Towards real-time embeded system design for video capturing and privacy protection

摘要

Real-time video processing is a challenge in many embedded system applications. They can be general-purpose processor based (software) or Custom single purpose processor based (hardware). Software processor based rely mostly on real-time operating system layer to allow multiple software threads to run concurrently. Hardware processor based don't need these layers because they are not executing any programs, but rather operate in finite state machines. Video Capturing, Color space conversion, and Privacy protection are considered in this article as three fundamental components for effective and secure real-time video streaming system. The proposed model is based on concurrent finite state machines which is inherently concurrent. After careful verification, the proposed system is compared with a base point system. The base point system is built using a soft-core Nios II processor on a reconfigurable hardware, i.e., FPGA chip, for performance evaluations. Then a Custom Sequential Machine is designed and modeled in Verilog HDL, which utilizes integer calculations and has concurrent interfaces to input and output memory devices for synchronization purposes. The Custom Sequential Machine design uses integer approximation rather than floating point for color space conversion. A speed-up factor of 4 orders of magnitude higher than the base point system for color space conversion only. The base point system was not powerful enough to be serve real-time video capturing, processing and privacy protection, but was useful for illustration of the need of proposed system model.