A Faster Single-Channel SSVEP-Based Speller Using Peak Filter Extended Canonical Correlation Analysis

Steady-state visual evoked potential (SSVEP) based brain-computer interface (BCI) can provide an effective speller for disabled people. With the attempts in building user-friendly BCIs, training-based systems using a single electroencephalogram (EEG) channel attract attention increasingly. One important step for training-based BCIs performance improvement is to construct clean templates. However, since band-pass filters can not precisely extract the fundamental and harmonic frequency components, there remains much noise in the templates extracted by existing methods. A novel peak filter extended canonical correlation analysis (PF-eCCA) was proposed in this work. Firstly, a peak filter strategy was developed to construct templates, which can emphasize the specific frequencies and suppress unrelated components. Then, eCCA was employed for classification. This method is evaluated on the Benchmark dataset and achieved the highest information transfer rate (ITR) of 138.7 bits/min, which significantly out-performed the state-of-the-art method. The proposed peak filter strategy can also be used to improve several other existing methods with low additional computation costs.