This paper proposes a novel, robust, and efficient Wavelet Neural Network (WNN) technique to remove EOG artifacts by combining the approximation capabilities of both wavelet and neural network methods. In WNN, EOG recordings are not required once the NN being trained and the WNN algorithm can perform artifact correction in a single channel data. The method (1) decomposes the contaminated EEG signals to a set of wavelet coefficients, (2) passes the coefficients located in low frequency wavelet sub-bands through a trained artificial neural network (ANN) for correction and (3) reconstructs a clean version of EEG signals based the corrected coefficients. We applied the method to EEG data contaminated by EOG artifacts and compared the results with those obtained by other state-of-the-art methods including ICA and a wavelet thresholding method.