LAUSR

Detection of intruders and predicting their activities are the first and foremost needs of surveillance. An embedded system employing geophone, adaptive event extraction, and robust machine learning algorithms have made it possible not only to detect the presence of a potentially harmful intruder but also to predict to a high degree of accuracy, his state of motion, and to take counter action at the earliest. This paper aims to be an in-depth study of this simple yet effective technique of intruder detection and its subsequent predictive analysis of motion which should come as a handy aid to security solutions all-around. The proposed event extraction technique detects footfall events and extracts portions of the signal that correspond to an event. Using the classifier SVM-RBF and the proposed event extraction technique the presence of an intruder can be predicted with an accuracy of 86% from a signal of length 2 s and its state of motion with an accuracy of 77% from a signal of length 6 s.