Implementation of gesture recognition technology optimised by neural networks in OpenMV

Sign language serves as a vital communication medium for the deaf and hard-of-hearing community; yet, existing gesture recognition systems face challenges such as high costs, limited accuracy in complex environments, and hardware dependence. This study presents a novel gesture recognition system leveraging the OpenMV platform, TensorFlow, and EdgeImpulse to address these issues. The proposed system achieves real-time translation of gestures into text with an accuracy exceeding 98%, demonstrating robustness in varying lighting and background conditions. By integrating machine vision capabilities with cost-effective hardware, this system overcomes the limitations of prior methods, such as the reliance on expensive equipment and poor adaptability to real-world scenarios. These findings highlight the system's potential for widespread application in assistive technologies, offering an affordable and efficient solution for improving communication accessibility.