The brain-computer interface (BCI) based on steady-state visual evoked potential (SSVEP) has been one of the most stable BCI that is able to transfer commands effectively by the aids of canonical correlation analysis (CCA) for frequency recognition. Nevertheless, CCA-based SSVEP detection encounters the spectral non-uniformity of spontaneous background EEG activity which deteriorates the performance. In this study, we found the performance of SSVEP-based BCI highly predictable by the standard deviation of resting-state response using CCA. Therefore, the proposed normalized CCA (nCCA) aims to calibrate the spectral non-uniformity of background activity which causes bias in classification, and outperforms standard CCA significantly in simulated online tests. In addition, nCCA features near-zero calibration making it a nice fit to practical applications of SSVEP-based BCI spellers.