視覚の誘導場によるテキスト独立形筆者識別法の提案(<論文小特集>人間の視覚・聴覚情報)

概要

論文の詳細を見る
The purpose of this study is to develop a text-independent writer-recognition method using feature extraction by the induction field model in vision. We propose using the strain of the hadwritten character for the standard character to recognize writers. To perform that, it is necessary to quantify the strain as the values of features. The induction field model in vision it possible to quantify the strain. In this method, we first calculate the induction field in vision and we extract a constant potential line for each character image. Next, we calculate the distance between the potential line and a central point in every direction. Then we make the ratio between the distance in case of the handwritten character and that in case of the standard character to be the values of features. Finally, we recognize writers from Euclid distances between the values of features in case of the handwritten character and that in case of the standard character. To verify the effectiveness of this method, we experimented for 20 persons and we obtained 95% recognition rate.
2000-09-15