Visual human action classification for control of a passive walker

Research output: Chapter in Book/Report/Conference proceedingConference contribution

2 Citations (Scopus)

Abstract

Human action/behavior classification plays an important role for controlling systems having interaction with human users. Safety and dependability of such systems are crucial especially for walking assist systems. In this paper, upper body joint model of a user of a walking assist system is extracted using a depth sensor and a probabilistic model is proposed to detect possible non-walking states that might happen to the user. The 3D model of upper body skeleton, is reduced in dimension by applying Principal Component Analysis (PCA). The principal components are tested to have a normal distribution allowing a multivariate normal distribution fitting for walking data. The model is shown to be capable of recognizing four different falling scenarios and sitting. In these non-walking states, the motion of a passive-type walker called 'RT Walker', is controlled by generating brake force to assure fall prevention and sitting/standing up support. The experimental data is gathered from an experienced physical therapist capable of imitating different walking problems.

Original languageEnglish
Title of host publication2017 7th International Conference on Modeling, Simulation, and Applied Optimization, ICMSAO 2017
PublisherInstitute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)9781509054541
DOIs
Publication statusPublished - 2017 May 26
Event7th International Conference on Modeling, Simulation, and Applied Optimization, ICMSAO 2017 - Sharjah, United Arab Emirates
Duration: 2017 Apr 42017 Apr 6

Other

Other7th International Conference on Modeling, Simulation, and Applied Optimization, ICMSAO 2017
CountryUnited Arab Emirates
CitySharjah
Period17/4/417/4/6

Keywords

  • Fall detection
  • PCA
  • Visual classification
  • Walker robot

ASJC Scopus subject areas

  • Computer Networks and Communications
  • Applied Mathematics
  • Control and Optimization
  • Modelling and Simulation

Fingerprint Dive into the research topics of 'Visual human action classification for control of a passive walker'. Together they form a unique fingerprint.

  • Cite this

    Taghvaei, S., Hirata, Y., & Kosuge, K. (2017). Visual human action classification for control of a passive walker. In 2017 7th International Conference on Modeling, Simulation, and Applied Optimization, ICMSAO 2017 [7934895] Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/ICMSAO.2017.7934895