A preliminary study for remote healthcare system: Activity classification for elder people with on body sensors

Kocak S., Artug T., Tulum G.

6th International Conference on Control Engineering and Information Technology, CEIT 2018, İstanbul, Turkey, 25 - 27 October 2018, (Full Text) identifier identifier

  • Publication Type: Conference Paper / Full Text
  • Doi Number: 10.1109/ceit.2018.8751859
  • City: İstanbul
  • Country: Turkey
  • Keywords: Classification, Feature selection, K-NN, MLP, On body sensors, PCA, RBF, ReliefF, SVM
  • Yıldız Technical University Affiliated: No

Abstract

Development of intelligent care system for elder people have been investigated in recent years. In this study, to detect emergency situations for elder people, activity classification was aimed using on body sensor data. Multi-layer perceptron, radial basis function networks, k- nearest neighbor and support vector machines were used in classification. In feature selection process principal component analysis and ReliefF were used. Accuracy of classification was above 85% for every classifier and the best performance was acquired with 3-NN with 99.8% accuracy. When feature selection was applied 5- NN was showed the highest performance with 99.4%. This study shows that it is possible to develop remote care system by using sensors and classifiers for a more secure life for elder people.