Paper Title
Speak with Your Hands-AI Powered Hand Gesture Recognition

Abstract
Interpreting hand gestures is critical in developing accessible human-computer interaction systems, in particular to enable communication for people who are hearing impaired or speech impaired. In this research, we propose a real-time sign language translation system, which captures hand gestures and converts those gestures unto text. Our system uses a hybrid approach of computer vision and machine learning, using a combination of Media Pipe for detecting hand landmark positions in real-time, and a custom-trained machine learning model and rule-based logic for detecting gestures. The machine learning piece is trained on 3D coordinates of each of the hand landmarks whereas the rule-based logic allows for valid detection of at least certain pre-established gestures. The back- end of the system is built using Flask, while the front-end has an interactive web-based interface that utilizes a camera in real- time and displays translated gestures immediately. In order to test the efficacy of the system, a series of tests was performed using 11 variable complexity gestures pre-defined gestures. Overall, the results show very high recognition accuracy ranging from 80%-90% and low latencies under 100 milliseconds per frame, suggesting that it is applicable in real-time. This research advances gesture-based interaction systems, and presents a modular and scalable solution for future work. Keywords - Hand gesture recognition, Sign language translation, Media Pipe, Machine learning, Flask, Real-time systems, Human-computer interaction, Accessibility, Rule- based detection, Web interface.