Applied Some New Features in Off-line Recognition of Totally Unconstrained Handwritten Numerals Using Neural Network

摘要

In this paper some new features in recognition of totally unconstrained handwritten numerals are applied. The new features are base on image projection and do wavelet transformation, then at the deferent scale calculate the projection's fractal dimension, use the fractal dimension as feature applied to neural network input. The new features have some advantages, it is rotate invariant, and it can represent the image's characteristic at deferent scale. In order to verify the performance of the new features experiments with a handwritten numericla data base collected from Beijing Postal center were performed. The correct recognition rate in training set was 99.55