@article{oai:yamagata.repo.nii.ac.jp:00001120,
 author = {渡邊,  洋一 and 猿田,  和樹},
 issue = {1},
 journal = {山形大学紀要. 人文科学 = Bulletin of Yamagata University. Humanities},
 month = {Feb},
 note = {論文(Article), Handwritten Kanji characters were assessed by Hayashi's Quantification Method: a multivariate analysis on categorical variables. Sample characters were 528 handwritten characters selected from signatures written by 230 students. Thirty three subjects were asked to classify these characters into three levels of legibility: Good-Ordinary-Poor.Good-writing and Poor-writing characters were selected on the majority of this classification. As a result, 50 characters were selected as Good-writing group and 45 characters were selected as Poor-writing one. These characters were analyzed by the Quantification Method. The predictor variables were some global features and local features. First, a discriminant analysis of this method was proceeded between Good-writing group and Poor-writing group. The minimax hit rate of this analys is was 90.8%. The most contributory variable for discrimination was a global feature: the rate of the height and the width of character, and the second variable was a local feature: the presence/absence of the continuity of strokes. Thus, it was shown that we can assess quantitatively the legibility of handwritten Kanji characters by their global features and local features. Next, a structure analys is of this method was proceeded to map these features in a multidimensional space. As a result, it was observed that the features contributing for Good-writing were plotted in relatively narrow area. On the other hand, the features for Poor-writing scattered over wide area. These results suggest that the 'Poor-writing' should include multivariate factors.},
 pages = {85(224)--97(212)},
 title = {悪筆文字の定量化とその構造的特徴},
 volume = {15},
 year = {2002}
}