Fall Detection Using Depth Videos

Please enable javascript in your browser.

Hosseinzadeh, Matin | 2017

773 Viewed

Type of Document: M.Sc. Thesis
Language: Farsi
Document No: 49487 (05)
University: Sharif University of Technology
Department: Electrical Engineering
Advisor(s): Vosoughi-Vahdat, Bijan
Abstract:
Falls are one of the major causes leading to injury of elderly people. Using wearable devices for fall detection has a high cost and may cause inconvenience to the daily lives of the elderly. In this project, we present an automated fall detection approach that requires only a low-cost depth camera. Our approach combines two computer vision techniques, spatio-temporal fall characterization and a learning-based classifier to distinguish falls from other daily actions. A dense set of spatio-temporal feature vectors are computed from video to provide a localized description of the action, and subsequently aggregated in an empirical covariance matrix to compactly represent the action. Then, we utilize an improved extreme learning machine (ELM) classifier to identify fall from those of other actions. We used our method on an action dataset that consists of six types of actions (falling, bending, sitting, squatting, walking, and lying) from twenty subjects. Experimenting with the dataset shows that our approach can achieve up to 93% accuracy on two-class classification, and 86% accuracy on six-class classification of fall event
Keywords:
Depth Images ; Computer Vision ; Falling Detection ; Kinect Sensor ; Eldery Care