3d International ICST Conference on Pervasive Computing Technologies for Healthcare

Research Article

A theoretic algorithm for fall and motionless detection

Download647 downloads
  • @INPROCEEDINGS{10.4108/ICST.PERVASIVEHEALTH2009.6034,
        author={Shumei Zhang and Paul McCullagh and Chris Nugent and Huiru Zheng},
        title={A theoretic algorithm for fall and motionless detection},
        proceedings={3d International ICST Conference on Pervasive Computing Technologies for Healthcare},
        proceedings_a={PERVASIVEHEALTH},
        year={2009},
        month={8},
        keywords={Acceleration; Fall detection; Threshold; Phase angle; Motionless.},
        doi={10.4108/ICST.PERVASIVEHEALTH2009.6034}
    }
    
  • Shumei Zhang
    Paul McCullagh
    Chris Nugent
    Huiru Zheng
    Year: 2009
    A theoretic algorithm for fall and motionless detection
    PERVASIVEHEALTH
    ICST
    DOI: 10.4108/ICST.PERVASIVEHEALTH2009.6034
Shumei Zhang1,*, Paul McCullagh1, Chris Nugent1, Huiru Zheng1
  • 1: University of Ulster, Jordanstown, BT37 0QB,Northern Ireland, UK
*Contact email: zhang-s2@ulster.ac.uk

Abstract

robust method of fall and motionless detection is presented. The approach is able to detect falls and motionless periods (standing, sitting, and lying) using only one belt-worn kinematic sensor. The fall detection algorithm analyses the phase changes of vertical acceleration in relation to gravity and impact force using kinematic variables. A phase angle value was used as a threshold to distinguish between falls and normal motion activity. There are two advantages with this approach in comparison with existing approaches: (1) it is computationally efficient and theoretic (2) it is based on a single threshold value which was determined from a kinematic analysis for the falling processes. To evaluate the system, ten subjects were studied each of which performed different types of falls and motionless activities during a period of monitoring activity. These included: normal walking, standing, sitting, lying, a front bend of 90 degrees, tilt over 70 degrees and four kinds of falls (forward, backward, tilt left and right). The results show that 100% of heavy falling, 97% of all falls and 100% of motionless activity were correctly detected in a laboratory environment and the beginning and ends of these events were determined.