Multi-BAM and back propagation neural networks in handprinted character recognition

• Main Authors: Aberin, Ponciano A., Dy, Wellie G., Santos, Jaybee N., Suarez, Julius Q.
• Format: text
• Language: English
• Published: Animo Repository 1993
• Subjects: Neural networks (Computer science); Optical pattern recognition; Input design, Computer; Computer systems
• Online Access: https://animorepository.dlsu.edu.ph/etd_bachelors/16373
• Institution: De La Salle University

The group implemented two neural network models, namely, Multi-BAM and Back Propagation, on a personal computer. Both models were trained and tested for handprinted character recognition. At the same time, the group was able to establish an interactive handprinted character recognition system using the two neural network models. The system was divided into two parts. The first part was the character input module that handles scanning, separation, and scalling. The second part was the neural network module that handles learning and recognition. Either of the two neural networks models may be applied. The multi-BAM model was implemented using a two-layer recurrent network. On the other hand, a feedforward connection with one hidden layer was used for the implementation of the back propagation network. The two neural networks learning and recognition performance were evaluated using different configurations and parameters.