<![CDATA[Medication misidentification remains a safety risk for visually impaired and low-vision users who manage medicines independently. This systematic literature review synthesizes eligible studies on real-time medication recognition systems on real-time medication recognition systems that combine deep learning (DL), optical character recognition (OCR), and accessible feedback mechanisms. Following PRISMA 2020, searches were conducted in IEEE Xplore, Scopus, PubMed, Web of Science, ACM Digital Library, and ScienceDirect for studies published from 2019 to July 2025. Fifty studies met the eligibility criteria. Of these, 30 reported quantitative accuracy and/or latency metrics suitable for comparative extraction. The evidence was synthesized across platform type, recognition technique, evaluation dimension, multimodal integration, and deployment feasibility. Reported accuracies ranged from approximately 70% to 99%, but direct comparison remains limited by differences in datasets, number of medicine classes, image conditions, and metric definitions. Latency evidence indicates that several mobile and embedded systems support sub-second to near-real-time feedback, while cloud-assisted systems may require longer response times. Mobile, wearable, embedded, and multimodal approaches show complementary strengths, yet persistent gaps remain in public datasets, real-world validation with visually impaired users, robust multilingual OCR, transparent quality reporting, and healthcare-system integration. This review contributes a structured synthesis of DL-OCR techniques, accessibility features, and deployment constraints for assistive medication recognition.]]>
Copyrights © 2026
