Handwriting Recognition

by Liu, Zhi-Qiang; Cai, Jin-Hai; Buse, Richard; Cai, Jinhai

• ISBN13: 9783540401773
• ISBN10: 3540401776
• Format: Hardcover
• Copyright: 2003-10-01
• Publisher: Springer Verlag

Summary

This book takes a fresh look at the problem of unconstrained handwriting recognition and introduces the reader to new techniques for the recognition of written words and characters using statistical and soft computing approaches. The types of uncertainties and variations present in handwriting data are discussed in detail. The book presents several algorithms that use modified hidden Markov models and Markov random field models to simulate the handwriting data statistically and structurally in a single framework. The book explores methods that use fuzzy logic and fuzzy sets for handwriting recognition. The effectiveness of these techniques is demonstrated through extensive experimental results and real handwritten characters and words.

Table of Contents

1 Introduction	1	(16)
1.1 Feature Extraction Methods	2	(4)
1.1.1 Extracting Features from Binary Images	3	(2)
1.1.2 Extracting Features from Gray-Scale Images	5	(1)
1.2 Pattern Recognition Methods	6	(11)
1.2.1 Statistical Pattern Recognition	6	(3)
1.2.2 Structural Pattern Recognition	9	(2)
1.2.3 Neural Networks for Pattern Recognition	11	(1)
1.2.4 Soft Computing in Handwriting Pattern Recognition	12	(5)
2 Pre-processing and Feature Extraction	17	(44)
2.1 Pre-processing of Handwritten Images	17	(13)
2.1.1 Pre-processing for Handwritten Characters	17	(4)
2.1.2 Pre-processing for Handwritten Words	21	(9)
2.2 Feature Extraction from Binarized Images	30	(1)
2.3 Feature Extraction Using Gabor Filters	31	(26)
2.3.1 Skeletonization Using Gabor Filters	42	(7)
2.3.2 Results of Skeletonization	49	(8)
2.3.3 Extracting Oriented Segments Using Gabor Filters	57	(1)
2.4 Concluding Remarks	57	(4)
3 Hidden Markov Model Based Method for Recognizing Hand written Digits	61	(28)
3.1 Theory of Hidden Markov Models	61	(15)
3.1.1 Markov Process	61	(2)
3.1.2 Hidden Markov Models	63	(2)
3.1.3 Basic Algorithms for HMMs	65	(8)
3.1.4 Continuous Observation Hidden Markov Models	73	(3)
3.2 Recognizing Handwritten Numerals Using Statistical and Structural Information	76	(10)
3.2.1 Statistical Modeling	76	(6)
3.2.2 Structural Modeling	82	(4)
3.3 Experimental Results	86	(1)
3.4 Conclusion	87	(2)
4 Markov Models with Spectral Features for Handwritten Numeral Recognition	89	(18)
4.1 Related Work Using Contour Information	89	(2)
4.2 Fourier Descriptors	91	(4)
4.2.1 Feature Extraction	93	(2)
4.3 Hidden Markov Model in Spectral Space	95	(8)
4.3.1 Spectral Space	95	(3)
4.3.2 Semi-Continuous Markov Model	98	(2)
4.3.3 Evaluation, Re-Estimation and Initiation	100	(3)
4.4 Experimental Results	103	(1)
4.5 Discussion	104	(3)
5 Markov Random Field Model for Recognizing Handwritten Digits	107	(24)
5.1 Fundamentals of Markov Random Fields	107	(7)
5.1.1 One-Dimensional Markov Processes	107	(2)
5.1.2 Markov Random Fields	109	(3)
5.1.3 Markov Mesh Random Fields	112	(2)
5.2 Markov Random Field for Pattern Recognition	114	(8)
5.2.1 Maximum a posteriori Probability	115	(1)
5.2.2 Markov Random Fields for Modeling Statistical and Structural Information	116	(1)
5.2.3 Neighborhood System and Cliques	117	(1)
5.2.4 Minimizing the Likelihood Energy	118	(4)
5.3 Recognition of Handwritten Numerals Using MRF Models	122	(6)
5.3.1 Feature	122	(2)
5.3.2 Clique Function	124	(1)
5.3.3 Maximizing the Global Compatibility	125	(3)
5.3.4 Experimental Results	128	(1)
5.4 Conclusion	128	(3)
6 Markov Random Field Models for Recognizing Handwritten Words	131	(14)
6.1 Markov Random Field for Handwritten Word Recognition	131	(3)
6.1.1 Markov Random Field for Structural Modeling	132	(1)
6.1.2 Recognition based on Maximum a posteriori Probability	133	(1)
6.2 Neighborhood Systems and Cliques	134	(1)
6.3 Clique Functions	135	(3)
6.4 Maximizing the Compatibility with Relaxation Labeling	138	(2)
6.4.1 Relaxation Labeling	138	(1)
6.4.2 Maximizing the Compatibilities	139	(1)
6.5 Design of Weights	140	(1)
6.6 Experimental Results	141	(2)
6.6.1 Neighborhood Size	141	(1)
6.6.2 Iterations	142	(1)
6.7 Conclusion	143	(2)
7 A Structural and Relational Approach to Handwritten Word Recognition	145	(28)
7.1 Introduction	145	(1)
7.2 Gabor Parameter Estimation	146	(8)
7.3 Feature Extraction	154	(14)
7.3.1 Slant Correction	154	(8)
7.3.2 Part Extraction	162	(1)
7.3.3 Feature Extraction	163	(5)
7.4 Conditional Rule Generation System	168	(1)
7.5 Experimental Results	169	(3)
7.6 Conclusion	172	(1)
8 Handwritten Word Recognition Using Fuzzy Logic	173	(22)
8.1 Introduction	173	(1)
8.2 Extraction of Oriented Parts	173	(1)
8.2.1 Slant and Tilt Correction	174	(1)
8.3 System Training	174	(10)
8.3.1 Word Alignment	176	(4)
8.3.2 2-D Fuzzy Membership Functions	180	(4)
8.4 Word Recognition	184	(7)
8.4.1 Fuzzy Decision Making Process	186	(5)
8.5 Experimental Results	191	(1)
8.6 Conclusion	192	(3)
9 Conclusion	195	(28)
9.1 Summary and Discussions	195	(2)
9.2 Future Directions	197	(4)
9.3 References	201	(22)
Index	223