A Review of Hybrid Replica of Character Recognition Practices

Abstract

Linguistic resources such as shape extraction are critical for the development of language technologies such as script and speech recognition. The motivation behind this paper is to locate and implement the knowledge informants capable of inferring the shape of character object and recognizing it into appropriate character class through Simplified Fuzzy Adaptive Resonance Theory Map. Various technical and implementation issues regarding the selected set of features, tolerance to scale, distortion and style variation have been explored. A blend of font and size independent statistical methodologies such as Directional Distribution, Zernike Moments and Geometrical Moments have been presented to provide the one integrated feature vector to be fed as input to Fuzzy network. This review presents a modified approach in terms of scale and font invariance properties as the classifier is trained with well designed database of nine various fonts of each character with three varying sizes. The experimental results show that the proposed approach is fast and results in a higher performance regarding scale, rotation and font invariance properties.