This paper presents an Empirical mode decomposition (EMD) approach for achieving high-speed frequency-tagged steady state visual evoked potential (SSVEP) brain computer interface (BCI) system. Empirical mode decomposition (EMD) has been demonstrated as a local and fully data-driven technique for the data processing of nonlinear and non-stationary time-series. It allows the frequency and amplitude of a time-series to be evaluated with excellent time resolution. The proposed system utilized flickering sources with high flickering rate acting as cursor system for the purpose of alternatively communication system. Induced EEG signal were decomposed into different scale of oscillation components namely intrinsic mode functions. In order to find out the SSVEP-related IMFs, instantaneous frequency were computed by generalized zero crossing (GZC). After identification of the frequency power by quadrature detection, the result shows the proposed system achieved 51.46 bits/min of average ITR. Compare with FFT approaches, EMD method provide better temporal and frequency sensitivity in detection of cognitive state of neuron activities.