線上手寫注音符號辨識系統

Abstract

在最近的六個月內已有幾種不同的線上手寫中文辨識系統的商業產品上市 ，其重要性值得我們在未來的幾年內付出更多的研究心力。在這篇論文中 ，我們嘗試發展一個藉由注音符號而達成中文文字辨識的方法。在第一個 階段中，我們將所有可能的筆劃分類為11類基本筆劃，並建立了一個由基 本筆劃序列來代表每個注音拼音的資料庫。當接收從數位板來的資料時， 我們在每一筆劃的結束時辨識出其所屬的基本筆劃類別，在整個注音拼音 寫完時，根據所辨識出的基本筆劃，便可識別出此手寫字形所代表的拼音 。由於基本筆劃的辨識比較簡單，且在書寫完畢後只有少量的工作須處理 ，因此整個辨識的過程可快速完成。我們已將此線上手寫注音符號辨識系 統實作出來，基本上證明我們的想法是正確的，有一些問題是辨識率的瓶 頸，為使此系統進入完全實用階段，未來的改良仍是必要的。 Several commercial on-line handwritten Chinese characters recognition systems have become available in last six months. Its importance deserves more research efforts in next few years. In this thesis we try to develop a way of character recognition through its phonetic spelling. In the first place, we classified all possible strokes into 11 primitive strokes. Then we built a database representing each chracter by a sequence of primitive strokes corresponding to its phonetic spelling. When accepting input from tablet,at the end of each stroke writing,we recognize what primitive strokes it is. At the end of whole phonetic spelling writing,based on the primitive stroke recognized ,we can identify the character it represent. The point is that : primitive strokes are easy to identify and minimal work is to be finished at the end of writing. Thus recognition can be done at high speed. We have implemented such a system.It shows that our idea is basically correct. A few problems degrades its recognition rate. Further improvements are required to make it pratically useful.