Accurate Medicine Recognition and Audio Conversion for Handwritten Prescriptions

Even in today's digital age, many doctors continue to write prescriptions by hand, which could be misread by the patient or the chemist. As a result, accurate medical recognition is required for prescriptions. This research proposes a system for accurately recognizing medicines on handwritten prescriptions and converting them into audio format, benefiting visually impaired individuals. The system incorporates advanced techniques in image preprocessing, feature extraction, and machine learning, utilizing YOLOv4 model for prescription region identification and CRNN-CTC model architecture for medicine name and strength recognition. The model achieves a high accuracy rate of approximately 98 percent and a character error rate of 0.0029. By streamlining pharmacy workflows and improving medication safety, this system empowers individuals to efficiently manage their prescriptions while ensuring inclusive healthcare practices.