Integration of Extended Kalman filter and neural network for Foot-mounted Pedestrian Navigation

Guang Je Tsai, Yuan Rong Sih, Hone-Jay Chu, Kai-Wei Chiang

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

Abstract

Pedestrian Navigation with a Foot- mounted sensor is a common operation in indoor environments. The contribution in this paper is the step length estimation. The step lengths vary continuously according to the walking speed and walking frequency. In the general solution, the step length is modelled as a linear function in the regular gait in previous work, but is not applicable in the varied gait, thus we try to solve this nonlinear problem by using neural network. Furthermore, the estimation of the position of a person based on inertial sensors in Pedestrian Dead-Reckoning (PDR) solution greatly relies on heading calculation, thus, the heading drift reduction process is necessary. In this paper, we replace the compass with vision aiding, and also using other algorithms such like HDR, ZARU, vision-based and ZUPT to reduce the drift in Extended Kalman Filter (EKF) framework. Finally, we display the performance including each algorithms of the reduction methods and neural network.

Original languageEnglish
Title of host publication34th Asian Conference on Remote Sensing 2013, ACRS 2013
PublisherAsian Association on Remote Sensing
Pages4442-4447
Number of pages6
Volume5
ISBN (Print)9781629939100
Publication statusPublished - 2013
Event34th Asian Conference on Remote Sensing 2013, ACRS 2013 - Bali, Indonesia
Duration: 2013 Oct 202013 Oct 24

Other

Other34th Asian Conference on Remote Sensing 2013, ACRS 2013
CountryIndonesia
CityBali
Period13-10-2013-10-24

Fingerprint

Extended Kalman filters
Navigation
Neural networks
Sensors

All Science Journal Classification (ASJC) codes

  • Computer Networks and Communications

Cite this

Tsai, G. J., Sih, Y. R., Chu, H-J., & Chiang, K-W. (2013). Integration of Extended Kalman filter and neural network for Foot-mounted Pedestrian Navigation. In 34th Asian Conference on Remote Sensing 2013, ACRS 2013 (Vol. 5, pp. 4442-4447). Asian Association on Remote Sensing.
Tsai, Guang Je ; Sih, Yuan Rong ; Chu, Hone-Jay ; Chiang, Kai-Wei. / Integration of Extended Kalman filter and neural network for Foot-mounted Pedestrian Navigation. 34th Asian Conference on Remote Sensing 2013, ACRS 2013. Vol. 5 Asian Association on Remote Sensing, 2013. pp. 4442-4447
@inproceedings{0bf5636cec87414ea6cebcd1ba25ecce,
title = "Integration of Extended Kalman filter and neural network for Foot-mounted Pedestrian Navigation",
abstract = "Pedestrian Navigation with a Foot- mounted sensor is a common operation in indoor environments. The contribution in this paper is the step length estimation. The step lengths vary continuously according to the walking speed and walking frequency. In the general solution, the step length is modelled as a linear function in the regular gait in previous work, but is not applicable in the varied gait, thus we try to solve this nonlinear problem by using neural network. Furthermore, the estimation of the position of a person based on inertial sensors in Pedestrian Dead-Reckoning (PDR) solution greatly relies on heading calculation, thus, the heading drift reduction process is necessary. In this paper, we replace the compass with vision aiding, and also using other algorithms such like HDR, ZARU, vision-based and ZUPT to reduce the drift in Extended Kalman Filter (EKF) framework. Finally, we display the performance including each algorithms of the reduction methods and neural network.",
author = "Tsai, {Guang Je} and Sih, {Yuan Rong} and Hone-Jay Chu and Kai-Wei Chiang",
year = "2013",
language = "English",
isbn = "9781629939100",
volume = "5",
pages = "4442--4447",
booktitle = "34th Asian Conference on Remote Sensing 2013, ACRS 2013",
publisher = "Asian Association on Remote Sensing",

}

Tsai, GJ, Sih, YR, Chu, H-J & Chiang, K-W 2013, Integration of Extended Kalman filter and neural network for Foot-mounted Pedestrian Navigation. in 34th Asian Conference on Remote Sensing 2013, ACRS 2013. vol. 5, Asian Association on Remote Sensing, pp. 4442-4447, 34th Asian Conference on Remote Sensing 2013, ACRS 2013, Bali, Indonesia, 13-10-20.

Integration of Extended Kalman filter and neural network for Foot-mounted Pedestrian Navigation. / Tsai, Guang Je; Sih, Yuan Rong; Chu, Hone-Jay; Chiang, Kai-Wei.

34th Asian Conference on Remote Sensing 2013, ACRS 2013. Vol. 5 Asian Association on Remote Sensing, 2013. p. 4442-4447.

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

TY - GEN

T1 - Integration of Extended Kalman filter and neural network for Foot-mounted Pedestrian Navigation

AU - Tsai, Guang Je

AU - Sih, Yuan Rong

AU - Chu, Hone-Jay

AU - Chiang, Kai-Wei

PY - 2013

Y1 - 2013

N2 - Pedestrian Navigation with a Foot- mounted sensor is a common operation in indoor environments. The contribution in this paper is the step length estimation. The step lengths vary continuously according to the walking speed and walking frequency. In the general solution, the step length is modelled as a linear function in the regular gait in previous work, but is not applicable in the varied gait, thus we try to solve this nonlinear problem by using neural network. Furthermore, the estimation of the position of a person based on inertial sensors in Pedestrian Dead-Reckoning (PDR) solution greatly relies on heading calculation, thus, the heading drift reduction process is necessary. In this paper, we replace the compass with vision aiding, and also using other algorithms such like HDR, ZARU, vision-based and ZUPT to reduce the drift in Extended Kalman Filter (EKF) framework. Finally, we display the performance including each algorithms of the reduction methods and neural network.

AB - Pedestrian Navigation with a Foot- mounted sensor is a common operation in indoor environments. The contribution in this paper is the step length estimation. The step lengths vary continuously according to the walking speed and walking frequency. In the general solution, the step length is modelled as a linear function in the regular gait in previous work, but is not applicable in the varied gait, thus we try to solve this nonlinear problem by using neural network. Furthermore, the estimation of the position of a person based on inertial sensors in Pedestrian Dead-Reckoning (PDR) solution greatly relies on heading calculation, thus, the heading drift reduction process is necessary. In this paper, we replace the compass with vision aiding, and also using other algorithms such like HDR, ZARU, vision-based and ZUPT to reduce the drift in Extended Kalman Filter (EKF) framework. Finally, we display the performance including each algorithms of the reduction methods and neural network.

UR - http://www.scopus.com/inward/record.url?scp=84903468717&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84903468717&partnerID=8YFLogxK

M3 - Conference contribution

SN - 9781629939100

VL - 5

SP - 4442

EP - 4447

BT - 34th Asian Conference on Remote Sensing 2013, ACRS 2013

PB - Asian Association on Remote Sensing

ER -

Tsai GJ, Sih YR, Chu H-J, Chiang K-W. Integration of Extended Kalman filter and neural network for Foot-mounted Pedestrian Navigation. In 34th Asian Conference on Remote Sensing 2013, ACRS 2013. Vol. 5. Asian Association on Remote Sensing. 2013. p. 4442-4447