Online recognition of handwritten characters using differential angles and structural descriptors
Pattern Recognition Letters
A MDRNN-SVM hybrid model for cursive offline handwriting recognition
ICANN'12 Proceedings of the 22nd international conference on Artificial Neural Networks and Machine Learning - Volume Part II
Hi-index | 0.00 |
It is presented in this paper a new approach to the problem of feature extraction. The approach is based on the edge detection, where a set of feature vectors is taken from the source image. The images under this investigation are considered to be manuscript characters and the features are obtained by the distance from the contour of each character to several observation points placed around the image. Such observation points are arranged along different geometric polygons built in a way to surround the image. The approach is evaluated against the na茂ve bitmap matrix considering different types of polygon. The discrimination power of each method is computed using both statistic and neural network entries. The proposed approach provides also good response to the scale, rotation and translation problems in addition to discrimination.