A New Pattern Recognition Method Based on the Fusionof Multiple Feature Information
CHEN Wen jie, DOU Li hua, CHEN Jie, DU Li hui (Dept. of Automatic Control, Beijing Institute of Technology, Beijing100081, China)
Studies methods of pattern recognition fusing multi feature information in the presence of imprecise knowledge. Imprecise knowledge related with the object is first expressed with fuzzy logic rules, and raw information about object class expressed as BPA is obtained through applying fuzzy inference to multiple features. The final recognition result is then achieved making use of the Demster Shafter (D S) theory to eliminate the imprecision of BPA as much as possible. The simulation results prove that the recognition method presented here is effective to solve multi feature fusion recognition problems with the existence of imprecise knowledge. The combination of fuzzy logic and D S theory can effectively utilize the imprecise knowledge to classify the objects.