This article presents a new computational intelligence technique for pattern recognition of graphic elements (e.g. event-related potential, auditory evoked potential, k-complex, spindle) embedded in electro-encephalographic signals. More precisely, we have extended the learning vector quantization (LVQ) algorithm by Kohonen to non-identity assignment to robustly detect evoked potentials in a noisy electro-encephalographic signals for brain-computer interfaces. The improved LVQ is obtained by optimizing its assignment layer through the minimum norm least square algorithm, the same scheme found in Extreme Learning Machine (ELM). The proposed LVQ is evaluated using the Wadsworth BCI datasets on P300 speller. The experimental results show that the proposed LVQ improved the performance with less computational units.