TEXT RECOGNIZING METHOD USING SECOND-DIMENSIONAL CAPABILITY MODEL

摘要

PURPOSE: To recognize an unclear character or word by expressing a character with a pseudo imbedded Markov model, and comparing it with a model by using a viterbi algorithm. ;CONSTITUTION: Five superstates 210-214 are shown in state graphs 200 for a pseudo second-dimensional imbedded Markov model. The superstates are made correspond to five horizontal areas 110-114 in a pixel map 100. For example, the superstate 211 includes three states 220-222 expressing a white picture element, black picture element, and white picture element in whole rows in the area 111. Allowable transition among the superstates is shown by arrows 230 and 231, and an arrow 232 indicates that transition may be generated from a certain state to itself. In the same way, transition among the states is indicated by arrows 240 and 241, and the transition to the same state is indicated by an arrow 242. The capability of each state of the imbedded Markov model is calculated related with each observed value, and N×T array (N is the number of states, and T is the number of observed values in a permutation) is obtained according to a viterbi algorithm.;COPYRIGHT: (C)1994,JPO