A Signal Subspace Approach for Single-trial Evoked Potential Estimation

Evoked potentials (EP) are electrophysiological signals generated by the central nervous system in response to stimulation. Subspace methods in EP analysis commonly relate to the application of the singular value decomposition (SVD) to reveal the principal features of the signals. In single-channel EP analysis, the measurement matrix is constructed based on an ensemble of singletrial EPs, obtained from repeated presentation of stimuli. In this paper, a new subspace based method for EP enhancement is presented. The method exploits ensemble as well as within single sweeps waveform information. The proposed single-trial estimation method is applied to auditory EPs and outperforms the ensemble approach.