Unicode OCR using Artificial Neural Network methods

This project will recognizes typed latin optical characters and outputs the corresponding Unicode value and character. It is implemented using Artificial Neural Network methods and can be trained for any available font type.

The central objective of this project is demonstrating the capabilities of Artificial Neural Network implementations in recognizing extended sets of optical language symbols. The applications of this technique range from document digitizing and preservation to handwritten text recognition in handheld devices.

The classic difficulty of being able to correctly recognize even typed optical language symbols is the complex irregularity among pictorial representations of the same character due to variations in fonts, styles and size. This irregularity undoubtedly widens when one deals with handwritten characters.

Hence the conventional programming methods of mapping symbol images into matrices, analyzing pixel and/or vector data and trying to decide which symbol corresponds to which character would yield little or no realistic results. Clearly the needed methodology will be one that can detect €Ëproximity€„¢ of graphic representations to known symbols and make decisions based on this proximity. To implement such proximity algorithms in the conventional programming one needs to write endless code, one for each type of possible irregularity or deviation from the assumed output either in terms of pixel or vector parameters, clearly not a realistic fare.

An emerging technique in this particular application area is the use of Artificial Neural Network implementations with networks employing specific guides (learning rules) to update the links (weights) between their nodes. Such networks can be fed the data from the graphic analysis of the input picture and trained to output characters in one or another form. Specifically some network models use a set of desired outputs to compare with the output and compute an error to make use of in adjusting their weights. Such learning rules are termed as Supervised Learning.

One such network with supervised learning rule is the Multi-Layer Perceptron (MLP) model. It uses the Generalized Delta Learning Rule for adjusting its weights and can be trained for a set of input/desired output values in a number of iterations. The very nature of this particular model is that it will force the output to one of nearby values if a variation of input is fed to the network that it is not trained for, thus solving the proximity issue. Both concepts will be discussed in the introduction part of this report.

The project has employed the MLP technique mentioned and excellent results were obtained for a number of widely used font types. The technical approach followed in processing input images, detecting graphic symbols, analyzing and mapping the symbols and training the network for a set of desired Unicode characters corresponding to the input images are discussed in the subsequent sections. Even though the implementation might have some limitations in terms of functionality and robustness, the researcher is confident that it fully serves the purpose of addressing the desired objectives.