A Hybrid Approach to Sign Language Recognition Using MediaPipe and Machine Learning

This paper provides an efficient and accurate sign language recognition system in real time that understands gestures employed using MediaPipe and Random Forest in American Sign Language (ASL). The system captures and processes hand gestures in real time, allowing for optimized responses on the trained dataset with the extended feature for a user to create its own database in just a few clicks. MediaPipe framework is designed to provide powerful hand tracking abilities and is used to capture and refine landmarks which in turn contributes in improving gesture recognition accuracy. The processed data which is captured with OpenCV library is fed into the model trained on the dataset created by authors resulting in 99.81% accuracy in recognizing hand gestures. The findings emphasize the system’s potential applications in communication, education, and accessibility, emphasizing its importance for people who have hearing impairments. The combination of MediaPipe with Random Forest and CNN provides a practical solution for a precise real-time sign language detection with a user-friendly model.