Human-Perception Handwritten Character Recognition Using Wavelets

The human vision system effortlessly recognizes familiar shapes despite all changes and distortions found in the retinal images. This work proposes a novel approach for recognition of handwritten characters which is based on human perception. The wavelet transform is used to simulate the multiresolutional capability of the vision and to extract features such as fixation points and image details at horizontal, vertical and diagonal directions.A previous system [1] which uses wavelet directional features yielded a recognition rate of 98.25% using NIST numerals database. 1 Objective To develop a system based on human visual perception which is able to efficiently recognize handwritten characters. 2 Human Visual Perception [2] Eyes move and fix successively at the most informative points of an image; Eyes actively perform a selective and problem-oriented collection of information from the visible world; Neurons in visual cortex perform direction oriented selectivity by the detection of edges and local bars. 3 System Modelling Principles Pre−processed Image Extraction Directional Features Fixation Points Detection Attention Window Attention Windows Select all