This paper reports on application of Complete Ensemble Empirical Mode Decomposition (CEEMD) technique to pre-process Electro-Oculogram (EOG) signals before eye blink detection technique is implemented. EOG is a non-stationary signal which is affected by different kinds of interferences. During the time of recording EOG signal gets contaminated by Electromyography (EMG) signal. In this paper CEEMD is used to decompose the EOG signal into several intrinsic mode functions (IMFs). After thresholding each IMF the signal is reconstructed using all of the IMFs. The resulting denoised signal is then used to detect eye blink