利用模糊筆劃鑑定之線上手寫字辨識

Abstract

在本篇論文中，我們提出一個利用模糊筆劃鑑定(fuzzy stroke type identifiction)和筆劃模式比對(model-based stroke string matching)之線上手寫中文字辨識系統。根據筆劃書寫的順序，每一個參考字可以表示成一個一維的模型(1-D string model)。書寫的字可以有尺寸大小上及其它不同的變異，而且限制很小。這個系統主要包含兩個部分：標準筆劃型式鑑定(primitive stroke type identification)以及利用筆劃序列比對(character recognition by stroke string matching)來辨識文字。在第一部份，首先從輸入筆劃中將筆劃線段(stroke segments)正確抽取出來，接著輸入字的所有筆劃將利用我們所提之模糊方法，鑑定成為一串標準筆劃型式的序列。最後一部分，這個字將藉由筆劃模型比對而被辨識出來。辨識的結果是經由測試六百零五個手寫字而得，其中，每個字都是由六個不同的輸寫型式。所得到的辨識率是百分之九十二點一，倘若以前兩名的辨識字來計算，則辨識率可達百分之九十六點七。

This thesis presents an on-line Chinese character recognition system based on fuzzy stroke type identification and model-based stroke string matching. According to the writing stroke sequence, each character is described by a 1-D stroke string model. The written characters can be loosely constraints, which are quite flexible on size and allow various variations. The proposed recognition system consists of two phases: primitive stroke type identification and character recognition by stroke-string matching. In the former phase, the stroke segments of imput strokes are extracted firstly and the strokes of input character are then identified as a sequence of primitive stroke types by our fuzzy approach. In the latter phase, the character is recognized by the stroke string matching. The recognition results are tested based upon the 605 handwritten characters about 6 variations for each one. The obtained recognition rate is 92.1%, and the cumulative classification rate of choosing the second most similar characters is up to 96.7%.