This paper describes how the last cutting-edge advances in Computational Intelligence are being applied to the field of biometric security. We analyze multimodal identification systems and particularly the score fusion technique and some issues related with it. Fundamentally, the paper deals with the scores normalization problem in depth, which is one of the most critical issues with a dramatic impact on the final performance of multibiometric systems. Authors show in this paper the results obtained using a number of fusion algorithms (Neural Networks, SVM, Weighted Sum, etc.) on the scores generated with three independent monomodal biometric systems (the modalities are Iris, Signature and Voice). The paper shows the behavior of the most popular score normalization techniques (z-norm, tanh, MAD, etc), and proposes a new score normalization procedure with an optimized performance harnessing tested fusion techniques and outperforming previous results through a proof-of-concept implementation.