Optical Braille Recognition using Circular Hough Transform
This work addresses a domain-specific problem for visually impaired individuals by providing an incremental improvement in Optical Braille Recognition.
The paper tackles the problem of converting Braille documents to natural language to bridge communication gaps for blind students, achieving an accuracy of 98.71% on a dataset of 54 Braille scripts using a smartphone-based method.
Braille has empowered visually challenged community to read and write. But at the same time, it has created a gap due to widespread inability of non-Braille users to understand Braille scripts. This gap has fuelled researchers to propose Optical Braille Recognition techniques to convert Braille documents to natural language. The main motivation of this work is to cement the communication gap at academic institutions by translating personal documents of blind students. This has been accomplished by proposing an economical and effective technique which digitizes Braille documents using a smartphone camera. For any given Braille image, a dot detection mechanism based on Hough transform is proposed which is invariant to skewness, noise and other deterrents. The detected dots are then clustered into Braille cells using distance-based clustering algorithm. In succession, the standard physical parameters of each Braille cells are estimated for feature extraction and classification as natural language characters. The comprehensive evaluation of this technique on the proposed dataset of 54 Braille scripts has yielded into accuracy of 98.71%.