CONVOLUTIONAL NEURAL NETWORK (CNN) BASED APPROACH FOR BRAILLE BLOCK RECOGNITION WITH IMPROVED NOISE REDUCTION

Abstract

This paper proposes a Convolutional Neural Network (CNN) based approach for Braille block recognition with a focus on improved noise reduction. Braille images are susceptible to various noise sources that can hinder character recognition accuracy. This work introduces a novel denoising pre-processing stage utilizing a dedicated CNN architecture designed to remove noise while preserving Braille cell features. The pre-processed images are then fed into a separate CNN for Braille block recognition [1]. The proposed approach is evaluated on a publicly available Braille image dataset and compared with existing methods. The results demonstrate that the denoising pre-processing stage significantly improves Braille block recognition accuracy compared to traditional pre-processing techniques and other CNN-based approaches [2].