The Handwriting recognition based on a-Si:H detector(FOQUOD)

• Main Authors: Shu,Su-Frang, 許菽芳
• Other Authors: Lee,Si-Chen
• Format: Others
• Language: en_US
• Published: 1994
• Online Access: http://ndltd.ncl.edu.tw/handle/23019837549381932847

碩士 === 國立臺灣大學 === 光電(科學)研究所 === 82 === Handwriting recognition is the present trend for optical character recognition (OCR) system. Its development follows two kinds of approaches, i.e., hardware and software. Hollow FOur QUadrant Orientation Detector (FOQUOD) developed in our lab is a new type of detector which can detect the orientation and position of an image edge. Because of the direct extraction of these informations,structural analysis which is difficult but has high distortion tolerance in feature extraction is possible. A structural handwriting recognition algorithm based on an 8 x 8 hollow FOQUOD array is developed to do preprocessing of the input data: noise elimination, thinning, structural feature extraction and classification. In this study, the algorithms for structural recognition of handwriting based on FOQUOD array have been constructed. They can recognize characters which are rotated, translated parallelly or with distortion. They are able to distinguish the end, angle, fork, cross and curve in a character even if the edge orientation contains noises. A four path checking rule is built to search a straight line along the same orientation on a pattern which has not been skeletonized yet. It combines thinning, searching and recording processes to save extraction steps. A local summation rule is also proposed to get local feature from irregular values of detector output. After numerous test of simulation and experimental data, this method has been proved to have high potential for application in handwriting recognition.