which improved to 6.15 characters/min when suggestions were utilized,
demonstrating the effectiveness of our design.
Keywords: BCI Speller, Virtual keyboard, Vietnamese keyboard.