觸控面板筆劃萃取之線上中文手寫辨識系統設計

On-Line Chinese Characters Recognition System Design Based on Stroke Feature Extractor

10.30167/JOIT.200906.0004

蕭如宣(Jue-Hsuan Hsiao)；陳彥廷(Yan-Ting Chen)

VHDL ； SOPC ； OLCCR ； ANN ； VHDL ； SOPC ； OLCCR ； ANN

亞東學報

29期（2009 / 06 / 01）

25 - 35

繁體中文

本文之文字庫以教育部公佈的5401個常用中文繁體字，作爲研究的基礎對象。首先，我們提出一個以8個方位區分筆段定義中文字基本筆劃特徵，並以VHDL Soft-IP設計技術，實現筆劃特徵萃取器來抽取中文31個基本筆畫特徵並建立特徵資料庫。我們再提出以17個起手筆劃分類，再以中文文字部首分群的方法建立中文字筆順資料庫。配合這種特殊結構的資料庫，每當我們取得一組筆順序列碼，就可經由一種樹狀搜尋演算法，快速正確的搜尋資料庫達成辨識效果。 在資料庫搜尋方面，我們進行最短及最長搜尋路徑的分析，驗證我們已經能提供一個辨識率和辨識速度兼具的雛型平台設計。

In this thesis, we utilize SOPC chip to design an on-line Chinese character recognition system. SOPC chip has easy learning and use, fully design elasticity and fit to fast software and hardware co-design, etc. Firstly, we proposed eight directions distinguish stroke's segment concept, and used it to define the fundamental strokes of Chinese character with number feature. A VHDL Soft-IP has been designed to accomplish stroke features extraction and thirty one fundamental stroke features was extracted to build Chinese Stroke Feature (CS Feature) database. Whenever we obtain a stroke feature, we can use Binary Tree Searching algorithm to search out the stroke code. Chinese characters consist of strokes in ordering, so utilize strokes flow of sequence code can define Chinese character concept. We propose a Chinese Stroke Flow (CS Flow) database to describe Chinese characters with sequence codes of stroke flow. This database classifies into 16 classes by the start's stroke of Chinese characters, and then grouping each class by radicals of Chinese characters. Whenever we obtain a sequence code of stroke flow, we can use Binary Tree Searching algorithm to search out the Big 5 code of Chinese character. Under the CS Flow database with such special structure, a binary tree searching algorithm can fast accuracy searching database that come to a recognizable result. The Chinese characters library of this work uses 5401 daily Chinese characters, Ministry of Education announced, as the basic target of studying. In the database of CS Feature, we proceed to search analysis of route shortest and longest. The analysis results are 3 comparisons and 198 comparisons respectively. We also proceed to search analysis of route shortest and longest in the database of CS Flow. The analysis results are 12 comparisons and 402 comparisons respectively. This hardware software co-design proved that we have already offered a platform of good recognizable rate and speed.

1. 觸控面板基本原理
2. 電容式觸控技術原理介紹
3. 觸控面板產業概況
4. 余國華(2002)。利用混合式的辨識法來辨認線上手寫中文字的字首。國立中興大學／應用數學系。
5. 李宜仲(2000)。利用文字結構來辨識線上手寫中文字之字首。國立中興大學／應用數學系。
6. 林世勇(2005)。銅活字中文辨識之序列比對演算法研究。中華大學／資訊工程學系碩士班。
7. 林慶昇(2001)。利用階層分析法來分類線上手寫中文字之字首。國立中興大學／應用數學系。
8. 陳茂太(1988)。類神經網路線上手寫中文字辨識系統。國立中央大學／資訊及電子工程研究所。
9. 蕭如宣(2003)。SOPC系統設計。儒林。
10. 蕭如宣(2002)。VHDL數位電路設計。儒林。