An Expansion Method for Off-Line Hand-Printed Character Recognition Using On-line Character Writing Features

In off-line hand-printed character recognition, improving the recognition of deformed patterns is an important area of research. Although many recognition techniques have been proposed, none of these methods perform well enough for practical use. Our research examined how to improve recognition of various deformed hand-printed characters by extending conventional recognition techniques. Our proposed method expands existing techniques using on-line character writing information that contains writing pressure information. To evaluate the effectiveness of this expanded recognition method, expanded recognition systems based on both simple pattern matching and the Hidden Markov Model (HMM) were constructed. As a result of using on-line character writing information to expand these methods, both methods performed better in recognizing alphabet characters and Japanese Hiragana characters. Our proposed method also required fewer character samples when constructing the HMM recognition system.