Processamento colaborativo ARM-FPGA no cyclone V SoC estudo comparativo de arquiteturas com córtex-A9 e raspberry Pi 5 para reconhecimento de gestos

Abstract

Embedded systems play a central role invarious applications that require real-time processing, combining hardware and software to perform specific functions. In this con- text, hybrid architectures, which integrate diferente Technologies such as microcontrollers, FPGAs, and SoCs, stand out for their ability to combine processing flexibility and energy efficiency. An example of such architectures is the DE1-SoC development kit, which combines na ARM Cortex-A9 processor with a Cyclone VFPGA. This work presentes a comparative analysis of hybrid embedded system architectures Applied to digital image processing (DIP), using gesture recognition as a case study. Three main configurations were evaluated: an approach based exclusively on the Raspberry Pi5,a collaborative architecture between the Raspberry Pi5 and the Cyclone VFPGA, and a fully integrated solution using theDE1-SoC. The central objective was to investigate the advantages and limitations of each configuration in terms of response time, processing efficiency, and scalability. The results high lighted that the collaborative architecture between the Raspberry Pi5 and the FPGA achieved superior performance, with a reduction ofapproximately89.1%intotalexecutiontimecomparedtotheRaspberryPi 5 used independently.I ncontrast, the integrated solution with the Cortex-A9 showed an execution time about 29.8 times greater in data transfer, revealing bottle neck sin input and out put operations. The isolated RaspberryPi5,in turn, proved insufficient for real-time demands, high light ingits limitations in scenarios with strict deadlines. Based on these results, the work emphasizes the importance of architectural planningin embedded systems and suggests future research focused on optimizing communication protocols and improving operating system management for high-performance applications.