Sign Language Detection Using Deep Learning

To provide equal opportunity in society, communication is an important aspect. The goal of this real-time American Sign Language (ASL) static alphabet detection system based on AI is to offer enhanced accessibility for the hard-of-hearing community by accurate detection of the static alphabet of ASL. Our approach utilizes the Mediapipe framework to detect 21 landmarks, producing a 63 -dimensional feature vector, which is passed through a custom Multi-Layer Perceptron(MLP) model. As experimental results show, the system is a useful device for real applications since it is more precise and has less latency. Our approach completes the communication gap by using automatic sign recognition to make the space more accessible. Future research will add dynamic gesture detection to the framework, which can further make natural human-computer interaction possible.