設定形狀規則用於手寫字辨識

Abstract

在本篇論文中我們研究如何設定形狀規則，並將他們轉換為邏輯規則來篩選掉不正確的標準字元，目的是為了減少在空間拓樸失真量測方法中的候選標準字元 [1]，如此一來能大量減少不必要的計算，此外，利用形狀限制加入自組織網路訓練流程來提高辨識率，這些篩選規則與限制是讓辨識演算法減少執行時間與增加辨識率最關鍵的兩個步驟。

This work shows how to set shape rules and convert them into logical rules to skip incorrect templates and reduce the number of candidate templates in the spatial topology distortion measurement method [1]. This will drastically reduce the number of computations with improved recognition. Also, the recognition rate is improved by including shape constraints in the self-organizing matching process. The rules and constraints compose the two key steps to reduce the running time and improve the recognition rate.