FEATURE EXTRACTION FOR CHARACTER RECOGNITION

The m-files inside this zip file extracts features of single characters of english language based on their geometric properties from the input image. Two approaches are explained for extracting feature vectors. The final feature vector generated for my purpose had more 120 elements. The features are based on the basic geometric shapes that comprises a single character. It calculates how many basic line segments are there in the given character. Basic line segments like horizontal, vertical and diagonal are what looked for. Their length and area normalized to the entire image are used as features. The number of holes in a image is another feature. SImilar properties are used for feature extraction.

The algorithm used for this work is written and uploaded at

http://arxiv.org/abs/1202.3884

A set of sample images I used for feature extraction has been given here

http://www.mathworks.com/matlabcentral/fileexchange/26158-hand-drawn-sample-images-of-english-characters