Improvement of identification algorithms in the freehand curve identifier FSCI

Abstract

SUMMARYWe already proposed the freehand curve identifier FSCI, based on fuzzy theory, as a freehand figure recognition engine. We next investigated the improvement of the recognition rate of FSCI by learning optimization of parameters. This paper proposes improvements of the FSCI algorithm itself, namely, improvement of the hypothetical fuzzy model formulation, and introduction of necessity into the fuzzy inference rules. It was shown by experiments that these improvements have effects which cannot be obtained by conventional parameter adjustment.