Multiple Classifier Fusion Based on Testing Sample Pairs

A new multiple classifier fusion approach is proposed, which use the classification based on Testing Sample Pairs (CTSP) as member classifiers. To make use of the potential information provided by the classifier, in this paper, CTSP classifier's output is modeled with the fuzzy membership function. Then, in multiple classifier fusion, the fuzzy-cautious ordered weighted averaging approach with evidential reasoning (FCOWA-ER) is used to combine the membership functions generated by different member classifiers. Experimental results show that the proposed multiple classifier fusion approach can effectively improve the classification accuracy.