Two-Stage Frequency Recognition Method Based on Correlated Component Analysis for SSVEP-Based BCI

Abstract

A canonical correlation analysis (CCA) is a state-of-the-art method for frequency recognition in steady-state visual evoked potential (SSVEP)-based brain-computer interface (BCI) systems. Various extended methods have been developed, and among such methods, a combination method of CCA and individual-template-based CCA has achieved the best performance. However, the CCA requires the canonical vectors to be orthogonal, which may not be a reasonable assumption for the EEG analysis. In this paper, we propose using the correlated component analysis (CORRCA) rather than CCA to implement frequency recognition. CORRCA can relax the constraint of canonical vectors in CCA and generate the same projection vector for two multichannel EEG signals. Furthermore, we propose a two-stage method based on the basic CORRCA method (termed TSCORRCA). Evaluated on a benchmark data set of 35 subjects, the experimental results demonstrate that CORRCA significantly outperformed CCA, and TSCORRCA obtained the best performance among the compared methods. This paper demonstrates that CORRCA-based methods have a great potential for implementing high-performance SSVEP-based BCI systems.

Hierarchical feature fusion framework for frequency recognition in SSVEP-based BCIs

“…COCA has been used to investigate crosssubject synchrony of neural processing [34], and inter-subject correlation in evoked encephalographic responses [33,39]. Recently, it was introduced for frequency recognition in SSVEP-based BCIs [31]. Below we provide a brief description of the standard CORRCA method.…”

“…Furthermore, we used the r-square value to evaluate the discriminability of the features obtained by each method. In the current study, the r-square value was computed using feature values of the attended target stimulus and the maximal feature values of the non-attended stimuli [38,31], as the following formula [41].…”

“…N 1 and N 2 are the numbers of features in these two sets. As the sign of this difference is important, we adopt the signed r-square value calculated as follows [41]: Previous studies have demonstrated that the number of channels and training trials can significantly influence performance [38,31]. Here we further in- vestigate the classification accuracies of HFCORRCA and CORRCA methods at different number of channels and training trials.…”

“…For EEG signal analysis, the COCA method maybe more efficient and practical in real-world applications [31,32].…”

“…Both the standard CCA method and CORRCA method provide multiple correlation coefficients to measure the correlation between two multi-dimensional signals. It is worth noting that traditionally only the largest coefficients is selected as feature while all others are ignored [27,28,38,31]. This inevitably results in the loss of the discriminative information.…”