The EMD-CCA with Neural Network Classifier to Recognize the SSVEP Frequency

To increase the number of stimulation frequencies in the Steady-state visual evoked potential (SSVEP)-based brain-computer interface, we are forced to broaden the frequency range due to the frequency resolution restriction. This will enter frequencies with harmonic relation into the stimulation frequency range and lead to increase in frequency recognition error. In this paper, a three-stage method including the empirical mode decomposition (EMD), the canonical correlation analysis (CCA) and neural network classifier has been proposed that can solve the recognition error problem for wide frequency range including frequencies with harmonic relation. Visual stimulus ranged from 6-16 Hz with an interval of 0.5 have been generated using Matlab and the psychophysics toolbox. The SSVEP signal was recorded from ten subjects via one electrode placed at Oz. After extracting the intrinsic mode functions (IMFs) of the signal by EMD and reconstructing the combined signals, the CCA has been applied. Two features including the detected frequency and the correlation value in this frequency have been extracted and they were given to the neural network classifier. For eight-second time window, the average accuracy of the CCA for N=1 was 78% and for N=2 was 74%, while the corresponding values of the proposed method were 82% and 77% respectively. For four-second time window, the accuracy was increased from 78% to 83% for N=1 and it was increased from 78% to 80% for N=2. N is the number of harmonics in the generation of the reference signal in the CCA. For wide frequency range, the proposed method has been able to improve the frequency recognition accuracy compared to the standard CCA method. according to this, by broadening the stimulation frequency range, the possibility of increasing the number of frequency options and thus increasing the information transfer rate are provided.