IMPLEMENTATION OF REJECTION STRATEGIES INSIDE MALAYALAM CHARACTER RECOGNITION SYSTEM BASED ON RANDOM FOURIER FEATURES AND REGULARIZED LEAST SQUARE CLASSIFIER

Abstract

Robust and reliable recognition are indeed necessary requirements for optical character recognition systems. Distortions present in the document image and the pre-processing errors cause the optical character recognition system to apply rejection policies to achieve reliable recognition in computer assisted applications. The objective of this paper is to implement a robust and reliable character recognition system for Malayalam language. Random Fourier features classified with Regularized Least Square loss function based Regression classifier can approximate the non-linear kernel machines. Baseline Malayalam character recognition based on Random Fourier features and Regularized Least Square regression classifier is implemented in this paper. Up on this baseline character recognition system, rejection strategies are applied and are experimented with real world document images. An improvement in recognition accuracy is achieved with the simulated Malayalam character recognition system at the cost of rejecting character images having low classification score.

Keywords

• Character recognition
• Random Fourier features
• Regularized least square classifier
• Rejection approach
• Accuracy - rejection curve