Implementation of the Extended Gaussian Mixture Model Algorithm for High Definition Video

Abstract—Background subtraction (BGS) and background identification algorithms can perform quite fast, these are used in many real time video processing systems to identify the moving objects. But they are not robust enough to be used in various situations. So, to detect such moving objects from the complex scene, a method called Gaussian Mixture Model (GMM), a background identification algorithm is used. In this paper, first the performance of GMM is analyzed and implemented in field-programmable gate-array (FPGA). Secondly, it is implemented in GPDK-90nm CMOS standard cell technology in two various forms. The first form is to reduce energy by constant voltage scaling and second form is to reduce silicon area utilization by using folding technique.
Index Terms—Gaussian mixture model (GMM), Background subtraction,field programmable gate-array (FPGAs), application specific integrated circuits (ASICs) and moving object detection.