Military Hand signals are a method of visual communication for field use and are now the most common forms of communication during operations. It is important that communications in missions are clear, distinct and understandable. Hand Sign Language (SL), have become a standard part of communication in the military, especially when voice communication is not desirable or silence has to be maintained for security. With the advent of drones and high-precision cameras, communication using hand signals can be very effective, as it is quick and the support device does not need to be in proximity. A trained human operator may be required to translate the meaning of these hand signals. Instead, a machine learning algorithm can be developed to recognize and translate these messages, saving time for translation and the requirement of specialized training for the interpreters. This project aims to create a method that can translate army hand signals into their respective labels, which can be used by drones for reconnaissance and safety. Our model, which is based on Convolution Neural Networks, achieves a cross-validation accuracy of 98.32% and test accuracy of 97.94%.
[Dataset and Model will be released soon]
To be presented at Springer's International Virtual Conference on Industry (IVCI)