Katakana character recognition using neural networks
Noriyuki Sakuma · ThinkTech (Texas Tech University) · 1993
This thesis presents a new preprocess1ng model for character recognition and simulation experiments. The goal of the model is to recognize Japanese Katakana characters by dealing with segments having vectors with 9 blocks. The JK9BVL preprocessor deals with a segment of a character as a vector and presents the position, length, and direction of the vector with the output value of each block which consists of 9 blocks. The simulation experiments consist of two parts. The first part involves learning a character set and testing two noise data sets. The second part involves the recognition test of two sets of unknown character sets.