Indoor human fall detection system based on automatic vision using computer vision and machine learning algorithms

The detection of unnatural falls among older adults living alone at home is an essential requirement of the day. Injury after a fall is common, and immediate medication is often necessary. A fall detection system using the following features such as orientation angle, aspect ratio, Motion History Image and object below the threshold line is proposed. The combination of these features can lead to a robust fall detection system. These features are used as an input for fall detection systems using machine-learning techniques such as Support Vector Machine, K-Nearest Neighbors, Stochastic Gradient Descent, Decision Tree and Gradient Boosting. The effect of using different classifiers and feature sets for the performance of fall detection is observed. This approach provided a promising F-measure value of 96% for Decision Tree algorithm on Coffee room environment