跳到主要內容

臺灣博碩士論文加值系統

(44.201.97.0) 您好!臺灣時間:2024/04/13 10:49
字體大小: 字級放大   字級縮小   預設字形  
回查詢結果 :::

詳目顯示

我願授權國圖
: 
twitterline
研究生:周遵儒
研究生(外文):Chou, Tzren Ru
論文名稱:線上草寫體中文字變形模型及其應用研究
論文名稱(外文):Deformation Models of On-Line Cursive Chinese Characters and Their Applications
指導教授:陳文村陳文村引用關係
指導教授(外文):Chen Wen Tsuen
學位類別:博士
校院名稱:國立清華大學
系所名稱:資訊科學學系
學門:工程學門
學類:電資工程學類
論文種類:學術論文
論文出版年:1995
畢業學年度:83
語文別:中文
論文頁數:128
中文關鍵詞:線上中文字辨識變形模型彈性比對貝濟耳曲線隨機逼近自動簽名鑑識文字產生
外文關鍵詞:On-Line Chinese Character Recognition (OLCCR)Deformation
相關次數:
  • 被引用被引用:1
  • 點閱點閱:483
  • 評分評分:
  • 下載下載:0
  • 收藏至我的研究室書目清單書目收藏:0
一個能夠妥善處理文字變形的方法是提高文字辨識系統效能的關鍵所在,
我們稱這種變形的處理架構為文字變形模型(Character Deformation
Model, CDM),該模型是由文字的表示法和一套將文字變形加以模組化的
機制所構成,在一般的辨識系統內部,文字變形模型所扮演的角色是參考
字,而其中的輸入文字則被視為此模型的個案 (Instance),我們根據這
樣的基本想法設計用以建立高效能辨識系統所需的模型。在本論文中我們
提出了兩種文字變形模型,一是限制型拋物線變換(Constrained
Parabola Transformation, CPT)模型,一是隨機式三次貝濟耳曲線(
Stochastic Cubic Bezier Curve, SCBC) 模型,在限制型拋物線變換模
型中的變換是用來描述文字變形的,參考曲線經由這些限制型拋物線變換
作用後即視為對所觀察到之輸入文字的合理逼近,另一方面,在隨機式三
次貝濟耳曲線模型中參考字是一連串的隨機式三次貝濟耳曲線和雜訊的組
合,而輸入字便是其相對應之參考字隨機模型的量測值,這兩個模型都能
精確地掌握住文字變形的特性,並作為建立高效能辨識系統的基礎。除了
文字變形模型本身,論文中還提出了一個曲線的比對方法,稱作曲線調整
程序(Curve Alignment Procedure),用來建立參考字與輸入字之間的正
確對應關係,整個程序分作兩個比對步驟,轉折點的彈性比對 (Elastic
Match of Turn Points)與合併-拆解過程 (Merge-and-Split Process)
, 透過這個曲線調整程序,輸入文字曲線被分割成適當的小段曲線,而
個別小段曲線和參考字模型的每一段隨機式三次貝濟耳曲線取得一對一的
對應關係,這樣的對應關係可以保證獲得較可靠的文字距離度量以及較精
確的參數估計結果,並且大幅度降低曲線比對時的計算成本。我們就所提
出的方法進行了一連串實驗,根據實驗結果證實,不論是以限制型拋物線
變換模型或隨機式三次貝濟耳曲線模型為基礎所建立的辨識方法都遠比早
期的典型方法更為優異,在此,我們認為這種新的辨識方法已強化至足以
有效應付草寫體文字的變形,而作為實用辨識系統的核心。本文共分為八
章:第一章是概論;第二章說明階層式比對的觀念並提出曲線調整程序;
第三章提出變換型文字變形模型;第四章提出隨機型文字變形模型;第五
章對所提出模型的效能作進一步分析;第六章強化所提出的模型以適應個
別的文字書寫方式;第七章建議模型的其他應用;第八章做一結論並指出
未來的發展方向。
A good method to manipulate the character deformations is
essential to improving the performance of a recognition system.
This manipulation framework is called a character deformation
model (CDM). The CDM is composed of two components; one is a
character representation and the other is a mechanism to model
the deformations. In this framework, the CDM''s serve as the
reference patterns in a recognition system, and the input
characters are considered as the instances of these reference
models. Based on the concept of manipulating the character
deformations, we are devoted to designing the remarkable CDM''s
for the construction of a high performance recognition system.
In this thesis, two CDM''s, named constrained parabola
transformation (CPT) model and stochastic cubic B\''{e}zier
curve (SCBC) model, are proposed. In the CPT model, the input
character is represented by the piecewise reference curves with
the CPT''s that approximate the deformations. On the other hand,
the reference character in the SCBC model is represented by a
sequence of SCBC''s with some random noises, and the input one
is treated as the measurement of its related reference model.
Both of the CDM''s have the ability to accurately describe the
character deformations and are the foundation of constructing a
high performance recognition system. Some experiments have been
performed. As observed from these experimental results, the
recognition methods based on the CPT and the SCBC models are
more effective and efficient than those proposed in some
typical previous works. It is believed that the proposed
methods are robust enough to act as the kernel of a practical
recognition system.
Cover
Contents
1 Introduction
1.1 Research Motivation
1.2 Ccharacter Defomation Modesl(CDM)
1.3 Overview of Proposed Approaches
1.3.1 Hierarchical Match of Curves
1.3.2 Transformational Character Deformation Model
1.3.3 Stochastic Character Deformation Model
1.3.4 Adaptation to Individual Writing Styles
1.3.5 Other Applications
1.4 Thesis Organization
2 Hierarchical Match and Curve Alignment Procedure
2.1 Input Character Repesentation
2.2 Hierachical Curve Match
2.3 Curve Alignment Procedure
2.3.1 Elastic Match of Turn Points
2.3.2 Merge-and-Split Process of Piecewise Bezier Curves
3 Transformational Character Deformation Model
3.1 Constrained Parabola Transformation Model
3.2 Feasible Deformation
3.3 Constrained Parabola Transformation(CPT)
3.4 parameter Estimation
3.5 Experimental Results
4 Stochastic Character Deformation Model
4.1 Stochastic Cubic Bezier Curve Model
4.1.1 Stochastic Cubic Bezier Curves(SCBC)
4.1.2 Tangent of an SCBC
4.2 Parameter Estimation
4.3 Experimental Results
4.3.1 Recognition Scheme
4.3.2 Classifaction of 8-Scharacters
4.3.3 Classification of Similar Characters
4.4 Improvements of Distance Measure
4.4.1 Control-Point-Based Distance Measure
4.4.2 Weighted Distance Measure
4.4.3 Comparison
5 Further Analysis of Model Performance
5.1 More about Recognition Rate
5.1.1 Stroke Difference
5.1.2 Warping Factor
5.2 quality of Distance Measures
5.2.1 Mean and Standard Deviation
5.2.2 Fisher Criterion
5.2.3 Discrimination Ration
5.3 MIsclassification Analysis
5.3.1 Reason Discussion
5.3.2 Possible Improvements
6 Adaption to Individual Writing Styles
6.1 Updating Rule for COT Model
6.2 Updating Rule for SCBC Model
6.3 Experimental Results
6.3.1 Adapted Results of CPT Model
6.3.2 Adapted Results of SCBC Model
7 Other Applications
7.1 Automatic Signature Verification
7.2 Charcter Generation
8 Conclusions and Further Researches
8.1 Conclusions
8.2 Further Researches
A Piecewise Approximation of Cubic Bezier Curves
B Attributed String Editing Algorithm
C Approximate Conversion of Cubic Bezier Curves
Other
QRCODE
 
 
 
 
 
                                                                                                                                                                                                                                                                                                                                                                                                               
第一頁 上一頁 下一頁 最後一頁 top