Hardware/software co-design for accelerating human action recognition

Abstract

Action recognition is gaining more importance and attention due to its wide applications. But the need of implementing it in real time is essential. This paper suggests a hardware/software co-design which accelerates the whole recognizing function. In this setup we are using Microsoft© Kinect device, Intel© Atom system and FPGA based SoC device from Altera©. The initial experiments show that it is possible to recognize actions in real time using the proposed setup by processing the skeleton data. This promising result would encourage designers to take the step forward of using mixed CPU and FPGA based SoC platform to increase the performance of image and video processing operations especially in recognition systems.