Abstract
This project proposes a real-time system for American Sign Language (ASL) recognition, utilizing webcam-based video capture, computer vision, and machine learning techniques. By detecting and interpreting ASL gestures captured from a laptop's webcam, the system translates gestures into text or speech, enhancing communication accessibility. Key components include real-time video processing using OpenCV, hand and gesture detection algorithms, and a machine learning model for gesture classification. The project aims for high accuracy by training on a comprehensive ASL gesture dataset and addressing challenges such as background noise, lighting variations, and real-time processing requirements. The system is designed to recognize single-handed ASL gestures initially, with plans to extend support for two-handed gestures in future iterations. Leveraging Python for prototyping and TensorFlow for machine learning, the project incorporates state-of-the-art models to ensure reliable and scalable performance. By focusing on modular design and adaptability, this project also provides a framework that can be expanded to support other sign languages or integrated into broader accessibility platforms.
The resulting ASL recognition tool holds potential applications for aiding the deaf and hard-of-hearing community, as well as broader applications in accessibility technologies. Through this work, we aim to contribute to advancements in human-computer interaction and accessible communication, laying the groundwork for further development and scalability in ASL recognition systems.