A Novel SSVEP-based Word Speller Based on Sliding Multi-Window Strategy

Abstract

Speller has been the strong illustration of steady-state visual evoked potential (SSVEP) brain-computer interfaces (BCIs) for being an efficient communication interface. However, the main limitation of the SSVEP speller is the low communication rates caused by extra input time for gaze shifting and waiting. In this study, a novel spelling method using the word as an input unit is proposed, which saves the waiting time between two consecutive characters. SSVEP features are extracted from continuous EEG signals containing different frequencies using filter bank canonical correlation analysis (FBCCA) with sliding multi-window strategy. Both offline and online experiments have been conducted with ten participants by a 40-character speller. The average accuracy and information transfer rate (ITR) of the proposed word speller reached $90.72\pm 7.27\%$ and $220.40\pm 22.68$ bits/min respectively at a 1.2 s time window, higher than the traditional character-based speller with $87.52\pm 6.19\%$ accuracy and $155.60 \pm 16.01 \text{bits}/\text{min}$ ITR. The improvements demonstrated the efficiency of the sliding multi-window strategy for an SSVEP-based word speller.