On-chip Object Recognition System Using Random Forests

This paper presents how hardware-based machine learning models can be design for the task of object recognition. The process is composed of automatic representation of objects as covariance matrices follow by a machine learning detector based on random forest (RF) that operate in online mode. First, in more general terms, the problems of in-accuracy, limited precision, and robustness are treated. Then describe the algorithmic and architecture of our digital random forest (RF) classifier employing logarithmic number systems (LNS), comprises of several computation modules, referred to as 'covariance matrices', 'tree units', 'majority vote unit', and 'forest units'. experiments of the object recognition are provided to verify the effectiveness of the proposed approach, which is optimized towards a system-on-chip (Soc) platform implementation. Results demonstrate that the propose model yields in object recognition performance comparable to the benchmark standard RF, AdaBoost, and SVM classifiers, while allow fair comparisons between the precision requirements in LNS and of using traditional floating-point.