Sevilla, Lieka J.
Recognition of handwritten numbers using backpropagation - Cebu City CIT-U 2012
Sevilla, Lieka J.; College of Computer Studies, Cebu Institute of Technology-University; March 2012, Recognition of Handwritten Numbers Using Backpropagation.
Adviser: Prof. Cherry Lyn Sta. Romana
Handwritten Characters Recognition (HCR) is one of the vast topics in the fields of image processing. Until now, researchers are still finding the beat ways on how to recognize handwritten data efficiently and effectively. According to previous researches, the algorithm that could highly recognize handwritten data, although not perfectly, is the neural-network-based algorithm. Handwritten Characters Recognition can either involve numbers, alphabet, or alphanumeric data recognition. The subjects for recognition in this system were numeric data only.
The system uses the Backpropagation algorithm of the Artificial Neural Network (ANN) for recognizing handwritten numeric data. There were three main processes involved in this system: segmentation, feature extraction, and recognition. The approach used in the segmentation of the image into individual characters was the growing regions blob detection algorithm. Prior to feature extraction, the segmented image has undergone character segmentation and pre-processing stage. The pre-processing stage involves binarization, noise reduction, and normalization.
The system was not able to produce a very satisfactory recognition rate. The possible reason for the recognition error in the system are the lack of training samples, limited pre-processing and normalization ability, and the feature extraction approach used was not applicable to all kinds of image data. It is recommended that the number of training samples is increased and the image is thoroughly pre-processed and normalized. Finding another approach for feature extraction is also recommended.
T Se84 2012
Recognition of handwritten numbers using backpropagation - Cebu City CIT-U 2012
Sevilla, Lieka J.; College of Computer Studies, Cebu Institute of Technology-University; March 2012, Recognition of Handwritten Numbers Using Backpropagation.
Adviser: Prof. Cherry Lyn Sta. Romana
Handwritten Characters Recognition (HCR) is one of the vast topics in the fields of image processing. Until now, researchers are still finding the beat ways on how to recognize handwritten data efficiently and effectively. According to previous researches, the algorithm that could highly recognize handwritten data, although not perfectly, is the neural-network-based algorithm. Handwritten Characters Recognition can either involve numbers, alphabet, or alphanumeric data recognition. The subjects for recognition in this system were numeric data only.
The system uses the Backpropagation algorithm of the Artificial Neural Network (ANN) for recognizing handwritten numeric data. There were three main processes involved in this system: segmentation, feature extraction, and recognition. The approach used in the segmentation of the image into individual characters was the growing regions blob detection algorithm. Prior to feature extraction, the segmented image has undergone character segmentation and pre-processing stage. The pre-processing stage involves binarization, noise reduction, and normalization.
The system was not able to produce a very satisfactory recognition rate. The possible reason for the recognition error in the system are the lack of training samples, limited pre-processing and normalization ability, and the feature extraction approach used was not applicable to all kinds of image data. It is recommended that the number of training samples is increased and the image is thoroughly pre-processed and normalized. Finding another approach for feature extraction is also recommended.
T Se84 2012