Abstract
The World Health Organization (WHO) reports that most common cause of injuries to elderly people increases every year due to fall events. Human fall events are one of the most important health problem among the elderly people whos aged 65 and above, which could probably result in a significant barrier to their independent living. This chapter presents a method for fall activity detection based on Motion Projection Profile (MPP) features extracted from temporal difference image to represent a various levels of a person's posture. Falls are detected by analyzing the projection profile features consist of the measure of motion pixel of each row, column, left diagonal and right diagonal of the temporal difference image and gives adequate information to recognize the instantaneous posture of the person. The experiments are carried out using publicly available fall detection dataset and the extracted MPP feature set are modeled by the various machine learning methods like Support Vector Machine (SVM) with polynomial kernel, SVM with Radial Basis Function (RBF) kernel, K-Nearest Neighbor (KNN) and the Decision tree (J48) algorithm are used to classify the fall activies. Experimental results show that SVM with RBF kernel is an efficient to recognize the fall activity with an overall recognition accuracy of 89.55% on the fall detection dataset, which outperforms other machine learning methods.