A Support Vector Machine based pedestrian recognition system on resource-limited hardware architectures
Alessandro Ghio, Stefano Pischiutta · 2007
We present here a hardware-friendly structure for the support vector machine (SVM), useful to implement its feedforward phase on resource limited devices, such as field programmable gate arrays (FPGAs), on which a floating-point unit is seldom available. We tested our proposal using an artificial machine-vision benchmark dataset for automotive applications.