FPGA-based implementation of a dynamic hand gesture recognition system
Yaw‐Ying Tsai, Yu-Fan Lai, Chao Xu, S. -J. Ruan · IET conference proceedings. · 2024
This paper presents a low-power dynamic hand gesture recognition system based on an FPGA platform. The system consists of two main components: hand tracking and gesture recognition. A concatenation of image processing methods is proposed for hand tracking to achieve accurate tracking with minimal computational resources. Gesture recognition is performed using a Convolutional Neural Network (CNN) model trained on the NIST dataset. The implementation is done on the PYNQ-Z2 board, utilizing its Python packages to efficiently program image processing algorithms. The power consumption of the system is measured at 2.15 W, which is lower than similar FPGA-based implementations. This research demonstrates the feasibility of achieving low power consumption in hand gesture recognition systems while maintaining high accuracy and performance.