Design of hand skeleton extraction accelerator for a real-time hand gesture recognition

Abstract

Applications such as automobiles, robots and games require a real-time operation in embedded systems. However, since the accurate hand gesture recognition requires a large amount of computation, it is difficult a real-time operation. In this paper, we propose a hand skeleton extraction accelerator for real-time hand gesture recognition. We analyze the hand gesture recognition algorithm to find the parts with high computational complexity and determine which routines that are difficult a real-time operation. And the hardware accelerator is implemented using HLS method for embedded system. Implemented hand skeleton extraction accelerator circuit was tested its operation using Xilinx's Zynq-7000 FPGA (XC7Z020) device. Our circuit operates in real-time in an embedded system and recognition success rates is 86.8%.