A NOVEL STRING GRAMMAR FUZZY CLUSTERING

Abstract

Methods of pattern recognition have a broad range on real-world applications. The techniques used to solve pattern recognition problems can be grouped into two main fundamental approaches, i.e., a numerical approach and a structural approach. This thesis only focus on the clustering method to treat the problems of pattern recognition and its applications by using the structural approach. We proposed a variety of new fuzzy clustering techniques for domain of string. We also proposed fuzzy median string by incorporating the idea of fuzzy to median string for prototype calculation of these string grammar fuzzy clustering algorithms. To evaluate the performance of our string grammar fuzzy clustering methods, we run our experiment on standard real-world data sets and real applications, i.e., Thai sign language translation and identification of cardio-pulmonary resuscitation activity in medical simulation videos. From the results, our string grammar fuzzy clustering algorithms can detect noisy data, and cluster overlapping data as well.