Project DetailsHand gesture recognition facilitates intuitive, hands-free human-computer interaction in applications such as smart environments, healthcare, and virtual reality. This study explores the use of Ultra-Wideband (UWB) radar signals for classifying dynamic hand gestures, leveraging UWB’s fine range resolution, low power consumption, and capability to detect close-range targets. To enhance classification accuracy, we apply clutter removal and systematic data augmentation as preprocessing steps. Several recent deep neural networks are evaluated and compared, with the most effective model selected for final implementation. Experimental results demonstrate that advancements in preprocessing and neural network architectures enable a reduction from three radar units to one while maintaining performance, along with improved classification accuracy compared to prior studies.
