This paper describes an approach where posterior-based features are applied in those recognition tasks where the amount of training data is insufficient to obtain a reliable estimate of the speech variability. A template matching approach is considered in this paper where posterior features are obtained from a MLP trained on an auxiliary database. Thus, the speech variability present in the features is reduced by applying the speech knowledge captured on the auxiliary database. When compared to state-of-the-art systems, this approach outperforms acoustic-based techniques and obtains comparable results to grapheme-based approaches. Moreover, the proposed method can be directly combined with other posterior-based HMM systems. This combination successfully exploits the complementarity between templates and parametric models.