| HOME | RESEARCH | TEACHING | PUBLICATIONS | STUDENTS | GRIFFITH | SCHOOL |
|
|
|
| Handwriting Recognition Application |
Some research has already demonstrated the success of ANN-based pattern recognition systems for the recognition of handwritten zip codes [1], [2]. Research in the field has been ongoing for a substantial amount of time. The ANN-based zip code recognition system is very significant due to the excellent recognition rates attained. The zip codes used for training and testing were sampled from an actual U.S. post office. The authors' work proved the accuracy and speed which could be attained using a backpropagation neural network. It was also demonstrated that their work could be extended for use with larger tasks.
This research attempts to extend their research focusing on another difficult task which is handwritten postal address recognition. Sample handwritten addresses were collected and stored in a large database. The aim was to develop a system that could segment, preprocess, and classify the handwritten addresses acquired. An ANN was used for the actual classification of characters and addresses [3], [4].
The proposed system is shown in Fig 1 (below). This research is ongoing and future work will use a benchmark database for experimentation purposes.
Figure 1 (12 Kb)
[1] Denker J. S. et al., 1989, Neural Network Recognizer for Hand-Written Zip Code Digits, Neural Information Processing Systems, 1, 396-493.
[2] Le Cun, Y. et al., 1990, Handwritten Digit Recognition with a Back-Propagation Network, Neural Processing Systems, 2, 323-331
[3] B. K. Verma, M. Blumenstein, 1996, An Intelligent Neural System for a Robot to Recognize Printed and Handwritten Postal Addresses, in Fourth IASTED International Conference on Robotics and Manufacturing, IASTED RM'96, Hawaii, USA, 80-84.
[4] M. Blumenstein, and B. Verma, 1997, A Segmentation Algorithm used in Conjunction with Artificial Neural Networks for the Recognition of Real-World Postal Addresses, International Conference on Computational Intelligence and Multimedia Applications, Gold Coast, Australia, 155-160.