Braille recognition and text-to-speech (TTS) systems enhance accessibility for visually impaired individuals by converting tactile inputs into digital and spoken formats. This survey analyzes over 50 studies on image processing, deep learning (e.g., CNNs), and speech synthesis models like Tacotron and LPCNet. While progress has been made, key challenges persist—such as high costs, limited accuracy in noisy settings, portability issues, and the scarcity of annotated multilingual datasets. The paper highlights current trends, including hybrid models, real-time processing, and wearable-based solutions. It concludes with future directions focused on lightweight, adaptive, and cost-effective systems to improve accessibility in underserved regions.