Quick Recognition and Relative Minimum Distances Filtering Assisted Recognition Based on Noisy-robust Rough Set

Abstract

With the development of rough set theory and it’s strengths and weaknesses in the recognition field, the rule and recognition fusing method of noisy-robust rough set (NRRS) are proposed based on full normalized deposal, the non-uniform companding and simple dynamic clustering coding. The robustness of NRRS is improved by weighted reliability during training and recognition to dispose the accidental samples and conflict samples. At the same time, this paper gives the quick recognition and relative minimum distances between classes filtering assisted recognition algorithm based on NRRS. The recognition simulation shows that the method has a good anti-noise performance, processing efficiency and recognition effect.