<p>Braille is a tactile writing system that enables blind and visually impaired individuals to access literature, education, and communication. Braille character recognition, however, remains challenging due to input quality variations, complex patterns, and language-specific adaptations. This paper introduces a novel partitioning framework that systematically decomposes Braille cells into structured subregions, combined with a partial-occupancy feature extraction method that reduces class complexity and enhances discriminability. To support both blind users, through automatic Braille-to-text conversion, and sighted users who need to interpret Braille without expertise, the proposed system aims to deliver robust, lightweight recognition across diverse conditions. Evaluated on cross-lingual datasets including the DSBI (Chinese Braille with overlapping dots) and Angelina (Russian Braille with real-world distortions), the method achieves accuracy gains of up to 99% compared to conventional classifiers. These improvements demonstrate the system’s enhanced robustness to noise, dot deformation, and variations in illumination. Overall, the results confirm that the proposed partitioning and feature-encoding strategy significantly improves recognition accuracy and adaptability across heterogeneous Braille datasets.</p>