Session-to-Session Transfer in Detecting Steady-State Visual Evoked Potentials with Individual Training Data

The information transfer rate (ITR) of steady-state visual evoked potential (SSVEP)-based brain-computer interfaces (BCIs) has been significantly improved in the past few years. Recent studies have demonstrated the efficacy of advanced signal processing methods, which incorporate preliminarily recorded individual training data in SSVEP detection. However, conducting experiments for collecting training data from each individual is cumbersome because it is time-consuming and may cause visual fatigue. To simplify the training procedure, this study employs a session-to-session transfer method, which uses transfer templates obtained from datasets collected from the same subjects on a different day. The proposed approach was evaluated with a 40-class SSVEP dataset from eight subjects, each participated in two sessions on two different days. Study results showed that the proposed transfer method achieved significantly higher performance than conventional method based on canonical correlation analysis (CCA). In addition, by employing online adaptation, the proposed method reached high performance that is comparable with the most efficient approach in previous studies. These results indicate the feasibility of a high-performance SSVEP-based BCI with no or little training.