異なる特徴抽出法を併用した手書き類似文字の識別

概要

論文の詳細を見る
To extract unique feature vectors of handwritten characters and thereby identify them efficiently, we now propose and make available a means to use feature vectors known as P-DC and P-LOVE. These feature vectors tend to be used separately in general systems. However, we sought to achieve a more effective way of recognition by simultaneously using both sets of unique feature vectors. After some experiments using an integral type of neural network (NN), we found that we were able to truly accomplish a better way of recognition if we used both feature vectors together rather than using each of them separately. Based on the judgment obtained from such a recognition system using an integral NN, we also found that it became possible to process rejections in an effective manner.
明石工業高等専門学校の論文