Score Normalization by Dynamic Time Warping

Multi-expert approach is a well-known paradigm to support complex decisions and several decision fusion techniques have been considered. In this field score normalization is a fundamental step to allow high performances and hence several techniques have been proposed so far. This paper addresses the problem of score normalization and presents a new technique based on Dynamic Time Warping. The experimental results, carried out in the field of pattern classification, show the superiority of the new technique with respect to other approaches in the literature, based on MIN-MAX, z-score and characteristics functions.