Title :
Segmentation and recognition of unconstrained numerals on Chinese bank-check
Author :
Su, Hui ; ZHAO, Bin ; Ma, Feng ; XIA, Shaowei
Author_Institution :
Dept. of Autom., Tsinghua Univ., Beijing, China
Abstract :
The methods of segmentation and recognition of unconstrained numerals on Chinese bank check are discussed for building a recognition system. The combination-segmentation method of connected domain and projection is proposed to segment each digit. This segmentation method provides a way to solve the problem of segmentation of the connected digits with a connecting stroke. Three classifiers are designed independently with different methods using different features of handwritten digits. They can compensate each other when confusing digits are met. Under the limitation of fault-tolerant rate, and with the three classifiers of high isolated digit recognition rate, the check recognition rate of the system is practical in routine work of bank
Keywords :
bank data processing; character recognition; image classification; image segmentation; neural nets; statistical analysis; Chinese bank-check; character recognition; connected digits; connecting stroke; handwritten digit recognition; hierarchical neural networks; image segmentation; multiple structural feature classifier; principal component analysis classifier; unconstrained numeral recognition; Fault tolerant systems; Handwriting recognition; Image segmentation; Joining processes; Neural networks; Office automation; Principal component analysis;
Conference_Titel :
Systems, Man, and Cybernetics, 1996., IEEE International Conference on
Conference_Location :
Beijing
Print_ISBN :
0-7803-3280-6
DOI :
10.1109/ICSMC.1996.565501