Abstract
ObjectiveFrequency recognition methods based on spatial filtering have been widely studied to enhance the classification performance of steady-state visual evoked potential (SSVEP)-based brain-computer interface (BCI) systems. This paper proposes a novel canonical correlation analysis (CCA)-based method to improve the recognition performance of SSVEPs. MethodsThe proposed method is based on a novel framework in which two spatial filters are trained using individual training data and SSVEP templates constructed from sine-cosine reference signals for SSVEP denoising and correlation coefficient generation, and then two correlation coefficients are integrated for target frequency recognition. The proposed method was compared with five frequency recognition algorithms requiring spatial filter training. Besides, the ensemble version of the proposed method was further developed and evaluated. Main resultsThe experimental results on two 40-class benchmark datasets recorded from 35 subjects and 70 subjects respectively show that the proposed method outperformed the five competing methods significantly in most cases. ConclusionThe proposed method achieved higher classification accuracy and information transmission rate, shorter training time, and more stable performance. Thus, the proposed CCA-based method is more suitable for SSVEP-based BCI systems.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call