A Multi-Format Document-to-Speech Reader on Raspberry Pi for Accessible Digital Learning

Abstract

Blind learners frequently receive educational materials in heterogeneous digital formats, yet accessing those files can require a conventional computer, a screen reader, and multiple navigation steps. This paper presents a Raspberry Pi-based document-to-speech reader that directly processes TXT, PDF, DOCX, and PPTX files from removable storage through a compact Braille-oriented key interface. The Python implementation separates file navigation, format-specific text extraction, segmentation, keyword filtering, presentation control, file transfer, and speech synthesis into functional modules. Candidate extraction libraries were assessed with samples containing formatted prose, hyperlinks, Greek characters, equations, images, a resume layout, and presentation slides. The comparison showed that docx2python retained the broadest range of DOCX content, including hyperlink and media markers, although equation conversion was imperfect. PyPDF2 produced the most faithful mathematical-text output among the tested PDF libraries, while python-pptx extracted the evaluated slide text without observed errors. Google Text-to-Speech was selected because playback could be interrupted within the implemented interaction flow. Functional verification confirmed file selection, sequential reading, keyword search, slide navigation, and file relocation using the assigned keys. The prototype demonstrates a practical architecture for portable access to common learning documents, while internet dependence, legacy DOC/PPT incompatibility, mathematical-symbol pronunciation, and conversion delay remain priorities for further development.