Handprint to text.
Date of Publication
2000
Document Type
Bachelor's Thesis
Degree Name
Bachelor of Science in Electronics and Communications Engineering
Subject Categories
Systems and Communications
College
Gokongwei College of Engineering
Department/Unit
Electronics and Communications Engineering
Abstract/Summary
The thesis entitled, Handprint to Text , is basically a software enabling the user to convert a scanned handprinted text document into a text file. This text file may then be opened in word processors for printing or further refining. The acceptable scanned document should be a BMP file and the handprint recognizing engine is the neural network, Bidirectional Associative Memories (BAM) developed by Bart Kosko. By doing so, the researcher will not have to worry about typing his work anymore. He would just have to scan his research draft, process the scanned work and open the text file in MS-Word and he instantly has his work typewritten. That way, the researcher has more time doing more important things other than typing.
The neural network model was implemented in Turbo C and was trained to recognize all 26 uppercase letters. These letters were each represented as a 24 X 18 matrix consisting of either a value of 1 or -1. The matrix was then passed to the neural network for proper classification. A single weight matrix representing all the letters was not sufficient, therefore multiple weight matrices were used, one per letter. To further improve recognition, five sets of letters were used to train the neural network. That way, the neural network has more exemplar pairs to base its decision. The output of the neural network is a vector consisting of 8 bits. These bits were used to properly encode the text file.
Prior to feature extraction, median filtering was employed to reduce noise in the image. After filtering, thresholding was done to set the image into two level colors only for ease of feature extraction.
Abstract Format
html
Language
English
Format
Accession Number
TU09602
Shelf Location
Archives, The Learning Commons, 12F, Henry Sy Sr. Hall
Physical Description
56 numb. leaves ; Computer print-out.
Keywords
Optical character recognition--Computer programs; Text processing (Computer science)
Recommended Citation
Ching, J., Chua, C., Lee, B., Lee, W., & Lim, J. (2000). Handprint to text.. Retrieved from https://animorepository.dlsu.edu.ph/etd_bachelors/10408
Embargo Period
9-16-2021