Template-based analysis techniques are good candidates to robustly detect transient temporal graphic elements (e.g. event-related potential, k-complex, sleep spindles, vertex waves, spikes) in noisy and multi-sources electro-encephalographic signals. More specifically, we present the impact on a large dataset of a wavelet denoising to detect evoked potentials in a single-trial P300 speller. Using coiflets as a denoising process allows to obtain more stable accurracies for all subjects.