Reading Characters for Blind and Visually Impaired People

Books and documents are key sources of knowledge, but access is limited for visually impaired or blind individuals who cannot read. This study proposes a Character Recognition System (CRS) to improve reading accessibility for such individuals using advanced technology.A new dataset, Printed English Characters (PEC), was created, comprising 5400 images (100 samples for each character: lowercase letters, uppercase letters, and spaces). The images undergo preprocessing operations such as binarization, segmentation, and resizing before being input into a Deep Convolutional Neural Network (DCNN).The DCNN, composed of multiple sub-networks and layers, demonstrates high performance, achieving a recognition accuracy of 99.57%, along with excellent precision, sensitivity, specificity, and F1-score. The system integrates hardware (webcam, PC, sound amplifier) and software for image processing and deep learning, enabling it to acquire, recognize, and audibly read characters and words.By recognizing space characters, the system constructs words and reads them aloud, offering significant assistance to blind and visually impaired individuals in accessing textual information.