<abstract xmlns="http://www.w3.org/1999/xhtml"><p> According to several studies, the European population is rapidly aging far over last years. It is therefore important to ensure that aging population is able to live independently without the support of working-age population. In accordance with the studies, fall is the most dangerous and frequent accident in the everyday life of aging population. In our paper, we present a system to track the human fall by a visual detection, i.e. using no wearable equipment. For this purpose, we used a Kinect sensor, which provides the human body position in the Cartesian coordinates. It is possible to directly capture a human body because the Kinect sensor has a depth and also an infrared camera. The first step in our research was to detect postures and classify the fall accident. We experimented and compared the selected machine learning methods including Naive Bayes, decision trees and SVM method to compare the performance in recognizing the human postures (standing, sitting and lying). The highest classification accuracy of over 93.3% was achieved by the decision tree method.</p></abstract>

According to several studies, the European population is rapidly aging far over last years. It is therefore important to ensure that aging population is able to live independently without the support of working-age population. In accordance with the studies, fall is the most dangerous and frequent accident in the everyday life of aging population. In our paper, we present a system to track the human fall by a visual detection, i.e. using no wearable equipment. For this purpose, we used a Kinect sensor, which provides the human body position in the Cartesian coordinates. It is possible to directly capture a human body because the Kinect sensor has a depth and also an infrared camera. The first step in our research was to detect postures and classify the fall accident. We experimented and compared the selected machine learning methods including Naive Bayes, decision trees and SVM method to compare the performance in recognizing the human postures (standing, sitting and lying). The highest classification accuracy of over 93.3% was achieved by the decision tree method.

Comparison of Machine Learning Methods for the Purpose Of Human Fall Detection

Slovak University of Technology, Faculty of Materials Science and Technology in Trnava, Institute of Applied Informatics, Automation and Mathematics, Hajdóczyho 1, 917 24 Trnava, Slovak Republic

Research Papers Faculty of Materials Science and Technology Slovak University of Technology

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 3.0 License.

{"article-title":"Comparison of Machine Learning Methods for the Purpose Of Human Fall Detection"}

According to several studies, the European population is rapidly aging far over last years. It is therefore important to ensure that aging population is able...

Comparison of Machine Learning Methods for the Purpose Of Human Fall Detection

Online veröffentlicht: 06. Feb. 2015

Seitenbereich: 69 - 76

DOI: https://doi.org/10.2478/rput-2014-0037

Schlüsselwörter
fall detection, human fall, comparison of machine learning methods, Kinect camera

© by Maximilián Strémy

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 3.0 License.

Comparison of Machine Learning Methods for the Purpose Of Human Fall Detection

Online veröffentlicht: 06. Feb. 2015

Seitenbereich: 69 - 76

DOI: https://doi.org/10.2478/rput-2014-0037

Schlüsselwörterfall detection, human fall, comparison of machine learning methods, Kinect camera

© by Maximilián Strémy

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 3.0 License.

Schlüsselwörter
fall detection, human fall, comparison of machine learning methods, Kinect camera