Fly-LeNet: A deep learning-based framework for converting multilingual braille images.

Area of Science:

• Computer Science
• Artificial Intelligence
• Assistive Technology

Background:

• Braille-assistive technologies improve accessibility for blind individuals.
• Current Optical Braille Recognition (OBR) systems lack multilingual conversion capabilities.
• This limits sighted individuals' ability to learn braille for self-study.

Purpose of the Study:

• To develop a novel system for converting braille images into multilingual text.
• To address the limitations of existing OBR systems in multilingual braille translation.
• To facilitate braille learning and communication for sighted individuals.

Main Methods:

• A segmentation and deep learning approach named Fly-LeNet was proposed.
• The method incorporates image acquisition, preprocessing, and segmentation using Mayfly optimization and thresholding.
• A braille multilingual mapping step and the LeNet-5 deep learning model for braille cell recognition are utilized.

Main Results:

• Fly-LeNet achieved high classification accuracies of 99.77% and 99.80% on two distinct braille image datasets.
• Experiments were conducted on datasets containing braille signs for alphabets, numbers, and punctuation.
• The model demonstrated robust performance in recognizing braille cells.

Conclusions:

• The Fly-LeNet system shows significant potential for multilingual braille transformation.
• This technology can enhance communication between blind and sighted individuals.
• It offers a valuable tool for self-learning braille and promoting inclusivity.