Pattern recognition using fuzzy inference with lacked input data

In pattern processing, it is difficult for all features to be extracted correctly. For a system using fuzzy inference, if input datum is lacking, the system does not work well. In the paper, a framework of a modified fuzzy inference method with lack of input data is introduced, and experimental results are provided. In the proposed method that is modified from Mamdani's fuzzy inference, a result of each rule is the adjustment for the purpose of protection from influence of lacking input data. The adjustment of the resulting fuzzy labels at each rule is used as the degree of importance which is set up in the rules by humans. In the results of a simple experiment, the system can infer well with lack of input data using this method. In results of simple experiments using a set of five fuzzy rules (three input and three output), when one or two input data are lacking, the system infers correctly.