Estimating a visual evoked potential (VEP) from the human brain is challenging since its signal-to-noise ratio (SNR) is generally very low. An eigendecomposition-based subspace approach originally proposed for enhancing speech corrupted by colored noise, has been investigated and tested in the single trial extraction of VEPs. This scheme arbitrarily labeled as an eigen-decomposition (ED) method has been compared with a third-order correlation (TOC) method, using both realistic simulation and real human data. The results produced by the ED algorithm show much cleaner waveforms, and higher degree of consistency in detecting the P100, P200, and P300 peaks. copyright by EURASIP.