Feature Extraction Methods for Character Recognition

Feature extraction is an important stage in the recognition of characters particularly for handprinted characters. Based on a study of the extensive literature in handwritten character recognition, this review paper describes the commonly used feature extraction techniques using global features, distribution of points, geometrical and topological features, linguistic descriptions, use of contexts and fuzzy sets. The preliminary considerations like pre-processing operations, i.e., isolation, smoothing, thinning, normalization and centering are discussed first. Features of different characters like numerals, alphanumeric, FORTRAN, Katakana, Kanji and characters of some Indian languages (Devanagri, Bengali, Telugu and Tamil) are given. In the end, the scope of future research in the area is indicated.