NON LINEAR PATTERN RECOGNITION USING VOLTERRA SERIES AND HOG

Abstract

This master thesis addresses issues related to non linear pattern recognition. More specifically we are interested in various non linear patterns from the one dimensional images for example recognition of analog signals, detection can be done in two dimensional also in case of human beings poses. In this thesis feature extraction vector is used to detect the non linear patterns for example HOG (Histogram of oriented gradients) and volterra series. As in our daily life there are various types of non linear patterns and recognition of those non linear patterns will be essential. For example in power quality event detection, signal detection by robotic machines.