A Study on Handwriting Analysis by OCR

First Author*, Second Author**,
Third Author**


Institute Name

Institute Name, if any


– OCR, commonly known as
Optical Character Recognition also known as Optical Character Reader. It is
used to collect information from handwritten documents or manuscripts, printed
paper data records. It is a common method of digitizing handwritten manuscripts
so that they can modified or used digitally in modern technology.

Terms – Acknowledgement,
Conclusion, OCR Procedure, Text Processing.


THE recognition and conversion from images of text have
always been a challenging task for automatic data processing and information
retrieval and services. In particular, the task of scanning human handwriting
and making them not only digital readable, but also searchable and digitally
editable, is important to retrieve and collect information.

this way, the information of old manuscripts or any handwritten document can be
a valuable and interesting source to build a strong and complete information
network. Different organizations are interested in the mass scale digitization
of historic manuscripts or handwritten documents with a focus on offering
improved full-text searching.


digital collections and information systems digitalize handwritten documents,
such as old manuscripts of any historic civilization or very old manuscript or
any handwritten document. However, optical character recognition of old handwritten
manuscripts often poses different challenges. In the following phrases we
summarize the main issues.

OCR problems of Handwritten

with handwritten documents, we face different problems.

problems: One of the major
problems is the quality of the original manuscript and the quality of the scan.
This includes issues such as curled pages, blurred fonts, or manually edited
pages (e.g. stamps or hand-written notes).

type and layout: Human
handwritings are not supported by standard OCR software. The fonts or
characters are different with the change of person. Old manuscripts are even
harder to recognize, because they use font styles that are totally new to
modern human society. Thus, also the spacing between words and characters is
often not consistent. Additionally, historic papers often use different and
inconsistent layout structures.

knowledge base: Traditional OCR
software uses knowledge bases based on contemporary dictionaries and
grammatical structures to enhance the OCR procedure and does not provide manuscripts
documents. Additionally, historic manuscripts often do not follow specific
orthographic structures and rules, thus words can be written differently in the
same text.

manuscript or handwriting is different from each other. Thus, very specific and
unique problems can occur for every project. In the next section, we take a
closer look at the overall OCR process, and possibilities to improve the
different steps.

OCR Process

technology develops by time the necessity of recording data in handwritten
format decreases by time, now a days all records are being kept in format of
digitized text document or media, so it is necessary to focus on these specific
problems in the OCR process. In the following section, we describe the
procedure of OCR with a focus on creating a learning / feature base for
handwritten documents, which can be used for improving machine learning
algorithm. To improve the accuracy of the OCR process, different actions can be
taken in every single step of the process.

The first phase of Optical Character Recognition is the scanning phase. This
phase is one of the most important phases. If possible, scans should be made of
well-preserved and clean originals. The scanning resolution should be at least
300 dpi and the output image a lossless image format (e.g. tiff).

Pre-processing: This
is the second phase of Optical Character Recognition. In this step, the scanned
document can be manually optimized for the OCR process. This includes image
editing processes such as increasing the contrast, reducing noise, or
simplifying the colors.

In this phase of Optical Character Recognition, the chosen OCR system reads the
images and applies an algorithm to recognize the characters.  It is crucial to choose OCR software that fits
the current problem and supports a training/learning algorithm.

Create learning base: To improve the OCR-process it is very important to
create and improve the learning base for training the OCR system. This base
consists of a dictionary fitting the document improved character pattern.

In this phase, knowledge can be applied which had not yet been available to the
OCR system. In a final step, the output can be corrected manually.


In this paper, we tried to throw light on the process
of OCR of scanned old documents, historic books, manuscript of a very old
document. To compare the accuracy of the OCR methods, a normalized version of
the Levenshtein distance can be used. Since every historic book is different
and poses its own and new challenges, the most important step of an OCR process
is building a learning base. The main contribution of this work is a model for
OCR processes of historic books with old fonts. With such a model, with respect
to preened post-processing, the accuracy of OCR of manuscripts or human
handwriting can be improved significantly, compared to related approaches.


This work would not be possible without the help of Prof.
Sukanya Roy, Lecturer at University of Engineering & Management, Kolkata. We
would also like to thank the other group members for their hard work and
dedication to complete the case study on Optical Character Recognition in
Handwriting Analysis

                   VII.        REFERENCES

[1] Holley, R., “How good can it get? Analyzing
and improving OCR accuracy in large scale historic newspaper digitization
programs.” D-Lib Magazine 15.3/4 (2009).

[2] “The challenges of historical materials and an
overview on the technical solutions in IMPACT”

[Online]. Available: https://impactocr.wordpress.com/2010/05/07/anoverview-of-technical-solutions-in-impact/

[3] Mori, S,, Ching Y. S., and Kazuhiko Y.,
“Historical review of OCR research and development.”

Proceedings of the IEEE 80.7 (1992): 1029-1058.

[4] Feng, S., and Manmatha, R., “A hierarchical,
HMM based automatic evaluation of OCR accuracy for a digital library of
books.” Digital Libraries, 2006. JCDL’06. Proceedings of the 6th

Conference on. IEEE, 2006.

[5] Gupta, M. R., Jacobson, N. P., and Garcia, E. K.,
“OCR binarization and image pre-processing for searching historical
documents.” Pattern Recognition 40.2 (2007): 389-397.

[6] Wikipedia, Basic Idea of Optical Character
Recognition, [Online]

Leave a Reply

Your email address will not be published. Required fields are marked *