An enhanced handwriting recognition tool for the visually impaired

• Main Authors: Huzaimi, Muhammad Zikry, Mohd Ramli, Huda Adibah, Saidin, Norazlina
• Format: Proceeding Paper
• Language: English; English
• Published: IEEE 2024
• Subjects: TK7800 Electronics. Computer engineering. Computer hardware. Photoelectronic devices; TK7885 Computer engineering
• Online Access: http://irep.iium.edu.my/115621/7/115621_An%20enhanced%20handwriting.pdf; http://irep.iium.edu.my/115621/8/115621_An%20enhanced%20handwriting_Scopus.pdf; http://irep.iium.edu.my/115621/; https://ieeexplore.ieee.org/document/10652433; https://doi.org/10.1109/ICOM61675.2024.10652433

Handwritten text serves as an essential means of conveying ideas and messages. It is often characterized by diverse handwriting styles, variations in character shapes, as well as the presence of overlapping strokes and characters. However, for visually impaired individuals, this poses significant hurdles as existing recognition tools may not reliably provide accurate information. To address this, an enhanced handwriting recognition tool powered by Optical Character Recognition (OCR) is proposed. This tool integrates a Raspberry Pi microcontroller and a camera module for image capture, along with a text-to speech engine to empower the visually impaired. Moreover, the tool employs Artificial Neural Network (ANN) and a hybrid Artificial Neural Network + Hidden Markov Model (ANN+HMM) classification methods to enhance recognition performances. In addition to the functionality test, a series of accuracy and recall rate tests for different handwriting styles was conducted to assess the tool's performance. The results demonstrated the superiority of the hybrid ANN+HMM model over the standalone ANN, achieving an impressive 46.3% improvement in accuracy and a perfect 100% recall rate, particularly for cursive handwriting. This groundbreaking innovation contributes to fostering a more inclusive and accessible world for all.